13Th - 14Th September 2017 Sandown Park Racecourse, Esher, Surrey

13th - 14th September 2017 Sandown Park Racecourse, Esher, Surrey

Abstracts

These papers are published with permission of the authors who have granted IPEM a non– exclusive right to publish, reproduce, distribute, display and store the abstract worldwide in all forms (including on the Institute’s website), formats and media now known or as developed in the future, including print, electronic and digital forms. Full copyright of the abstracts remains with the author.

Institute of Physics & Engineering in Medicine Fairmount House 230 Tadcaster Road York YO24 1ES

ISBN 978-1-903613-64-1

2 CONTENTS

Wednesday 13th September 2017 ...... 4 Plenary Session – Woolmer Lecture ...... 4 EFOMP Session ...... 5 Clinical Measurement ...... 9 Cardiovascular Computational Modelling I ...... 14 Student/Associate/Trainee I ...... 19 Clinical Imaging ...... 22 Plenary Session ...... 27 Influencing decision makers: a researcher’s guide ...... 28 Pre Clinical Imaging ...... 29 Cardiovascular Computational Modelling II ...... 32 Student/Associate/Trainee II ...... 37 Image Guided Therapy ...... 41 Making the leap: becoming a medical engineering entrepreneur ...... 44 Musculoskeletal Biomechanics ...... 45 Cells and biocompatible materials ...... 52 Student/Associate/Trainee III ...... 58 Radiotherapy Imaging ...... 65 Thursday 14th September 2017 ...... 72 Statistics Teaching Session ...... 72 Plenary session: IOP MPG sponsored lecture ...... 72 Radiotherapy Advances ...... 73 Virtual Reality Workshop ...... 79 Ultrasound Workshop ...... 79 Author Workshop ...... 79 Research culture: visions of 2035 ...... 80 Diagnostic Radiology ...... 81 Industry session: Future Innovations in Imaging and Radiotherapy ...... 86 Cardiovascular Engineering ...... 88 Medical Robotics ...... 93 Radiotherapy Dosimetry I...... 97 Innovative and Sustainable Healthcare Technology ...... 100 Engineering Design...... 103 Radiotherapy Dosimetry II ...... 106 Professional Session ...... 110 Clinical Engineering ...... 113 Magnetic Resonance ...... 118 POSTERS ...... 124 INDEX OF PRESENTING AUTHORS ...... 160

3 Wednesday 13th September 2017

Wednesday 13th September 2017, 09.45 – 10.30 Plenary Session – Woolmer Lecture

Why do rigid tumours contain soft cancer cells? Franziska Wetzel1,, Anatol W. Fritsch1,, Dapeng Bi7, Roland Stange1, Steve Pawlizak1, Tobias R. Kießling1, Mareike Zink1, Lars-Christian Horn3, Klaus Bendrat4,9, Maja Oktay5, Axel Niendorf4, John Condeelis6, Michael Höckel2, M. Cristina Marchetti8, M. Lisa Manning8, Josef A. Käs1 1Soft Matter Physics Division, Leipzig University, 04103 Leipzig, Germany. 2Department of Obstetrics and Gynecology, Women's and Children's Center, Leipzig University, 04103Leipzig, Germany. 3Division of Breast, Urogenital and Perinatal Pathology, Institute of Pathology, Leipzig University, Germany. 4Pathology Hamburg-West, 22767 Hamburg, Germany. 5Montefiore Medical Center, Bronx, NY 10467, USA. 6Albert Einstein College of Medicine, Bronx, NY 10461, USA. 7Center for Studies in Physics and Biology, Rockefeller University, USA. 8Department of Physics, Syracuse University, USA. 9Institute for Biochemistry and Molecular Cell Biology, University Hospital Hamburg- Eppendorf, Martinstrasse 52, 20246 Hamburg, Germany.

As early as 50 AD, the Roman medical encyclopaedist Celsus recognized that solid tumours are stiffer than surrounding tissue. However, cancer cell lines are softer, which facilitates invasion. This paradox raises several questions: Does softness emerge from adaptation to mechanical and chemical cues in the external microenvironment, or are soft cells already present inside a primary tumour? If the latter, how can cancer tissue be more rigid than normal tissue and yet contain more soft cells? Here we show that in primary samples from patients with mammary and cervical carcinomas, cells do exhibit a broad distribution of rigidities, with a higher fraction of softer and more contractile cells compared to normal tissue. Mechanical modelling based on patient data reveals that tumours with a significant fraction of very soft cells can still remain solid. Moreover, in tissues with the observed distributions of cell stiffness, softer cells spontaneously self-organize into multicellular streams, possibly facilitating cancer metastasis.

4 Wednesday 13th September 2017, 11.00 – 12.00 EFOMP Session

European Examination Board – report on the first initiative 1,2John Damilakis 1EFOMP President, EEB Chair 2Department of Medical Physics, University of Crete, Greece

EFOMP’s Examination Board (EEB) has been established to facilitate the harmonization of Medical Physics education and training standards throughout Europe. EEB introduces the European Diploma of Medical Physics (EDMP) and the European Attestation Certificate to those Medical Physicists that have reached the Medical Physics Expert level (EACMPE). EEB examinations are tests of excellence in Medical Physics. They are designed to assess the knowledge, skills and competences requisite for the delivery of high standard Medical Physics services.

Currently medical physicists in European countries face difficulties in providing the necessary qualification evidence when they seek employment in other EU Member States or other countries. The EDMP will facilitate mobility of medical physicists in Europe and beyond. Furthermore, EEB provides an attestation certificate to those medical physicists that have reached the Medical Physics Expert level to be recognized by the relevant competent authorities of the EU according to the EU Directive 2013/59/EURATOM laying down the basic safety standards for protection against the dangers arising from exposure to ionising radiation (EU BSS). Information about EDMP and EACMPE, examination dates, venue, application forms and other documents can be found in http://efomp.org/index.php/eeb The EEB examinations are voluntary. EEB diplomas will not replace any national certificates. However, they will be a common European qualification for medical physicists and will help to standardise training and expertise in Medical Physics across Europe.

The 1st EEB exams in the field of 'Diagnostic and Interventional Radiology' were organized successfully in Prague, Czech Republic, on the 6th and 7th of July, 2017. Twenty one (21) applications were submitted by the deadline. The number of eligible applications were 18 i.e. 14 for the EACMPE and 4 for the EDMP. Nine (9) candidates passed the EACMPE exams and three (3) candidates passed the EDMP exams. The next EEB exams in all 3 sub-fields of Medical Physics (Diagnostic and Interventional Radiology, Nuclear Medicine, Radiation Oncology) will be held in Copenhagen before the European Congress of Medical Physics (August 2018).

Education and Training initiatives - Eutempe-RX, ESMPE 1Brambilla M, 1Medical Physics, University Hospital of Novara, IT.

Introduction Promoting of education and training programmes is one of the three columns of the mission of EFOMP. To achieve this goal, EFOMP is engaged in a series of activities such as the production of a number of documents (Policy Statements) proposing guidelines for education, training and accreditation programs, workshops, seminars, or the organisation of a European Network of Medical Physics Schools. In the following, school related activities are described with respect to the European School of Medical Physics Experts (ESMPE), and the participation in the Network “European Training and Education for Medical Physics Experts in Radiology” (EUTEMPE-RX) The European School for Medical Physics Experts - ESMPE In partnership with the Czech Association for Medical Physics, EFOMP has organized in 2013 the first module which has been specifically targeted towards Medical Physicists who would like to achieve Medical Physics Expert status in the Nuclear Medicine subspecialty. Since then, eight editions of the school have been completed (4 in Nuclear Medicine, 3 in Diagnostic Radiology and 1 in Radiotherapy) with more than 300 participants coming from all European countries.

Duration Edition Year Title Domain (Hours) Clinical Medical Device Management: Specification, Acceptance testing, Nuclear 1st 2013 19 Commissioning, QC and Advanced applications Medicine in Whole-body PET/CT Advanced Kinetic Modeling and Parametric Methods. Advanced SPECT and PET Nuclear 2nd 2014 19 Applications in Cardiology, Neurology and Medicine Oncology Diagnostic 3rd 2015 Digital Mammography and Quality Controls 18 Radiology Nuclear 4th 2015 Radiopharmaceutical dosimetry 17 Medicine Computed Tomography Imaging: Dosimetry, Diagnostic 5th 2016 Optimization and Advanced Clinical 20 Radiology applications. Practical aspects of Radiation Dosimetry in Nuclear 6th 2016 18 Targeted Radionuclide Medicine Therapy Medicine 7th 2017 Imaging in Radiotherapy Radiotherapy 19 Magnetic resonance Imaging: Advanced clinical Diagnostic 8th 2017 18 applications- Safety Aspects - Quality controls Radiology

The European Training and Education for Medical Physics Experts in Radiology (EUTEMPE- RX) The EUTEMPE-RX is a network of 12 excellent teaching centres which established dedicated courses at EQF level 8, with course content driven by the “Guidelines on the MPE projects” [1], with a balanced e-learning and a face-to face session, both of 40 hours’ workload. The educational tools and methods involves Online communication & feedback between teachers and learners, Interactive self-assessment test Up-to-date multimedia material, Solutions for practical training, Interactive exercises, Databases of Material, Effective sharing of material, software tools and knowledge, Cross-fertilization between course modules. Bibliography [1] EC. (2014 ) RP174 European Guidelines on the Medical Physics Expert

EFOMP Communication activities: New website and social media 1Koutsouveli E 1 Department of Medical Physics, Hygeia Hospital, Athens, Greece EFOMP Internet Manager, Communications and Publications Committee

The European Federation of Organisations for Medical Physics (EFOMP) serves as an umbrella organisation for 34 national societies which together represent more than 8000 medical physicists and clinical engineers working in the field of medical physics. EFOMP communication strategy has as core aim to promote the roles of Medical Physicists through digital communication channels, promotional documents, publications and presence in International events. The Company website serves as one of the organization’s main communication channels to share information about the scientific, educational, professional activities of the organisation and its national members and affiliated international societies. An appealing website is considered as essential for an organisation with a vision to convey messages in an effective and open way not only to the scientific community but also to the wider public. The existing EFOMP website had several drawbacks from the user point of view, such as poor functionality and usability, outdated aesthetics, unorganised content, bad navigation. On September 2017 a new website is scheduled to be launched: simple, fast, easily browsed, with consistent layout. The new website along with the EFOMP social media accounts (LinkedIn, Twitter, Facebook), that were inaugurated earlier in September 2016, will further foster the contribution of Medical Physicists, connect and engage

6 with health professionals, scientists, companies, decision makers, trainees, students, and general public in and outside Europe. Information on EFOMP activities (European Examination Board, European School for Medical Physics Experts, biennial European Congress of Medical Physics), publications (European Journal of Medical Physics, electronic European Medical Physics News, Protocols, Policies, Guidelines) and projects (EURAMED platform, EUTEMPE-RX, BSS Transposition, ENETRAP III, ENEN) are constantly shared through the digital communication channels. EFOMP presence in International events (booths, invited talks, leadership meetings, joint sessions, workshops) is being disseminated through communication platforms and contributes in establishing a two way communication and interaction between all interested parties. EFOMP new website and social media accounts will be further used to facilitate the exchange of knowledge, training, education, best practices, research, development and improve information distribution to the global audience.

Scientific workgroups – Mammo dosimetry, Cone Beam CT protocol, etc 1,2Mika Kortesniemi 1Chair of the EFOMP Science Committee 2HUS Helsinki Medical Imaging Center, Helsinki University Hospital, Finland

One of the aims of the European Federation of Organisations for Medical Physics (EFOMP), as stated in its constitution, is to encourage exchanges between the National Member Organisations and disseminate professional and scientific information through publications and meetings. The purpose of the quality assurance protocol series is to develop a minimum set of easily implemented quality control tests on diagnostic and therapeutic systems that can be used to assure the performance of a system within a set and acceptable range. It is intended that these be implemented as part of the daily routine of medical physicists and system users throughout Europe in a harmonised way so allowing results to be compared. EFOMP pursues to update these protocols every few years taking into consideration the experience gained through their implementation in different settings.

EFOMP Mammo Protocol (2015) is the first of the series on quality control protocols. It proposes tests that can be applied by each country with the resources available locally, accepting the fact that there are different technological levels across the European countries. The main topics of the protocol will be presented, focusing on radiation dose and image quality factors in mammographic systems.

The most recent EFOMP working group document is the Cone Beam CT (CBCT) protocol, currently under publication (2017), presenting procedures for quality control of CBCT systems used for dental, radiotherapy, interventional radiology and guided surgery applications. Publication is the second of the series on quality control protocols. It was created as a collaborative effort with ESRO and IAEA, hence called the EFOMP-ESTRO-IAEA protocol, as a result of the experience and knowledge of an international group of leading medical physics experts. The main topics of the

7 protocol will be presented.

There is also work in progress regarding a new universal breast dosimetry under the EFOMP- AAPM Task Group 282. The development aims to a new model and corresponding methodology to estimate the breast average glandular dose (AGD) from x-ray based image acquisitions including standard mammography, contrast-enhanced mammography, spot mammography, magnification mammography, and breast tomosynthesis.

Big data and deep learning will profoundly change various areas of professions and research in the future. This will also happen in medicine and medical imaging. As physicists, we should pursue beyond the concept of quality, to extend our methodology and competence towards measuring and optimizing the diagnostic value in terms of outcome. Functional implementation requires data utilities starting from data collection and management and culminating in the data analysis methods. In order to make sensible system from the vast heterogeneous big data, data quality control and validation are prerequisites for further analysis, classification, interpretation, probabilistic and predictive modeling. The topic will be touched as a potential future focus area.

Project Activities 1 Tsapaki V. 1EFOMP

The European Federation of Organisations in Medical Physics (EFOMP) is a Company Limited by Guarantee in England and Wales. Consequently EFOMP has the opportunity to participate in European Projects and benefit from the financing of such projects from the European Union (EU) to facilitate the development of Medical Physics as a Science and as a Profession. This activity requires a dedicated team of knowledgeable persons to monitor the announcements of new projects, as well as the preparation of the proposals themselves, their submission to the EU and to involve and support medical physics groups and institutions in the submission of proposals. The Projects Committee is responsible to the Council of the Federation for all these tasks. In this respect, the Projects Committee actively pursues to engage in and support those projects that: 1) improve the status of the medical physicist in Europe and internationally, 2) improve the education and training of the medical physicist, 3) improve the education and training of medical and other healthcare professionals in medical disciplines involving ionising and non-ionising radiation, 4) pursue scientific research that could have a positive impact on healthcare and 5) those that advance the radiological safety of patients, staff and members of the public. Engagement in projects, on behalf of EFOMP, is undertaken by members of the Project Committee, by EFOMP Officers, or past Officers, as well as delegated representatives from the National Member Organizations. Other expert specialists may also be appointed to join a project team, particularly where there is a perceived gap in the experience required of the team to be most effective in pursuance of any project tendering process. It is also recognised that women bring many positive advantages to the workforce such as different insights and values, motivation and methods of working. Positive encouragement for women to work on projects is therefore strongly supported by EFOMP. EFOMP has participated on more than 7 European projects the last years mainly related to education and training such as the European Network on Education and Training in Radiological Protection (ENETRAP III), the Medical Physics Expert (MPE) project or the European Training and Education for Medical Physicists to achieve MPE status in Diagnostic and Interventional Radiology (EUTEMPE-RX) project. The Federation is also a founding member of the European Alliance for Medical Radiation Protection Research (EURAMED) platform. The EURAMED platform has the vision to lead the European research activities in medical radiation protection (RP) and to assume an umbrella function for the harmonisation of practice to advance the European RP safety culture in medicine. EFOMP is also an active member of European Institute for Biomedical Imaging Research (EIBIR). EIBIR is a non- profit organisation that supports biomedical imaging researchers. It offers expert advice along with proposal preparation services, and assumes project coordination, management and dissemination roles in international research projects.

8 Wednesday 13th September 2017, 11.00 – 12.00 Clinical Measurement

Machine Learning for Wearable Physiological Monitoring David A. Clifton, Associate Professor of Engineering Science, Group Leader - Computational Health Informatics (CHI) Laboratory, Fellow of Balliol College, University of Oxord

Wearable healthcare sensors are becoming increasingly well-described in the technical literature, and yet few (if any) systems penetrate for use into clinical practice - primarily because the resulting data, which are prone to artefact and other noise, are insufficiently robust for use in clinical practice. Machine learning techniques based on large-scale probabilistic inference offer a principled means of obtaining robust estimates of physiological data in the presence of the substantial noise associated with wearables - and which can cope with "missingness" of data in a robust manner. This talk describes work at Oxford into the field of machine learning for healthcare sensor informatics, and demonstrates how machine learning can make wearable sensing usable in clinical practice.

Acknowledgements DAC acknowledges the support of the Royal Academy of Engineering; Balliol College, Oxford; NIHR, EPSRC, NERC, DfID, and the Wellcome Trust.

Shape Analysis of Patients' Arms Suffering from Unilateral Lymphoedema Kevork Karakashian (1) Lawrence Shaban (1) Cheryl Pike (2) Raoul Van Loon (1) Swansea University, Swansea, UK (1) Lymphoedema Network Wales, Neath, UK (2) Background: Lymphoedema is a condition in which a patient suffers from chronic swelling of subcutaneous tissue due to deficient drainage of interstitial fluid3. Due to the lack of information and tools available to analyse shape of oedemic limbs, the main objectives of this study were to 1) Examine the use of 3D cameras to scan upper limbs and capture shape information. 2) Identify different shape related metrics that could be used to monitor/diagnose lymphoedema. Other studies that have used 3D cameras were merely interested in finding accurate volume measurements without analysing shape further in depth1,2,4,5. Methods: After acquiring ethical approval, upper limbs of 24 patients were scanned using the ASUS Xtion Pro 3D camera, producing triangulated 2D manifolds of their arms. Results: Results below are visualised on the scans after removing artifacts and post-processing in MATLAB.

(a) (b) (c) Figure 1: a) Volume error calculation (ml) for all 24 patients when comparing disc volume method vs more accurate integral method (divergence theorem). b) Circumference measurements differentiated between healthy and affected arms across the orange shaded region, while circularity measurements in regions shaded with red and orange (p=0.05, t-test) c) Relative radial changes between healthy and affected arm displaced as a θ-z graph (swelling map).

9 Discussion: Circular geometric formulae used in clinical settings, such as disc method, overestimated limb volume by about 100 ml, with high errors at regions of least circularity (Fig. 1a). The circularity metric provided the ability to localise positions across the arm most sensitive to oedema i.e. elbow and wrist (Fig. 1b). Lastly, figure 1c clearly shows that the swelling is not homogeneously distributed along the arm, and can be used to target localised oedema. Conclusion: A fast method was developed for analysing the shape of oedemic arms that gives complimentary information to the current metrics such as volume. The shape related metrics could be used to monitor, or even diagnose the incidence of lymphoedema. Key references: [1] Buffa, R., E. Mereu, P. Lussu, V. Succa, T. Pisanu, F. Buffa, and E. Marini. A New, Effective and Low-Cost Three-Dimensional Approach for the Estimation of Upper-Limb Volume. Sensors 15:12342-12357, 2015. [2] Hoevenaren, I. A., A. Verhulst, M. Hameeteman, R. Vreeken, T. J. Maal, and D. Ulrich. Three-dimensional stereophotogrammetry as an accurate tool for analyzing lymphedema of the hand. JPRAS Open 10:40{46, 2016. [3] Lawenda, B. D., T. E. Mondry, and P. A. S. Johnstone. Lymphedema: a primer on the identication and management of a chronic condition in oncologic treatment. CA Cancer J. Clin. 59:8-24, 2009. [4] Lu, G., G. N. DeSouza, J. Armer, B. Anderson, and C.-R. Shyu. A system for limb volume measurement using 3D models from an infrared depth sensor. In: 2013 IEEE 15th International Conference on e-Health Networking, Applications and Services, pp. 64-69. [5] Öhberg, F., A. Zachrisson, and Å. Holmner-Rocklöv. Three-dimensional camera system for measuring arm volume in women with lymphedema following breast cancer treatment. Lymphat. Res. Biol. 12:267-274, 2014.

Measurements of water content in ex-vivo porcine skin using initial design of pulsed LED and Transimpedance amplifier system Meha Qassem (1) Panayiotis Kyriacou (1) City University of London, London, UK (1) Background. The current standard measurement of skin hydration is based on electrical capacitance probes, but these suffer certain drawbacks as a result of their indirect relationship with dermal water content and need of controlled working environments [1] . Alternatively, we have been using Near Infrared (NIR) spectroscopy to analyse skin hydration and have suggested the possible capability of NIR measurements in detecting changes in skin barrier function following treatments [2], [3]. The objective of this study was to explore the possibility of using simple circuitry incorporating LEDs and photodiode, designed to work in the NIR region, as a custom-made device to measure skin moisture against a standard spectrophotometer using measurements obtained during a desorption test. Methods. Custom made circuitry, based on monochromatic LEDs of NIR light of centre wavelengths at 1450 nm, 1850 nm and 2350 nm, and an InGaAs photodiode, were used in a light source transmission and transimpedance amplifier configuration. The device was used to determine water content in porcine skin during desorption tests, and compared against a benchtop high end commercial spectrophotometer. The selection of these wavelengths was based on previous spectral measurements of porcine skin during a desorption test which highlighted the bands of significance and the changes expected at these particular bands as a result of dehydration. Weight measurements were included and used as a reference method against both techniques. The correlation and agreement of circuit outputs, first in relation to spectral measurements, then to weight readings were investigated. Results. A strong correlation was found between outputs of the custom device and spectral absorbance i.e. r = -.933, p = < .000 at 1450 nm, r = -.939, p = < .000 around 1900 nm, and r = -.946, p = < .000 at 2310 nm. This was true at each selected wavelength, as well as with gravimetric measurements. Furthermore, percentages of water Figure 1. Bland-Altman plots of % water content at content calculated using custom measurements 1460 and 1900 nm bands. Differences between were found to be in agreement with percentages weight and spectral measurements (top), and weight calculated using spectral and gravimetric and voltage measurements (bottom). techniques. Discussion. Beyond these promising results, efforts will be directed to multiplexing the operation of the LEDs, and building the appropriate algorithms for estimation of skin moisture as well as a calibration formula for resetting the instrument. This can be done by repeating desorption tests using gravimetric, and both commercial and in-house optical systems. Conclusion. Multi-wavelength optical sensing and instrumentation can be used to determine skin moisture accurately, and can be optimised to work in in vivo mode. This will allow measurements to be made in a non-invasive and rapid manner. Key references [1] E. Berardesca and European Group for Efficacy Measurements on Cosmetics and Other Topical Products (EEMCO), ‘EEMCO guidance for the assessment of stratum corneum hydration: electrical methods’, Skin Res. Technol., vol. 3, no. 2, pp. 126–132, 1997. [2] M. Qassem and P. A. Kyriacou, ‘Reflectance near-infrared measurements for determining changes in skin barrier function and scattering in relation to moisturizer application’, J. Biomed. Opt., vol. 20, no. 9, pp. 095008–095008, 2015. [3] M. Qassem and P. Kyriacou, ‘Use of reflectance near-infrared spectroscopy to investigate the effects of daily moisturizer application on skin optical response and barrier function’, J. Biomed. Opt., vol. 19, no. 8, pp. 087007–087007, 2014.

Opto-impedance measurements of blood lithium levels for patients with bipolar disorder Loukas Constantinou (1) Meha Qassem (1) Iasonas Triantis (1) Panayiotis Kyriacou (1) City University of London, London, UK (1) Background. Bipolar Disorder (BD) is a common mental disorder, characterized by alternating episodes of mania and depression, with life-threatening risks to the patient [1,2]. Suffers of BD require long-term treatment plans to manage the condition, and in most cases, this involves regular intake of mood-stabilising drugs such as lithium. Lithium preparations are considered the most reliable mood stabilisers for patients with BD, and the most effective at reducing the risk of suicide [3]. However, maintaining blood lithium concentration within the narrow therapeutic range of 0.5-1.0 mEq is crucial. This is because toxicity can occur at slightly elevated concentrations (>1.5 mEq), leading to serious complications or even death [4]. The aim of this work is to develop a personal lithium blood level analyser for Bipolar patients taking lithium medication. Methods. Using a custom-made tetrapolar probe for measurements of electrical impedance, and a spectrophotometric technique, opto-impedance measurements were conducted on samples of blood (plasma) which had been spiked with lithium carbonate to make up various concentrations between 0.6-2.0 mEq. Flame photometry measurements were used as reference. Multiple Linear Regression (MLR) was used to analyse acquired data, and compare the results of combined versus optical-only measurements. Results. MLR analysis of feature optical wavelengths (599 & 546 nm) resulted in a model with cross-validated R^2 value of 0.0121342 and Root Mean Square Error RMSE of 0.360189. In the case of combined opto-impedance analysis, it was possible to predict the concentration of lithium with an accuracy of >80% (R^2 CV: 0.852395 and RMSEC: 0.105129). Figure 1. Calibration curves of optical (A), and combined (B) measurements The calibration curve resulting from both models is shown in Fig 1. Discussion. Results are most complementary of combined electrical impedance and optical evaluations, and there is clear benefit in providing unmatched accuracy and sensitivity in measurements of lithium and perhaps the possibility of other analytes. Conclusion. Future work will now focus on the technical development of an integrated and miniaturized system for measurement of lithium levels in blood with an acceptable level of accuracy and sensitivity. Key references. In alphabetical order, numbered. [1] R. H. Belmaker, ‘Bipolar Disorder’, N. Engl. J. Med., vol. 351, no. 5, pp. 476–486, Jul. 2004. [2] I. M. Anderson, P. M. Haddad, and J. Scott, ‘Bipolar disorder’, BMJ, vol. 345, p. e8508, Dec. 2012. [3] R. W. Licht, ‘Lithium: still a major option in the management of bipolar disorder’, CNS Neurosci. Ther., vol. 18, no. 3, pp. 219–226, Mar. 2012. [4] R. Oruch, M. A. Elderbi, H. A. Khattab, I. F. Pryme, and A. Lund, ‘Lithium: a review of pharmacology, clinical uses, and toxicity’, Eur. J. Pharmacol., vol. 740, pp. 464–473, Oct. 2014.

12 Pulsatile Monte Carlo model for Reflectance Pulse Oximetry Subhasri Chatterjee (1) Justin Paul Phillips (1) Panayiotis Kyriacou (1) City, University of London, London, UK (1) Background: Pulse oximetry (PO), a non-invasive method for estimation of arterial oxygen saturation from photoplethysmographic signals, has become a standard of care in anaesthesia and related medical specialties presently [1,3]. Despite its wide ranges of applications, the precise nature of light-tissue interaction underlying the technique is not yet fully understood. Information on optical path distribution and reflectance is crucial for more accurate, precise and advance PO applications, as mentioned in several studies [1,2,6]. To address these requirements, interaction of light with a pulsatile monolayer slab of perfused dermal skin tissue was simulated using one of the most reliable and convenient computational modelling approach, Monte Carlo technique (MC) [5]. Method: The steps of MC algorithm and the tissue optical-properties at two PO wavelengths 660 (r) and 940 nm (ir) used in the present model are described in our previous work [2]. The ‘pulse’ in this model was achieved by increasing the diastolic (dia) arterial blood volume, i.e. 5% in systole (sys) by 2%. The venous blood volume, i.e. 5% was assumed to be unaffected by the pulse, with a venous oxygen saturation lesser by 10% than SaO2. It was considered that the slab thickness was 1 cm and source-detector separation was 4 mm. Gaussian beam of 1 mm radius was made incident to the model. The detector diameter was 1 mm. An in-built MC program written in MATLAB was executed to find out the optical path in PO at different oxygen saturation (SaO2=10-100%) and produce the calibration curve [4] by calculating the ‘ratio of ratios’ [ RIIII(1 sys / dia |)/(1 r  sys / dia |) ir ]. Results and Discussion: It is shown in Fig.1(a) that mean optical path of red light is greater than infrared for all SaO2. Also, diastolic optical path is always greater than systolic optical path. Thus, in theory, the assumption normally made that optical paths are equal for different wavelengths and diastole-systole is not appropriate. Fig.1(b) shows the decreasing ratio of ratios with increasing SaO2 which is expected and agrees with the empirical calibration curves [6].

Fig.1. Mean optical path (a) and Ratio of ratios (b) with arterial oxygen saturation (10-100%) Conclusion: A pulsatile monolayer Monte Carlo model describes the optical path differences at different perfusion states and oxygen saturation, raising concerns regarding the accuracy of pulse oximetry readings and their interpretation in the clinical setting. The present work also shows a way to measure R for very low values of SaO2, which is a well-known limitation of PO that is attributed to ethical constraints in obtaining such data in humans. Future work shall be carried out with multilayer multilayer tissue model. Such studies are invaluable for basic understanding and technological advancement in the field of non-invasive physiological monitoring and measurement. References: [1] Abay T Y and Kyriacou P A 2015 Reflectance photoplethysmography as non-invasive monitoring of tissue blood perfusion IEEE Trans. Biomed. Eng. 62 2187-95. [2] Chatterjee S et al 2016 Monte Carlo investigation of the effect of blood volume and oxygen saturation on optical path in reflectance pulse oximetry Biomed. Phys. Eng. Express 2 065018. [3] Kyriacou P A 2006 Pulse Oximetry in the oesophagus Physiol. Meas. 27 R1-R35. [4] Reuss J L 2005 Multilayer Modeling of Reflectance Pulse Oximetry IEEE Trans. Biomed. Eng. 52 153-159. [5] Wang L et al. 1995MCML-Monte Carlo Modeling of light transport in multi-layered tissues Comput. Methods Prog. Biomed. 47 131–46. [6] Webster J G (ed) 1997 Design of Pulse Oximeters (Medical Science Series) 1st edn (London: Taylor and Francis).

13 Wednesday 13th September 2017, 11.00 – 12.30 Cardiovascular Computational Modelling I

Understanding Heart Tissue through Waves David Nordsletten, King’s College London

Personalised models of cardiac mechanics have evolved into a powerful tool for studying the human heart in health and disease. Combining detailed information on the heart kinematics extracted from medical images, with mathematical models of cardiac function, patient-specific models provide a mathematical representation of individual hearts. As model parameters are linked to intrinsic tissue properties such as stiffness and contractility, unique and accurate parameter estimates are a prerequisite for the potential translation of personalised models to the clinic.

In parallel, magnetic resonance elastography (MRE) has evolved into a powerful tool for interrogating material stiffness. MRE has been exploited in the liver, breast, and brain using measured periodic waves to extract material stiffness properties. Transducing waves via external vibrations, MRE provides a more direct measure of the characteristics of tissues and their potential diseases. Integrating this work into the heart is complicated by a myriad of challenges including the inherent cardiac motion of the heart, evolving material properties due to the contractile state of the muscle, and the nonlinear effects of deformation on the apparent stiffness of the material.

In this talk, we discuss the merger of these two worlds, bridging between the nearly quasi-static model-based assessment of heart function and the high frequency wave-based assessment. In particular, we discuss the simple influence of deformation on apparent stiffness, explaining both from the realm of traditional wave mechanics and biomechanical theory.

A finite element model to optimise the fatigue life of a new polymeric heart valve prosthesis Marta Serrani(1) Eugenia Biral(1) Joanna Stasiak(1) Geoff Moggridge(1) University of Cambridge, Cambridge, UK (1) Background. In the literature, many studies have reported promising short-term outcomes for polymeric heart valve prostheses, typically made of polyurethane-based polymers; nevertheless, none of these prototypes reached a clinical evaluation stage due to the lack of long-term reliability [2]. Our group is currently developing an aortic polymeric heart valve made of a styrenic block copolymer [1]; orienting the cylindrical microstructure of this polymer along the most stressed direction allows enhancement of the valve hydrodynamics and lifetime. Within this context, predicting the valve fatigue life for different designs, materials and microstructures is crucial for the development of a reliable polymeric heart valve. The aim of this work is the implementation of a modelling tool which, combined with experimental data on the polymer fatigue properties, allows prediction of valve lifetime and optimisation of the design.

Methods. A finite element model of the polymeric valve was developed. The block copolymer mechanical behaviour was described by an anisotropic hyperelastic constitutive law and the material parameters estimated by uniaxial tensile tests [3]. The material orientation, which is mainly circumferential within the valve leaflets, was defined by means of a custom MATLAB routine. Physiological pressure and boundary conditions were applied to simulate the valve closing, since during this phase the leaflets must bear the biggest load. Different valve designs were simulated; also, the effect of inaccuracies in valve manufacture (e.g. non-uniform leaflet thickness) on its fatigue performance was considered. To investigate the fatigue behaviour of the material, crack propagation tests were performed on polymeric planar samples. The experimental data were post- processed to find the relationship between number of cycles to failure and Strain Energy Density (SED) for the polymer. In fact, the SED is one of the most widely used criterion to predict elastomers’ lifetimes. To validate this approach, valve prototypes were manufactured by injection moulding and tested at 20 Hz in a commercial fatigue tester until failure.

14 Results. The crack propagation tests allowed to find the number of cycles to failure at different SED for the block copolymer, while the computational model was used to find the value and the location of the maximum SED for each prototype. Combining these two results, it was possible to predict the valve lifetime for different designs (Fig.1). 1.00E+09

1.00E+08 Fig. 1. Predicted (squares) and 1.00E+07 experimental (circles) lifetime of polymeric valves with different 1.00E+06 designs. The red line represents Cycles the minimum durability specified in 1.00E+05 the ISO 5840, while the blue line 1.00E+04 was obtained by crack growth 1.00E-02 1.00E-01 1.00E+00 tests. SED [MPa]

Discussion. The comparison between predicted and experimental durability showed good agreement; in particular, the predictions were closer to the real fatigue life when manufacturing inaccuracies were taken into account.

Conclusion. The approach presented has proven to be a powerful tool to improve the design and manufacture of a new polymeric heart valve prosthesis and, more in general, of polymeric devices subjected to cyclic loads.

Key references. 1. De Gaetano et al, Int J Artif Organs, 2015, 38(11): 600–606 2. Kheradvar et al, Ann Biomed Eng, 2015, 43(4):833-43 3. Serrani et al, J Biomech Eng, 2016, 138(6):061001

Identifying haemodynamic determinants of pulse pressure: an integrated numerical and physiological approach Samuel Vennin (1) Ye Li (1) Marie Willemet (1) Henry Fok (1) Haotian Gu (1) Jordi Alastruey (1) Phil Chowienczyk (1) King's College London, London, UK (1) Background The focus of numerical modelling in understanding the haemodynamics of hypertension has been largely to develop ever more detailed distributed models of the circulation [1]. Although these models provide in-depth understanding of how pressure-flow relationships can potentially be influenced by the arterial tree, they are of limited value when seeking to determine to what extent a given blood pressure (BP) components depends upon properties of the heart, proximal and peripheral arteries. Methods We examined the ability of a simple “reduced” model comprising a proximal characteristic impedance linked to a Windkessel element [2] to accurately predict central pulse pressure from aortic blood flow, verified that parameters of the model corresponded to physical properties and applied the model to examine pulse pressure dependence on cardiac and vascular properties. Pulse pressure obtained from the reduced model was compared with theoretical values obtained using a distributed multi-segment model in a “virtual population” of subjects in which cardiovascular properties were varied over the pathophysiological range (n=3,095), in normotensive subjects during modulation of physiology with vasoactive drugs with divergent actions on cardiac and vascular properties (n=13, 10 men, age 47±10 years, BP 111±16/69±9 mmHg, mean±SD) and in hypertensive subjects (n=156, 83 men, age 46±17 years, BP 130±23/83±13 mmHg). Results Central pulse pressure derived from the reduced model agreed with theoretical values (mean difference±SD, -0.1±1.9 mmHg) and with measured values (mean difference±SD, -1.9±3.7 and -

15 1.1±3.7 mmHg for normotensive and hypertensive subjects respectively). Parameters extracted from the reduced model agreed closely with theoretical and measured physical properties. Central pulse pressure was seen to be determined mainly by total arterial compliance (inversely associated with central arterial stiffness (PWV)) and ventricular dynamics: the volume of blood ejected up to time of pulse pressure and the rate of ventricular ejection up to this time point. Increased flow and/or volume accounted for 20.3 mmHg (56%) of the difference in pulse pressure between the upper and lower tertiles of the hypertensive subjects. Discussion That such a simple model can explain pressure-flow relationships without invoking the complexities of pressure wave propagation and reflection can be explained by backward pressure waves maintaining an approximately constant relationship to forward waves [3]. Conclusion We present a simple model, verified numerically and against extensive physiological data that defines the relationship central aortic pulse pressure, arterial properties and ventricular dynamics. Total arterial compliance, closely related to aortic PWV, dominates in determining the contribution of arterial tree, and stroke volume in the contribution of ventricular dynamics but there is an important component of rate of ventricular ejection that influences pulse pressure. Ventricular dynamics account for a relatively large proportion of the increased pulse pressure in hypertension. Key references [1] Reymond, Philippe, et al. "Validation of a patient-specific one-dimensional model of the systemic arterial tree." American Journal of Physiology-Heart and Circulatory Physiology 301.3 (2011): H1173-H1182. [2] Westerhof, Nico, Jan-Willem Lankhaar, and Berend E. Westerhof. "The arterial windkessel." Medical & biological engineering & computing 47.2 (2009): 131-141. [3] Li, Ye, et al. "Forward and Backward Pressure Waveform Morphology in Hypertension." Hypertension 69.2 (2017): 375-381.

Modelling Haemodynamics and Thrombosis in Cerebral Aneurysms Treated with Flow- Diverter Devices Thomas Peach (1) Malebogo Ngoepe (2) Katerina Spranger (1) Yiannis Ventikos (1) University College London, London, UK (1) University of Cape Town, Cape Town, South Africa (2) Background: Cerebral aneurysms are a local weakening and ballooning of the arterial wall in the vasculature of the brain. Growth and rupture of such lesions carries a mortality rate of around 33% and morbidity rate of a further 33%, often resulting in severe stroke and significant disability.[2] An increasingly popular minimally invasive treatment for the management of cerebral aneurysms is the flow-diverter (FD): a collapsible stent placed in the affected artery, covering the aneurysm neck and leading to aneurysm thrombosis and stabilisation.[3] Improvements in computational efficiency have advanced modelling of virtual flow-diverter treatment to a speed and standard now relevant to the clinical setting.[4] In this study a complete modelling pathway is presented from clinical imaging through to virtual intervention and prediction of treatment outcome. The method is presented in the context of bifurcation aneurysm geometries; as such cases present a unique challenge to clinical decision-making. Methods: Segmentation of angiographic patient imaging is completed in Osirix(v.8.0.2, Freeware) with the 3D aneurysm geometry and local vasculature extracted. Virtual device deployment is performed with a fast method developed in Matlab(MathWorks, Natick, MA) and employing a spring-mass approach to modelling device strut stiffness with a rigid vessel wall.[1] The haemodynamics of the aneurysm and surrounding vasculature are computed in the CFD package CFD-ACE+(ESI Group, Paris) assuming a rigid vessel wall and Newtonian blood model. The underlying biochemistry of the environment is modelled assuming a Wagenvoord cascade[6], which accounts for the unsteady, three-dimensional transport of proteins involved in the process of coagulation. Tissue factor is considered the sole-initiator of thrombus development and is expressed at low shear rate.[5] Thrombus is modelled as a porous zone that alters the local flow pattern in the aneurysm dome leading to progressive growth of the thrombus region. Results: Reductions in aneurysm inflow rate and Wall Shear Stress (WSS) before and after virtual

16 treatment by FD are compared for each aneurysm case. The initiation site and rate of thrombus formation between different treatment scenarios are also considered. Discussion: Virtual treatment of bifurcation aneurysms with flow-diverter devices shows variation in performance dependent upon device positioning in the (approximately symmetric) vessel geometry alone. Non-trivial variations in haemodynamic properties including aneurysm inflow rate, Wall Shear Stress (WSS), and thrombus formation are seen and may have significant clinical implications for treatment choice and device placement. Conclusion: This study presents a computational pipeline capable of offering clinically relevant insight for virtual treatment planning for FD treatment of bifurcation aneurysms in the cerebral vasculature. Many observed phenomena are non-trivial and arise from subtle haemodynamic changes that can only be elucidated with a computational model. Key References [1] Blom, F. Considerations on the Spring Analogy, Int. J. Numer. Meth. Fluids 2000; 32:647–68 [2] Brisman, J. et al. Cerebral Aneurysms, N Engl J Med 2006; 355:928-39 [3] Byrne, J. & I. Szikora. Flow Diverters in the Management of Intracranial Aneurysms: A Review, EJMINT 2012; 1225000057 [4] Chung, B. & J. Cebral. CFD for Evaluation and Treatment Planning of Aneurysms: Review of Proposed Clinical Uses and Their Challenges, Ann Biomed Eng 2015; 43(1):122-38 [5] Ouared, R. et al. Thrombosis Modeling in Intracranial Aneurysms: a Lattice Boltzmann Numerical Algorithm, Comput. Phys. Commun 2008; 179(1–3):128–131 [6] Wagenvoord, R. et al. The Limits of Simulation of the Clotting System, J Thromb Haemost 2006; 4(6):1331–38

17 The Effects of Material Properties and Transmural Pressures on Lymphatic Valve Sufficiency Daniel Watson (1) Igor Sazonov (1) David Zawieja (3) James Moore Jr. (2) Raoul van Loon (1) Swansea University, Wales, UK (1) Imperial College, London, UK (2) Texas A&M, Texas, USA (3) Background. The lymphatic system plays an important role in maintaining fluid homeostasis. Fluid is transported from the interstitial through a series of valved compliant structures called lymphangions to the venous system. Valve function is essential to fluid transport, lymphatic insufficiency leads to an accumulation of fluid within the tissues called lymphedema. Attempts to develop a lumped model for the lymphatic system have progressed in recent times[1][3]. Finite element analysis of idealised valve geometries is currently based on assumptions made concerning material properties have yet to be validated experimentally[5]. However, experimental understanding of lymphangions is complicated by their small size. This work presents a finite- element analysis of a subject-specific valve geometry. The sensitivity of assumptions in idealised modelled are assessed against existing experimental data[2][4]. Methods. A lymphangion taken from the mesentery of a rat was subjected to confocal imaging and reconstructed to produce a geometry. Experimental data was used to recreate representative wall motion through an elastic-plastic model. The lymphatic leaflets were represented as a series of neo-Hookean shells. An increasing pressure was applied in a steady-state analysis until the valve had closed and compared to a rigid wall model. This process was repeated for various valve stiffnesses. Results.

The effect of initial shear modulus on the pressure required to close a subject-specific lymphatic valve with flexible wall the black lines represent experimental data found in [2]. Discussion. The presence of a flexible wall has a detrimental affect on valve closure. Representative shear moduli would appear to lie between 10 and 20KPa. Conclusion. The appearance of torsional motion during valve closure should be reviewed further as these aspects are not presently included in modelling. Current shear modulus assumption of 45- 50KPa should be re-evaluated[5]. Key references. [1] C. D. Bertram, C. Macaskill, M. J. Davis, J. E. Moore Jr., Development of a model of a multi- lymphangion lymphatic vessel incorporating realistic and measured parameter values, Biomechanics and Modeling in Mechanobiology 13 (2) (2013) 401–416. [2] M. J. Davis, E. Rahbar, A. A. Gashev, D. C. Zawieja, J. E. Moore Jr., Determinants of valve gating in collecting lymphatic vessels from rat mesentery, American Journal of Physiology - Heart and Circulatory Physiology 301 (1) (2011) 48–60. [3] S. Jamalian, C. D. Bertram, W. J. Richardson, J. E. Moore Jr, Parameter sensitivity analysis of a lumped-parameter model of a chain of lymphangions in series, American Journal of Physiology - Heart and Circulatory Physiology 305 (12) (2013) H1709–H1717. [4] E. Rahbar, J. Weimer, H. Gibbs, A. T. Yeh, C. D. Bertram, M. J. Davis, M. A. Hill, D. C. Zawieja, J. E. Moore Jr., Passive pressure diameter relationship and structural composition of rat mesenteric lymphangions, Lymphatic Re- search and Biology 4 (10) (2012) 152–163. [5] J. T. Wilson, R. van Loon, W. Wang, D. C. Zawieja, J. E. Moore Jr., Determining the combined effect of the lymphatic valve leaflets and sinus on resistance to forward flow, Journal of Biomechanics 48 (13) (2015) 3584 – 3590.

18 Wednesday 13th September 2017, 11.00 – 12.30 Student/Associate/Trainee I

Model-based Image Analysis of Left Atrial Wall Thickness Using Direct Measurement from CT Lauren-Emma Blake (1) Jonathan M. Behar (1,2) R. James Housden (1) Ronak Rajani (1,2) Kawal Rhode (1) Rashed Karim (1) Division of Imaging Sciences and Biomedical Engineering, King's College London, London, UK (1) Department of Cardiology, Guy's and St. Thomas' NHS Foundation Trust, London, UK (2) Background. Atrial fibrillation (AFib) is the most common cardiac arrhythmia globally, affecting 1.0-1.5% of the general population. Changes in atrial architecture, including wall thickening, are often seen in AFib. However it is poorly understood whether these changes predispose to fibrillation development, occur as part of the disease process or are correlated to other factors such as age. Atrial ablation is used to treat drug resistant or refractory AFib. To increase procedure safety and efficiency, ablation energy can be titrated: increasing in thicker regions where ectopic depolarisation is more prevalent, reducing for thinner areas to reduce the likelihood of complications e.g. perforation. The aim of this study was to use advanced image analysis to compute and construct a left atrial wall thickness atlas for predicting wall thickness in patients undergoing ablation procedures. As part of this, a wall thickness investigation was carried out using a novel approach of high density data. Methods. Patient cardiac CT datasets (n=10 aged 43-77) were first delineated using semi- automatic segmentation. A maximum intensity projection was then used to map patient-specific atrial thickness to 3D anatomical shells. From this, an atlas was created from an average of patient-specific wall thickness meshes fused using surface-to-surface registration to a mean left atrium. Validation was then carried out using the atlas: leave-one-out cross-validation to predict thickness on new and unseen patient data, and comparison with randomly-generated wall thickness to investigate whether the atlas’ prediction improved above chance. Results. A 3D left atrial wall thickness atlas was created (Figure 1). A statistically significant difference was found between the predictive ability of the mean atlas compared to randomly- generated thickness on unseen patient data (p<0.0001). Patient anatomical shells created as part of the atlas procedure were divided into four regions for measurement, following protocols found in literature. Average thickness values were then determined for each region, making direct measurements from CT: anterior 1.90 +0.41mm, posterior 1.77 +0.51mm, posterior-superior 1.84 +0.59mm and posterior-inferior 1.60 +0.40mm. Figure 1: Created 3D left atrial wall thickness atlas Discussion. The constructed wall thickness atlas has potential to be used to estimate left atrial wall thickness. The main source of error came from the surface-to-surface registration of the patient anatomical shell to the mean left atrial shell. Thickness measurements were comparable with figures from recent literature. No statistically significant difference was found between anterior and posterior wall thickness, however the cohort is too small to extrapolate this result to a larger population. Conclusion. A left atrial wall thickness atlas was successfully constructed, and a novel method presented to measure 3D wall thickness, providing results that were comparable with literature. A robust workflow was created, allowing future repetition of the study with a larger cohort to further validate the results. Key references. 1. Beinart et al. Left atrial wall thickness variability measured by CT scans in patients undergoing pulmonary vein isolation. Journal of Cardiovascular Electrophysiology. 22 (11), pp.1232-1236 (2011) 2. Hayashi et al. Left atrial wall thickness and outcomes of catheter ablation for atrial fibrillation in patients with hypertrophic cardiomyopathy. Journal of Interventional Cardiac Electrophysiology. 40 (2), pp.153-160 (2014) 3. Karim et al. Segmentation Challenge on the Quantification of Left Atrial Wall Thickness. Unpublished conference paper at: Segmentation of Left Atrial Wall Thickness Challenge Statistical

19 and Computational Modelling of the Heart (STACOM), October 17, 2016, Athens, Greece 4. Lip et al. Atrial fibrillation – the growing epidemic. Heart. 93 (5), pp.542-543 (2007) 5. Whitaker et al. The role of myocardial wall thickness in atrial. Eurospace. Available from: http://europace.oxfordjournals.org/content/early/2016/05/30/europace.euw014.long (2016)

Technique for measurement of dose area product in panoramic dental radiography Nadia Latif (1) Stephen McCallum (1) NHS Grampian, Aberdeen, Scotland, UK (1) Background. The radiation output for dental panoramic examination is measured as dose-area product (DAP). There are challenges in accurately measuring radiation doses due to the number of uncertainties in the commonly used methods of DAP measurement [1, 2, 3, 4, 6]. The aims of this study were to propose a quick and robust method of DAP measurement by using a DAP meter; to compare the results of the proposed method with that using an in-beam solid state detector and film; and to suggest the size of ionization chamber of DAP meter. Methods. Tests were performed on 15 panoramic units to evaluate the difference in the two DAP measurement techniques. The dose measurement techniques used in this study were: a. DAP meter using an ionisation chamber (PTW Diamentor E2, chamber measuring area 141mm x 141mm). b. A solid state detector (Raysafe X2 solid state detector) and film (Gafchromic XR QA2) Further tests were also performed by exposing the film at the tube output to find size of ionisation chamber of DAP meter that can be used on all the panoramic machines.

OPT Ceph

a.set up - DAP meter b.set up-detector c.adult and child beam size d. beam size at and shape at sec collimator prim. collimator Results. DAP The height of beam at tube output 2 Size Min Max Mean Median 1st quartile 3rd quartile (mGycm ) for panoramic examination was Adult 19 84 58 60 49 70 measured as 43mm and for Technique A Child 16 56 40 46 32 47 cephalometric examination 22mm Adult 18 80 51 46 43 66 (figure d) for largest size detector Technique B Child 13 56 34 36 32 42 with active area height of 148mm. Discussion. The results obtained with these two techniques for third quartile values agreed within + 6% for adult patient and within + 11% for child patient. The DAP meter method measured higher DAP for most of the units and the detector-film method has higher discrepancies due to misalignment of detector and beam area measurement errors. Five out of 15 units tested during this study had v shaped beam. Indirect methods of measuring DAP suggested by previous studies are valid for rectangle shaped beam only [1, 2, 3, 4, 6]. Previously we assumed a uniform rectangular beam shape which has resulted in error of up to 45% if the beam was in fact V shaped. Conclusion. The technique using DAP meter was found to be quick, simple and most reliable. The use of DAP meter is proposed for accurate and immediate measurement. Due to small beam area at the tube output, the DAP meter that is used for checking calibration of other radiology equipment can be used. Key references. 1. Doyle P et al. Techniques for measurement of dose width product in panoramic dental radiography. British Journal of Radiology 2006; 79:142-7. 2. Gulson AD et al. Doses to patients from dental X-ray examinations in the UK, 200-2004. Health Protection Agency, 2007; HPA-RPD-022. 3. Hart D et al. Doses to patients from radiographic and fluoroscopic X-ray imaging procedures in

20 the UK- 2010 review. Protection Agency 2012; HPA-CRCE-034. 4. Holy J R. The measurement of X-ray beam size from dental panoramic radiography equipment. Health Protection Agency 2012; HPA-CRCE-032, ISBN 978-0-85951-713-3. 5. Tierris CE et al. Dose area product reference levels in dental panoramic radiology. Radiation Protection Dosimetry 2004;111:283–7 6. William JR and Montgomery A. Measurement of dose in panoramic dental radiology. British Journal of Radiology 2000; 73:1002-6.

In vitro assessment of monocytes to macrophages differentiation using macromolecular crowding Adil Rashid (1) Pramod Kumar (1) Amir Ghaemmaghami (1) University of Nottingham, Nottingham, UK (1)

Background. Monocytes monitor changes in their environment. Upon tissue damage, monocytes are rapidly recruited to the site where they can differentiate into tissue macrophages. Macrophages play a central role in inflammation and restoring homeostasis. Although macrophages secrete matrix metalloproteinases that breakdown the extracellular matrix (ECM), they also secrete ECM components such as collagen. ECM components facilitate monocyte to macrophage differentiation and aid tissue repair. In vivo, the extracellular microenvironments of monocytes are occupied by high concentrations of macromolecules, termed macromolecular crowding (MMC). In vitro, MMC can be mimicked using synthetic macromolecules at levels that correspond to the physiological fractional volume occupancy. Recently, synthetic macromolecules used in culture medium have been shown to significantly accelerate ECM deposition; thus, increasing the physiological relevance of the in vitro microenvironment. This paper aims to assess the impact of MMC on monocyte to macrophage differentiation and ECM deposition. Methods. THP-1 cells, used as a surrogate for monocytes, were stimulated with PMA to differentiate into macrophages in the presence of the synthetic macromolecules, Ficoll and Dextran Sulphate. An alamarBlue®Assay, Sirius Red staining, Microscopy and Flow Cytometry were used to assess cell viability, collagen deposition, morphology and surface expression of THP-1 macrophage markers at days 2 and 6. Results. At day 2, both Ficoll and Dextran had no impact on ECM deposition or differentiation. At day 6, Ficoll caused a significant 30% increase in collagen deposition and a significant two-fold increase in CD14 and CD11b expression whereas Dextran had no impact. There were no associated cytotoxic effects. Discussion. Ficoll accelerated ECM deposition and macrophage differentiation. These observations support the notion that the ECM has a role in controlling monocyte differentiation. Ficoll mediated MMC may therefore be a bioengineering strategy with the potential to accelerate the formation of an ECM rich microenvironment that can be used to modulate macrophage function. However, further investigations are required to confirm the findings. Key references. Daigneault, M., Preston, J. A., Marriott, H. M., Whyte, M. K. B., & Dockrell, D. H. (2010). The Identification of Markers of Macrophage Differentiation in PMA-Stimulated THP-1 Cells and Monocyte-Derived Macrophages. PLoS ONE, 5(1), e8668. https://doi.org/10.1371/journal.pone.0008668 Kumar, P., Satyam, A., Fan, X., Collin, E., Rochev, Y., Rodriguez, B. J., … Zeugolis, D. I. (2015). Macromolecularly crowded in vitro microenvironments accelerate the production of extracellular matrix-rich supramolecular assemblies. Scientific Reports, 5, 8729. Retrieved from http://dx.doi.org/10.1038/srep08729

Wednesday 13th September 2017, 11.00 – 12.30 Clinical Imaging

The Relationship between [18F]fluoride PET Kinetic Parameters and Bone Density Measurements from Lumbar Spine Quantitative CT Inderbir Jassel (1) Glen Blake (2) Musib Sidique (2) Walsall Manor Hospital, Walsall, UK (1) King's College London, London, UK (2) Background: [18F]fluoride PET can be used to obtain the physiological parameter of plasma clearance, or Ki, which can be used to represent the rate of bone remodelling within a given region [5] of interest . Ki has been shown to correlate well with the gold standard of bone biopsy derived remodelling parameters in pigs and it has already been used to help to further our understanding of the pathological processes of bone diseases such as Paget’s disease and the mechanisms of drugs in osteoporosis [1, 2, 3, 4].

The Ki parameter is also theoretically proportional to bone mineral density; it is therefore not truly an accurate measure of bone turnover. Evidence which favours this is that when Ki was corrected using DEXA derived bone density measurements it correlated stronger with the gold standard [2] when compared to uncorrected Ki . 18 If the correlation between Ki derived from [ F]fluoride PET and bone density measured in Hounsfield Units (HU) from CT could be demonstrated then this would justify correcting Ki for bone density using HU. In this study we aim to investigate the presence of this correlation using the trabecular bone of lumbar vertebrae as regions of interest. Methods: Analysis was performed on the images from 63 human PET/CT scans where the radiotracer [18F]fluoride was used. In each image regions of interest (ROIs) were created for the trabecular bone of lumbar vertebrae L1-L4. A novel variation of the Patlak method was used to obtain the Ki from these ROIs from single frame PET data [5]. Hounsfield Unit data was also obtained from these ROIs from the corresponding CT scan images. For each scan these parameters were averaged so that a single Ki and single HU value could be attributed to each scan.

Results. The Ki and HU of each scan were plotted on the graph above which consists of 63 data points. Spearman’s Rank correlation was found to be 0.427 (P < 0.05).

Discussion. Our results demonstrate a positive and statistically significant correlation between Ki derived from [18F]fluoride PET and Hounsfield Units on CT scan. This is the first study to demonstrate this relationship, which other studies using smaller populations have failed to show [3] 18 Conclusion. Our results demonstrate a positive correlation between Ki derived from [ F]fluoride PET and Hounsfield Units from CT scans when using lumbar vertebrae as regions of interest. This means that it can be justified to correct Ki using Hounsfield Units to create a novel parameter which is more representative of the gold standard of bone biopsy derived parameters of bone remodelling. Key references. 1 Cook GJR, Blake GM, Marsden PK, Cronin B, Fogelman I. Quantification of skeletal kinetic indices in Paget’s disease using dynamic 18F-fluoride positron emission tomography. J Bone Miner Res. 2002 May;17(5):854–9 2 Frost ML, Moore AE, Siddique M, Blake GM, Laurent D, Borah B, et al. 18F-fluoride PET as a noninvasive imaging biomarker for determining treatment efficacy of bone active agents at the hip: a prospective, randomized, controlled clinical study. J Bone Miner Res. 2013 Jun;28(6):1337–47. 3. Frost ML, Siddique M, Blake GM, Moore AEB, Schleyer PJ, Dunn JT, et al. Differential effects of teriparatide on regional bone formation using (18)F-fluoride positron emission tomography. J Bone Miner Res. 2011 May;26(5):1002–11. 4 Piert M, Zittel TT, Becker GA, Jahn M, Stahlschmidt A, Maier G, et al. Assessment of porcine bone metabolism by dynamic. J Nucl Med. 2001 Jul;42(7):1091–100.

22 5 Siddique M, Blake GM, Frost ML, Moore AEB, Puri T, Marsden PK, et al. Estimation of regional bone metabolism from whole-body 18F-fluoride PET static images. Eur J Nucl Med Mol Imaging. 2012 Feb;39(2):337–43

Same Day Cerebral Perfusion and Dopamine Receptor Imaging for Patients with Suspected Dementia Thomas Biggans (1) NHS Tayside, Dundee, UK (1)

Background The underlying cause of cognitive impairment is often unclear. When impairment is first evident it can be difficult to distinguish between Parkinson plus syndromes. Imaging both cerebral perfusion and dopamine receptor function can provide a valuable insight into the patient’s physiology at an early stage [1-5].

Imaging on the same day is challenging due to the close proximity of the photon energies emitted from Tc-99m (140 keV) and I-123 (159 keV). This work considered whether it was possible to obtain diagnostic quality images from a DaTSCAN acquired a few hours after a HMPAO scan on the same day. Methods A small phantom was filled with I-123 in the form of sodium iodide and diluted with water. The activity of I-123 used (7 MBq) produced a similar count rate to the average clinical case. This phantom was placed inside a larger cylindrical phantom containing Tc-99m pertechnetate and water. As for the I-123, the amount of technetium (9 MBq) was determined by the average clinical count rate and subsequently adjusted for the expected decay in the proposed technique. SPECT acquisitions were taken with various energy windows (159 keV ± 10%, 159 keV ±5% & 167 keV ± 10%). 120 views with 30s per view and 1.14 zoom to cover 360o. Data was reconstructed using 3D-OSEM with and without resolution recovery. Results Resultant images qualitatively look comparable and provide similar contrast to noise ratios. Use of LEHR + 7.5% collimators, 167- 10% keV energy window and reconstructing without resolution recovery provided the highest quality images. Discussion From initial phantom results it appears achievable to obtain diagnostic quality images using a same day protocol. There may be some loss in contrast and resolution due to the method and this would have to be noted when reporting on the images however this loss is in the same order of magnitude as for example poor patient setup. Conclusion The work completed so far has provided a fundamental proof of concept which is encouraging. Further work will be conducted in the near future with more anatomically similar phantoms to cement this evidence base. It is the aim that both patients and the department will benefit from implementation of this technique. Key References [1] Chia-Jung Chang et al, Separation of Two Radionuclides in Simulataneous Dual-Isotope Imaging with independent Component Analysis, Biomed Eng Appl Basis Comm, 2006; 18:264-269 [2] G El Fakhri et al, Quantitative simultaneous (99m)Tc-ECD/123I-FP-CIT SPECT in Parkinson’s disease and multiple system atrophy, Eur J Nucl Med Mol Imaging, 2006 Jan; 33(1):87-92 [3] Y Du & EC Frey, Quantitative evaluation of simultaneous reconstruction with model-based crosstalk compensation for 99mTc/123I dual-isotope simultaneous acquisition brain SPECT, Med. Phys. 36 (6)

23 [4] K Van Laere et al, Dual-Tracer Dopamine Transporter and Perfusion SPECT in Differential Diagnosis of Parkinsonism Using Template-Based Discriminant Analysis, J Nucl Med 2006; 47:384-392 [5] M Lorberboym et al, 123I-FP-CIT SPECT Imaging of Dopamine Transporters in Patients with Cerebrovascular Disease and Clinical Diagnosis of Vascular Parkinsonism, J Nucl Med 2004; 45:1688-1693

Virtual Histology Intravascular Ultrasound-based Plaque Structural Stress Estimation Improves Prediction of Future Major Adverse Cardiovascular Events Zhongzhao Teng (1) Adam Brown (1) Patrick Calvert (1) Yuan Huang (1) Stephen Hoole (2) Martin Goddard (2) Nick West (2) Martin Bennett (1) Jonathan Gillard (1) University of Cambridge, Cambridge, UK (1) Papworth Hospital NHS Trust, Cambridge, UK (2) Background. Around two thirds of all spontaneous thrombotic coronary events resulting in myocardial infarction (MI) or sudden cardiac death are due to rupture of an atheromatous plaque1,2. Ruptured plaques exhibit large necrotic lipid cores, thin overlying fibrous caps and evidence of microcalcification3. The precursor lesion for rupture, termed a 'thin-cap fibroatheroma' (TCFA), displays several of these compositional features3. However, prospective studies have shown that future clinical event rates attributable to these 'high-risk' plaques were <10% over three years4, highlighting the need of novel markers to improve plaque-based risk stratification. Under physiological condition, plaque is subject to mechanical loading due to pulsatile blood pressure and heart beating. Plaque rupture occurs when such mechanical loading, e.g., plaque structural stress (PSS), exceeds its material strength5. We therefore hypothesized that PSS may be increased in plaques responsible for cardiovascular event and that integration of PSS with high-risk virtual histology intravascular ultrasound (VH-IVUS)-defined characteristics could improve our ability to predict future events. Methods. In the Virtual Histology in Vulnerable Atherosclerosis (VIVA) study, patients (n=170) undergoing percutaneous coronary intervention (PCI) with either stable angina or acute coronary syndrome (ACS) were prospectively enrolled. VH-IVUS imaging was performed in the culprit vessel before PCI and in all three major coronary arteries after PCI. Major adverse cardiovascular events (MACE) data were collected by telephone interview and clinic follow-up (median 1,115 days). Independent IVUS analysis was performed to identify fibrous, fibrofatty, necrotic core and dense calcium. Non-culprit lesions responsible for MACE (n=22) were matched to control lesions (n=22) based on plaque characteristics that are known to affect PSS, including luminal area, plaque burden, necrotic core/dense calcium volume and plaque classification. Predicted probabilities for every non-culprit lesion were determined by logistic regression analysis, allowing propensity matching to be performed between plaques responsible for MACE and a control arm. Results. MACE was associated with plaque burden (PB) ≥70% (hazard ratio (HR) 7.6, 95%CI 2.0- 28.7, p=0.003), minimal luminal area (MLA) ≤4mm2 (HR 3.5, 95%CI 1.1-11.0, p=0.036) and VH- defined thin-cap fibroatheroma (VH-TCFA) (HR 5.9, 95%CI 1.3-26.6, p=0.022), although absolute event rates for high-risk lesions remained <10%. Maximum PSS was normalised by the systolic blood pressure of each patient. It showed marked heterogeneity across and between similar lesions, but was significantly increased in MACE lesions at high-risk regions, including PB≥70% (13.9±11.5 vs. 10.2±4.7, p<0.001) and VH-TCFA (14.0±8.9 vs. 11.6±4.5, p=0.02). Furthermore, PSS improved the ability of VH-IVUS to predict MACE in plaques with PB≥70% (adjusted log-rank, p=0.003) and MLA≤4mm2 (p=0.002). Plaques responsible for MACE had larger superficial calcium inclusions, which acted to increase PSS (p<0.05). Discussion. Atherosclerosis plaques are typically demonstrable at different stages of development throughout the coronary tree. Consequently, non-culprit lesions greatly outnumber culprit lesions in patients undergoing PCI, and underlie the majority of subsequent MACE. PSS is a proposed driver of rupture that integrates plaque anatomy and composition with physical forces that plaques experience. Moreover repeated cycles of subclinical rupture and repair also underlie rapid plaque growth6, leading to luminal encroachment and symptoms of progressive angina. The integration of PSS and baseline plaque characteristics has the potential in identifying higher risk lesions. Conclusion. Baseline PSS is increased in plaques responsible for MACE, and improves the ability of intra-coronary imaging to predict events. Biomechanical modeling may complement plaque

24 imaging for risk stratification of coronary non-culprit lesions. Key references. 1. Davies, MJ, New England Journal of Medicine 310, 1137-40, 1984. 2. Finn, AV, Arteriosclerosis, Thrombosis and Vascular Biology 30, 1282-92, 2010. 3. Virmani, R, Arteriosclerosis, Thrombosis and Vascular Biology 20, 1262-75, 2000. 4. Stone, GW, New England Journal of Medicine 364, 226-35, 2011. 5. Richardson, PD, Lancet 2, 941-4, 1989. 6. Burke, AP, Circulation 103, 934-40, 2001.

Ultrasound measurement of brain tissue motion: a cross-sectional healthy volunteer study Georgina de Vries (1,2) Poppy Turner (1) Caroline Banahan (1,2) Jyoti Nath (1) Imane Bnini (1) Stefanie Berger (1) James Campbell (1) Kirk Beach (3) Mark Moehring (4) Asanka Dewaraja (4) Kumar Ramnarine (1,2) Emma Chung (1,2) Dept. of Cardiovascular Sciences, University of Leicester, Leicester, UK (1) Dept. of Medical Physics, University Hospitals of Leicester NHS Trust, Leicester, UK (2) University of Washington, Seattle, WA, USA (3) Broadview Laboratories, Seattle, WA, USA (4)

Background. It is well known that the brain ‘pulses’ with the cardiac cycle (1-2) but typical distances associated with brain tissue motion have yet to be fully explored (3-5). This study uses transcranial Doppler (TCD) ultrasound to measure brain tissue pulsations in healthy adults sampled from a cross- section of ages. The aim of the study was to characterise tissue pulsations in healthy subjects as a function of age, sex, heart-rate (HR), and arterial blood pressure (BP) to inform future research. Methods. Brain tissue pulsation (BTP) estimates were obtained from healthy volunteers with no known brain injuries or heart conditions using a commercially available TCD ultrasound system equipped with a 2 MHz transducer (Spencer Technologies, MA, USA). The probe was positioned above the right eyebrow and held in place by an elasticated headband. Eight second recordings of brain tissue motion were obtained from 30 overlapping 3 mm gates ranging in depth from 2 to 8 cm along a single beam-line. Pulsation measurements were synchronised to HR using a 3-lead Electrocardiogram (Nihon Kohden, Japan) and analysed in MatLab to obtain an estimate of distance moved by the tissue over the cardiac cycle. Blood pressure readings were obtained using a standard arm-cuff device before and after the TCD recordings. Data were entered to Excel and an exploratory statistical analysis performed in Stata to examine the distribution of measured BTPs across subjects and investigate possible relationships between parameters. Results. 111 adults (58 men) aged from 19 to 80 years consented to take part in this study. We show that good quality recordings, showing obvious pulsations and an absence of motion artefacts, can be obtained through the forehead. Average pulsation magnitude over all depths ranged from a median (IQR) value of 3 (6) to 79 (65) µms between subjects, and tended to increase linearly with depth (from 11 (11) µms near the surface of the brain to 26 (24) µms near the ventricles). A linear regression model with log transformed maximum pulsation amplitude as the dependent variable and age, heart rate and blood pressure as explanatory variables suggested age (p>0.004) and heart rate (p>0.003) are significant predictors of maximum pulsation amplitude. Discussion. This study demonstrates the feasibility of measuring brain tissue motion through the skull using ultrasound. We provide reference data for measurements of the right forehead position of healthy volunteers. This dataset can be used as a basis for developing models of BTP and to develop physiological experiments aimed at understanding the origins of brain tissue pulsations and relationship to cerebrovascular physiology. Conclusion. Further physiological experiments in healthy volunteers are underway to better understand the contributions of BP, CO2, tissue compliance, and other parameters to brain tissue pulsations.

25 Key references. [1] Waghul, M.E., Eide, P.K., Madsen, J. R., 2011. The pulsating brain: A review of experimental and clinical studies of intracranial pulsatility. Fluids and Barriers of the CNS, 8(5). [2] Greitz, D., Wirestam, R., Franck, et al., 1992. Pulsatile brain movement and associated hydrodynamics studied by magnetic resonance phase imaging. Neuroradiology, 34(5), pp. 370-80. [3] Kucewicz, J. C., Dunmire, B., Giardino, N. D., et al., 2008. Tissue pulsatility imaging of cerebral vasoreactivity during hyperventilation. Ultrasound Med Biol.,33(5), pp. 1200-8 [4] Desmidt, T., Hachemi, M. E., Remenieras, J. P., et al., 2011. Ultrasound Brain Tissue Pulsatility is decreased in middle aged and elderly type 2 diabetic patients with depression. Psychiatry Res., 193(1), pp. 63-4 [5] Kucewicz, J. C., Dunmire, B., Leotta, D. F., et al., 2007. Functional tissue pulsatility imaging of the brain during visual stimulation. Ultrasound Med Biol., 33(5), pp.681-90.

26 Wednesday 13th September 2017, 13.30 – 14.00 Plenary Session

Subject-specific modelling in healthcare: Now’s the time Prof Marco Viceconti, Executive Director, Insigneo Institute, University of Sheffield

Subject-specific modelling is reaching in various specific applications a level of maturity sufficient to consider its clinical adoption in routine healthcare. This presentation revises the state of the art in this field, with particular reference to the work being conducted at the Insigneo institute for in silico Medicine of the University of Sheffield. We consider three families of translational of targets: digital patient, where the end user is a clinical specialist that use the model as decision-support system; in silico clinical trials, where the model is used by a company in the development and regulatory assessment of a new biomedical product; and personal health forecasting, where the model is used by a patient who needs to self-manage a chronic condition. Recent advances suggest now is the time to see these technologies widely adopted.

27 Wednesday 13th September 2017, 14.00 – 15.00 Influencing decision makers: a researcher’s guide

Influencing decision makers: a researcher’s guide Royal Academy of Engineering

Have you considered how public policy is made and how it is relevant to your day-to-day work? This session will provide a valuable introduction to the policy and decision-making landscape. Hear policy professionals from across government and biomedical engineering policy field describe the mechanisms and key players in the area. You will learn about the role of research in influencing decision makers, how research can have an impact on national policy, and health policy in particular, and in turn how their work impacts the biomedical engineering field. By the end of this session, you will have gained an insight into where you can find out more and get further involved in the policy sphere.

28 Wednesday 13th September 2017, 14.00 – 15.00 Pre Clinical Imaging

Radiology + Pathology = Noninvasive Pathology: Leading the way of integration in mouse models of cancer Dr Yann Jamin, Children with Cancer UK Research Fellow, Division of Radiotherapy and Imaging, The Institute of Cancer Research, London.

Moving away from the "one-size-fits-all” approach to cancer therapy into the new era of precision medicine, there is a continuing need for validated diagnostics to characterise tumour pathophysiology for the selection and scheduling of safe and effective therapies for individual patients. Magnetic resonance imaging (MRI) affords quantitative biomarkers to non-invasively visualise the regional variation and temporal evolution of relevant structure-function relationships in tumours in vivo, and which inform on treatment response and resistance. Susceptibility-contrast MRI, oxygen-enhanced T1-weighted MRI and MR elastography are examples of advanced MRI methodologies that provide biomarkers that map important aspects of the tumour microenvironment including tumour blood supply, oxygenation, cellular and stromal integrity and their therapeutic modulation. Rigorous cross validation of MRI biomarkers and histopathology is an essential step for the translation of these biomarkers into clinical practice. Integrating digitized pathology into preclinical multi-parametric MRI investigation in clinically-relevant mouse models of cancer can help calibrate MRI into a noninvasive pathology tool that can complement diagnostic molecular pathology with longitudinal (time) noninvasive imaging (space) of the tumour phenotype and improve our position to deliver precision medicine.

Blast lung injury quantification through imaging Kalpani Vitharana (1) Joseph Sherwood (1) Thuy-Tien Nguyen (1) Andrew Bodey (2) Angelo Karunaratne (1) Hari Arora (1) Imperial College London, London, UK (1) Diamond Light Source, Didcot, UK (2)

Background. Blast lung, which is caused by the direct impact of a blast wave to the thorax, is one of the most prevalent forms of blast injuries which can lead to mortality if not treated on time. Symptoms include shortness of breath, respiratory distress and haemoptysis1. On a macroscale, blast lung induces oedema and parenchymal haemorrhage. However, there is limited knowledge on the damage mechanisms of different types of blast wave profiles in the lungs. The aim of this study is to quantify damage using micro-CT imaging in order to assess the morphological changes of the alveoli which are responsible for gas exchange. The morphological parameters investigated are alveolar volume, sphericity and shape factor. Methods. Cadaveric Sprague Dawley rats were ventilated at a fixed pressure and were then blasted using a shock tube which simulates real life explosions2. The lungs were excised and placed in a mounting platform for scans to be taken. The raw images were analysed using Avizo to quantify pre-and post-blast alveoli. The images were filtered in order to remove background noise which facilitated with segmentation. Segmentation binarized the images so that the alveoli can then be quantified. Results.

Control

Blasted

Figure 1: Example of a control and a blasted lung tissue with their respective morphological quantified parameters

Discussion. The injury mechanisms of blast lung and how different blast wave profiles damage the lung have not been elucidated yet. Micro-CT imaging was used to investigate the effects of blast at a microscale in order to quantify damage in terms of alveolar volume, shape factor (surface area to volume ratio) and sphericity. It was found that blasted tissues had significantly greater volume and sphericity compared to the control, whilst shape factor was found to be lower. Conclusion. Blast lung injury was quantified using micro-CT imaging in terms of alveolar volume, sphericity and shape factor. Hopefully this could further elucidate the pathophysiology of blast lung and provide potential routes for therapeutic interventions. Key references. (1) Ferraro, D.M. & Hiles, P.D., Blast-related lung injuries, 2016, Curr Pulmonol Rep, 5(2), 70-76 (2) Eftaxiopoulou T et al., Prolonged but not short-duration blast waves acute inflammation in a rodent model of primary blast limb trauma, 2016, Injury, 47(3), 625-632 (3) Mackenzie, I et al., What the intensive care doctor needs to know about blast-related lung injury, 2013, JICS, 14(4), 303-312

In vivo MRI Cell Tracking of Autologous Mesenchymal Stem Cells in an Ovine Osteochondral Defect Model. Alicia El Haj (1) Hareklea Markides (1) Karen Newall (2) Heike Rudorf (2) Lia Blokpoel Ferreras (3) James Dixon (3) Frances Henson (2) Keele University, Stoke-on-Trent, UK (1) University of Cambridge, Cambridge, UK (2) The University of Nottingham, Nottingham, UK (3)

Background: The application of mesenchymal stem cells (MSCs) in treating cartilage damage and osteoarthritis (OA) has been made possible by the immunosuppressive, differentiation and homing abilities of these cells. A non-invasive means of assessing cell integration, tissue regeneration and cellular bio-distribution is fundamental in evaluating the success of this therapy (Zhou et al., 2006). Previous studies have focused on the development of a standardised SPION (superparamagnetic iron oxide nanoparticle) & MRI (Magnetic resonance imaging) based protocol to image and track stem cells in vivo (Markides et al., 2013). In this study the suitability of this protocol in evaluating cell homing was assessed in an osteochondral ovine model over 7 days.

Methodology: Autologous ovine MSCs were isolated, expanded and labelled overnight with Nanomag; a 250nm dextran coated SPION using a novel transfection agent; P21-8R. In vitro and ex vivo MRI detection thresholds were determined prior to in vivo studies. Cell viability, proliferation and differentiation potential pre and post Nanomag labelling were also evaluated. A single 8mm diameter osteochondral defect was created in the medial femoral condyle in the left knee joint of each sheep with the contralateral joint serving as the control. 10x106 Nanomag labelled-MSCs (P3) were subsequently labelled with a fluorescent lipophilic dye (DII) and delivered by intra-articular injection at either 1 week or 4.5 weeks post defect creation. Sheep were then sacrificed 7 days post implantation and immediately MR imaged using an ESAOTE 0.2T scanner and validated using a Siemens 3T MRI scanner. Joints were then processed for histology.

Results: In vitro and ex vivo MRI data demonstrated significant increase in MRI contrast as a

30 result of P21-8R:Nanomag uptake. Cell viability, proliferation and differentiation capabilities were not affected by Nanomag-labelling. In vivo MRI data revealed the presence of Nanomag-labelled cells within the synovial joint 7 days’ post implantation however, cells were not seen to home to the site of injury when implanted either 1 nor 4.5 weeks after the defect was created. MRI results were validated by histology with the presence of implanted cells visible by fluorescent microscopy. Serum CRP (c-reactive protein) levels were measured by ELISA with no obvious increase in CRP levels observed as a result of P21-8R:Nanomag delivery.

Conclusion: This study clearly demonstrates the potential of Nanomag and P21-8R as an effective means of imaging and tracking cells in an ovine osteochondral defect model with great implications in the clinical translation of a wide range of stem cell based therapies.

We would like to Acknowledge the ARUK center for Tissue Engineering and the UKRMP Acellular delivery hub for support and funding.

References:  Zhou R, Acton PD, Ferrari VA: Imaging stem cells implanted in infarcted myocardium. J Am Coll Cardiol 2006, 48:2094–2106.  Markides, H., Kehoe, O., Morris, R. H. & El Haj, A. J. 2013. Whole body tracking of superparamagnetic iron oxide nanoparticle-labelled cells--a rheumatoid arthritis mouse model. Stem Cell Res Ther, 4, 126

31 Wednesday 13th September 2017, 14.00 – 15.00 Cardiovascular Computational Modelling II

Non-invasive AAA detection method based on arterial pressure/velocity waveform analysis Igor Sazonov (1) Ashraf Khir (2) Perumal Nithiarasu (1) Swansea Univerisity, Swansea, UK (1) Brunel University London, London, UK (2) Background. Abdominal aortic aneurysm (AAA) is often asymptomatic. To detect an AAA on time, a continuous screening programme is necessary. All existing and used in practice AAA detection methods (US echography, CT, MRI) have drawbacks. We report here a novel approach for AAA detection, based on measurement and analysis of the arterial pressure and velocity waveforms. Methods. Reflection of waveforms from artificial aneurysm was studied experimentally and numerically. The waveform analysis is based on the 1D mathematical modelling of the systemic circulation and on the 0D approximation of the aneurysm. The typical length of AAA is 10-12 cm which is much smaller than the shortest wavelength, ~1m, contributing to the wave pulse. We propose that the aneurysm can be characterised by its lumped parameter: aneurysm compliance, which is the difference between the vessel compliance with AAA and the compliance of same vessel without it. The aneurysm compliance can be considered as AAA severity indicator: it is large if the AAA has a large diameter and/or it has weak vessel wall. The least square method enables evaluation of the aneurysm compliance and location. Results. Results of the numerical simulation in a human arterial network with embedded AAA and the retrieval of the AAA parameters are presented in the Table below.

3 t, ms CA, MPa/cm AAA-2 AAA-3 AAA-2 AAA-3 Ref. Num. Num. Ref. Num. Ref. Num. RCA 194 151 161 25 13 78 48 LCA 189 145 161 25 16 78 71

Here CA is the AAA compliance and t is the wave propagation time from the waveform measurement site to the AAA and back. The parameters are evaluated using velocity waveforms computed in left (LCA) and right (RCA) carotids. Here AAA-2 and AAA-3 refer, respectively, to two and three fold wider AAA than the healthy vessel. “Ref.” denotes the reference value calculated directly using geometrical and mechanical parameters of the arterial network. Discussion. An analogous AAA detection method has been proposed in [1] which is based on calculation of reflection coefficient by measuring pressure and flow rate waveforms in the aorta. However, the sensitivity of this method is extremely low. The method also requires a non-invasive pressure measurement in the aorta, which is a rather complicated task. Our method [2] is essentially more sensitive and it can use measurements performed in accessible parts of the human body. The severity coefficient introduced here can be more useful for predicting AAA rupture than its size. Conclusion. A novel method of detecting aortic aneurysm is proposed based on measurement and analysis of pressure/velocity waveforms. The method has been verified (i) experimentally in a simplified network with interchangeable artificial aneurysms of different diameter and (ii) using numerical modelling of the experimental network within the framework of a 1D approach. The method has the potential to provide clinicians with a non-expensive, easily employable tool for aneurysm screening, monitoring and the intervention. Key references. [1] A. Swillens, L. Lanoye, J. De Backer, N. Stergiopulos, P.R. Verdonck, F. Vermassen, and P. Segers. Effect of an abdominal aortic aneurysm on wave reflection in the aorta. IEEE Transactions on Biomedical Engineering, 55:1602-1611, 2008 [2] I. Sazonov, A.W. Khir, W.S. Hacham, E. Boileau, J.M. Carson, R. van Loon, C. Ferguson, and P. Nithiarasu. A novel method for non-invasively detecting the severity and location of aortic aneurysms, Biomechanics and Modeling in Mechanobiology. (2017) doi:10.1007/s10237-017- 0884-8.

32 Development of a cardiovascular network model during pregnancy in humans Jason Carson (1) Michael Lewis (1) Dareyoush Rassi (1) Simon Emery (2) Raoul Van Loon (1) Swansea University, Swansea, UK (1) Abertawe Bro Morgannwg University Health Board, Swansea, UK (2)

Background. The human cardiovascular system undergoes substantial changes over the course of pregnancy, including an increase of blood volume by approximately 30%, significant reductions in vascular resistances, and numerous vessel adaptations. If inadequate or excessive adaptation occurs, health problems such as hypertension, hypotension, placental insufficiency, or pre- eclampsia can develop. The aim of this research is to use a computational model of the cardiovascular system to explore the mechanisms involved in 'normal' (healthy) cardiovascular adaptation during pregnancy and how poor adaptation can lead to pathology. Measurements from several of our group's previous studies (D’Silva et al., 2014; Carpenter et al., 2015; 2016a; 2016b) were used to tune and validate the model. Methods. The computational model of the cardiovascular system is described by: the 1D blood flow equations, which are given by the conservation of mass, conservation of momentum and a viscoelastic constitutive law; and lumped models for vascular beds, valves and the heart. The enhanced trapezoidal rule method (ETM) by Carson and Van Loon (2016) is used to solve the resultant system of equations. Results. The figure shows the model output (line), with the standard error of the mean of the measured data (shaded region) for the first, second, and third trimesters of pregnancy, and for the post- partum period. (CO: Cardiac Output; TPR: Total Peripheral Resistance). The model gives solutions consistent with the measured data and with other literature, including the volume of blood reaching the placenta. Discussion. The model can capture notching in uterine arterial flow waveforms, which is often present in patients with problems such as pre-eclampsia and placental insufficiency. This could allow the model to be used in the investigation of various health problems, and would potentially help discriminate between high-risk and low-risk pregnancies. Conclusion. Our model is the first to capture the main characteristics of cardiovascular adaptation that occur during human pregnancy. We are further developing this model in our ongoing studies into the mechanisms of action of common cardiovascular health issues in pregnant woman. Key references. [1] R. Carpenter, S. Emery, D. Rassi, O. Uzun, M. Lewis. Recruitment of pregnant woman to an exercise-intervention study: a flexible randomisation approach. Journal of Obstetrics and Gynaecology, 36(2):200-207, 2016, doi:10.3109/01443615.2015.1049988. [2] R. Carpenter, S. Emery , O. Uzun, D. Rassi , M. Lewis. Influence of Physical Exercise on Baroreceptor Sensitivity during Pregnancy. The Journal of Maternal-Fetal & Neonatal Medicine. 30(5):514-519, 2016 . Doi: 10.1080/14767058.2016.1179275. [3] R. Carpenter, S. Emery , O. Uzun, L. D’Silva , M. Lewis. Influence of antenatal physical exercise on haemodynamics in pregnant women: A flexible randomisation approach. BMC Pregnancy and Childbirth.15:186, 2015. doi: 10.1186/s12884-015-0620-2 [4] J. Carson and R. Van Loon. An implicit solver for 1D arterial network models. Int J Numer Meth Biomed Engng, 2016, doi:10.1002/cnm.2837. [5] L. D’Silva , R. Davies, S. Emery , M. Lewis. Influence of somatic state on cardiovascular measurements in pregnancy. Physiological Measurement.35(1):15-30, 2014. doi: 10.1088/0967- 3334/35/1/15.

33 The use of reduced order models to determine fractional flow reserve using non-invasive techniques Jason Carson (1) Sanjay Pant (1) Perumal Nithiarasu (1) Swansea University, Swansea, UK (1)

Background. Fractional flow reserve (FFR) is the current standard of care for determining the severity of coronary artery stenosis (Lofti et al. 2014). The procedure is invasive, and requires a catheter to be inserted from the femoral artery to the site of the stenosis; the patient is injected with a product such as adenosine to induce hyperaemic conditions (maximal blood flow); a pressure sensitive wire is then used to measure the pressure drop across the stenosis; the FFR value is calculated as the ratio of the pressure distal to the stenosis divided by the proximal pressure. FFR can be used to guide the treatment strategy and can avoid unnecessary stenting, improve health outcomes, and is cost saving when compared to the use of percutaneous coronary intervention (PCI) based on quantitative coronary angiography (QCA) (Pijls et al. 2007, Tonino et al. 2009). Recently, minimally invasive techniques to estimate FFR have been proposed. These minimally invasive procedures generally require the use of coronary angiography or coronary computed tomography angiography (CCTA). Patient-specific models are then created by segmenting images of the coronary arteries, and are then simulated in a model to solve for flow and pressure. Methods. The technique used in this study involves the use of CCTA and reduced order modelling (both geometry and the flow field) of the coronary arteries to determine fractional flow reserve. Segmentation of the CT scans is performed, and the centre lines are extracted in order to produce a mesh based on length and diameter. This can then be used to produce a fine 1D spatial mesh for the reduced order model; which uses the 1D equations of pulse wave propagation for the larger coronary vessels, and a lumped 0D model to represent the vascular bed. Results. Following work from Boileau et al. (2017), the reduced-order model was compared with a rigid-wall 3D solution for various stenotic configurations. The solutions of the two methods were consistent for the majority of cases. Discussion. The solutions generally gave good agreement between 1D and 3D cases. However, for more severe stenoses, the FFR tended to be over-estimated by the 1D model when compared to the 3D solution. Conclusion. The reduced order model allows a significantly faster solution than that of the 3D model. Our 1D model does not require any stenosis or pressure drop models that depend on the manual identification of the stenotic lesion(s). Work is underway to compare the 1D-0D model with clinical invasive coronary catheter angiography-based measurement of FFR. Key references. [1] Boileau E, Nithiarasu P. One-dimensional modelling of the coronary circulation. Application to noninvasive quantification of fractional flow reserve (FFR). Computational and Experimental Biomedical Sciences: Methods and Applications, LN Comput. Vis. Biomech., vol. 21, Tavares J, Jorge R (eds.). Springer International, 2015; 137–155. [2] Boileau E, Pant S, Roobottom C, Sazonov I, Deng J, Xie Xianghua, Nithiarasu P. Estimating the accuracy of a reduced-order model for the calculation of fractional flow reserve (FFR). Int. J. Numer. Meth. Biomed. Engng., e02908. 2017. doi: 10.1002/cnm.2908. [3] Lotfi A, Jeremias A, Fearon W, Feldman M, Mehran R, Messenger J, et al. Expert consensus statement on the use of fractional flow reserve, intravascular ultrasound, and optical coherence tomography: A consensus statement of the society of cardiovascular angiography and interventions. Catheter Cardiovasc. Interv. 2014; 83:509–518. [4] Pijls N, van Schaardenburgh P, Manoharan G, Boersma E, Bech J, van’t Veer M, et al. Percutaneous coronary intervention of functionally nonsignificant stenosis: 5-year follow-up of the DEFER Study. J. Am. Coll. Cardiol. 2007; 49:2105–2111. [5] Tonino P, De Bruyne B, Pijls N, Siebert U, Ikeno F, van’t Veer M, et al. Fractional flow reserve versus angiography for guiding percutaneous coronary intervention. N. Engl. J. Med. 2009; 360:213–224, doi:10.1056/NEJMoa0807611.

Patient-specific estimation of aortic blood pressure using one-dimensional blood flow modelling: development and testing using in silico and in vivo data Jorge Mariscal-Harana (1) Arna van Engelen (1) Torben Schneider (2) Mateusz Florkow (1,2) Peter Charlton (1) Bram Ruijsink (1) Hubrecht de Bliek (3) Israel Valverde (1) Marietta Charakida (1) Kuberan Pushparajah (1) Spencer Sherwin (4) Rene Botnar (1) Jordi Alastruey (1) Division of Imaging Sciences and Biomedical Engineering, King's College London, London, UK (1) Philips Healthcare, Guildford, UK (2) HSDP Clinical Platforms, Philips Healthcare, Best, The Netherlands (3) Department of Aeronautics, Imperial College London, South Kensington Campus, London, UK (4)

Background. Current clinical evidence indicates that central (aortic) blood pressure is a better marker of hypertension and cardiovascular risk than brachial pressure [4]. However, central blood pressure (CBP) can only be accurately measured in the clinic via an invasive procedure known as catheterisation. At present, Magnetic Resonance Imaging (MRI) provides accurate non-invasive measurements of aortic geometry and blood flow. We propose an approach to estimate CBP from MRI using one-dimensional (1-D) blood flow modelling. Several studies [1,3] have shown that 1-D modelling can accurately reproduce the main features of the pressure, flow and area waveforms. Methods. We created a population of 22 virtual (computed) healthy subjects following Willemet’s approach [6]. Each subject was generated by varying cardiac (stroke volume, cardiac period and ejection time) and arterial (stiffness and peripheral resistance) parameters of a 1-D model of the larger systemic arteries [5] within a wide range of physiological values. This population was used to develop our algorithm for the 1-D aortic model in silico. We then tested its accuracy in a clinical population of 8 aortic narrowing patients. The 1-D numerical simulations were run using our validated in-house computational framework of arterial haemodynamics [2]. Results. From our in silico study, estimations using the aortic model for ± 30% cardiac and arterial parameter variations showed maximum relative errors for systolic, mean and diastolic CBP of 4.5%, 3.6% and 4.2%, respectively. Corresponding average relative errors from our in vivo study were 5.4%, 1.5% and 8.0%.

Fig. 1: CBP estimation for a single virtual subject (left) and systolic CBP estimation using the 1-D aortic model compared to the reference systolic CBP from in silico and in vivo data (right) Discussion. In silico data offers an alternative to in vivo data for the optimisation and testing of the aortic model algorithm. This virtual population could be expanded to include cardiovascular diseases affecting CBP. The aortic model estimates CBP using information which is readily available in the clinic, such as MRI and brachial pressure measurements, and could provide valuable insight into cardiovascular health. Conclusion. We provide a proof of concept for the patient-specific estimation of CBP from data measured non- invasively using aortic MRI combined with a non-invasive peripheral pressure measurement.

Key references.

35 [1] Alastruey, J. et al. J R Soc Interface, 13(119):20160073, 2016. [2] Boileau, E. et al. Int J Numer Method Biomed Eng, 31(10):1-33, 2015. [3] Huberts, W. et al. J. Biomech., 45:1684-1691, 2012. [4] McEniery, C.M. et al. European Heart Journal 35(26), 17191725, 2014. [5] Mynard, J. P. et al. Ann. Biomed. Eng., 43(6):1443-1460, 2015. [6] Willemet, M. et al. Am. J. Physiol. Heart Circ. Physiol., 309(4):H663-75, 2015.

36 Wednesday 13th September 2017, 14.00 – 15.00 Student/Associate/Trainee II

MRI NHS Breast Screening Program: a multi-centre study Steven Jackson (1) Michael Dubec (1) Christie NHS Foundation Trust, Manchester, UK (1)

Background. The NHS Breast Screening Program (NHSBSP) Report 68[4] gives technical guidelines for the provision of an MRI based breast cancer screening service for women at higher risk of developing breast cancer. The report recommends weekly QA of the scanner and breast coil combination comprising a signal to noise ratio (SNR) and fat suppression efficiency calculation[3]. Draft guidelines exist for three MRI system manufacturers[1,2,5,6] detailing sequence parameters and expected tolerances for these measurements. The guidelines have been followed at six sites covering MRI scanners from two manufacturers (five 1.5T and one 3T). The aim of the study was to investigate practicalities of performing the recommended testing regimen in addition to assessing in-house software developed for analysing the acquired data.

Method. Images of bottle phantoms positioned in clinically used breast coil were acquired. A single-slice T1-weighted spin echo image and two multi-slice T1-weighted spoiled gradient echo acquisitions with frequency-selective fat suppression on and off were acquired for SNR and suppression efficiency calculation respectively. A MATLAB script was written to automatically analyse datasets arriving from six NHSBSP sites involved. Three methods of SNR calculation have been implemented; one uses the noise calculated from an ROI outside the bottle phantoms, a second determines signal and noise from five ROIs within the phantom. A third method involving image subtraction was recently introduced. The angle of the bottle phantom inside the breast coil was plotted against measured SNR. Plots of SNR and suppression efficiency over time were prepared for each site.

Results. Of the SNR methods attempted, five ROIs placed within the phantom gave least variability of results (coefficient of variation (COV) = 0.07/0.05 (L/R coil), n=37, compared to 0.16/0.15 (L/R), n=37 with noise ROI outside phantom and 0.24/0.27 (L/R), n=4 with subtraction method at one site). Corresponding results have been obtained at other sites. The five ROI method is less sensitive to bottle position within the field of view as it avoids ghosting artefact. The angle at which the bottles stand inside the breast coil was found not to correlate strongly with 5-ROI SNR (Correlation coefficient = 0.31/0.34 (L/R)). Suppression efficiency was found to behave consistently (COV 0.01/0.01 (L/R) same site as above). Occasional problems with frequency specific fat suppression led to some unusable images.

Discussion. The analysis of NHSBSP MRI QA data represents a significant on-going workload for MR physics departments. We have evaluated a number of options for performing the calculations and drawn conclusions as to the suitability of each. Where unusable images were received from sites extra training was put in place for radiographers.

Conclusion. Existing NHSBSP MRI QA scan guidelines have been implemented and advice is provided for overcoming problems that other users may encounter. The data has been analysed successfully and efficiently using in-house software with resulting recommendations for the most suitable SNR calculation method.

Key references. 1. Charles-Edwards, G., 2013. Appendix B: Philips Breast Screening MRI QA. 2. Graves, M., 2013. GE 8-Channel Breast Array QA. 3. IPEM, 1998. Report 80: Quality control in magnetic resonance imaging. York: IPEM. 4. NHS Breast Cancer Screening Programme, 2012. Report 68: Technical Guidelines for the Magnetic Resonance Imaging for the Surveillence of Women at Higher Risk of Developing Breast Cancer, Sheffield: NHS Cancer Screening Programme. 5. Wilson, D., 2013. MR Quality Assurance Procedures for NHS Breast Screening Programme (NHSBSP) for SIEMENS systems. 6. Wilson, D., 2013 Appendix C: SIEMENS BREAST QA.

Pre-Surgical Planning for the Robot-Assisted Transoesophageal Echocardiography used in Cardiac Procedures Areeb Zar (1) Ruchi Gandecha (1) Shuangyi Wang (1) Richard Housden (1) Kawal Rhode (1) King's College London, London, UK (1)

Background. Transoesophageal echocardiography (TOE) was introduced to aid diagnosis and surgical decision-making (1). TOE is performed manually, often with fluoroscopy, with high levels of operator radiation (2). Protective clothing can lead to orthopaedic diseases (3). A robot to hold the probe and allow remote control has been developed (4), with an automatic workflow demonstrated (5). Complex sensing components are needed for this. A manual workflow is proposed to allow operator control of the robot. Simulation experiments were needed to assess feasibility of this method. Methods. This study focused on the heart structures that needed monitoring for ICD insertion, ASD repair, and Left atrial ablation. The structures were manually defined using a phantom MRI and a view-planning platform was used to plan probe positions based on two parameters. P1 is the percentage of points defining the structure in the beam. P2 is the distance of the centroid of the points from the centre of the US beam. A virtual US image was obtained using 0o and 90o image plane steering. An experiment was performed using the pre-planned position. Once the structure was found, the image was registered with the virtual US image. The transformation matrix mapping probe pose to MR coordinates was found and pose parameters measured. Results. The average probe distance error was 13.0mm (σ=4.5mm). The average orientation error was 7.5o (σ=4.5o). Avg P1 = 100%, Avg P2 difference was = 6.9mm Discussion. There is some error between pre-planned an actual probe pose. But the distances are relatively small compared to a 1cm movement of the heart throughout the cardiac cycle. Breaking down the error components shows significant error in one plane for each view. Depth error is due to manual insertion of the probe. The x, y axis errors can be attributed to the oesophagus phantom not tight to the probe like a real oesophagus; the error it due to limitations of our set-up. Deeper structures were not visible due to the gap between the transducer and the phantom leading to attenuation – this is because only a TTE heart phantom is available. Using a TOE phantom should improve US image quality and the registration. Conclusion. The study shows that a manually controlled robot can be beneficial in surgery. Easy workflow to follow. Results are comparable to automatic control. Next stage is to test on cadavers. Key references:.(1) Kapoor PM, Muralidhar K, Nanda NC, Mehta Y, Shastry N, Irpachi K, et al. An update on transesophageal echocardiography views 2016: 2D versus 3D tee views. Ann Card Anaesth. 2016 Oct;19(Supplement):S56–S72. (2) McIlwain EF, Coon PD, Einstein AJ, Mitchell CKC, Natello GW, Palma RA, et al. Radiation safety for the cardiac sonographer: recommendations of the Radiation Safety Writing Group for the Council on Cardiovascular Sonography of the American Society of Echocardiography. J Am Soc Echocardiogr. 2014 Aug;27(8):811–816. (3) Ross AM, Segal J, Borenstein D, Jenkins E, Cho S. Prevalence of spinal disc disease among interventional cardiologists. Am J Cardiol. 1997 Jan 1;79(1):68–70. (4) Wang, S., Housden, J., Singh, D., Althoefer, K., and Rhode, K. Design, testing and modelling of a novel robotic system for trans-oesophageal ultrasound. Int J Med Robot. 2016 Sep;12(3): 342-54. (5) Wang, S., Singh, D., Johson, D., Althoefer, K., Rhode, K., and Housden, J. Design, Robotic ultrasound: view planning, tracking, and automatic acquisition of trans-esophageal echocardiography. IEEE Robot Autom Mag. 2016 Dec;23(4): 118-127.

38 Evaluation of the Positioning Accuracy of the Robotic-assisted Transoesophageal Echocardiography used in Cardiac Procedures Ruchi Gandecha (1) Areeb Zar (1) Shuangyi Wang (1) Richard Housden (1) Kawal Rhode (1) King's College London, London, UK (1)

Background: 3D Transoesophageal Echocardiography (TOE) is widely used to monitor the heart during cardiac interventional procedures [1]. The TOE probe is manually controlled and the skilled operator is required to hold the probe which is tedious and harmful for the duration of the longer interventional procedures, especially when TOE is used in conjunction with X-ray [2]. Recently, a robot has been designed that is able to hold the TOE probe and auto-position it, which has mitigated the disadvantages of TOE’s current use [3-5]. However, the probe positioning accuracy required for TOE remains unknown as ultrasound scans are feature-based approaches, making it difficult to evaluate the performance of the proposed robot. Simulations were carried out using mitral valve repair, tricuspid valve repair and transmyocardial laser revascularisation on two patients’ MR data to investigate the necessary accuracy for probe positioning by the TOE robot. Methods: The acoustic window for selected procedures was first determined based on the targeting structures required to be monitored. This was achieved by manually finding the best possible probe posture in a custom-designed view-planning platform so most of the targeting structures are located in the centre of field of view (FOV) [4].100,000 different probe postures were generated with random 6-DOF error offsets within certain ranges from the defined probe posture for each view. Influences of the positioning errors to the ultrasound image space were analysed based on percentage of points that are within the FOV (P1) and distance from the mass centroid of the targeting structures to the centre of the 3D ultrasound FOV (P2). Results: The influence of the probe positioning errors at different translation (depth) and orientation (rotation) offset levels were assessed, with P1 and P2 calculated for each procedure. The results are shown below. Level Probe Positioning Accuracy Range P1 (%) P2 (mm) Mitral Valve Repair Tricuspid Valve Repair Transmyocardial Laser Revascularisation Translation Orientation Translation Orientation Translation Orientation error (mm) Error (degrees) error (mm) Error (degrees) error (mm) Error (degrees) 1 10 10 Not possible Not possible 10 5 >95 <15 2 10 20 10 10 20 10 9040 >40 >50 >45 >50 >30 ≤70 ≥40 Discussion: The probe positioning error has different consequences depending on the procedure. A small change in probe position will have a greater influence on P1 with mitral valve repair compared with transmyocardial laser revascularisation. Conclusions: There are variations in P1 and P2 for a given translation error and orientation error depending on the procedure, therefore requirements are in levels. For all selected procedures and both patients, a probe positioning error less than 20mm and 10 degree will achieve level 3 accuracy and it was therefore established to be the basic preferential requirement for the robotic TOE positioning accuracy. References: [1] American Heart Association. Transesophageal Echocardiography (TEE). 2015; Available at: http://www.heart.org/HEARTORG/Conditions/HeartAttack/SymptomsDiagnosisofHeartAttack/Transesophag eal-Echocardiography-TEE_UCM_441655_Article.jsp#.WAnXU_krIqM. Accessed 10/21, 2016. [2]McIlwain EF, Coon PD, Einstein AJ, Mitchell CK, Natello GW, Palma RA, et al. Radiation Safety for the Cardiac Sonographer: Recommendations of the Radiation Safety Writing Group for the Council on Cardiovascular Sonography of the American Society of Echocardiography. Journal of the American Society of Echocardiography 2014;27(8):811-816. [3]Wang S, Housden J, Singh D, Althoefer K, Rhode K. Design, testing and modelling of a novel robotic system for trans-oesophageal ultrasound. Int J Med Robotics Comput Assist Surg 2015 18th June 2015:31/10/2016. [4]Wang S, Singh D, Johnson D, Althoefer K, Rhode K, Housden RJ. Robotic Ultrasound: View Planning, Tracking, and Automatic Acquisition of Transesophageal Echocardiography. IEEE Robotics & Automation Magazine 2016;23(4):118-127. [5]Wang S, Singh D, Lau D, Reddy K, Althoefer K, Rhode K, et al. Probe Tracking and its Application in Automatic Acquisition Using a Trans-esophageal Ultrasound Robot. 2011

39 Shear Wave Elastography (SWE) and Velocity Vector Imaging (VVI) as predictive markers of arterio-venous fistula outcomes: Preliminary results Conor MacDonald (1) Rose Ross (2) John Graeme Houston (0) Clinical and Molecular Medicine, University of Dundee, Dundee, UK (1) Vascular Laboratory, NHS Tayside, Dundee, UK (2) Background: Aterio-venous fistulae (AVF) are the preferred option for vascular access in haemodialysis patients, however their failure rate remains high. Many factors are associated with their failure to mature to a state useable for dialysis. One factor is arterial stiffness (AS), which is common in patients with chronic kidney disease [1]. The evidence regarding AS is conflicting: low arterial elasticity has been shown to correlate with stenosis and failure to mature in AVFs [2], but a recent study found no such link [3]. AVF creation has been shown to lower global AS [4]. AS can be quantified using ultrasound (US) shear wave elastography (SWE) [5] as the velocity of a shear wave is directly related to the stiffness of the tissue. Velocity vector imaging (VVI) can quantify changes in vessel wall motion, which may be related to AS [6]. Few studies have been conducted using SWE or VVI in patient groups but it is believed that this method could provide a simple measure of AS to accompany existing pre-surgical US scans in patients undergoing AVF creation. Methods: SWE and VVI were recently introduced as part of the pre-operative assessment for patients awaiting creation of an AVF in Ninewells hospital. From NHS databases, VVI data and SWE images will be obtained. Patient follow-up will allow linkage to AVF outcome. A comparison study on healthy volunteers is underway, aiming to recruit 40 healthy adults. Volunteers will undergo US imaging identical to patients awaiting AVF creation. The healthy volunteers will be split into two age groups in order to assess differences relating to arterial stiffness. This healthy cohort will also act as a comparator to the AVF group. Results. The retrospective study will include patients undergoing pre-operative assessment until September 2018. Currently, 8 patients have been identified. The volunteer study will run until August. Preliminary results will be available by the conference date. Discussion: If a link between AS, as measured by SWE or VVI, and the outcome of the AVF were to be found, these methods could potentially be added to existing pre-surgical US protocols as a simple method of pre surgical assessment. If post-surgical changes are observed, SWE may yield further information about the global effects of AVF creation and failure. Key references. [1] S. B. Prenner, J. A. Chirinos. Arterial stiﬀness in diabetes mellitus, Atherosclerosis (2015) 238 [2] M. Kheda, L. Brenner, M. Patel, J. Wynn, J. White, L. Prisant, S. Jones, W. Paulson, Influence of arterial elasticity and vessel dilatation on arteriovenous fistula maturation: a prospective cohort study, Nephrol Dial Transplant (2010) 25 (2) [3] A. Masengu, J. B. Hanko, A. P. Maxwel, Arterial stiffness and arteriovenous fistula failure of maturation, J Vasc Access 2016; 17(6) [4] M. S. Utescu, A. LeBoeuf, N. Chbinou, S. Desmeules, M. Lebel, M. Agharazii, The impact of arteriovenous fistulas on aortic stiffness in patients with chronic kidney disease, Nephrol Dial Transplant (2009) 24 [5] E. Messasa, M. Pernotb, M. Couade, Arterial wall elasticity: State of the art and future prospects, Diagnostic and Interventional Imaging (2013) 94 [6] Ma, X. J., Duan, Y. Y., Yuan, L. J., Cao, T. S., Wang, Y., Yang, H. G., & Chen, S. (2012). Quantitative assessment of maternal common carotid artery mechanics using velocity vector imaging in pre-eclampsia. Eur J Obstet Gynaecol Reprod Biol, 160(1)

40 Wednesday 13th September 2017, 14.00 – 15.00 Image Guided Therapy

Overview of the Development of a Radiopaque Drug-eluting Beads for Image-guided Intra- arterial Delivery into Liver Malignancies Andrew Lewis (1) Biocompatibles UK Ltd, Camberley, UK (1) Background. Transarterial Chemoembolisation (TACE) is a minimally invasive image-guided therapy for the treatment of liver malignancies. Drug-eluting Beads (DEB) have been developed that allow intra-arterial administration via a microcatheter to enable vascular occlusion of the tumour feeding arteries whilst providing controlled and sustained locoregional delivery of the chemotherapeutic drug. This maximises drug dose in the tumour, minimising systemic exposure and reducing side effects1. One drawback is that DEB technology is based on synthetic polymers that are X-ray lucent and hence the location of the beads is at best only inferred by lack of contrast flow and retention. We sought to develop Radiopaque DEB (ROB) that could be visualised under conventional X-ray imaging modalities to provide the physician with both intra- and post-procedural feedback on the presence (or absence) of ROB in target and off-target locations. Methods. We investigated many different methods for adding radiodense atoms such as iodine, barium, bismuth into DEB. Covalent attachment to the preformed beads using an aromatic iodinated species similar to that used as the basis of soluble contrast agents (Fig 1 A) offered the best balance between radiopacity and ability to handle and deliver the ROB. Optimisation of iodine content was established in vitro by imaging assessment using micro-CT, fluoroscopy and MDCT, using both ROB suspended in agarose phantoms and packed into small tubes of various diameters (line phantoms) to mimic packing in blood vessels or different sizes. Effects of bead modification on physical properties were evaluated in vitro using mechanical property analysis, penetration under flow, suspension times and microcatheter delivery and robustness assessments. Drug elution performance was evaluated using a number of standardised drug elution methods that model release in the early and later phases of administration (during delivery and once occluded in a vessel). ROB distribution, imaging and long-term biocompatibility was established with rabbit and sheep renal embolisation studies, a rabbit VX2 tumour model and a long-term swine hepatic artery embolisation study using micro-CT, fluoroscopy and CBCT and MDCT. Results. ROB which contained ~150mg Iodine/mL sedimented volume offered the best balance of X-ray visibility without artefacts and whilst remaining able to suspend/deliver. Size ranges 40-90, 70-150 and 100-300 m size were obtained and could be loaded with doxorubicin (recommended dose 37.5 mg/mL, Fig 1B (unloaded are gold in colour (inset), loaded are red)), which released at a rate within the specification of the unmodified DEB size ranges. Long-term implants showed the ROB to be visible in vivo for at least 90 days post-procedure (likely longer) and highly biocompatible in nature. ROB have now been used in the clinic and are proving useful for treating patients with liver malignancies (Fig1 (C-E) images of an HCC tumour under different modalities).

CHO (B) Figure 1: (A) OH OH HO OH HO O (A) Method of ROB HO O OH O HO Triiodobenzyl O preparation moiety HO OH O OH OH O (B) Unloaded ROB (inset) Drug-eluting Bead (DEB) Radiopaque (Based on Polyvinyl alcohol) Bead and doxorubicin loaded ROB (C) (D) (E) (C-E) Fluoroscopy, CBCT MIP immediately post- procedure and MDCT image at 48h of a human hepatocellular carcinoma Fluoroscopy CBCT/MIP MDCT (48h) tumour containing ROB

Discussion and Conclusion. The ability to see where the ROB are going will enable the physician to more precisely target the area of treatment, identify areas of potential under-treatment and improve patient safety by detecting any off-target embolization – together with the benefits of

41 the sustained localised release of chemotherapeutic agent.

[1] Lewis, A.L. & Dreher, M., J. Controlled Release, 161 (2), 338-350, 2012.

Validating Scar Quantification for Guiding Cardiac Resynchronisation Therapy Maria Panayiotou (1) R. James Housden (1) Daniel Toth (3) Peter Mountney (4) Alexander Brost (5) Jonathan M Behar (2,1) C Aldo Rinaldi (2) Athanasius Ishak (1) Onik Chowdhury (1) Kawal S Rhode (1) Rashed Karim (1) Division of Imaging Sciences and Biomedical Engineering, King’s College London, London, UK (1) Department of Cardiology, Guy's and St. Thomas’ Hospitals NHS Foundation Trust, London, UK (2) Siemens Healthcare, Ltd, London, UK (3) Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA (4) Siemens Healthcare, GmbH, Germany (5) Background. The use of implantable cardiac devices has increased in the last 30 years. Cardiac resynchronisation therapy is a procedure which involves implanting a pacemaker for reversing heart failure. The pacemaker leads must be implanted into cardiac myocardial tissue, and the optimal site for implantation is highly patient-specific. However, most implanters use empirical lead placement. One region that has been identified to have a poor response rate in lead placement is tissue with transmural scar. Studies that precisely measure transmurality of scar tissue in the left ventricle (LV) are few [1,3,4], and most studies lack proper validation of their transmurality measurement technique. To our knowledge, no phantom studies have been performed to validate transmurality. In this work we propose: 1) validation of our image analysis based transmurality measurement technique with phantoms and artificially created data, and 2) open sourcing the phantom imaging data so that it may be used by future methods as ground truth for transmurality. Methods. Six separate phantoms in the shape of the LV were created using 3D rapid prototyping. Each phantom contained concentric circular chambers for holding PVA gel [4] which imitated transmural scar at six different configurations. The phantoms were imaged under MRI. The LV anatomy and scar were reconstructed from short-axis slices of MRI [5] and measured using our new transmurality quantification technique. Transmurality was quantified using a 360-degree ray- tracing technique, in which scar thickness could be obtained from the intersection of each ray with myocardium and scar pixels in the MRI image. Transmurality as a percentage 0-100% was represented on the standard 2D Bull’s eye 16-segment American Heart Association LV model. Results. The transmurality quantification technique was validated on all six phantoms, each with a different transmurality configuration covering varying amounts of transmural scar. Figure 1(d) illustrates the output of scar transmurality along with percentage scar volume.

Figure 1. (a) 6 scar phantom configurations. (b) long and short axis LGE MR (c) short axis scar segmentation (d) scar distribution, burden and trasmurality of an example phantom configuration.

Discussion. In the absence of ground truth, it is challenging to validate algorithms that measure indices for cardiac scar. In this work, we created LV scar phantoms of known transmurality that can be imaged with MRI and also reconstructed into the standard 2D bull’s eye plot. The imaging data obtained for each phantom has been made open source. It is an invaluable resource providing a ground truth for future algorithms to validate transmurality measurements. Conclusion. This work presents an approach for validating techniques that measure transmurality of scar in the LV. A new 360-degree ray-tracing algorithm for transmurality computation was validated on short-axis MRI images of the phantoms.

42 Key references. [1] El Berbari, R. et al. An automated quantification of the transmural myocardial infarct extent using cardiac de-MR images, EMBC 2009, 4403–4406. [2] Jolly, M.-P. et al., Combining registration and minimum surfaces for the segmentation of the left ventricle in cardiac cine MR images, MICCAI 2009, 910–18, 2009. [3] Heiberg, E. et al. Automated calculation of infarct transmurality, Comput Cardiol, 2007, 165–68 [4] Schuijf, J.D. et al. Quantification of myocardial infarct size and transmurality by contrast- enhanced magnetic resonance imaging in men, Am J Cardiol, vol. 94, no. 3, 2004, 284–88. [5] Surry, K et al., Poly(vinyl alcohol) cryogel phantoms for use in US and MR, PMB, 2004, 5529.

43 Wednesday 13th September 2017, 15.30 – 17.30 Making the leap: becoming a medical engineering entrepreneur

Making the leap: becoming a medical engineering entrepreneur Royal Academy of Engineering Keynote speaker: Professor Andrew Lewis, Director of Research and Development, Biocompatibles UK Ltd

Ever thought you have an inner entrepreneur? Want to spin out a company but don’t know how to get your foot on the ladder? Come and hear from those who have already begun their journey and those who have done it time and time again. In an informal environment, this session is an opportunity to hear tips and advice on writing business cases, how to protect IP and finding the right funding and support.

This session is supported by the Royal Academy of Engineering’s Enterprise Hub which harnesses the expertise, insight and networks of Academy Fellows, who include some of the UK’s most successful entrepreneurs and business leaders, to support the country’s most promising engineering entrepreneurs.

During this session Andy will share his experiences of medical product development, how the landscape is changing, things that are essential to consider and some pitfalls to avoid. He will also provide some insight into what has made the entrepreneurs he has met in this area special and why they have succeeded.

44 Wednesday 13th September 2017, 15.30 – 17.30 Musculoskeletal Biomechanics

Production and Characterisation of a PLA-HA composite 3D-Printing Filament for Bone Regeneration Applications in Maxillofacial Reconstruction C. Amnael Orozco-Diaz (1,2) Cheryl A. Miller (1) Gwendolen Reilly (1) Robert D. Moorehead (1) Fiona Gilchrist (1) University of Sheffield, Sheffield, UK (1) Consejo Nacional de Ciencia y Tecnología, Mexico City, Mexico (2)

Background. Current treatment options in maxillofacial reconstruction of complex defects – such as those found in Cleft Palate or extensive instances of Orbital Floor Reconstruction – requires cancellous-bone autografts, that promote bone formation but lack structural integrity and can result in lower-than-optimal bone forming at the lesion site. The material here proposed promotes is aimed to promote the formation of new functional bone on the site through 3D-printed implant that fits the surrounding structures and loads appropriately, while avoiding the side-effects and risks of grafting surgery [1]. Implementing advanced manufacturing technologies could allow for improving of this treatment through the production of patient-specific, image-study based implants [2], [3]. We are exploring the development and characterisation of a suitable polylactic acid-hydroxyapatite (PLA-HA) composite material the construction of anatomically-accurate bioactive implants. Methods. Analysis of the base PLA after 3D-printing was carried out according to ASTM D638-03 for tensile specimen production and mechanical testing, and to ISO-10993-13-2010 for degradation profile. Degradation tests were conducted according to the accelerated degradation protocol described in the standard, subjecting tensile specimens to phosphate buffered saline at 70° C for a period of 23 days[4]. Mass loss during the experiment was analysed through dry-weighing of the specimens. Crystallinity was determined through differential scanning calorimetry with entalphy of fusion and entalphy of crystallisation data. In vitro biocompatibility tests for five material variants were carried out using the osteosarcoma-like MG63 cell line through 21 days on 3D-printed disks of 1cm in diameter. Cell survivability was determined using the resazurin reduction essay. Evidence of calcium and collagen deposition was sought after using the alizarin red and the sirius red essays respectively. Results. The mechanical integrity of the 3D-printed specimens fails as soon as at 24 hours, while other variables behave similar to non-3D-printed PLA. Biocompatibility tests show higher concentrations of HA promote better collagen deposition, but have little effect on cell survivability, which is overall favourable. Discussion. Rapid loss of mechanical integrity in 3D-printed PLA during degradation tests is probably due to erosion of the layer-layer interfaces of the specimens, which should have no effect on the other measured variables. The composite showed arguably better bone-regeneration properties at higher concentrations of HA, but they come at the expense of practicality in producing and 3D-printing the material. Conclusion. Production of a composite practical PLA-HA 3D-printing filament is possible, with an optimal concentration of HA ranging below 20% by weight to promote bone formation without compromising processing properties of the material. 3D-printed PLA implants are likely to resorb faster than those shaped through other methods, which could be of clinical relevance for bone regeneration applications in growing children. Key references. In alphabetical order, numbered. [1] S. Puwanun, “Developing a tissue engineering strategy for cleft palate repair.,” University of Sheffield, 2014. [2] P. H. de Moraes, S. Olate, M. Cantín, A. F. Assis, E. Santos, F. de O. Silva, and L. de O. Silva, “Anatomical Reproducibility through 3D Printing in Cranio-Maxillo-Facial Defects,” Int. J. Morphol., vol. 33, no. 3, pp. 826–830, 2015. [3] A. L. Jardini, M. A. Larosa, R. M. Filho, C. A. D. C. Zavaglia, L. F. Bernardes, C. S. Lambert, D. R. Calderoni, and P. Kharmandayan, “Cranial reconstruction: 3D biomodel and custom- built implant created using additive manufacturing,” J. Cranio-Maxillofacial Surg., vol. 42, no. 8, pp. 1877–1884, 2014.

45 [4] N. A. Weir, F. J. Buchanan, J. F. Orr, D. F. Farrar, and A. Boyd, “Processing, annealing and sterilisation of poly-L-lactide,” Biomaterials, vol. 25, no. 18, pp. 3939–3949, 2004.

Glenohumeral contact forces during activities of daily living to aid implant design and rehabilitation planning Christian Klemt (1) Joe Prinold (1) Daniel Nolte (1) Anthony M J Bull (1) Imperial College London, London, UK (1)

Background. The stability of the shoulder in the mid-range of motion is achieved through compression of the humeral head into the glenoid labral concavity due to the contraction of the muscles that cause movement [5] as well as through active control of humeral head translation through co-contraction of the rotator cuff muscles that act to resist shear forces [6]. Therefore, the precise knowledge of glenohumeral shear and compression forces during activities of daily living (ADLs) is essential to define realistic preclinical test procedures for shoulder replacements to improve implant design and fixation [4]. Furthermore, this information will aid rehabilitation planning as those data enable advice to be given to patients to avoid joint overloading [1]. There is currently no detailed knowledge about shoulder forces during ADLs with existing studies only focusing on individual tasks and presenting results as resultant joint reaction force, thereby not quantifying the shear component of force [2,3]. The aim of this study is to analyse shoulder contact forces during ADLs to aid implant design and rehabilitation planning.

Methods. Shoulder kinematics of seven male participants was recorded using a 10-camera optical motion tracking system (Vicon, UK). 27 ADLs were measured, including: lift block to head height, clean back, sit to stand, perineal care, reach back of head, lift shopping bag from floor, drink from mug, eat with spoon, reach opposite axilla, driving, abduction and forward flexion. These data served as input to the UK National Shoulder Model (UKNSM)[3,7] which calculates shoulder contact forces [3]. The contact forces were normalised by defining start and end point of each motion and the data were interpolated between these points using a cubic spline function.

Results. 41% of the ADLs (11/27) produced shear forces of at least a third of the glenohumeral contact force with the shear force for a pick and place task exceeding 50% of the contact force. The shear forces during the other ADLs such as eating and drinking were negligible.

Discussion. This is the first study to analyse shoulder contact forces during all essential activities of daily living. The results demonstrate that the shear forces for 11 of 27 activities are up to 40% of the stabilising joint compressive force. These values represent novel insight into shoulder loading and will improve implant design and rehabilitation planning.

Conclusion. This study provides novel insight into the loading of the shoulder during activities of daily living and will aid implant design and rehabilitation planning.

Key references. [1] Anglin et al. (2000). Proc Mech Eng. 214:637-644. [2] Bergman et al. (2007). J

46 Biomech. 40:2139-2149. [3] Charlton and Johnson (2006). Proc Mech Eng. 220:801-812. [4] Geraldes et al. (2016). J Orthop Res. 24:1-10. [5] Lazarus et al. (1996). J Bone Joint Surg Am. 78:94-102. [6] Lee et al. (2000). J Bone Joint Surg. 82:849-857. [7] Pandis et al. (2015). Clin. Biomech. 30:839-846.

Mechanical behaviour of human ACL and PCL insertion sites as function of strain rate investigated at the macro-to-nano scale level by SAXS/WAXS techniques Erica Di Federico (1) Nick J. Terrill (2) Antony M.J. Bull (1) Centre for Blast Injury Studies, Department of Bioengineering, Imperial College London, London, UK (1) Diamond Light Source Ltd., Diamond House, Harwell Science and Innovation Campus, Didcot, UK (2)

Background. Ligaments attach to bone through a functionally ordered structure (enthesis) with spatial gradients in composition, organization, and mechanical properties. This structural arrangement minimizes stress concentration permitting the transmission of load between mechanically different materials and varies, dependent on loading rate [1, 2]. Entheses injuries are of special concern as to obtain an appropriate recoupling of the ligament to the bone that is effective in enabling stress transfer in a physiological manner, the restoration of the insertion structure is required [3, 4]. The absence of integrative solutions for the restoration of this interface have directed much of the current effort toward tissue engineering strategies focused on the in vitro production of a suitable ligament-to-bone insertion [5]. However, further studies are required to gain a more detailed knowledge of the enthesis structure and the mechanisms by which load is transferred across the native tissue-to-tissue interfaces, to enable the design of suitable scaffolds for enthesis replacement and ultimately for the clinical translation of these tissue engineering approaches to the functional and integrative repair of ligament injuries. The aim of this study is to analyse the structural organization and tissue mechanics of human ACL and PCL insertion sites as a function of strain rats using in-situ mechanical tensile testing combined with SAXS/WAXD imaging techniques.

Methods. Human ligament-bone specimens were harvested from the cadaveric tibial and femoral insertion site of both ACL and PCL, gripped in a micro-tensile tester with video extensometry [5] for tissue strain observation and tested in tension at 0.001/s, 0.005/s and 0.01/s. Fibril and mineral strain were measured from the percentage shift of the azimuthally integrated intensity of 1st order collagen peak and (002) mineral peak of SAXS/WAXD spectra, respectively (I22, Diamond Light Source, UK). Macro and Nano-mechanical parameters such as tissue and collagen fibril modulus, collagen fibril and mineral crystal strain and mineral degree of orientation were defined.

Results. Tissue and collagen-fibril modulus were both higher in bone compare to fibrocartilage and ligament. The highest strains were found at the ligament mid-substance which progressively attenuated at the fibrocartilage and bone. A similar trend was observed for the mineral phase, with increased mineral strain to tissue strain ratio at higher strain rate. A progressive increase in both collagen fibril and mineral particles degree of orientation with increasing strain magnitude was also observed. The ACL and PCL insertions exhibit the same trends as described, however, these varied between tibial and femoral insertions.

Discussion. The gradual increase in elastic modulus and fibre alignment from the non-mineralized to the mineralized regions of the insertion are consistent with a nano-structural system that promotes load transfer between these dissimilar tissues. Human ACL and PCL mechanical properties are sensitive to strain rate at different length scales. Dissimilarities in the mechanical behaviour of femoral and tibial insertions of the cruciate ligaments may be attributed to significant differences in their structure and composition, with the femoral enthesis characterised by a higher fibrocartilage content and a more acute ligament attachment angle than the tibial insertion site [7].

Conclusions. Findings of this study demonstrated that the macro and nano-mechanical properties and related performances of the insertion are region-dependent, providing a deeper understanding of the loading environment within the ligament-to-bone interface. To our knowledge this is the first study to investigate tissue mechanics and organization of the human enthesis across strain rates

47 at multiple length scales.

Key references. [1] Momersteeg TJA, et al. J Biomech, 1995. 28:745 -752, [2] Bonner TJ, et al. J Mech Behav Biomed Mater, 2015. 41:261–270, [3] Spalazzi JP, et al. J Orthop Res, 2006. 24:2001–10, [4] Pendegrass CJ, et al. J Orthop Res, 2004. 22: 1316–1324, [5] Helen HL, et al. Annu Rev Biomed Eng, 2013. 15:201–226, [6] Karunaratne et al. J Bone Miner Res, 2012.27:876– 890. [6] Momersteeg TJA, et al. J Biomech, 1995. 28:745 -752, [7] Beaulieu ML, et al. J Orthop Res, 2015. 33: 1811-17.

In Silico Size Effects in Cancellous Bone Carl Muscat (1) Marcus Wheel (2) Philip Riches (1) Department of Biomedical Engineering, University of Strathclyde, Glasgow, UK (1) Department of Mechanical Engineering, University of Strathckyde, Glasgow, UK (2)

Background: Cancellous bone can be said to exhibit several important length scales which must be included in order for it to be modelled appropriately. Classical elasticity does not take into consideration these length scales; thus, there is a need for different continuum theories which enable the incorporation of such length scales so that the behaviour of cancellous bone can be better understood and modelled [2,3,4]. The aims of this study are to test the hypothesis that, due to its microstructure, cancellous bone will exhibit size effects when loaded in silico, and also to test whether any size effects are consistent with non-local continuum elasticity theories such as micropolar (Cosserat) theory. Such data is vital in understanding whether a suitable continuum theory can be used to model cancellous bone. Methods: A large sample of trabecular bone was taken from a bovine distal femur and a micro- computed tomography (µ-CT) scanner (Skyscan 1172, Bruker) was used to obtain a high resolution segmented image. Morphometric analysis was used using the Skyscan software; CTAn. This determined the average trabecular length, 푙푡. A large surface mesh was then generated and used to create three groups of 10 different samples of size 푑 × 푑 × 2푑 with 푑 = 2푙푡, 3푙푡 and 4푙푡 maintaining a constant aspect ratio between samples. Solid meshes were generated from the surface meshes using 3-Matic, which were tested in silico under unconfined uni-axial compression and uni-axial torsion using finite element analysis (ANSYS 16.1) assuming that the individual trabeculae exhibit isotropic elasticity (퐸 = 12 퐺푃푎, 휈 = 0.3). Normalised compressive and torsional stiffness were calculated to ascertain any size effects. Results: Normalised torsional and compressive stiffness vary with size as shown in Figure 1 (ANOVA, p < 0.002). In both testing modalities, the largest sample was significantly stiffer than the smallest two samples (Tukey, p < 0.002). Discussion: Cellular solids, foams and Voronoi structures are known to exhibit size effects in shear, torsion, bending and compression [1,2,6] and have been used to model cancellous bone in order to find couple-stress constants [4]. It can be observed that Figure 1: Torsional and compressive stiffness against as size increases, the samples size. Data represent mean +/- one standard error. become stiffer, which is inconsistent with micropolar (Cosserat) theory. Furthermore, the fact that a size effect is observed in compression is also inconsistent with this theory (no size effects are predicted under compression). It can also be noticed that as the size of the sample increases, the overall effect of the microstructure starts to decrease and it can be theorised that the stiffness would approach an asymptote. In the studies by Goda et al. (2014 and 2015) [3,4], it is assumed that micropolar and couples stress theories respectively can be used as a continuum theory to model the size effects of cancellous bone. However, in this study, it has been found that cancellous bone exhibits size

48 effects that are contradictory to the effects predicted my micropolar and couple stress theories Conclusion: Our study suggests that micropolar and couple stress theories do not appropriately model the constitutive behaviour of cancellous bone. References 1. Chen et al, J Mech Phys Sol,50 (5): 955-977, 2002 2. Eringen A., Microcontinuum Field Theories, Theory of Micropolar Elasticity 101-248, 1999 3. Goda et al, Biomech Model Mechanobiol, 13:53-83,2014 4. Goda et al, J Mech behav Biomed Mater, 51:99-118, 2015 5. Onck et al, Int J Mech Sci, 43 (3): 681-699, 2001 6. Tekoglu et al, J Mat Sci, 40: 5911-5917, 2005

The ionic contribution of proteoglycans to meniscal load transmission Fahd Mahmood (1,2) Jon Clarke (1,2) Philip Riches (1) University of Strathclyde, Glasgow, UK (1) Golden Jubilee National Hospital, Glasgow, UK (2)

Background. Load transmission is a primary function of the meniscus1. In articular cartilage and intervertebral disc, proteoglycans play a key role in maintaining hydration via negatively charged moieties which generate Donnan osmotic pressures2. A similar role for proteoglycans in meniscal tissue has not been established. We aimed to investigate the role of proteoglycans in meniscal tissue using confined compression experiments in solutions of varying ionic concentration. Methods. Whole bovine menisci were excised on day of slaughter and frozen at -20°C. 8mm diameter, ~5mm thick circular samples were punched out, wrapped in clingfilm to prevent swelling and allowed to defrost for 2 hours. Each sample was placed within a confined compression chamber, permeable both top and bottom. The apparatus was bathed in either distilled water, 0.14M PBS or 3M PBS before being subjected to 5% ramp compressive strain. 3M PBS solutions are predicted to be negate all Donnan osmotic pressures whilst deionised water is predicted to negate mobile ion gradients3. The hold phase lasted for 300 seconds. FEBio (v2.4, Univ of Utah) software was used to fit results to a non-linear poroviscoelastic model with strain dependent Holmes-Mow permeability4. Analysis was conducted using one-way ANOVA with Tukey post-hoc analysis. Significance was set at p <0.05. Results. 10 samples were tested in each solution. The resultant relaxation curve is shown below:

0 0 50 100 150 200 250 300

-0.5

-1 Force Force (N) -1.5

-2 Time (secs)

0.14M PBS Deionised water 3M PBS

Mean values (S.D) E k (strain M (exponential Beta (exponential G1 T1 (Young's dependent strain dependent stiffening (viscoelastic (relaxation Solution Modulus) permeability) coefficient) coefficient) coefficient) time) 0.14% 0.42 68.85 PBS (0.32) 0.53 (0.39) 0.00 (0.00) 0.11 (0.19) 0.55 (0.26) (24.39) Deionised 0.38 52.30 water (0.23) 0.89 (0.59) 0.00 (0.00) 0.00 (0.00) 0.59 (0.18) (19.78)

49 0.05* 1088.89 3% PBS (0.08) 0.01* (0.02) 0.00 (0.00) 0.05 (0.09) 0.00* (0.00) (2324.25) *p<0.05 compared with 0.14% PBS/deionised water Discussion. These results suggest that ionic effects play a significant role in modulating the mechanical behaviour of meniscal tissue. It is hence important to include the influence of ionic effects when developing mathematical models of this tissue. In keeping with other tissues, it is likely these effects are mediated via proteoglycans. Conclusion. Proteoglycans significantly influence the mechanical behaviour of meniscal tissue through ionic effects. Key references 1 - Shrive NG, O’Connor JJ, Goodfellow JW. Clin Orthop Relat Res 1978;(131):279–87. 2 - Heneghan P, Riches PE. J Biomech. 2008;41(11):2411–6 3 - Mow, VC, Schoonbeck, JM, 1984. Trans Orth Res Soc: 9, 262 4 - Holmes, MH, Mow, VC, J Biomech 1990. 23, 1145–1156

An Automated Workflow for Short Hip Stem Placement and FEA analysis in Total Hip Arthroplasty Zahra Asgharpour (1) Amelie Sas (2) Pim Pellikaan (2) Sjoerd Kolk (1) Thierry Scheerlinck (3) Harry van Lenthe (2) Materialise N.V., Leuven, Belgium (1) Biomechanics Section, KU Leuven (KUL), Leuven, Belgium (2) University Hospital Brussels, Brussels, Belgium (3)

Background. Total hip arthroplasty (THA) is one of the most successful orthopaedic procedures performed today. The operative technique is standardized and aims to restore the hip joint after damage or joint diseases such as osteoarthritis, inflammatory arthritis or trauma [1]. It is estimated that over 300,000 total hip arthroplasties are performed each ear in the United States alone. A large set of different designs are available which can be divided based on the fixation method or size. Although THA is a commonly performed procedure, still a characteristic set of complications occur. Recently short stems have been designed with the aim to spare soft tissues and proximal bone, allowing more bone to be available in case of a revision surgery. This stem design would be helpful for the increasing group of younger patients undergoing THA [2]. Due to the anatomical variation in the population and variety of stem design and positioning in the hip the studies are getting much larger and hence the need for automatization is increasing. In this study, Finite Element method has been employed to study the mechanical parameters such as stress-strain distribution in the implanted femur compared to healthy femur taking into account the variability in femur geometry, position of the implant and loading conditions. In order to minimize the computational effort, a structured and automated process has been employed. Methods. Numerical method is used to analyze the effect of implant placement in the bone. The Models of the femur were reconstructed from CT-images using the Mimics Innovation Suite (Materialise N.V. Leuven, Belgium). Three main anatomical parameters were automatically measured using a script including the neck-shaft angle (NSA), neck length (NL) and anteversion angle (AV). These parameters are often used to represent the anatomical shape of the proximal femur and play an important role in the biomechanics of the hip [3], [4]. Using scripting functionality in the software, CAD design files of the stems were automatically sized and aligned into the femurs to restore an anatomical hip rotation center. Stem size and position could be fine-tuned manually. The segmented femurs were further preprocessed for finite element analysis. The Finite Element analysis proceeds in Abaqus (Dassault Systemes) where the femurs are subjected to peak joint/muscle forces during walking and stairclimbing.

Figure 1: An illustration of the automated workflow Results, Discussion & Conclusion. The results of the finite element simulation have been evaluated for six healthy samples and six implanted samples. The strains in the femur with a large NSA were much higher than the corresponding strain values in the femur with a small NSA. In order to achieve the optimal implant position there is a need for an automated and structured process which will facilitate the pre surgical planning. This study showed that the finite element models of the femur can be reconstructed from CT-images in an automated manner, hence reduce the computational time and efforts for a large number of population with anatomical variations. Key references. 4. Garellick G., Kärrholm J., Lindahl H., Malchau H., Rogmark C., Rolfson O., The Swedish National Hip Arthroplasty ,et al. Annual Report 2014. In Register, SHA (ed), Göteborg, Sweden. 5. Saito, S. Complications of Total Hip Arthroplasty and Their Prevention and Management. Japan Medical Association Journal, et al. 44(4): 165-170 6. G. Duda, C. Küning, G. Bergmann, S. Tohtz, C. Perka, and M. Heller. Biomechanics of the artificial hip, et al. Surgery of the hip, pages 19–33. Elsevier, 2013. 7. R. U. Kleemann, M. O. Heller, U. Stoeckle, W. R. Taylor, and G. N. Duda. The loading arising from increased femoral anteversion and offset may lead to critical cement stresses, et al. J Orthop Res, 21(5):767–74, 2003

51 Wednesday 13th September 2017, 15.30 – 17.30 Cells and biocompatible materials

Poro-elastic properties of the Actifit meniscal scaffold in confined compression Fahd Mahmood (1,2) Jon Clarke (2) Philip Riches (1) University of Strathclyde, Glasgow, UK (1) Golden Jubilee National Hospital, Glasgow, UK (2) Background. The Actifit meniscal scaffold is designed to augment meniscal repair through encouraging ingrowth of meniscal tissue (1). The implant is sutured into place within a defect in the meniscus and the influx of blood encourages deposition of meniscus-like tissue within the scaffold (2), which subsequently degenerates (3). It is composed of polyester and polyurethane components (4) and is designed to encourage the ingrowth of meniscal tissue (5) in damaged or torn menisci. There has been no independent characterisation of the mechanical properties of the scaffold. We aimed to determine the mechanical properties of the Actifit in confined compression. Methods. 5mm diameter, 4mm thick circular samples of an Actifit scaffold were placed within a confined compression chamber, permeable top and bottom. The apparatus was bathed in distilled water before being subjected to 5% ramp compressive strain. The hold phase lasted for 3600 seconds. FEBio (v2.4, Univ of Utah) software was used to fit results to a non-linear poroviscoelastic model with strain dependent Holmes-Mow permeability. Results. Five samples were tested. The mechanical parameters derived using the finite element analysis software are presented in the table below. Parameter Value (mean [standard deviation]) Young’s modulus 0.62 (0.15) MPa Strain dependent permeability 0.66 (0.23) x10-15 m4/Ns Exponential strain coefficient 0.006 (0.00) Exponential stiffening coefficient 1 (0.88) Viscoelastic coefficient 0.15 (0.09) Relaxation time 138.38 (77.63) seconds Equilibrium modulus 0.84 (0.2) MPa Discussion. Whilst the values for Young’s modulus and permeability presented here are similar to those present in the literature for meniscal tissue tested in confined compression (6,7), the value for the equilibrium modulus is somewhat higher. Conclusion. Similar mechanical properties to the native meniscus are likely to help encourage ingrowth of meniscal tissue into the scaffold, aiding regeneration of injured meniscal tissue. Key references. 1. Leroy A, Beaufils P, Faivre B, Steltzlen C, Boisrenoult P, Pujol N. Actifit® polyurethane meniscal scaffold: MRI and functional outcomes after a minimum follow-up of 5 years. Orthop Traumatol Surg Res. 2017 2. Heijkants RGJC, van Tienen TG, Ramrattan NN, van Calck R V, de Groot JH, Veth RPH, et al. Biological evaluation of porous degradable polyurethan scaffolds as meniscus replacement. University of Groningen; 2004. 3. van Minnen B, van Leeuwen MBM, Kors G et al, In vivo resorption of a biodegradable polyurethane foam based on 1,4-butanediisocyanate.J Biomed Mater; 2008;85A(4):972–82. 4. Heijkants RGJC, van Calck R V., de Groot JH, Pennings AJ, Schouten AJ, van Tienen TG, et al. Design, synthesis and properties of a degradable polyurethane scaffold for meniscus regeneration. J Mater Sci Mater Med; 2004 Apr;15(4):423–7. 5. de Groot JH, Zijlstra FM, Kuipers HW, Pennings AJ, Klompmaker J, Veth R, et al. Meniscal tissue regeneration in porous 5050 copoly implants. Biomaterials; 1997;18(8):613–22. 6. Joshi MD, Suh JK, Marui T, Woo SL. Interspecies variation of compressive biomechanical properties of the meniscus. J Biomed Mater Res.1995 Jul;29(7):823–8. 7. Martin Seitz A, Galbusera F, Krais C, Ignatius A, Dürselen L. Stress-relaxation response of human menisci under confined compression conditions. J Mech Behav Biomed Mater. 2013;26:68–80.

Characterization and control of a biocompatible macro-emulsion system Julian Dye (1) Matt Potter (2) Hua Ye (1) Zhangfeng Cui (1) Xuxin Lim (1) University of Oxford, Oxford, UK (1) Oxford University Hospitals NHS Trust, Oxford, UK (2)

Background. Biocompatible emulsions have a wide range of applications in the food, pharmaceutical and medical industries. However, literature on control of parameters affecting biocompatible emulsions are limited. Moreover, the mechanisms by which proteins interact with emulsion systems are not well elucidated. In this study, we describe factors which can be readily manipulated for control of emulsion droplet size and stability, and propose a novel approach for studying protein-emulsion interaction. Methods. A mixture of mineral oil, excipient, surfactants and aqueous buffer were used to create a model physiologically compatible emulsion system. Manufacture control variables were as follows: surfactant hydrophile-lipophile balance (HLB) & concentration, shear rate, excipient concentration and temperature. Emulsion stability was assessed using Turbiscan LAB Stability Analyzer (Formulaction). Emulsion viscoelastic properties were measured using the DWS Rheolab Microrheology machine (LS Instruments). Droplet sizing was characterized by measuring droplet diameter via light microscopy and ImageJ. To study protein adsorption on interfacial boundaries, FITC-albumin was added to the emulsion system and imaged using confocal microscopy. Quantification of protein interaction within the aqueous and oil phase was done by relative fluorescence measurement using the Spectramax i3x Multi-Mode Detection Platform. Qualitative assessment of protein conformation changes was performed with FTIR and tryptophan fluorescence. Statistical analysis was performed using ANOVA (GraphPad Prism (GraphPad Software Inc.). Results. There was a mean droplet diameter significant decrease in as HLB value increases. Temperatures between 4 and 37 °C did not result in significant changes to mean droplet sizes. Increase in surfactant and excipient concentration resulted in significantly smaller droplet sizes. Additionally, at low concentrations of surfactant, increase in mixing speeds resulted in decrease in mean diameter of emulsion droplets down to the fragmentation limit (f) whereby all available surfactants are already at the interfacial region of the emulsions. Increasing viscosity of the aqueous phase of the emulsions system by using high excipient concentrations also resulted in smaller mean droplet sizes. Emulsion stability was also improved by increasing surfactant and excipient concentrations. FITC-albumin preferentially adsorbed forming a boundary layer around the emulsion interface, rather than being evenly distributed within the continuous phase. However, breaking the emulsion by centrifugation, resulted in recovery of label in the aqueous phase and tryptophan fluorescence confirmed preservation of native albumin conformation. No protein was detected in the cream phase by fluorescence or FTIR. Discussion. A reversible interaction between globular proteins and the aqueous emulsion interface may arise from entropic exclusion from free solution or attractive between surfactant-protein interactions. Figure 1. FITC-Albumin The combination of sorbitan and polyethylene oxide surfactants distribution at the emulsion function as an effective barrier to protein denaturation at the oil interface. Scalebar 500 µm. interface. This property is an important requirement for emulsion biocompatibility.

Conclusion. We demonstrate the parameters for control of a novel biocompatible emulsion system and characterize the interaction between the protein-emulsion at interfacial boundaries. This reversible adsorption process may have a role in drug delivery or biomedical applications in the future.

53 A bio-mimetic highly porous scaffolding system for in situ post-irradiation characterization of pancreatic cancer cellular behaviour Stella Totti (1) Mark Allenby (2) Susana Brito Dos Santos (2) Roger Webb (3) Andrew Nisbet (1,4) Athanasios Mantalaris (2) Eirini Velliou (1) University of Surrey, Guildford, UK (1) Imperial College London, London, UK (2) Advanced Technology Institute, Ion Beam Centre, Guildford, UK (3) Royal Surrey County Hospital, Guildford, UK (4) Background. Pancreatic cancer, the fifth leading cause of all cancer related deaths in the UK and the eighth worldwide1, is a highly malignant solid tumour with poor prognosis and lack of efficient treatment screening options. The development of advanced bioinspired and biomimetic in vitro platforms a better understanding of pancreatic cancer cellular behaviour of specific individuals to treatment, therefore accelerating the application of novel and/or personalised therapies from bench to bed. In the emerging field of tissue engineering, three-dimensional (3D) scaffolding systems2-4 can recapitulate more accurately features the in vivo tumour microenvironment, i.e., structure, architecture, porosity, biochemical features as well as oxygen, nutrient and treatment (chemotherapy-radiotherapy) distribution/gradients. We have previously reported the development of a 3D highly porous polyurethane scaffolding system which enabled the long-term cultivation (up to 5 weeks) of pancreatic cancer cells.2,4 In our current work we explore the effect of different extracellular matrix proteins (RGD, collagen -I and a blend of collagen-I, laminin and fibronectin) on the cancer growth kinetics in the 3D system. Furthermore, we perform in situ cellular characterisation of the ex vivo 3D tissue, in order to map areas of different cellular properties that could potentially affect the resistance of the tumour to treatment. Additionally, we evaluate the response of the pancreatic cancer 3D model to different types of irradiation treatment (classical radiotherapy, i.e., X-rays and novel proton therapy). Methods. The pancreatic adenocarcinoma cell lines PANC-1 and AsPC-1 were cultivated for 5 weeks within the polyurethane scaffolds2,4. The scaffolds were coated with different extracellular matrix proteins, i.e. fibronectin, laminin, collagen-I and a mixture of them (collagen-I, laminin, fibronectin) and uncoated scaffolds were used as controls. Scanning Electron Microscopy was conducted to evaluate cell adhesion and distribution within the scaffold. In situ confocal microscopy of representative scaffold sections took place for visualization of extracellular matrix production (collagen-I and laminin-2), proliferation (Ki-67) and stress response markers (HIF-1a, LC3). Furthermore, quantification of extracellular laminin and fibronectin production was performed by measuring the protein concentration in the spent medium. Irradiation experiments were conducted with an absorbed dose of 8 Gy for both protons and x-rays. Results and Discussion. The pancreatic cancer cells were viable up to week 5 for all coating conditions. The specific cell kinetics were protein coating cocktail dependent. Scanning electron microscopy verified that pancreatic cells adhere within the scaffold pores. With confocal microscopy we confirmed that cells in the scaffold are Ki-67 positive up to the culture end point. The cell population that was closer to the edges of the scaffolds was HIF-1a negative and a small region of hypoxic cells was present at the centre of the scaffold, which was also culture period dependent. Cell clusters were observed within the 3D scaffolds the size and density of which were time and location dependent, i.e., bigger/denser clusters were formed at the edges of the scaffold. Furthermore, significant amount of extracellular matrix production was detected for both cell lines. Finally, for the irradiation experiments, cell viability significantly dropped only 24 days post irradiation, whereas cellular detachment was more evitable 24 hours post proton irradiation, as compared to photon irradiation indicating the potential of proton therapy for pancreatic cancer. Conclusion. Our findings show that our 3D pancreatic tumour like system allows the maintenance of cellular properties in terms of proliferation and extracellular matrix production. Normoxic versus hypoxic areas in our 3D platform are able to mimic the in vivo solid tumour oxygen distribution. Additionally, our 3D scaffolding system allows long term post irradiation cell survival monitoring, which enables the application of fractionated irradiation that is commonly used for in vivo pancreatic cancer treatment. Therefore, our platform is a very promising, versatile, low cost tool for in vitro studies and for high throughput treatment screening of pancreatic cancer. Key references. 1. American Cancer Society. Cancer Facts and Figures 2016. 2. Totti, S., et al., Eur Cell Mater, 31 (Supp 1):441, 2016.

54 3. Totti, S., et al., Drug Discov Today, 22 (4), 690-701, 2017. 4. Totti, S., et al., Eur Cell Mater, 33 (Supp 2):0249, 2017.

An Agent-Based Model of Interacting Stem Cells and Cancer Cells Marzieh A. Tehrani (1) Cecile Perrault (1) Dawn Walker (1) University of Sheffield, Sheffield, UK (1) Advancements in tissue engineering combined with the disease seeking nature of stem cells have provided new grounds for targeted therapy of cancer. However, the discrepancies found in existing literature on the role of un-modified stem cells at tumour sites [1], indicates the need for further research. Computational modelling coupled with cell culture experiments can provide further insight into the mechanisms behind stem cell and cancer cell interactions and ultimately assess the efficacy of stem cell therapy of cancer. Through the development of an agent-based model (ABM), the resulting effects of known and hypothosised rules regarding individual cell characteristics and cell-to-cell interactions are simulated. An agent-based model is a bottom-up approach that uses information regarding cell behaviour at the single cell level to generate emergent cell population results. Where possible, the model rules are informed and the final model predictions validated using results and observations obtained from cell culture experiments run simultaneously, allowing for a one-to- one mapping of in vitro and in silico results (figure 1). In vitro approaches provide an insight into actual cell behavior under given conditions. However, these methods are limited by factors such as cost, time and technological advancements in available protocols. In silico tools provide means for quantitative analysis of accumulated data in addition to exploring scenarios and queries otherwise impossible to create in the lab. However, these tools can lack in accuracy and realistic correlation with actual biological behaviour. The combination of both in vitro and in silico methods results in a powerful tool that compensates for the limitations of both approaches, and can be used to address issues such as the efficacy of stem cell mediated therapy of cancer.

Figure1: An example of one-to-one mapping of in silico and in vitro results. Top: Emergent pattern of ABM showing the expansion of mixed population of stem cells (red) and cancer cells (green). Each iteration is equivalent to 20mins of real time. Bottom: Fluorescent microscopy images of co- cultured stem cells (red) and cancer cells (green). [1] Klopp, A. H et al,. 2011. Concise review: Dissecting a discrepancy in the literature: do mesenchymal stem cells support or suppress tumor growth? Stem Cells, 29(1), pp. 11-19: 6119-6126.

Impacting cancer cells via mechanical waves: can we change cellular behaviour? Marlies Glatz (1) Marco Fiorito (1) Gilbert Fruhwirth (1) Ralph Sinkus (1) King's College London, London, UK (1)

Background. The majority of cancer deaths are caused by cancer metastasis, where cancer cells leave the primary tumour and form outgrowths at a secondary site. Microenvironmental signals of (bio)chemical and mechanical nature play a key role in this process. [3,4] Work has shown that mechanical forces (stresses), displacements (shear) and deformations can translate into biochemical signals affecting cancer cell survival. [1,2,5] Here we study a novel approach of utilizing oscillatory low frequency shear waves to non-invasively impact cancer cell growth and invasion. We validated the method in vitro, with the ultimate aim to investigate in vivo the changes in metastatic potential in a therapeutic setting. Methods. The metastatic human breast cancer cell line MDA-MB 231 was used to form tumour spheroids in a bovine collagen type I matrix (Fig.A). The stiffness of the collagen matrix was measured using magnetic resonance elastography (MRE) and the maximum area of shear stress was simulated (Fig.B). To investigate the effect of focused shear waves on cancer cells, the tumour spheroids were exposed to shaking under the following conditions: 3 h at 10 Hz with 100 µm displacement. This treatment was repeated on three consecutive days and the spheroids were monitored by brightfield microscopy throughout the experiment. Quantification of both spheroid growth and cell invasion were performed by excluding pixels underneath a certain intensity threshold; after exclusion the software creates a map distinguishing between the mother core and the migrating cells of the tumour spheroid. Results. Spheroid growth was accompanied by invasion of MDA-MB-231 cells into the surrounding collagen matrix (Fig.C/green) when samples were left static. When spheroids were exposed to shaking once per day (cf. Methods), the spheroids grew to a smaller extent and invasion into the matrix was reduced (Fig.C/red). Image quantification demonstrated a reduction of spheroid growth by a factor of ~2 as well as a reduction of invasion by a factor of ~3 (Fig.D and E).

Discussion. The experiments showed that repeated (A) Tumour spheroids are embedded in bovine collagen type 퐈-gel. Non shaken wells serve as control, whereas the shaken wells undergo the 3h treatment at 10 Hz, 100µm displacement in uniaxial direction. (B) Mapping the maximum shear strain pattern within the well predicts the optimum position of the tumour spheroid. Therefore Magentic Resonance Elastography (MRE) was performed to characterise the stiffness of the collagen gel followed by the simulation of the maximum shear strain pattern for 10Hz and 100µm displacement within the well. (C) Representative images of the tumour spheroid development in bovine collagen type 퐈 over the whole experimental duration. Images framed in green represent the “baseline” non shaken group; images framed in red show the growth and invasion of the shaken tumour spheroid. Scale bar: 500µm (D) Growth comparison of the mother core of the tumour spheroid between the two conditions. The growth of the shaken spheroid is reduced by a factor of 1.97. (E) Plot showing the proportion of invading cells between shaken and non shaken tumour spheroids. The invasion in the shaken group is reduced by a factor of 2.98. (shaken N=4;non exposure to a defined low frequency shear stress (10 Hz, shaken N=3) P by 2way ANOVA 100 µm) reduced the ability of MDA-MB-231 spheroids to expand and invade into the surrounding collage matrix. Conclusion. The results indicate that the duration of the 3 h treatment is enough to lower the growth and invasion rate of breast cancer cells. As such, MRE could be used not only as a monitoring tool, but also as part of a therapeutic treatment. Key references. In alphabetical order, numbered. 1.Ahmadzadeln, PNAS2017,14:1617-1626 2.Barney, Chemical Engineering 2016, 11:85–93; 3.Cheng, PlosOne2009;4(2):e4632; 4.Polacheck, PNAS 2014;11(7) no. 72447–2452; 5.Valastyan, Cell2011;147(2):275-92

56 Mechanobiology of Human Stem Cell Derived Osteocytes in Bone-on-Chip Elisa Budyn (1,2) Morad Bensidhoum (3) Bertrand Cinquin (1) Patrick Tauc (1) Ecole Normale Superieure de Cachan, Cachan, France (1) University of Illinois at Chicago, Chicago, IL, USA (2) University Paris 7, Paris, France (3)

Background. Osteocytes are long-term living cells [1] responsible for bone mineralization and characterized by a body bearing numerous processes. Osteocytes are at the center [2] of bone homeostasis and coordinate the interactions between the osteoclasts and MSCs during remodeling. Osteoblasts differentiate in osteocytes or bone lining cells [3] that cover non- remodeling endosteal bone surfaces, are recruited during the activation phase to seal the site of bone formation by osteoblasts, modulate the osteoclastic resorption and further differentiate into osteoblasts [4]. However fundamental knowledge of human cells in situ biology is challenging. Organs-on-chip offer the opportunity to develop in situ 3D imaging and modelling of concurrent chemo-mechanical phenomena in controlled conditions and long-term culture. The application of mechanical load to live systems is known to influence the cell differentiation, matrix formation and biological response [5]. In situ measurements of the chemico-mechanical interactions between cells and both native and neo-formed ECM contribute to the understanding of tissue regeneration mechanism to design successful repair of massive bone loss. Methods. Human mesenchymal stem cells were reseeded in decellularized human bone to create bone-on-chip that were cultured for over 20 months. The cell morphology, expressed genes, and their production of proteins and mineral using PCR, histochemistry, immunohistochemistry, in situ immunofluorescence and the conditioned medium were analyses. The cells were mechanically stimulated by the application of micro-load to the bones where the cells reside in custom-made and customized testing machines. The live systems were imaged under confocal microscopy allowing concurrent measurement of variations in calcium cytoplasmic concentrations under mechanical load. Results. The bone-on-chip produced after 109 days an ECM of which the strength was nearly a quarter of native bone and that contained type I collagen at 256 days and calcium minerals at 39 days [6]. The morphology of the cells and the newly-formed matrix were further characterized by immunofluorescence under confocal microscopy at 547 days. Different cells populations were identified and one type was organized in network. F1: Differentiated Discussion. The cytoplasmic calcium concentration variations seemed to adapt to the expected in vivo mechanical load at the successive stages of MSCs in bone- cell differentiation in agreement with studies using fluid flow stimulation [7]. on-chip Conclusion. Different cells populations were identified and one type was organized in network indicating differentiation, reorganization as in vivo in the newly formed matrix and some degree of repair over a long preriod of time. Key references. 1- bone, 2- in vitro, 3- long term culture, 4- mechano-biology, 5- MSCs. [1] L. S. Bella, M. Kayser and C. Jones. The mineralized osteocyte: A living fossil. American J. of Phy. Anthrop., 137:449456, 2008. - [2] L.F. Bonewald, M. Johnson, and M. Kneissel. Preface: The osteocyte. Bone, 54:181, 2013. - [3] S.C. Miller. The bone lining cell: A distinct phenotype? Cal. Tis. Int., 41:1-5,1987. - [4] D.N. Menton, D J Simmons, S L Chang, and B Y Orr. From bone lining cell to osteocyte. Anatom. Rec., 209:29-39, 1984. - [5] W.R. Thompson, G. Uzer, K.E. Brobst, Z. Xie, B. Sen, S.S. Yen, M. Styner and J. Rubin, Osteocyte specific responses to soluble and mechanical stimuli in a stem cell derived culture model, Sci. Report, 5:11049, 2015. - [6] E. Budyn, M. Bensidhoum, S. Sanders, E. Schmidt, P. Tauc, E. Deprez, H. Petite, Bone on-chip to study osteocyte mechanotransduction and ECM formation, Eu. Cell and Mat., 32(Suppl. 4), 32, 2016. - [7] X.L. Lu, B. Huo, M. Park, and X.E. Guo. Calcium response in osteocytic networks under steady and oscillatory fluid flow. Bone, 451:466-73, 2012

57 Wednesday 13th September 2017, 15.30 – 17.30 Student/Associate/Trainee III

Radiobiological Analysis of Treatment Plans for Improving Clinical Outcome Thomas McMullan (1) NHS, Edinburgh, UK (1) Background: Two approaches are used to calculate the normal tissue complication probability (NTCP) as a function of the dose delivered to a critical organ volume. The first is the Lyman- Kurcher-Burman (LKB) [3,4] phenomenological model. The other approach is the relative seriality (RS) model, which considers the function of the organ in relation to the volume irradiated when arranged in series, parallel or a combination of series and parallel [2]. A cohort of non-small cell lung cancer patients (n=69) were selected to investigate the performance of the LKB and RS NTCP models using various parameter sets [1,5] from the literature for predicting Radiation Pneumonitis (RP). The dose summary measures V5, V20, V30, max dose, and mean lung dose (MLD) as useful predictors of RP were also assessed. Methods: Dose volume histograms (DVHs) for the non-small cell lung cancer patients were extracted from the Eclipse treatment planning system. The DVH data was imported into Matlab for data analysis, where NTCPs were calculated using the LKB and Relative Seriality models. The models and effects of summary measures were evaluated using univariate, and ROC analysis. Results: The dose volume metrics, V5, V10, and Mean Lung Dose (MLD) were Table 1: Results of univariate analysis. found to be useful predictors V5, V10, and Mean Lung Dose (MLD – paired lung) were found to be useful predictors of RP. None of the NTCP values were statistically of RP. Furthermore, the LKB and Relative Seriality models significant in predicting RP. were found not to predict the clinical outcome very well. The LKB model using the Burman parameter sets performed the best in ROC analysis, with an AUC of 0.69.

58 Discussion: In the cohort studied, the results demonstrate that NTCP calculations alone are not sufficient to predict RP. Other factors such as V5, V10, and Mean Lung Dose (MLD) should also be considered when evaluating treatment plans. Conclusion. V5 was found to be the most useful predictor of RP. The NTCP models are not powerful predictors on their own. Future work could incorporate the summary measures such as V5, V10, and Mean Lung Dose (MLD) directly into the NTCP models to improve their predictive capability. References :1. Burman C, et al. Fitting of normal tissue tolerance data to an analytic function. Int J Radiat Oncol Biol Phys 1991;21:123-135. 2. Kallman P, et al. Tumour and normal tissue response to fractionated non-uniform dose delivery. Int J Radiat Oncol Biol Phys 1992;62:249-262. 3. Kutcher GJ, et al. Histogram reduction method for calculating complication probabilities for three dimensional treatment planning evaluations. Int J Radiat Oncol Biol Phys 1991;21:137-146. 4.Kutcher GJ, et al. Calculation of complication probability factors for non-uniform irradiation. The effective volume method. Int J Radiat Oncol Biol Phys 1989;16:1623-1630 5. Seppenwoolde Y. et al. Comparing different NTCP models that predict the incidence of radiation pneumonitis. Int J Radiat Oncol Biol Phys 2003;55:274-735

Steady-State Strain Rate as a Measure of Damage Accumulation in Tendon Rowan Nicholls (1) Mark Thompson (1) University of Oxford, Oxford, UK (1)

Background: Tendinopathy is a debilitating and widely prevalent problem for which there is currently little mechanistic understanding. Insight into this area might be made by considering the factors that influence tendon damage accumulation, such as diabetes [1] and functional role [2]. However, while both a loss of elasticity [3] and the development of an unrecoverable elongation [4] have been adopted by the field as quantifications of tendon damage, these are merely phenomenological markers. Experiments that incorporate the viscoelastic nature of tendons, such as stress relaxation [5], cyclic [4] and static fatigue [3] tests, are often used to try elucidate more direct links between the microscale mechanisms and macroscopic phenomena. This particular investigation follows on from a pilot study (preliminary results presented at MEIBioEng 2016, Oxford, UK) into the effects of both functional role and of an STZ-model of diabetes on the susceptibility of murine tendon to damage. A viscoelastic property - steady-state strain rate - was measured during creep tests to failure and found to have a significant negative correlation with STZ treatment vs control (p = 0.001 in tail tendons, p = 0.01 in Achilles tendons) and likewise with energy-storage vs positional functional roles (p = 0.05 in STZ-treated tendons, p = 0.02 in control tendons). The present study aimed to establish a link between this steady-state strain rate and traditional measures of tendon damage. Methods: Material testing apparatus (Zwick UK) was used to apply cyclic tensile load to murine tail tendons (n = 5) from a single specimen (Wistar; Charles River Laboratories, UK). A loading pattern of 1 minute at full load followed by 10 minutes at no-load (sufficient to allow full viscous relaxation) was used. 'Full-load' differed for each specimen, ranging from 24 to 32 MPa in increments of 2 MPa, while 'no-load' was consistently 750 kPa. Elasticity was measured during each loading phase and

59 permanent strain measured during each no-load phase. The data from each full-load phase was then extracted and concatenated to create a pseudo-static fatigue test - strain vs time data for the sample under a constant load. Steady-state creep rate was then calculated as the gradient at the point of infection. Results: Steady-state strain rate showed a strong positive linear relationship (r = 0.88) with change in permanent strain over time, but only a weak relationship (r = 0.48) with change in stiffness. Discussion: This study is thought to be the first to investigate steady-state strain rate as a measure of tendon damage accumulation rate. It has the potential to inform future investigations into the links between macro- and microscale properties - necessary for the development of more robust models of damage - and into the treatment and prevention of damage. Further testing that incorporates microscale imaging should be performed to confirm the link between strain rate and damage. Conclusion: Steady-state strain rate is indeed correlated with an increase in permanent strain over the course of a creep test, but not with an increase in loading modulus. This suggests that the damage mechanisms that cause permanent strain are related to those that influence strain rate. Key References: [1] Holmes, G. B., & Lin, J. (2006). Etiologic factors associated with symptomatic Achilles tendinopathy. Foot & Ankle International, 27(11), 952–959 [2] Ker, R. F., Alexander, R. M., & Bennett, M. B. (1988). Why are mammalian tendons so thick? Journal of Zoology, 216(2), 309–324. [3] Wang, X. T., & Ker, R. F. (1995). Creep rupture of wallaby tail tendons. Journal of Experimental Biology, 198(3), 831–845. [4] Fung, D. C., Wang, V. M., Laudier, D. M., Shine, J. H., Basta-Pljakic, J., Jepsen, K. J.,Flatow, E. (2009). Subrupture tendon fatigue damage. Journal of Orthopaedic Research, 27(2), 264–273. [5] Screen, H. R. C., Toorani, S., & Shelton, J. C. (2013). Microstructural stress relaxation mechanics in functionally different tendons. Medical Engineering and Physics, 35(1), 96–102.

60 3D Printing as an alternative production process for the manufacture of cranial plates Georgina Curwen (1) University of Cambridge, Cambridge, UK (1)

Background. Cranial plates are traditionally manufactured by planishing a piece of metal around a treated and coated skull segment. Cambridge University hospital has been looking into the use of 3D printing as one of a number of options for alternative techniques for producing titanium patient-specific cranial plates. 3D printing is an emerging technology within the medical industry and has been used increasingly for models, teaching, guides and implants among a number of other applications3. This has facilitated mass customisation for medical devices. This method offers advantages in saving operating room procedure time, it can enable a less invasive procedure, 2 reduce cost as well as risk of infection . 5 Cranial plates are classed as custom-made devices due to their patient specificity . As a result the regulatory requirements imposed on the manufacturer differ to most medical devices. However, the latest definition of a custom-made device specifically excludes devices that are produced via mass customisation4. The loss of qualification as a custom-made device would result in significant alterations to the required process. This study was carried out to look at the roles of departments involved in the production process. This was carried out with particular focus on the regulatory requirements, and how these are expected to change1 and market analysis6. Methods. This study was predominantly based on a review of regulatory documentation and secondary sources. After establishing the current processes for cranial plate production, based on internal documentation, regulatory review enabled gap analysis. A horizon scan allowed the evaluation of future regulatory developments and targeted scanning of competitors in the sector enabled market analysis to be carried out. Results.

The result of the project was a series of process flow diagrams, ranging from the high and low level workflows, and the resultant processes for possible competitive positions. Additionally, a roadmap of standards for 3D printing as an emerging technology in the medical device industry was produced. Discussion. The results proved to be particularly interesting along the theme of future regulatory developments. Three key areas which may cause substantial impact are: the development of the definition of a custom-made device such that cranial plates are excluded; the introduction of 3D printing specific regulations; the impact of Brexit on regulatory requirements and the market. Conclusion. 3D printing provides an excellent alternative manufacturing process for custom-made cranial plates. However, there are a number of considerations that affect the commercial prospect of the project and will need to be continually evaluated. Adaptations to the current process will be required for current and future regulatory compliance. Key references. In alphabetical order, numbered. 1Abass, M. (2017). "The potential impact of Brexit on the medical devices sector." Retrieved 10/05/17, 2017, from https://www.wrighthassall.co.uk/knowledge/legal- articles/2017/03/13/potential-impact-brexit-medical-devices-sector/. 2Bonda, D. J., et al. (2015). "The Recent Revolution in the Design and Manufacture of Cranial Implants: Modern Advancements and Future Directions." Neurosurgery 77(5): 814-824 3 Deradjat, D. and T. Minshall (2016). "Implementation of Rapid Manufacturing for Mass Customisation." Journal of Manufacturing Technology Management (Additive manufacturing special issue: part 2). 4 The European Parliament and The Council of the European Union (2017). Medical Device Regulation. Official Journal of the European Union 5Manning, L. (2011). "Custom skull implants on demand? Exactly." Retrieved 16/05/17, 2017, from http://medicaldesign.com/materials/custom-skull-implants-demand-exactly. 6Medical_Expo (2017). "Cranial Implants." Retrieved 25/05/17, from http://www.medicalexpo.com/medical-manufacturer/cranial-implant-7047.html.

61 Quantification of the forces exerted on to the handrail of a staircase during stair ascent and descent and handrail use strategies Maighread Ireland (1) Thomas O'Brien (2) Costis Maganaris (2) Azzam Taktak (1) Dept of Medical Physics and Clinical Engineering, Royal Liverpool University Hospital, Liverpool, UK (1) Faulty Sport and Exercise Sciences, Liverpool John Moore’s University, Liverpool, UK (2)

Background Stair negotiation presents a daily, yet challenging and hazardous task, especially for older people. In the UK, over 300,000 injuries and 550 deaths annually result from falls on domestic staircases [1]. Previous studies of falls risk on stairs have examined the lower limb biomechanics during ascent and decent, including 3-D motion and force plate analysis. However, we know little about how people use handrails to assist with stair negotiation. One reason for this is that until recently there have been no instrumented systems to quantify the forces applied on the handrails. This project sought to develop a calibration protocol for a novel instrumented handrail system, with the aim of identifying handrail negotiation strategies during staircase ascent and decent

Method The purpose-made instrumented handrails are built with strain gauges, and load cells to read loads in three-dimensions. The project aim was achieved in three stages. The first stage conditioned the handrail system output signal to remove or account for sources of error. The second stage was developing a calibration protocol for the system, allowing the handrail system output to be quantified. The final stage was identification of staircase negotiation patterns throughout a gait cycle, of staircase ascent and descent for 6 participants.

Results During staircase ascent, three different strategies were identified from three participants. These strategies were similar in the vertical loading direction, however variation in strategy was Figure 2: Photograph of the purpose- observed in the anterior-posterior loading direction. Peak built instrumented staircase normalised force was recorded for all 3 participants during 2nd double support phase, average peak vertical force was 1N/Kg. In staircase descent, three participants used the same handrail strategy, however subtle variations on location of centre of mass relative to hand placement resulted in different versions of the strategy.

Discussion The main sources of error that were accounted for or reduced were in the system were; background noise, cross-talk, and hysteresis. The calibration protocol devised in the second stage allowed the handrail system output to be converted to a functional value representing force, from Volts to Newtons. In Stage 3, some handrail strategies were identified from the handrail system. These demonstrated different techniques for supporting and off-loading body weight from the legs during stair negotiation.

Conclusion This study validated the handrail system as a tool to identify handrail use strategies. Future studies will look at whether those at high-risk to falls utilise similar or alternate strategies. The advantages and disadvantages of these techniques, in relation to staircase safety needs further analysis.

Declaration of Conflicting Interest: No conflicts of interest. References [1] Industry., U.D.o.T.a., DTI. Home and Leisure Accident Surveillance System (HASS), 24th (Final) report.

62 Towards Objective Medical Device Funding Strategies – The Knapsack Problem Method Thomas Taylor (1) Oxford University Hospitals NHS Foundation Trust, Oxford, Oxfordshire, UK (1) Background. The investment strategy for Medical Devices is one of the most critical components for any healthcare provider’s operational objectives, but one which is often misinformed or lacking in organisational context [1]. The NHS 2014 Five Year Forward View recognises the need for “informed decisions about future investment, as the economy improves” [2]. Several approaches have been described to address this issue, typically aiming to provide prioritisation models for maintenance and replacement strategies [3] [4], but scarcely for larger investment decisions. Methods. This work describes the application of a “0-1 Knapsack Problem” (KP) computational algorithm [5] to inform a capital funding allocation process, in order to maximise the total clinical, financial and operational risks mitigated per unit of spend, within a constrained budget.

Figures 3 and 2 – Map of funding thresholds for increasing budget (left) and risk value against residual budget spend (right)

The model was first evaluated using a Python-based dynamic programming method across a large, artificial, random dataset. Figures 1 and 2 show the output for allocation thresholds, risk value and residual spend, with increasing budget/spend. The same programme was then configured to determine the optimised project funding allocations for a historical capital bidding process dataset. Results. The output comparison against the traditional capital bidding process (CBP) showed an 18.3% increase in total risk mitigated, for an equivalent spend, for funding points allocated using the KP algorithm, as shown in figure 3. Alternatively, setting a constrained risk to that currently achieved by the CBP indicated a reduction in overall spend of 46.7% for KP allocations. Discussion. The results successfully displayed the applicability of this model to the allocation of funding for capital initiatives. The current limitations are recognised due to the lack of granularity and subjective biases associated with traditional 5x5 risk scoring Figure 3 – Comparison between the KP model and traditional CBP, for the 15/16 capital matrices. equipment vetting round Conclusion. This model forms the basis of a wider project roadmap, towards developing objective models for strategic decision making around the management of Medical Devices, by employing computational analytics to minimise risk at the most critical stage, improving quality and safety for patients and staff, while generating long-term efficiency savings for the organisation. Key References. [1] K. Willson, K. Ison and S. Tabakov, Medical Equipment Management, CRC Press, 2013. [2] NHS England, "Five Year Forward View," 2014. [3] N. Saleh and et al., "A New Approach for Preventive Maintenance Prioritization of Medical Equipment," in XIII Mediterranean Conference on Medical and Biological Engineering and Computing 2013, pp. 1059-1062, 2013. [4] N. Hamdi and et al., "An intelligent healthcare management system: A new approach in work-order prioritization for medical equipment maintenance requests," Journal of Medical Systems, vol. 36, no. 2, pp. 557-567, 2012.

63 [5] H. Kellerer, U. Pferschy and D. Pisinger, Knapsack Programs, Springer Science & Business Media, 2004.

A Low-Cost Informatics Approach to Improving Patient Safety through Automatic Incident Detection Haroon Chughtai (1,2) Tim Cross (1) Clinical Services Department, HCA Healthcare UK, London, UK (1) Medical Physics & Biomedical Engineering, UCLH NHS Foundation Trust, London, UK (2)

Background: No healthcare organisation is fully compliant with best practices and there will always be incidents that occur that could be learnt from [6]. Consequently, maintaining and improving patient safety requires monitoring; a prominent method being through incident reporting [5]. Through this individual cases can be investigated, lessons learnt, and feedback used to develop a patient safety culture with improved patient outcomes [4]. However, voluntary reporting is an incomplete source with many incidents not recorded [2]. A learning healthcare system would be improved if these unreported incidents were identified, allowing more lessons to be learnt, and strengthening a culture of safety. While it has been shown that while patient care data can be used to develop automatic event reporting triggers [3], this approach has previously been limited by data unavailability, lack of vendor support, and high cost of development.

We therefore propose an extensible, modular, informatics based approach [1] to tackle this reporting deficit using existing EHRS derived data sources, low-cost open-source software tools and rule-based event triggers. We used identification of unreported major haemorrhages from transfusion data as system proof of concept.

Methods: We inspected multiple EHRS data sources selecting, through exploratory data analysis and multidisciplinary expert knowledge, simple triggers that detected anomalous events. Using this insight we designed a rule-based analysis, identification and reporting pipeline in the R language. Known reported incidents were then compared to detected events to evaluate performance.

Results: Ultimately, we successfully prototyped a modular software system that accesses records from existing EHRS data feeds, identifies abnormal events, and reports these as possible incidents to follow up. Anomalous events from blood transfusion, theatre visit, and discharge data were identified, and a subset linked back to incident reporting records.

Discussion: From an engineering perspective, our developed approach is a simple information- based system that performs data aggregation, outlier detection, visualisation, and reporting. This approach helps address the challenge of unsatisfactory reporting rates, detecting potentially unreported incidents and communicating them in a way that facilitates action and change.

Conclusions: The developed pipeline forms the first steps in improving incident reporting through automated means, and so improving patient safety. Future work would identify new data sources and triggers as well as refine existing one, as well as introducing means of intelligently tuning the system through user feedback. Key References:

[1] P.M. Kilbridge, D.C. Classen, The Informatics Opportunities at the Intersection of Patient Safety and Clinical Informatics, J. Am. Med. Informatics Assoc. 15 (2008) 397–407. [2] A.B.-A. Sari, T.A. Sheldon, A. Cracknell, A. Turnbull, Sensitivity of routine system for reporting patient safety incidents in an NHS hospital: retrospective patient case note review, BMJ. 334 (2007) 79–79. [3] M.K. Szekendi, Active surveillance using electronic triggers to detect adverse events in hospitalized patients, Qual. Saf. Heal. Care. 15 (2006) 184–190. [4] The Health Foundation, Does improving safety culture affect patient outcomes?, 2011. [5] C. Vincent, Reporting and Learning Systems, in: Patient Saf., John Wiley & Sons, Ltd, Chichester, UK, 2010: pp. 75–95. [6] C. Vincent, The measurement and monitoring of safety, Heal. Found. (2013) 10.

64 Wednesday 13th September 2017, 15.30 – 17.30 Radiotherapy Imaging

Optimisation of Head and Neck Cone Beam CT for Image Guided Radiotherapy Anna Hughes (1) Michael Pearson (1) Barts Health NHS Trust, London, UK (1)

Background. Increased use of cone beam CT (CBCT) for Image Guided Radiotherapy, with some treatments involving CBCT every fraction, has resulted in the concomitant imaging dose accrued during treatment becoming more significant. Under IR(ME)R1 optimisation is fundamental in radiation protection and while there are manufacturer pre-set CBCT protocols for different tumour sites there is scope to optimise for the individual patient and purpose. The AAPM Task Group 75 Report2 gives specific guidance about imaging in radiotherapy. It states that imaging dose should not be considered negligible compared to treatment dose and describes how risk/benefit trade-offs are more complex than in diagnostic imaging as imaging can improve the therapeutic ratio. At Barts Health NHS Trust pelvis scans have been optimised for patient size using a method similar to that described by Wood et al.3 but manufacturer default settings are still in use for head and neck CBCT. Other studies have measured CBCT image quality in an anthropomorphic phantom4 and imaging dose using thermoluminescent dosimeters (TLDs) and CT dose index (CTDI) 5. The purpose of this work was to create an optimal CBCT mode for head and neck scans on Varian Clinac IX linacs and then measure the imaging dose. Image quality must be satisfactory for viewing the body outline and bony matching while keeping dose as low as reasonably practicable. Methods. Local head and neck patient size was analysed using the parameters set by the automatic exposure control on the planning CT scans of 119 patients. The difference in lateral separation from the neck to the shoulders was a bigger consideration than the difference in patient size therefore the CBCT mode was optimised for a single patient size. An anthropomorphic phantom was imaged with range of kV, mA and ms using a half fan technique to achieve a 45 cm diameter field of view. Quantitative image quality parameters (noise and contrast to noise ratio) were investigated. Noise defined as the standard deviation in HU of soft tissue regions of interest was found to be the most practical image quality measure. This was plotted against the exposure parameters and an optimal mode selected from the resulting curves. TLDs were calibrated for use in the imaging beam and used to measure absorbed dose from CBCT scans at positions of organs at risk in the anthropomorphic phantom. CTDIw was measured in the standard CTDI phantom. Results. The 45 cm field of view (FOV) allows the body contour to be outlined in more inferior slices compared to the 25 cm FOV of the manufacturer default mode. Noise was significantly higher in the shoulder slices than neck slices (21 HU vs. 43 HU for the manufacturer mode, 17HU vs. 42 HU for the new mode) but is similar in the manufacturer and new mode. Absorbed dose to the patient is lower for the new mode. At TLD positions within the imaging field there are dose reductions from 54-74% (estimated errors <5%). This is corroborated by the CTDIw measurements which show a 61% dose reduction. Imaging dose using the new mode is less than 1% of typical treatment point doses to that region. Discussion. The larger FOV of the new mode allows the body to be contoured in more inferior slices. This is useful for assessing the need to replan due to weight loss particularly for inferior PTVs. Image quality is far poorer in the shoulders then the neck due to the larger separation, so this is the limiting factor for dose reduction. Lower exposure parameters have caused the dose reduction but without an increase in noise, this may be explained by the difference in half and full fan reconstruction techniques. Conclusion. A new CBCT mode optimised for outlining the body contour and bony matching on the spine has been created for head and neck radiotherapy patients at Barts Health NHS Trust. This has a larger field of view, the same level of noise and 61% less dose than the manufacturer

65 default mode. Key references. 1 The Ionising Radiation (Medical Exposures) Regulations. 2000. 2 Medical physics, 34(10), pp. 4041-4063 3 British Journal of Radiology, 88(1055), pp. 20150364. 4 Journal of applied clinical medical physics / American College of Medical Physics, 17(2), pp. 6040. 5 International Journal of Radiation Oncology Biology Physics, 80(1), pp. 291-300.

Commissioning and Clinical Implementation of Stitched CBCT for Gynaecological Tumours Christina Agnew (1) Nithya Kanakavelu (1) Denise Irvine (1) Belfast Health and Social Care Trust, Belfast, UK (1)

Background. VMAT/IMRT can be used in the treatment of gynaecological tumours to create steep concave dose distributions to spare surrounding organs at risk such as the bladder, rectum and bowel (3). Organ motion is of particular concern for this treatment site as the CTV can comprise multiple structures that can move relative to each other and is impacted by bladder and rectal filling (1,2). Pre-treatment CBCT can be used to monitor OAR position and ensure adequate target coverage. Varian TrueBeam CBCT has a minimum length of 17.6 cm. Gynaecological treatments typically have length > 17.6cm. Using a single CBCT can only image a portion of the treatment site and therefore only verifies the position of a portion of the OARs. Furthermore, the best match over the entire treatment field may differ from the best match of the imaged portion, introducing an error into the treatment delivery. Margins to compensate for this uncertainty could negate the benefits of the IMRT/VMAT delivery. The ability to stitch multiple CBCTs together is now available as an advanced imaging feature in Varian Truebeam v2.5. The aim of this study was to commission the stitch CBCT feature and together with a radiographer led team, develop an imaging protocol to implement stitch CBCT for all VMAT gynaecological treatments. Methods. The accuracy of the multiple stitched reconstructed CBCT was assessed by comparison of a geometric bar and a pelvic phantom imaged with a single CBCT and a stitched CBCT. The noise level in the stitched overlap region was assessed and used as a surrogate for dose. An off- line no action level imaging protocol was developed as the stitch CBCT reconstruction is only available post treatment. Results. The stitch CBCT was acquired resulting in an overlap region of 1cm. Reconstruction errors <0.3mm were noted for both the simple geometric bar and the pelvis phantom. A decrease in the noise level of 10% was measured in the overlap region, relating to a 10% increase in dose in this region. The time required to reconstruct and save the stitched images is ~4 minutes.

1.2

0.7

0.2

30x30 noise Level NormalisedCT -20 -10 -0.3 0 phantom10 20 Phantom position (cm)

Figure 1: Stitched image of a simple geometric Figure 2: A comparison of the relative noise bar. level in a 30x30x20cm solid water phantom, imaged with a single CBCT and a stitched CBCT. Discussion. Stitching multiple CBCTs using Varian TrueBeam advanced reconstruction software provides an accurate representation of the imaged structure to <0.3mm. The increase in dose in a 1 cm overlap between CBCTs is of the order of 10%, which may be justified to ensure accurate positioning over the entire length of the tumour site. The time required to reconstruct and save the stitched CBCT from the treatment console must be taken into account during patient scheduling. Conclusion. Stitched CBCT is a useful tool for verifying the patient position for treatment sites of

66 length > 17.6 cm. Key references. 1) Heijkoop ST, et al. Radiother Oncol 2015; 117: 536–541 2) Lee JE, et al. Gynecol Oncol 2007;140:145–51 3) Roeske JC, et al. Radiother Oncol 2003;69:201e207

Towards automating respiratory gating QA Sergio Esteve (1) Sarah OS Osman (2) Denise M Irvine (1) Conor K McGarry (1,2) Radiotherapy Physics, Northern Ireland Cancer Centre, Belfast Health and Social Care Trust, Belfast, Northern Ireland, UK (1) Centre for Cancer Research and Cell Biology, Queen’s University, Belfast, Northern Ireland, UK (2)

Background. In preparation for local implementation of gating using the Varian RPM system, a quality assurance (QA) protocol for gating was developed and followed daily on two linear accelerators (linacs) [1]. The aim was to develop a fast semi-automated QA technique for monitoring the accuracy and reproducibility of gated treatment delivery and imaging over time. Methods. A phantom was constructed by attaching a ball-bearing to a Varian rotary gating phantom that converts the rotary motion to superior-inferior (G-T) and vertical (A-P) motion of the ball-bearing with a periodic breathing signal (amplitude = 0.94 cm, period = 5.3 s). The phantom was scanned on consecutive days on two Varian TrueBeam v2.5 linacs. In each session a fluoroscopy image was acquired to quantify and verify the full range of the ball-bearing motion. A kV triggered image was then acquired using amplitude-based gating (lower threshold 0.85 cm) at the start of beam-on. An integrated MV image was also acquired during the delivery of a 50 MU plan in amplitude-based gated mode. A MATLAB tool for analysis of the ball-bearing position on the acquired images was developed. The position of the centre of the ball-bearing in the triggered kV image and integrated MV image was quantified and validated over the first 3 fractions and used as a reference. Subsequent acquired images were then compared to the reference positions in the kV and MV images. Results. Figure 1 shows the user interface of the analysis tool developed. The program displays the triggered kV image and integrated MV image, the difference between the reference and the current ball-bearing position, the current couch position and the trend over time. Figure 1 shows a deviation greater than 1 mm due to an error in couch positioning.

Table 1 shows the difference between the reference values and results over 16 weeks. Linac 1 kV Linac 1 MV Linac 2 kV Linac 2 MV A-P (mm) 0.11 ± 0.26 0.06 ± 0.10 0.17 ± 0.33 0.08 ± 0.09 G-T (mm) 0.25 ± 0.30 0.22 ± 0.31 0.31 ± 0.69 0.33 ± 0.70 Discussion. In this study the accuracy and reproducibility of the RPM respiratory gated delivery system was assessed using a simple gating phantom and an in-house developed software tool. The position of the ball-bearing was highly reproducible with (on average) sub-millimetre accuracy. An initial tolerance of 1 mm was set to flag that gating required review. Couch positions were also recorded to ease trouble shooting as this was found to be the most probable cause of error. Further work will be presented which will allow tolerances to be set.

67 Conclusion. We have developed a semi-automated quality assessment protocol for gated treatments. This protocol provides a fast and accurate method to evaluate the gating accuracy on the linacs rather than inspecting the images on the workstation using the graticule. This protocol will be used to observe trends in the triggered images and set tolerances and action levels. Key references. [1] Keall, PJ., et al. "The management of respiratory motion in radiation oncology report of AAPM Task Group 76." Medical physics 33.10 (2006): 3874-3900

Dose determination of retrospectively 4DCT in radiotherapy Julian Liu (1) DhanaJayan Kothandan (1) Aysha Begum (2) Cancer Centre London, London, UK (1) Nottingham University Hospitals NHS Trust, Nottingham, UK (2)

Background: 4DCT had been used for radiotherapy treatment planning for dealing with organs movement caused by breathing, and the patient dose levels were different from those of 3D CT imaging. Many studies investigated patient dosimetry of imaging. The aim of this study is to determine the dosage of retrospectively gated scans using Siemens AS 4DCT. Scanning parameters, dose efficiency, noise levels, CT number accuracy, and the expected number of acquired images and applications were analysed, in order to develop dose reduction strategies. Methods. Three phantoms including CTDI head and body, RMI electron density and dynamic thorax were scanned using current clinical protocols for retrospectively gated 4DCT and 3D thorax scans, whilst controlling for range and image reconstruction locations. The dose measurements of air, CTDI head and body phantom scans were analysed and compared with manufacturer calculated CTDI and DLP. The noise levels of ROIs in an individual 4DCT image, a maximum intensity projection image and an averaged CT image were analysed and compared with those from routine 3D images. The scanning parameters of the protocols used were: 120 kV, 2 mm slice thickness for the RMI phantom and 3 mm for the dynamic phantom, 16 x 1.2mm beam, 0.5 seconds per rotation, B30f kernel, pitch of 1.0 for thorax/abdomen and 0.09 for 4DCT, and variable mAs with and without Caredose. Results: Over a scanning length of 24mm, using 26mAs and 26mAs/rot for 4DCT and Abdomen protocols respectively, the ratio of measured dose (air, head and body phantom), CTDI(W) and DLP between the two scans were 8.5±1.2, 9.21±0.26 and 8.66±1.73, respectively. Using 65mAs and 193mAs/rot with the addition of caredose, the ratio of measured dose for body phantom, CTDI(W) and DLP were 2.85, 3.19 and 3.5, respectively. Using the RMI electron density phantom, the average noise levels of the determined ROIs for all the inserts for both individual 4DCT and thorax images were consistent with the default mAs. With the addition of Caredose, the CT numbers of all RMI electron density inserts of the 4DCT images were within ±3 HU compared to those of the thorax images. Using a dynamic phantom, three ROIs were selected from the right and left lung, and the mediastinum. 10, 30 and 50 mAs were applied for the thorax protocol scans, and 10 and 17 mAs were applied for the 4DCT scans. The CT numbers were -800.1±2.0 for the lungs, and -13.3±1.3 for the mediastinum. Conclusion and discussion: In order to cover the retrospective 4DCT scanning range, a margin is needed which is approximately equal to the beam width. With the addition of Caredose, the reference mAs for 4DCT were significantly lower than those of standard 3D protocols, resulting in a three to four fold increase in CTDI(W) rather than an expected 10 fold. If the sole purpose of 4DCT is to generate averaged CT images and maximum intensity projection (MIP) for contouring, it is preferable to use the lowest available mAs, although the CT numbers would be of lower accuracy. Furthermore, the margin scan dose could be reduced significantly if manufacturers were able to make dynamically moving collimators. Key references 1. DeMarco J J et al, Evaluation of patient dose using a virtual CT scanner: Applications to 4DCT simulation and Kilovoltage cone-beam imaging. 2008 1742-6596 102 012006. 2. Li H S et al, Direct dose mapping versus energy/mass transfer mapping for 4D dose accumulation: fundamental differences and dosimetric consequences. 2014 Phys. Med. Biol. 59 173.

68 3. Milz S et al, A dose error evaluation study for 4D dose calculations. 2014 Phys. Med. Biol. 59 6401. 4. Roland T et al, A radiobiological analysis of the effect of 3D versus 4D image-based planning in lung cancer radiotherapy. 2009 Phys. Med. Biol. 54 5509. 5. Zhang Y et al, Improving 4D plan quality for PBS-based liver tumour treatments by combining online image guided beam gating with rescanning. 2015 Phys. Med. Biol. 60 8141.

Investigating patient movement in intracranial stereotactic radiosurgery treatments using machine log files Adam Agnew (1) Karl Jordan (1) Deirdre King (1) Mohamed Galal (1) Hermitage Medical Clinic, Dublin, Ireland (1)

Background. To account for intra-fractional patient movement during stereotactic radiosurgery, Cyberknife utilises two orthogonal ceiling mounted kV sources that can be used to track patient movement with a variable imaging interval of between 15 and 150 seconds according to the user's discretion. Imaging doses can be optimised by choosing an imaging interval which will keep imaging dose to a minimum while retaining delivery accuracy. Log files recorded during the treatment delivery can be used to establish both optimum imaging interval and also a suitable treatment margin to account for the patient movement that cannot be accounted for by imaging12. Methods. Log files for 56 intracranial patients (100 fractions, 4000 image pairs) were retrospectively assessed for patient movement. Movement between consecutive images was investigated, along with the movement of the patient over the duration of the treatment. Results. The standard deviation for the movement between patients was found to be 0.2 mm for sup/inf movement and 0.3 mm for ant/post and left/right movement. The rate of drift between the images was used to calculate that an interval of 28 seconds was required to restrict the translational movement of the patient to 0.1 mm for 95 % of the population. It was found that from image to image, patient movement did not vary with respect to interval times Discussion. As no relationship could be found between the interval times and movement, it cannot be assumed that the patient movement between images is linear. This movement cannot be accounted for by additional imaging and therefore can be accounted for with planning margins. To ensure target coverage, an additional margin of 0.6 mm (2SD patient movement) has been chosen. An imaging interval of 30 seconds has been chosen to ensure patient drift is kept to below 0.1 mm. Conclusion. Log files can be used to evaluate patient movement and used to create planning margins and imaging intervals. Key references. 1. Hoogeman, M., Prévost, J., Nuyttens, J., Pöll, J., Levendag, P. and Heijmen, B. (2009). Clinical Accuracy of the Respiratory Tumor Tracking System of the CyberKnife: Assessment by Analysis of Log Files. International Journal of Radiation Oncology*Biology*Physics, 74(1), pp.297-303. 2. Murphy, M. (2009). Intrafraction Geometric Uncertainties in Frameless Image Guided Radiosurgery. International Journal of Radiation Oncology*Biology*Physics, 73(5), pp.1364-1368.

Accuracy of 3D surface imaging system compared to conventional volumetric registration in patient positioning of radiotherapy pelvis treatment Oi-Ching Choi (1) Cancer Centre London, Wimbledon, London, UK (1)

Background. 3D imaging has shown advantageous results on breast patients in detecting set-up errors without any radiation (Deantonio et al 2011 & Alderliesten et al 2013). This study aims at studying the co-relation of the surface registration with the cone beam computed tomography (CBCT) in radiotherapy pelvis treatment. Methods. 12 pelvis patients with 267 fractions were selected which all of them had CT planning scans. During treatment, 3D surfaces were captured by a surface imaging system (AlignRT) prior to subsequent setup procedure. The set-up errors were verified by the CBCT before treatment beam delivery. The discrepancies were calculated when comparing with the original planning CT scans. The resulting errors were compared with linear regression analysis and Bland-Altman plots. Results. The Pearson correlation between setup errors were 0.57, 0.51, 0.65 in left-right (LR), craniocaudal (CC) and anterior-posterior (AP) directions respectively. For the differences between setup errors: The group means, systematic errors and random errors were 0.10cm, -0.30cm, - 0.09cm; 0.13 cm, 0.37 cm, 0.29 cm and 0.18 cm, 0.22 cm, 0.22 cm in LR, CC, AP directions respectively. The paired t-test for random errors showed a significant difference between the two systems along all direction (all with t<0.0001). Discussion. The setup measurements by the 3D surface imaging has good correlation with the setup errors detected by CBCT. It can be used to assess the setup reproducibility for pelvis patients and reduce the number of setup corrections while using the CBCT. Conclusion. AlignRT is radiation free, simple and fast method to analyse the setup alignment in 3DCRT of pelvis cancer patients. It is good at detecting large setup errors (i.e. wrong shift) before doing any CBCT. However, the systematic errors of using AlignRT to detect setup errors were too big and this should not reduce the routine use of CBCT clinically. Figure 1: Scatterplots with regression lines of the translation part of the setup errors

LR CC AP

2y = 0.5087x + 0.0173 y = 0.2675x - 0.1733 y = 0.8094x - 0.1468 -2 0 2 -3 -2 0 2

-2 0 2 -2 CBCT CBCT (cm) Align RT (cm) CBCT (cm) -3 AlignRt(cm) Figure 2: Bland-Altman plots to depict the translational part of the setup errors. Top and bottom dashed lines: 95% limits of agreement. Middle dashed line (bias): mean of the differences between setup errors.

LR CC AP

Upper LQA = - 0.473 Upper LQA = 0.429 Upper LQA = 0.347 Bias = 0.09 Bias = 0.100 Bias = - 0.300 Lower LQA = - 0.155 lower LQA = -0.652

Lower LQA = - 1.031 Difference Difference (cm) Average (cm) Key references. 1. Alderliesten T, Sonke J, Betgen et al 2011. Accuracy Evaluation of a 3-Dimensional Surface Imaging System for Guidance in Deep-inspiration Breath-Hold Radiation Therapy. Int J Radiation Oncol Biol Phys. 85(2) 536-542 2. Deantonio L, Masini L, Loi G et al (2011). Detection of Setup uncertainties with 3D surface registration system for conformal radiotherapy of breast cancer. Rpt of Pract Onco and radiother 16:77-81

Treatment Set-up Accuracy and Reproducibility for Patients Receiving Radiotherapy for Prostate Cancer – Effect of change in imaging practice Ruth McLauchlan (1) Katie Perkins (1) Deborah Brown (1) Stephen Mangar (1) Alison Falconer (1) Imperial College Healthcare NHS Trust, London, UK (1)

Background. The verification of radiotherapy treatment position based on bony anatomy for prostate cancer has been widely questioned. The position of the prostate within the pelvis is known to be affected by the filling and emptying of the bladder and the rectum which are also the organs at most risk of radiation-induced toxicity. Shifts in prostate position of >5 mm relative to the bony anatomy have been observed both in the literature and in our department when using implanted fiducial markers for visualization of the prostate on planar imaging1. Our current departmental CTV-PTV margins and imaging protocols are based on MV planar imaging matched with bony anatomy. Their applicability to treatment where imaging of the soft tissue target is now possible through the use of on-treatment CBCT was evaluated. Methods. The images for 20 consecutive patients receiving daily CBCT imaging matched to soft-tissue as part of a clinical trial protocol were reviewed by a single, experienced observer and the shifts in position of the prostate at treatment compared with at planning were recorded. The results were used to obtain population random and systematic errors following the method in the On Target2 document and entered into the van Herk3 margin recipe to obtain CTV-PTV margins for this patient group. Individual graphs of shift in target position at each fraction of treatment for each patient were generated to determine if more frequent imaging should be considered for prostate cancer patients than is current departmental practice. Comparison was made with a previous study performed in 2008 when standard imaging practice was MV planar imaging using bony anatomy. Results. Population random and systematic errors for our department in 2008 (MV EPID AP and Rt Lateral pair, bony anatomy match) and in 2017 (kV CBCT, soft tissue match) are given in the table below along with the resulting margins from the van Herk formula. AP [cm] SI [cm] LR [cm] 2008 2017 2008 2017 2008 2017 (MV) (CBCT) (MV) (CBCT) (MV) (CBCT) Population  [cm] 0.12 0.18 0.14 0.11 0.15 0.11 Population [cm] 0.22 0.28 0.22 0.17 0.29 0.23 Van Herk margin [cm] 0.45 0.64 0.51 0.39 0.58 0.44 Discussion. Both the population systematic and random errors in the AP direction are larger in the 2017 study than in 2008 but smaller in the SI and LR directions. This is due to bony anatomy being a poor surrogate for the soft-tissue prostate as was also demonstrated in our pilot study using fiducial markers1. All patients had mean shifts less than 0.5 cm but all had at least one fraction where the shift was up to 1.0 cm (except one patient who on one fraction had a SI shift of 1.3 cm) indicating that daily CBCT imaging should perhaps be considered in some cases, particularly with the move to hypofractionated regimes. However, the data demonstrates that our current CTV-PTV margins of 10 mm are sufficient with our standard imaging practice of CBCT for the first 3 fractions and weekly. Conclusion. Current departmental imaging practice and margins are sufficient to ensure accurate and reproducible treatment for patients receiving radiotherapy to the prostate at our centre but a move to daily CBCT imaging may be justified particularly with a move to hypofractionated treatment regimes. Key references. 1. Challapalli et al. Clinical Oncology 2012 24:590-591 2. On target: ensuring geometric accuracy in radiotherapy. Royal College of Radiologists: 2008. 3. Van Herk et al. Int. J. Radiation Oncology Biol. Phys. 2000 47:1121-1135

71 Thursday 14th September 2017

Thursday 14th September 2017, 08.00 – 09.00 Statistics Teaching Session

Statistics Teaching Session Azzam Taktak, The Royal Liverpool and Broadgreen University Hospitals NHS Trust

This session will briefly cover the subject of study design, sources of bias and how to avoid them. You will learn how to correctly formulate a research question in scientific and statistical terms and how to test a hypothesis. You will then learn how to choose the appropriate statistical test, check the validity of the tests, analyse and interpret the results.

Thursday 14th September 2017, 09.00 – 09.30 Plenary session: IOP MPG sponsored lecture

Moving towards wearable high-resolution brain scanners Barnes GR, Wellcome Centre for Human Neuroimaging, Institute of Neurology, University College London, UK.

Magnetoencephalography (MEG) is the measurement of magnetic fields due to electrical activity in the human brain. By measuring these fields outside the head we are able to track changes in neuronal activity at a millisecond time scale non-invasively. Traditional MEG has used superconducting sensors to measure these fields (which are about 10 million times smaller than the earth’s field). The main problem with superconducting sensors is that they have to be housed in a large cryogenic vessel. This means that MEG, like most neuroimaging systems, is optimized for scanning compliant adults who can remain still. I will talk about our first experiences with a new generation of sensors called Optically Pumped Magnetometers (OPMs). These sensors are not cryogenic and can be placed as close as 6mm to the scalp surface- offering 5-10 fold sensitivity increases (over cryogenic systems) in adults. As these sensors are relatively small they can be worn within a helmet allowing the subject to move naturally during scanning. I will discuss how this new technology will allow us to consider novel clinical and experimental paradigms optimized for traditionally challenging subject groups (such as young children).

72 Thursday 14th September 2017, 09.30 – 10.30 Radiotherapy Advances

Use of Hydrogels to Enhance Rectal Sparing during Stereotactic Abalative Body Radiotherapy using the Cyberknife Platform Ashley Richmond (1) Katrina Finnegan (1) Joshua Gesner (1) Jan Konieczek (1) The London Clinic, London, UK (1)

Background. The use of hydrogrel spacers to help spare the rectum is growing in popularity in the field of radiation therapy2,4 including brachytherapy3. In 2015 The London Clinic became the first UK hospital to use hydrogel SpacoarsTM by AugemixTM. Since then several patients have had them inserted prior to radiotherapy or SABR(CyberknifeTM) treatment. It also offers the possibility of retreating or potentially dose escalation of fractionation regimes. Methods. Three patients were selected who were having SBRT treatment. Fiducials and Spacoar were inserted in one session in day case procedure. The BKTM Ultrasound stepper unit was used transrectally to guide their insertion, which was performed via the transperineal route, whilst the patient was in the lithotomy position. A CT planning scan and MR T1 and T2 sequences were then performed 7 days later for each patient. The resulting scans were imported, fused and planned using MultiplanTM for treatment using PACE protocol(36.25Gy in 5 fractions) on the CyberknifeTM SBRT platform. Dose Volume Histogram (DVH) calculated and volumes are noted at the 37, 36.25, 35, 30, 25 and 20 Gy levels. Three past SBRT prostate patient treatment plans with Spacoar present were compared to the average data set obtained for Spacoar free plans. Results. Resulting data from the DVHs were compared with the volume in terms of percentage and cubic centimetres. Differences observed were 37Gy spacoar(n=3) 0.07% 0.07cc without(n=4) 1.53% 1.03cc, 35Gy 0.37% 0.27cc without 3.6% 2.881cc, 30Gy spacoar 1.93% 1.14cc without 9.55% 6.42cc and 20Gy spacoar 13.07% 8.29cc without 27.10% 18.48cc. Discussion. There was a reduction of dose to rectum as predicted1,5, particularly at the higher dose levels which are thought to be the areas where significant side effects occur. Typically a 1cm gap was generated between the prostate and the rectum. Some variability was observed between patient implants in terms or size and symmetry. Conclusion. The use of Spacoar hydrogels is a safe and effective way of reducing the dose received by ythe rectum. Injection of the Spacoar like fiducial markers is operator dependent and like all new techniques there is a learning curve as shown by the asymmetry of one of the early implants. Key references. 1. “Feasibility of and rectal dosimetry improvement with the use of SpaceOAR® hydrogel for dose- escalated prostate cancer radiotherapy”, Kirsten van Gysen et al, Journal of Medical Imaging and Radiation Oncology 58 (2014) 511–516 2. “Continued Benefit to Rectal Separation for Prostate Radiation Therapy: Final Results of a Phase III Trial”, Daniel A. Hamstra et al, Int J Radiation Oncol Biol Phys, Vol. 97, No. 5, pp. 976e985, 2017 3. Use of a rectal spacer with low-dose-rate brachytherapy for treatment of prostate cancer in previously irradiated patients: Initial experience and short-term results. Mahal et al, Brachytherapy. 2014 May 28. pii: S1538-4721(14)00527-3. doi: 0.1016/j.brachy.2014.05.001. 4. “Hydrogel Spacer Prospective Multicenter Randomized Controlled Pivotal Trial: Dosimetric and Clinical Effects of Perirectal Spacer Application in Men Undergoing Prostate Image Guided Intensity Modulated Radiation Therapy” Neil Mariados et al, Int J Radiation Oncol Biol Phys, Vol. 92, No. 5, pp. 971e977, 2015 5. “Low rectal toxicity after dose escalated IMRT treatment of prostate cancer using an absorbable hydrogel for increasing and maintaining space between the rectum and prostate:

73 results of a multi-institutional phase II trial”. Uhl et al, Radiother Oncol. 2013 Feb;106(2):215-9. doi: 10.1016/j.radonc.2012.11.009. Epub 2013 Jan 16.

Comparison of the PTV dose-volumetric parameters of early-stage cancer patients treated with the Pencil Beam Convolution (PBC) and Analytical Anisotropic Algorithms (AAA) and the association to the patient survival Natasa Solomou (1) Pinelopi Gkogkou (1) David Willis (1) Norfolk & Norwich University Hospitals NHS Foundation Trust, Norwich, UK (1) Background: In 2011 at NNUH, the decision was made to change the planning algorithm from Pencil Beam Convolution (PBC) to Analytical Anisotropic Algorithm (AAA). This was due to the introduction of RapidArc, which can only be used with AAA. It was decided that only one algorithm should be used for photon treatments. The aim of this retrospective study was (1) the comparison of the PTV dose-volumetric parameters of early-stage lung cancer patients planned with the PBC and AAA and (2) the association of the above PTV statistics to the survival of the patients of these two groups. Methods: All patients, treated at NNUH between 2009-2014, underwent 3D radical radiotherapy with 55 Gy/ 20 fractions over 4 weeks (no chemotherapy was received). The planning volume was set as Planning Target Volume (PTV)=Clinical Target Volume (CTV)+0.7 cm in the axial and CTV+1.2 cm in the longitudinal direction. Fourteen patients planned with PBC (Group Ia) and twenty-eight patients planned with AAA (Group II) were selected. All plans of the first group were also recalculated with AAA with pre-set Monitor Units (MUs) and referred to as Group Ib. The D50%, V95%, D2% and D98% PTV dose-volumetric parameters were evaluated for each patient for Groups Ia, Ib and II. The median follow-up period was 24 months post-treatment. AAA models the dose more accurately for lung patients in comparison to PBC [1], [2]. For study aim (1), Groups Ia and Ib were compared to confirm that the dose-volume parameters of Group Ia patients were statistically different when re-calculated with AAA with pre-set MUs. For study aim (2), Group Ib and Group II were then compared to demonstrate the difference in dose delivered to these two patient groups. The difference in survival of Groups Ib and II patients was further studied to investigate whether a change in algorithm had an effect on the patient survival. Statistical analysis was performed with SPSS (IBM Analytics, v. 23). The Wilcoxon signed-ranks test was selected for comparing Groups Ia and Ib. The Mann-Whitney U test and the Kaplan-Meier plot analysis were used for comparing Groups Ib and II. Values of P<0.05 were classed as statistically significant. Results: (1) The difference between each of the evaluated parameters between Groups Ia and Ib was found statistically significant (P=0.001). The difference in the average values between Group Ia and Ib was also calculated as: D50%= -4.9%, V95%=-32.3%, D2%=-3.2% and in D98%= -9.1%. (2) The AAA-recalculated DVHs (Group Ib) showed a greater spread, poor PTV coverage and steeper DVH gradients than Group II. A significant statistical difference was proved between each of the dose-volumetric parameters of Groups Ib and II (P=0.000). The difference in the average values of these evaluated parameters between Groups Ib and II was calculated as: D50%= -4.0%, V95%=- 31.4%, D2%=-2.5% and D98%=-9.2%. The median survival for Groups Ia/Ib was 29 months (Range: 1-77), against 36 months (Range: 8-59) for Group II. The Kaplan-Meier survival plots showed that Group II patients had higher probability of surviving 24 months post-treatment (78%), versus Group Ia/Ib patients (58%) (P=0.058). Discussion: The dose delivered to the patients planned with PBC (Group Ia) was shown to be significantly underestimated. This was further demonstrated by a drop in the survival of this group of patients. Although a non-statistically significant difference was calculated in the survival between the AAA-recalculated (Group Ib) and AAA-planned patients (Group II), a possible trend is shown towards significance. The results of this study are indicative but not conclusive due to the small sample size. Conclusion: The use of PBC is a strong indicator of reduction in the survival of PBC-planned patients. AAA planned-patients showed a significant improvement in coverage, which would imply that this would result in survival improvement. Key references: [1] Knoos T, et al. Limitations of a pencil beam approach to photon dose calculations in lung tissue. Physics in Med. and Biology. 1995; 40(9):1411. [2] Van EA, et al. Testing of the AA algorithm for photon dose calculation. Medical physics. 2006; 33(11):4130-48.

Biomechanical modelling for novel methods of tracking and accumulation of delivered dose to the rectum in prostate radiotherapy Leila Shelley (1,5) Michael Sutcliffe (1,5) Karl Harrison (2,5) Marina Romanchikova (3,5) Simon Thomas (3,5) Neil Burnet (4,5) Department of Engineering, University of Cambridge, Cambridge, UK (1) Cavendish Laboratory, University of Cambridge, Cambridge, UK (2) Department of Medical Physics and Clinical Engineering, Cambridge University Hosiptals, Cambridge, UK (3) Department of Oncology, Cambridge University Hosiptals, Cambridge, UK (4) Cancer Research UK VoxTox Research Group, Cambridge University Hosiptals, Cambridge, UK (5)

Background. Biomechanical finite element modelling (FEM) has been shown to successfully predict anatomical deformations within the context of image registration in radiotherapy [1, 2]. This work presents, for the first time, the application of a biomechanical model of the rectum for accumulation of delivered dose to the rectal wall during prostate radiotherapy. Accumulation of delivered dose, compared to planned dose, could improve understanding of normal tissue complication probability (NTCP), but methods of doing so have been notoriously challenging [3]. Methods. Prostate cancer patients treated with TomoTherapy within our Research Programme, had daily MVCT image guidance scans acquired prior to the delivery of each fraction of radiotherapy. Coordinates of the daily rectal contours, accounting for interfraction variation, were extracted from an in-house autocontouring system [4] and used as input for a 3D model within Abaqus (6.14-1). The rectal model and corresponding finite elements, or voxels, within the model, are grown to fit the autocontours, based on elastic deformation. The dose received by each voxel was calculated using a separate analysis tool [5]. Daily doses were accumulated throughout the course of radiotherapy to calculate total delivered dose, which was compared to planned dose. Data from a further 100 patients will be analysed. Results. Differences have been observed between planned and accumulated FEM-derived doses. Spatial analysis in 3D has shown hotspots of up to 13 Gy higher than accepted in the treatment plan.

Figures (T-B, L-R): Planned dose to rectum, accumulated dose on planning contour (limited by field of view), dose difference, dose-surface histogram; 2D dose surface maps (planned, accumulated, difference) for example patient Discussion. Biomechanical FEM of the rectum during prostate radiotherapy has enabled dose- history tracking at the voxel level, as well as spatial characterisation of dose in 3D, information that is lost when considering dose-volume histograms alone. In the future, we will link the FEM- calculated delivered dose with toxicity data to develop multivariate NTCP models, capable of on- treatment risk analysis to maximise therapeutic ratio. Conclusion. Biomechanical finite element modelling has been used to quantitatively accumulate dose to the rectum during prostate radiotherapy, offering spatial, voxel-level parametrisation of delivered dose, previously unachievable.

75 Key references. [1] Brock, K.K., et al., Med. Phys. 32:1647, 2005, [2] Brock, K.K., et al., Int. J. Radiation Oncology Biol. Phys. 64:4:1245, 2006, [3] Jaffray, D.A., et al., Int. J. Radiation Oncology Biol. Phys., 76:3:S135, 2010, [4] Sutcliffe, M.P.F., et al., CUED Technical Report CUED/C- MICROMECH/TR.100, 2015 [5] Thomas, S.J., et al., Med. Phys., 39:160, 2012.

Treatment plan optimisation for CyberKnife with MLC James Bedford (1) Peter Ziegenhein (1) Simeon Nill (1) Uwe Oelfke (1) The Institute of Cancer Research and The Royal Marsden NHS Foundation Trust, London, UK (1)

Background. The CyberKnife system (Accuray Inc., Sunnyvale, CA) includes a multileaf collimator (MLC), which allows maximal flexibility in field shaping and fewer monitor units in stereotactic radiosurgery than with a cone collimator [1]. The MLC consists of 26 leaf pairs, each of width 3.85 mm, giving a maximum field size of 115 mm x 100 mm at a nominal source-axis distance of 800 mm. In this study, a fast optimiser has been adapted and evaluated with a view to inverse planning for the CyberKnife with MLC. Methods. Two cases each of prostate and base of seminal vesicles, lung, liver and partial breast were considered, with hypofractionated dose prescriptions of 3-5 fractions. Treatment plans were imported into the in-house treatment planning system DynaPlan and dose influence matrices calculated using an Accuray dose model. Each treatment plan consisted of 110 beam orientations, using an average of two apertures per beam orientation (node). Fluence was optimised using the L-BFGS gradient descent method, sequencing was carried out using a standard sequencing method [2], and aperture optimisation was then carried out, also using a gradient descent method [3]. No attempt was made to optimise the numbers of beam directions, apertures or monitor units (MU) in the final plan. This method was implemented in a fast multi-threaded planning framework [3]. Plans were evaluated against the relevant clinical protocols (RTOG 0938, 0813, and 0413). Results. Figure 1 gives examples of the dose distributions for prostate, lung and breast cases. The optimisation method is able to meet the clinical constraints for the majority of cases. Inverse planning time is around 5-10 minutes per run for the cases considered. Median number of apertures is 301, median MU per prescribed Gy is 2406, and median conformality index, defined as VPRES DOSE / VPTV, is 1.02.

Figure 1. Example prostate, lung and breast cases. The dose levels as a percentage of the prescription dose are shown at the top left. Discussion. These results show that the fast optimisation scheme is able to produce plans of a clinical standard within a very practical timeframe. The use of a series of annular structures around the PTV provides dose distributions evenly balanced around the area treated. Further work is needed to optimise the treatment delivery time, which is substantially longer than with the current clinically available CyberKnife treatment planning system. Conclusion. The gradient descent method implemented on a multiple-core computation environment offers the possibility of fast optimisation for MLC-based delivery on the large number of nodes encountered in the CyberKnife system. Key references. 1. Asmerom G,Bourne D, Chappelow J, Goggin L M, Heitz R, Jordan P, Kilby W, Laing T, Maurer C R Jr, Noll J M, Sayeh S and Weber A 2016 The design and physical characterization of a multileaf collimator for robotic radiosurgery Biomed. Phys. Eng.

76 Express 2 017003 2. Xia P and Verhey L J 1998 Multileaf collimator leaf sequencing algorithm for intensity modulated beams with multiple static segments Med. Phys. 25 1424-34 3. Ziegenhein P, Kamerling C P, Bangert M, Kunkel J and Oelfke U 2013 Performance- optimized clinical IMRT planning on modern CPUs Phys. Med. Biol. 58 3705-15

MR for Lung Treatment Planning Glyn Coutts (1) Michael Dubec (1) Corinne Faivre-Finn (1,2) James O'Connor (1,2) Marcel van Herk (1,2) The Christie NHS Foundation Trust, Manchester, UK (1) School of Medical Sciences, University of Manchester, Manchester, UK (2)

Background. MRI provides excellent soft tissue contrast and functional information and can aid in the accurate delineation of primary tumour, involved lymph nodes and organs at risk (OAR) [1 - 3]. The addition of an MR planning scan could lead to more accurate planning volumes and improved overall survival rates. However, cardiac and pulmonary motion, low proton density and large magnetic susceptibility differences associated with lung parenchyma can render lung MR images unusable [2, 3]. The aim of this work is to develop a robust MR imaging protocol allowing tumour and OARs to be delineated more accurately than using CT and PET alone. Methods. Sequence optimisation has been carried out on 12 healthy volunteers using a 1.5T MR scanner (Magnetom Aera, Siemens Medical Systems), to allow visualisation of OARs associated with lung treatment planning. Sequences include: Turbo Spin Echo (TSE), STARVIBE, DIXON TSE, VIBE, STIR and TrueFISP. Several motion compensation techniques have been utilised. Image quality was assessed visually by a consultant radiologist and consultant thoracic oncologist and discussions led to development of sequence optimisation table and permitted further sequence optimisation. Results. The most promising sequences (table 1) for the OAR visualisations were found to be the TSE and STARVIBE. The DIXON TSE was found to be most suitable for visualisation of the brachial plexus.

Table 1. OAR visualisation rated following disucssion with consultant thoracic oncologist and consultant radiologist. Discussion. An MR imaging protocol has been developed for lung treatment planning. The TSE and STARVIBE appear to be most promising for visualising OARs. However, the longer scan times of such sequences may be difficult for lung cancer patients to manage. Hence, the feasibility of acquiring these sequences in patients needs to be evaluated. Conclusion. A set of sequences for optimal OAR visualisation have been determined, namely the STARVIBE, TSE and DIXON TSE. The optimised sequences will now be used in a study involving 35 lung cancer patients. The aim is to compare OAR and target volume delineation across multiple observers allowing inter and intra observer variation to be assessed. CT and PET will also be

77 included in the study to determine whether MR significantly changes how clinicians contour the target, lymph nodes and OARs. Key references. [1] Kauczor H et al. MRI of the Lung 2009. [2] Cobben C et al. TCRT 2015. [3] Puderbach, M Eur J Radiol 2007 64

78 Thursday 14th September 2017, 09.30 – 10.30 Virtual Reality Workshop

Virtual Reality Workshop Dr Jonathan Ashmore, NHS Highland

In the past 4 years virtual reality has become main stream where devices such as the google cardboard allow people to experience VR for as little as £5. One area where VR is expected to have a large impact is healthcare, particularly for its potential to relieve patient anxieties for medical procedures. In this workshop we will discuss, and the audience will get to experience, a VR based MRI preparation app developed by and currently in use at three UK hospitals. We will live demo how to film, edit and host 360 video as an example of how hospitals can develop their own VR preparation tools.

Ultrasound Workshop

Ultrasound Workshop Michael Lynn, Royal Berkshire NHS Foundation Trust

Ultrasound QA programmes, such as detailed in IPEM Report 102, generally concentrate on evaluating diagnostic imaging performance. This is of great importance for machines in Obstetrics/Gynaecology, Cardiology and Radiology. However, increasing use of ultrasound is being made for interventional procedures such as the placement of central venous catheters under ultrasound guidance as recommended by NICE. This workshop will examine the technical factors and image quality parameters which are important for this application. Delegates will use phantoms which emulate this procedure and learn how users can optimise images and avoid inflicting ‘needlestick injury’ on the lenses of probes.

Bladder scanners are increasingly used in healthcare settings and clinical decisions are based on the results. This workshop will demonstrate how to make functional checks of these devices to resolve user queries.

Thursday 14th September 2017, 09.30 – 10.30 Author Workshop

Best practice in writing for publication of scientific papers in a peer-reviewed journal: from an editor’s perspective Richard A. Black Editor-in-Chief, Medical Engineering & Physics

Research has been defined as “a process of investigation leading to new insights effectively shared” [1] hence writing for publication is an essential part of that process. The work of an academic goes hand-in-hand with communicating the results of research, but it is not simply about getting your work into the public domain: it is also about ensuring that a paper reaches the widest possible audience and is read by them; hence the increasing emphasis placed on bibliometrics as a measure of how effective we are at communicating the results of our research. Arguably, to be successful therefore, one also has to be able to communicate effectively in writing, which is a skill that takes time and practice to develop. Nevertheless, it is surprising how often problems occur.

In this presentation I shall remind authors (and reviewers) of their obligations, which are published in the author guidelines of most journal web sites [2-4]. With knowledge of the review process, and the criteria used to assess the quality of a paper submitted for publication, I believe many of these problems we as journal editors encounter when dealing with submissions might be averted.

Authors are required to make a series of disclosures about the work reported in a manuscript: declaring any conflict of interest, perceived or actual; acknowledging the source(s) of funding; and, for work involving human subjects, prior approval by an institutional review body on matters of ethics. Authors are obligated also to ensure also that any images or original artwork, even those that are generated by the authors themselves, and published in conference proceedings, for example, are not subject to copyright. In such cases permission to reproduce the material should be sought from the copyright owner prior to submission. Again, I refer readers to the author guidelines for details of the procedures that should be followed.

This presentation aims to offer practical advice and guidance on best practice when preparing manuscripts for submission to peer-reviewed journals across a range of disciplines [5-7], and to direct authors and reviewers to the resources freely available on publishers’ web sites. Participants will gain insights into the workings of the publication process, and learn how their research and practice fits into the wider research landscape.

References and bibliography [1] Research excellence framework: second consultation on the assessment and funding of research, HEFCE Research Excellence Framework, September 2009 (www.hefce.ac.uk) [2] Understanding the publishing process, Elsevier Ltd. http://cdn.elsevier.com/assets/pdf_file/0020/131816/Understanding-the-Publishing-Process.pdf [3] Elsevier for reviewers www.elsevier.com/reviewers/home [4] PLOS Medicine Guidelines for Authors/Reviewers: www.plosmedicine.org/static/reviewerGuidelines [5] Earnshaw, J.J. (2012) How to write a clinical paper for publication. Surgery (Oxford), 30(9) 437- 441 [6] Whitesides, G. M. (2004) Whitesides' Group: Writing a Paper. Adv. Mater., 16: 1375–1377 [7] Brand, R.A., Huiskes, R. (2001) Structural outline of an archival paper for the Journal of Biomechanics. J. Biomechanics, 34: 1371-1374

Thursday 14th September 2017, 11.00 – 12.30 Research culture: visions of 2035

Research culture: visions of 2035 Royal Society

Imagine that you are a member of the UK research community in 2035: what does your ‘idealised’ research culture look like? How does it support a vibrant and diverse workforce? And how does it maintain and improve research excellence? Discuss these and other issues at this Royal Society run workshop. This event is one of a series taking place across the UK throughout 2017, the outputs of which will be used to develop a vision of what future research culture could look like.

80 Thursday 14th September 2017, 11.00 – 12.30 Diagnostic Radiology

Dose audit approaches for those without a dose management system (DMS) P Charnock IRS Limited, Liverpool

Patient dose audit is a requirement under current IRMER2000 and is expected to remain in the new IRMER2018. Although patient dose audit itself is a requirement, there is also a requirement for optimisation and dose audit is only the beginning of any optimisation programme rather than a means to an end. Publications and initiatives such as the ICRP publication “Diagnostic Reference Levels in Medical Imaging” or the COMARE 16th report and the recommendation for Image Optimisation teams highlight this.

Many advances in data collection have been developed over recent years with a view to larger samples of data being used or complete sets of data being utilised more and more. In particular a number of manufacturers are supporting hospital Trusts and medical physics groups by offering dose management systems which promise to collect and collate data direct from PACS and display or report analyses for the purpose of informing a patient dose and optimisation programme.

However, while the potential for these systems is clear, many Trusts and medical physics groups are still unable to work with a DMS for a number of different reasons. And yet the requirement for dose audit and optimisation is still there and has been since 2000.

This talk will show attendees that until the manufacturers overcome the issues that are preventing some users form utilising their products, that there are still alternatives that can be and should be used.

Development and Implementation of a DQE technique to assess Image Quality for Digital Radiography Systems in the Saolta University Healthcare Group Matthew Donoghue (1) Brendan Tuohy (1) Emerald House (1) John McFadden (2) University Hospital Galway, Galway, Ireland (1) National University Of Ireland, Galway, Galway, Ireland (2)

The proper definition of a detector’s intrinsic image quality performance is required to ensure that the imaging system faithfully represents the anatomical features of the body part being imaged. The current medical physics practice involves the use of Leeds test object to qualitatively assess the performance of detectors in general radiology. The use of subjective testing can result in inter/intra operative bias which inherently results in uncertainty regarding to the accurate tracking of a detectors performance over a period of time.

Based on IEC standard 62220-1 and published literature [1-3 ]a technique was developed to measure Modulation Transfer Function (MTF), Normalised Noise Power Spectrum (NNPS) and Detective Quantum Efficiency (DQE) “in the field”. While the IEC standard is well designed [1], the following challenges were encountered: the acquisition of the appropriate filter material, sourcing of a suitable aluminium filter fitting the requirements set out in the IEC methodology and the availability of unprocessed raw data images. Typical clinical practice requires the measurements to be taken in as short a time as possible. In this poster or presentation we will give a summary of the methodology and some initial findings from the assessment of flat panel detector Digital Radiography (DR) systems from the Saolta University Health Care Group. A number of detectors from different manufacturers will be used for a comparison.

81 References 1) IEC 62220-1 (2003), Characteristics of digital X-ray imaging devices- Part 1: Determination of the Detective Quantum Efficiency 2) Samei, E Flynn, MJ. Reimann, A Method for Measuring the Presampled MTF of Digital Radiographic Systems using the Edge Test Device, Med. Phys. 1998, 25, p102-113 3) B. Donini, S. Rivetti, N. Lanconelli, Free Software for Performing Physical Analysis of Systems for Digital Radiography and Mammography. 2014, Med. Phys. 41 (5)

Evaluation of CDRAD and TO20 test objects in Digital Radiography Avril Weir (1) Elli-Noora Salo (1) Anna Janeczko (1) Jamie Douglas (1) Nick Weir (1) Department of Medical Physics, Royal Infirmary of Edinburgh, NHS Lothian, Edinburgh, UK (1)

Background. Image quality parameters that assess the visibility of low contrast objects (Threshold contrast, CNR) are a valuable tool to assess imaging performance in digital radiography for routine testing of x-ray systems and dose-image quality optimisation. The aim of this study was to evaluate two test objects containing discs of varying diameter and contrast (CDRAD [2] and TO20 [3]) for this purpose, together with their associated analysis software, with particular emphasis on sensitivity and practicality. Methods. Both test objects were imaged on the same digital radiographic equipment using two different geometries. First, the test object was placed directly on the detector in a scatter-free environment, and images acquired at 3 and 10 µGy receptor dose. Secondly, the test objects were positioned in the centre of 15 cm of Perspex and exposed under automatic exposure control, simulating a patient examination. In order to assess sensitivity of the image quality test results to changes in exposure conditions, the exposure using Perspex was repeated varying tube kilovoltage and mAs by ± 20 %. For each exposure condition five repeat images were obtained. CDRAD Analyser [2] was used to automatically calculate the image quality parameter IQFInv, total number of discs detected and contrast detail curves. AutoPIA software [1] was used to analyse TO20 images to give non-linearised CNR and total discs detected. Additionally, four observers manually determined the CNR and total number of discs detected in an anonymised, randomised subset of the TO20 images. Results. Representative example results showing variation in CNR (TO20) and Total discs detected (CDRAD) with changes in exposure parameters, as calculated by the software, are presented in Figures 1 and 2 respectively. When experienced observers manually scored TO20, the maximum intra- and inter-observer variation was 3 and 4.5 discs respectively. The variability in manually determined CNR was always less than 6 % with all observers.

Figure 1: Variation of CNR with change in exposure Figure 2: Variation of total discs detected with change in parameter (TO20). Error bars are 95 % CI. exposure parameter (CDRAD). Error bars are 95 % CI Discussion: CNR was found to be the most sensitive image quality parameter investigated and number of details detected in TO20 was the least sensitive. Both software programs are relatively easy to use, but CDRAD Analyser was quicker and offered more analysis options. This, together with its good sensitivity and objectivity makes CDRAD and software ideal for optimisation purposes. However, TO20 gives more interpretable information at the x-ray unit, allowing an instant indication of the visibility of low contrast objects, and is more portable. Manual assessment of TO20 was found to have acceptable sensitivity and inter- and intra-observer repeatability, and these are expected to improve through further experience. An advantage of manual analysis of

82 TO20 over using AutoPIA is that linearised CNR can be calculated. Conclusion: CDRAD and TO20 can be used to give a comprehensive image quality assessment of DR x-ray systems. Both can be imaged and analysed easily to give complementary information on low contrast visibility, which is sensitive to changes in equipment performance. Each test object and software has its own strengths. The results of this study have led us to employ TO20 in a scatter-free environment for routine QC testing and CDRAD in Perspex along with CDRAD Analyser for commissioning and optimisation purposes. Key references. [1] AutoPIA (Automatic Phantom Image Analysis), CyberQual, Italy. [2] Manual CDRAD 2.0 Phantom and Analyser software version 2.1, Artinis Medical Systems, The Netherlands. [3] T020, Leeds Test Objects LTD, United Kingdom.

Simulated low dose Cone Beam CT (CBCT) by addition of noise Abigail Bryce-Atkinson (1) Gillian Whitfield (2) Marcel van Herk (1,2) The University of Manchester, Manchester, UK (1) The Christie NHS Foundation Trust, Manchester, UK (2)

Background. It is useful to simulate the effect of reduced exposure settings on image quality to aid making decisions about imaging dose reduction. This has been done previously in diagnostic CT[1-4] and in mobile C-arm CBCT [5] by adding artificial noise into projection images. This work aims to develop a method to simulate low dose CBCT images for the Elekta XVI image guided radiotherapy (IGRT) system and validate its accuracy. Methods. CBCT images of a Catphan phantom were collected using the XVI 5.02 system at various exposures: 64mA/40ms, 32mA/40ms, 32mA/20ms, 16mA/20ms, 16mA/10ms, and 10mA/10ms. The signal-variance relationship was measured using the average standard deviation (SD) measurements from a 20x20 pixel homogeneous region from each projection at each exposure level. This relationship was used to compute the variance of the noise to be added to simulate low dose images from an initial high exposure scan, modifying the projections using a Matlab R2016a program. Using the highest non-saturated exposure (32mA/40ms) as the initial projections, projections were simulated at 0.5, 0.25, 0.125 and 0.078 of the exposure level to correspond and compare with the collected data. The accuracy of the simulation was assessed visually and by comparison of the mean pixel values, SD and contrast to noise ratio (CNR) using 10x10 pixel regions in the Catphan inserts. Results The relationship between signal intensity and variance was almost perfectly linear illustrating that the detector is quantum noise limited. Visually, the simulated exposures were in good agreement with the collected data, but showed slight underestimation of noise (Fig. 1), likely due to ghosting. The average differences between simulated and actual data across all measurements were 2.1%, 19.3% and 19.1% for mean pixel value, SD and CNR respectively. CNR measurements agreed with the visual assessment, showing slightly higher CNR for simulated data (Fig. 2).

Discussion. Our method can be easily applied to commercially available XVI software by replacing actual with simulated low dose projection images to predict image quality of low dose imaging. This is particularly relevant for developing low dose protocols for patients where imaging dose is of the

83 greatest concern, such as paediatrics. Conclusion. Low dose CBCT can be simulated to within reasonable accuracy by addition of Gaussian noise of the correct magnitude for each pixel’s signal to projection images prior to reconstruction. Key references. 1. Frush, D.P., et al., AJR Am J Roentgenol, 2002. 179(5): p. 1107-13. 2. Mayo, J.R., et al., 1997. 202(2): p. 453-7. 3. Massoumzadeh, P., et al., Med Phys, 2009. 36(1): p. 174-89. 4. Hanai, K., et al., J Comput Assist Tomogr, 2006. 30(6): p. 955-61. 5. Wang, A.S., et al., Med Phys, 2014. 41(7): p. 071915.

Estimating patient thickness from a single radiograph – proof of principle Mark Worrall (1) David Sutton (1) Sarah Vinnicombe (2) NHS Tayside, Dundee, UK (1) University of Dundee, Dundee, UK (2)

Background. Dose audit for paediatric patients is a key step in the optimisation of x-ray examinations1, however it is difficult due to the large variation in size between patients across the paediatric age range. In response, the National Radiological Protection Board (NRPB) suggested a method for paediatric patient dose audit in 20002 which involved directly measuring patient thickness or modelling it using measurements of height and weight. However, national patient dose audits performed in the UK in 20003, 20054 and 20105 did not contain enough data that included thickness or height and weight measurements to allow for the NRPB’s analysis methodology to be used.

This work aims to automate the estimation of patient thickness from the radiograph. Methods. Detector Dose Indicator (DDI) and Pixel Value (PV) calibrations were made on a dedicated cassette of a Fuji Computed Radiography (CR) system at varying beam qualities, achieved by varying the kVp and thickness of Solid Water High Equivalence (HE) attenuator. Using these calibrations, the image DDI or PV can be used to estimate the intensity at the detector. The initial intensity can be estimated using the examination Kerma Air Product (KAP) or mAs with a correction made for field non-uniformity. Effective linear attenuation coefficients can be directly measured at varying kVp for a single attenuator, in this case Solid Water HE.

Using these measurements to estimate final intensity I, initial intensity I0 and effective linear attenuation coefficient µ, the Beer-Lambert law can be used to estimate the object thickness. Results. The quality of the x-ray beam that exits a Solid Water HE attenuator of thickness 10cm or higher is very similar because of the presence of low energy scatter. Therefore, a single calibration for DDI or PV at each kVp can be accurately used to estimate the intensity at the detector. The uncertainties associated with the estimates of initial intensity I0 and the effective linear attenuation coefficient µ are low.

For images of stacks of Solid Water HE acquired at 60, 70 and 81kVp, the thickness estimated using this model had an average absolute error of 7.7mm and a maximum error of 11.7mm. Discussion. A model has been created that estimates the thickness of a uniform attenuator within 12mm of the known thickness. This uncertainty is comparable with the direct measurement of a patient and lower than that associated with an estimated thickness using height and weight measurements. This is a promising result for a proof of principle; if the model can be extended to materials of multiple composition with a comparable error this could be the means of solving patient thickness estimation for paediatric radiographic dose audit.

84 It is intended the model be incorporated into digital x-ray equipment; by populating look up tables with prior measurements and having the system make simple calculations, the estimated thickness could be available instantaneously after the image. Conclusion. This proof of principle has demonstrated that it is possible to accurately estimate an object’s thickness using only a single radiographic image and pre-measured values from look up tables. The model is now ready to consider attenuators of multiple composition with a view to making it a clinical tool.

Key references. 1. Hart D, Wall B F, Shrimpton P C, Bungay D R, Dance D R. R318-Reference Doses and Patient Size in Paediatric Radiology, NRPB 2000 2. IAEA human health series. Report 24 – Dosimetry in Diagnostic Radiology for Paediatric Patients, IAEA 2013 3. Hart D, Hillier M C, Wall B F. W14 – Doses to patients from medical x-ray examinations in the UK – 2000 review. NRPB 2002. 4. Hart D, Hillier M C, Wall B F. 29 – Doses to patients from radiographic and fluoroscopic x- ray imaging procedures in the UK – 2005 review. HPA 2007. 5. Hart D, Hillier M C, Shrimpton P C. 34 – Doses to patients from radiographic and fluoroscopic x-ray imaging procedures in the UK – 2010 review. HPA 2012.

85 Thursday 14th September 2017, 11.00 – 12.30 Industry session: Future Innovations in Imaging and Radiotherapy

Project InnerEye - enabling Radiation Oncologists with fast structure delineation Stefan Wijnen, Business Development Manager Microsoft Radiomics , Microsoft Research

Project InnerEye is a research project that uses state of the art artificial intelligence to build innovative image analysis tools to help radiation oncologists to treat cancer in a more targeted and effective way. It allows them to achieve 3D contouring of patients’ planning scans in minutes rather than hours. The assistive AI technology gives the experts full control of the output accuracy, while enjoying high levels of consistency, and enormous cost savings.

The 'dark art' of deep learning and radiology: demystifying the 'magic' Dr Christopher Austin, MBChB, MSc, Radiologist and Medical Director of Imaging AI and Analytics, GE Healthcare

Until recently, the concept of artificial intelligence has mostly been confined to academic discussions or science fiction. But in past few years, the combination of expanding digital data sets, increased computational power and advancements in data science techniques have made AI a reality, enabling machines to 'think' and 'see.' It's not surprising that various industries are embracing opportunities to develop and adopt AI applications that can demonstrate improvements in productivity, quality and safety.

With rising costs, workforce shortages, and a growing need for greater efficiency and effectiveness, it's no surprise that healthcare is preparing to take a seat at the AI table, positioning diagnostic imaging at the front of the queue. Is the arrival of AI in radiology hype or hope? And what are the potential implications to radiologists, specialists, GPs, and importantly patients?

This session will offer an understanding of the science of deep learning, how it is being applied to radiology images and the radiology workflow, and its potential to transform the specialty in exciting, not threatening new ways.

A Multi-professional, multi-organisational study to develop a device to immobilize the breast during radiotherapy: The SuPPORT 4 All study Heidi Probst PhD, MA, BSc(HONS), DCR(T) FCR Heath Reed MA, BSc(Hons) Keeley Rosbottom MSc, BSc(HONS), PgD, PgCert (ed) Andrew Stanton Helen Crank PhD, MSc, BA(Hons) Kirsty Bryan-Jones MSc (Sheffield Teaching Hospitals NHS Trust) Karen Collins PhD Clare Robertson (Panache Lingerie)

Breast cancer is a global problem(1) . Radiotherapy plays an important role in management. New radiotherapy techniques allow sparing of sensitive organs such as the lung and heart; but these require high precision to avoid undesirable outcomes(2) (3). Furthermore, women lie naked from the waist upwards when many are adjusting to an altered body image. An immobilisation prototype has been designed to improve radiotherapy accuracy, reduce potential dose to organs at risk (in

86 women with larger breasts) and improve patient modesty.

Aim To refine, produce and test a support bra for immobilising breast tissue during breast irradiation for women that have been diagnosed with breast cancer (and have undergone removal of the tumour leaving an intact breast).

The primary endpoint is a support bra that is technically acceptable to health-care professionals (HCPs) and aesthetically acceptable to patients. This presentation will describe the development and testing process of the SuPPORT 4 All (S4A) bra that has included collaboration across the NHS, industry and University sectors. The needs of the patient and HCPs have been at the heart of the design and refinement process, which has incorporated a participatory design methodology within the MRC framework for studies of complex interventions.

This presentation will highlight key challenges experienced through the design and implementation process as well as strategies employed to mitigate these difficulties.

Conclusions, Relevance and Impact A participatory design methodology has facilitated the design of a bra that should meet end user needs. There have been a number of challenges throughout the innovation pathway that have been overcome through collaborative team work across the multiple agencies involved in the project.

Funder: This study is funded by the NIHR Invention for Innovation Programme (i4i).

1. World Health Organisation GLOBOCAN data 2008. Available from: http://www.who.int/research/en/. 2. Chen SA, Schuster, DM, Mister D, Liu T, Godette K, Torres MA. . Radiation field design and patterns of locoregional recurrence following definitive radiotherapy for breast cancer. Int J Radiat Oncol Biol Phys. 2013;85(2):309-14. 3. Munshi A, Kakka S, Bhutani R, Jalali R, Budrukkar A, Dinshaw KA. Factors Influencing Cosmetic Outcome in Breast Conservation. Clinical oncology. 2009;21(4):285-93.

MR-Linac: Real-time, adaptive, image guided radiotherapy Dr David Roberts, Team Leader, Physics-Linac Platforms, Elekta Limited

Elekta are developing an MR-Linac, which is a combination of a state of the art radiotherapy delivery system, a 1.5 Tesla Philips MR imaging system and online adaptive workflow software.

The current state of the device will be presented with an overview of the workflow and the technical challenges of developing such a system. The system allows for the adaption of the patient’s treatment on a day by day basis ensuring conformity of the radiation to the treatment target irrespective of changes in the tumour’s shape or position.

Aspects of the system design will be presented along with the modifications to testing protocols and beam characteristics.

87 Thursday 14th September 2017, 11.00 – 12.30 Cardiovascular Engineering

Advances in cardiac implantable devices Stephen O'Connor (1) City, University of London, London, UK (1)

Advances in implantable pacemakers and defibrillators will be discussed together with their potential for improved patient outcomes. The first implantable pacemaker, 1958, in a 43 year-old patient extended his life to 86 years. The device had shortcomings especially in terms of longevity and pacing ability. Rapidly advancing technology increased longevity and allowed atrial pacing with consequential new modes including epicardial left ventricular, LV, pacing for cardiac resynchronisation therapy, CRT by 1994 (4) Device size, lead design and battery longevity have improved. The most recent drive has been to leadless pacing commencing with Nanostim (7), as proof of concept, followed by Micra for right ventricular, RV, pacing, however not complication free. There is reluctance to implant widely due to concerns regarding extraction and prior complications. Advancement in CRT is much needed as approximately 40% of patients currently fail to benefit with conventional epicardial systems. A conventional pacing lead has been placed in the LV for endocardial stimulation via an atrial transseptal approach (5). Patients have to be anticoagulated for life which is not the case with WiSE technology (8), an ultrasonic, leadless endocardial left ventricular, LV, pacing for CRT. This remains a first-generation product with a small, 12.7mm long, receiver placed on the LV free wall, but limited battery life, a requirement for a co-implant stimulating the RV and preferred 2 stage implant procedure. There are no technical restrictions on implementing a fully leadless CRT system combining the RV and LV leadless devices. A system has been developed (3), and first-in– man study completed, to pace the heart without any leads or hardware on or in the heart. The lead is placed in tissue above the heart and connected to a small pulse generator on the sternum. Whilst not approved in any geography, this could be helpful in emergency and time compromised situations for temporary pacing. Self-powered pacemakers are in development using a piezo electric energy harvester (2) The first implantable defibrillator, ICD, implanted in 1980, following the pioneering work of Mirowski and Mower, was non- programmable, shocking above 170 bpm, with a short battery life and peri- operative mortality, p-op, of 9.7%. Transvenous leads were developed with a reduction in p-op to <0.5%, systems were programmable and anti-tachy and brady pacing were available. Patients started to outlive their leads highlighting the Achilles’ heel of the system, the lead itself, with approximately 30% of shocks being inappropriate. The totally subcutaneous ICD, S-ICD, was developed (1) with no leads either on or in the heart, comprising an electrode with two sensing elements, a shocking electrode and pulse generator, PG. Whilst there have been no lead failures, pocket complications surprisingly increased restricting the uptake. An intermuscular technique (9) for PG placement, between serratus anterior and latissimus dorsi, has found favour, reducing the surgical complication 10-fold compared with truly subcutaneous placement. There are >33k patients implanted with the S-ICD worldwide. A recent advance in this area is the 44 F ‘string defibrillator’ (6), implanted subcutaneously, high right parasternal to left dorsal positions in an arc, caudally. This system does not have a conventional PG, but has sufficient energy for approximately 1 year, depending on usage, and can be recharged at a planned outpatient follow- up in 1 hour. First-in-man study was performed in 22 patients. Home monitoring for pacemakers and defibrillators is now recommended as ‘standard of care’ primarily for patient safety but also for cost reduction. MRI compatibility of implantable cardiac device is rapidly increasing. The area of implantable cardiac devices continues to move rapidly and the new technology, once proven, will improve patient outcomes. 1.Bardy G et al., N Engl J Med. 2010;363:36-44. 2.Chen Y, et al., Advanced Electronic Materials. 2017;3:3 1600540. 3. Knops R et al., HRS abstract. 2017. 4. Leyva et al., JACC. 2014;64:1047-58. 5. Morgan J et al., Eur Heart J. doi:10.1093/eurheart/ehv273. 6. Neuzil P, HRS 2017. Late breakers. 7. Reddy V. et al., N Engl J Med. 2015;373:1125-35. 8. Reddy V. et al.,

88 JACC. 2017;69: 2119-29. 9. Winter et al. Europace doi:10.1093/europace/euw297

Martin Black Prize for best paper in Physiological Measurement: An assessment of algorithms to estimate respiratory rate from the electrocardiogram and photoplethysmogram Dr Peter Charlton, Dr Timothy Bonnici, King's College London and University of Oxford

Over 100 algorithms have been proposed to estimate respiratory rate (RR) from the electrocardiogram (ECG) and photoplethysmogram (PPG). As they have never been compared systematically it is unclear which algorithm performs the best.

Our primary aim was to determine how closely algorithms agreed with a gold standard RR measure when operating under ideal conditions. Secondary aims were: (i) to compare algorithm performance with IP, the clinical standard for continuous respiratory rate measurement in spontaneously breathing patients; (ii) to compare algorithm performance when using ECG and PPG; and (iii) to provide a toolbox of algorithms and data to allow future researchers to conduct reproducible comparisons of algorithms.

Algorithms were divided into three stages: extraction of respiratory signals, estimation of RR, and fusion of estimates. Several interchangeable techniques were implemented for each stage. Algorithms were assembled using all possible combinations of techniques, many of which were novel. After verification on simulated data, algorithms were tested on data from healthy participants. RRs derived from ECG, PPG and IP were compared to reference RRs obtained using a nasal-oral pressure sensor using the limits of agreement (LOA) technique.

314 algorithms were assessed. Of these, 270 could operate on either ECG or PPG, and 44 on only ECG. The best algorithm had 95% LOAs of −4.7 to 4.7 bpm and a bias of 0.0 bpm when using the ECG, and −5.1 to 7.2 bpm and 1.0 bpm when using PPG. IP had 95% LOAs of −5.6 to 5.2 bpm and a bias of −0.2 bpm. Four algorithms operating on ECG performed better than IP. All high- performing algorithms consisted of novel combinations of time domain RR estimation and modulation fusion techniques. Algorithms performed better when using ECG than PPG. The toolbox of algorithms and data used in this study are publicly available.

This study is published at DOI: 10.1088/0967-3334/37/4/610

Catheter Alignment System for Vascular Access Mamadou A Diallo (1) Sorin Popa (2) Robert J Dickinson (1) Imperial College London, London, UK (1) Stent Tek Ltd, London, UK (2)

Current procedures to create fistula in the 2.5 million haemodialysis patients worldwide suffer from a high rate of failure.

An innovative vascular access solution has been developed that uses two intravascular catheters to direct a needle to cross between vessels to create a fistula without surgery. In order to guide the orientation of the needle deployment so it hits the other vessel, an asymmetric electrical field is generated that is detected by a catheter in the target vessel. Simulation and experimental evaluation of an electric field guidance system models results is shown and discussed.

The electric field was generated by two opposing electrodes mounted on the source catheter. Simulations were conducted (COMSOL). A model was created comprising two catheters 10mm apart in a cuvette filled with blood. The positive electrode on the source catheter has a boundary current source of 95A/m2 (Figure 1).

89 Then a bench model was built with two catheters placed in a cuvette filled with saline water at 0.9g/l to simulate blood. A 20kHz voltage applied to the two electrodes on the source catheter creates the electrical field. The second catheter with a receiver electrode placed 10mm away from the source catheter measures the electric potential.

Figure 1:2D COMSOL model. Figure 2: 2D simulation and experimental test results.

The electric potential detected by the receiver varies with the source catheter orientation such that when the positive source electrode is perfectly aligned with the receiver electrode, a maximum potential is observed (Figure 2). The maximum voltage as detected by the receiver indicates alignment with the centre of the source positive electrode allowing a crossing needle to be ejected from the artery to the vein to create the fistula. The targeting accuracy is measured as better than 5 degrees so ejecting a needle at this point intercepts with the target catheter.

This technique has been used to successfully deploy a guidewire between two vessels in-vivo.

Do spatial correlations of blood flow and pro-inflammatory changes support a role for inflammation in the development of atherosclerosis? Kai Riemer (1,2) Ethan Rowland (1) Eleni Bazigou (1) Meng-Xing Tang (1,2) Peter Weinberg (1,2) Imperial College London, London, UK (1) British Heart Foundation Centre of Research Excellence, London, UK (2)

Background. Atherosclerosis affects some sites within the arterial system but not others. This is widely assumed to be caused by blood flow dynamics producing local pro-inflammatory changes in the endothelial cells lining the arterial wall; however, it is unclear whether the inflammatory changes, are caused by or are unrelated to accelerated uptake of circulating cholesterol-carrying macromolecules. Here we present preliminary results to assess the spatial correlation between anatomic patterns of pro- inflammatory change, blood flow and uptake of circulating macromolecules. The latter two are known to change with age. Methods. Activation of the pro-inflammatory transcription factor NF-κB, arterial wall permeability and endothelial nuclear orientation (a surrogate for flow direction) were mapped in vivo at 12 locations around the branch of the left and right renal in a total of 24 young and mature male New Zealand white rabbits using high-resolution confocal microscopy. Flow conditions were altered by surgical nephrectomy or acute ligation of the left renal and compared to naive and sham ligation control groups. The intracellular distribution of NF-κB and levels of albumin uptake were quantified by a 3-

90 dimensional segmentation. Results. Intermediate results suggest a spatial relationship between anatomical regions where pro- inflammatory changes occur in arterial endothelium, with regions of elevated arterial wall permeability. Measurement of flow with microbubble-enhanced ultrafast ultrasound image velocimetry and computation of flow by numerical methods are currently in progress. Discussion. This will be the first time that the spatial correlations of blood flow and pro-inflammatory changes are mapped using high spatial resolution and non-invasive ultrafast ultrasound flow measurement. Conclusion. At this point in time, we cannot draw any conclusions from our data. Key references. [1] Leow et al. Flow Velocity Mapping Using Contrast Enhanced High-Frame-Rate Plane Wave Ultrasound and Image Tracking: Methods and Initial in Vitro and in Vivo Evaluation. [2] Bryan et al. Mechanoresponsive networks controlling vascular inflammation [3] Bailey et al. Mass transport properties of the rabbit aortic wall [4] Pfeiffer et al. Does low and oscillatory wall shear stress correlate spatially with early atherosclerosis? A systematic review [5] Rowland et al. Comparison of Statistical Methods for Assessing Spatial Correlations Between Maps of Different Arterial Properties [6] Mohamied et al. Change of Direction in the Biomechanics of Atherosclerosis

Method validation of an ultrasound (US) approach to assess velocity profile as a marker of atherosclerotic burden Mark Atherton, Carola Koenig, Brunel University London

Background. Cardiovascular disease (CVD) accounted for 27% of all deaths in the UK in 2014 [1]. CVD is associated with modifiable risk factors (dyslipidaemia, hypertension and diabetes) and treating each of these factors lowers the risk of CVD [2, 3]. It is impossible to estimate the benefit of risk factor modification in the individual patient and extrapolating data from multiple trials is difficult. It would be useful to have a marker of risk that accurately estimates real time risk by measuring blood flow factors associated with the pathogenesis of atheroma. The aim of this preliminary study is to validate a low-cost measurement technique for estimating blood flow velocity profiles and assess whether any of the measured / calculated factors known to be associated with atheroma was associated with coronary heart disease (CHD), thus establishing its feasibility and acceptability as a clinical tool and suggesting areas for future research (incl. sample size etc.).

Methods. The study was designed to recruit ≥ 52 patients (final recruitment: 30 without CHD, 27 with CHD) in order to 1) optimise non-invasive vascular imaging, 2) establish flow velocity profiles and 3) determine if any factors likely to be associated with atherosclerotic process were associated with CHD. The left-common carotid artery (L-CCA) was chosen as the measurement site. For each subject the velocity was measured at between 10-12 equi-spaced points across the L-CCA. Heart rate (HR), Peak Systolic (PS) and End of Diastole (ED) velocities were recorded. A Bezier curve was fitted through the measured velocity points for a normalised diameter. This velocity profile was then correlated against a number of computational fluid dynamics (CFD) outputs (see Results) such as wall shear stress (WSS, [4, 5]). We looked for differences in US data and CFD outputs between the two patients groups via unpaired t-tests. Associations between any significant US measures found above and CFD outputs was studied via multiple regression with factorised patient groups included in the models.

Results.

91 PS velocity was associated with CHD (without CHD, mean (SD): 62.8 (16.1) cm/s, with CHD, mean (SD): 53.6 (17.3) cm/s, p=0.042). None of the other factors was statistically different between the patient groups. Multiple regression showed that PS velocity was associated with the following CFD outputs: average pressure drop in the carotid bulb (p<0.001), area of average WSS in the location of the bulb less than 1 Pa (p=0.016), area of average WSS in the location of the bulb less than 2 Pa (p=0.006), area of average WSS in the location of the bulb less than 3 Pa (p=0.001) and area of average WSS upstream of the bulb less than 3 Pa (p=0.017).

Discussion. The results suggest that PS velocity may be an area of interest. Even with a small number of subjects a significant difference was observed. Many other factors of interest were associated with PS velocity and these associations must be studied in detail in a larger cohort stratified by treatment(s), age, gender, vascular pathology etc. It is apparent that current US technology lacks the resolution for optimal estimation of L-CCA velocity profile. However, newer technology entering the market promises to provide direct and more accurate velocity profile measurements.

Conclusion. This preliminary study with a small cohort suggests that further work using larger, more defined patient groups with better technology may ultimately lead to development of a non-invasive method of indicating real time CHD risk.

Key references. In alphabetical order, numbered. [1] Cardiovascular Disease Statistics 2015, BHF, Nuffield, University of Oxford [2] Martín-Timón, I., et al., World J. Diabetes. 2014 Aug 15; 5(4): 444–470 [3] Lorber, D., Diabetes Metab. Syndr. Obes. 2014; 7: 169–183 [4] Cecchi E. et al., 1, Atherosclerosis. 2011 Feb; 214(2):249-56 [5] Peiffer, V., Sherwin, S.J., Weinberg, P.D., Cardiovascular Research (2013) 99, 242–250

92 Thursday 14th September 2017, 11.00 – 12.30 Medical Robotics

Exploration of an electrorheological fluid based adaptable gripping device for medical technology Jeroen Bergmann (1) Ivo Radice (1) University of Oxford, Oxford, UK (1)

Background: Gripping devices enable people with disabilities to carry out everyday tasks and increase their independence. The most sophisticated gripping systems can be found in the field of prosthetics. However, these devices require ever increasing complex methods of control. There seems to be a lack of medical robotic gripping technologies on the market that can fully adapt to any possible shape. The development of a less complex, but more versatile device could greatly benefit patients globally, who have lost gripping functionality. The focus of this project is to develop an assistive technology device that can augment or replace the hand’s gripping function by fully adapting to the shape of an object, instead of requiring the environment to adapt to them. The aim of this study was to explore if a conceptual prototype of an electrorheological (ER) fluid system [1] can change the resistive forces when it was activated. Methods: A test rig was constructed with a moving part, which was set in silicone oil and could be displaced using a tensometer (Bose BioDynamic, USA). Silica particles were dispersed in the silicone oils (10cst and 1000cSt) by 10, 20, 30 and 40% of the total mass. A DC field with a voltage of 4k/mm was generated to activate the fluid [2] through the application of custom-made voltage multiplier circuit. The difference between the measured force when the device was displaced (a) without Silica particles and/or activation and when it was displaced (b) with particles and/or activation was captured in the coefficient μ. Each of the aforementioned mixtures of silicone oils and particles were tested 10 times and the order was randomized. An ANOVA with a Holm-Bonferroni method was applied to determine significant differences between conditions. Results: Significant differences were found among the conditions (F(14,135) = 19.80, p<0.001) and the post-hoc test revealed that the largest difference between activation and rest was found for 10cSt silicone oil with 30% silica particles (μrest=0.58±0.17 vs μactive=1.04±0.25, p<0.01). No difference was found in the measured force when the silica particles were removed from the oil (μrest=0.01±0.01 vs μactive=0.01±0.01, p=N.S.). This shows that the system was indeed reliant on the ER fluid for changing the resistive forces when activated. Discussion: The results show that the developed system can support an increased force when activated. The presented system was battery operated, showing the capability of making these devices wearable. Conclusion: This study found that an ER fluid system can be developed that can change resistive forces without any noticeable delay. An ER fluid based gripping system could drastically improve functional gripping of complex, natural shapes and create a new field of research for assistive technologies. References: [1] Conrad & Sprecher, Journal of Statistical Physics, 1073-1091, 64(5), 1991 [2] Stangroom, Physics in Technology, 290-296, 14(6), 1983

93 Control Strategy for a New Extra-corporeal Robotic Ultrasound System James Housden (1) Shuangyi Wang (1) Yohan Noh (1) Davinder Singh (2) Anisha Singh (2) Junghwan Back (1) Lukas Lindenroth (1) Alberto Gomez (1) Nicolas Toussaint (1) Caroline Knight (3) Jacqueline Matthew (3) Tara Fletcher (3) David Lloyd (4) John Simpson (4) Dharmintra Pasupathy (3) Joseph Hajnal (1) Reza Razavi (1) Hongbin Liu (1) Kaspar Althoefer (5) Kawal Rhode (1) King's College London, London, UK (1) Xtronics Ltd., Gravesend, UK (2) Guy’s & St Thomas’ NHS Foundation Trust, London, UK (3) Evelina Children’s Hospital, London, UK (4) Queen Mary University of London, London, UK (5) Background. Extra-corporeal ultrasound (US) scanning is routinely used in hospitals for diagnosis and screening programmes. In this standardized large-scale use, the challenges of manual scanning motivate the development of a robotic and automated scanning system. These challenges include the difficulty of finding standard views required by the imaging protocols, the risk of RSI and the need for experienced sonographers on-site. Several robotic systems have been introduced [2] since the 1990s to enhance different aspects of extra-corporeal US scanning: e.g. long-distance tele-operation, robot-operator interaction, and the possibility of automation based on advanced image processing [1]. Following these, we have designed a new arm-based robotic system with a dexterous end-effector for probe manipulation, and developed and tested both hybrid and manual remote control strategies. Methods. A robotic manipulator capable of positioning a Philips X6-1 probe in 6-DOF space has been designed and manufactured (Fig. 1). The probe holder mechanism includes multiple single- axis force sensors for the measurement of the 3D force vector applied by the probe to the patient. In addition, two proximity sensors are incorporated for the measurement of probe-to-skin distance. Using real-time monitoring of these forces and distances, a control strategy, summarized in Fig. 2, was implemented. Target poses of the probe are provided by the sonographer via control software, a haptic input device, or potentially from automatic analysis of images. In the control strategy, this input is adjusted by feedback from the two types of sensor and a model of the target surface to provide a modified target pose for the device. Thus, the probe automatically maintains contact with the scanning surface while moving freely tangentially to the surface.

Fig. 1 (left) Schematic drawing of the extracorporeal US robot scanning an abdominal phantom. Fig. 2 (right) Overview of the control strategy of the robot. Results. The robot has been tested on an abdominal ultrasound phantom and in a volunteer scan using a haptic joystick for input. Using proximity sensors, the control strategy was successful in moving the probe across the surface while maintaining an offset in the range 1–3 cm from the surface while the volunteer was breathing. Using force control, the probe was able to slide across while maintaining contact with and normal alignment to the surface at a target axial force of 2N. Discussion. The results demonstrate that the robotic scanning system can be successfully manipulated in a hybrid control strategy. A user remotely controls the position across the surface while the robot automatically maintains contact and orientation to this curved surface. Conclusion. An effective semi-automatic control strategy for robotic ultrasound scanning has been developed. References. [1] Baumgartner C.F., et al. Real-time standard scan plane detection and localisation in fetal

94 ultrasound using fully convolutional neural networks. MICCAI. 2016:203-211. [2] Priester, A. M., et al. Robotic ultrasound systems in medicine. IEEE Trans. Ultrason., Ferroelect., Freq. Control. 2013 Mar;60(3):507-23.

A novel soft, parallel end-effector for robot-assisted ultrasound interventions Lukas Lindenroth (1) Avinash Soor (2) Jack Hutchinson (2) Amber Shafi (2) Junghwan Back (1) Kawal Rhode (2) Hongbin Liu (1) King's College London, Department of Informatics, London, UK (1) King's College London, Department of Biomedical Engineering, London, UK (2)

Medical ultrasound imaging is one of the most accessible imaging techniques to date. Robotics systems have emerged with the aim to automate these procedures by using modified industrial or custom manipulators, often following the principle of active compliance to enable safe human robot interaction, which often comprises of complex control architectures and difficult hardware integration. To address these limitations, we investigate in this work the applicability of a soft robotic end-effector for the given purpose. A novel, parallel design is derived based on the medical requirements, which addresses common shortcomings in both robotic ultrasound systems and soft robotic devices. The platform can be actuated using three linear, compliant actuators to induce rotations and longitudinal translation of the ultrasound probe. Background. In robot-assisted or -guided ultrasound various systems have been developed which exhibit active compliance for human robot interaction. In [1], [2] and [3] industrial robots such as the Universal Robots UR5 and WAM lightweight anthropomorphic robot have been utilized to develop robotic platforms for ultrasound application. Moreover, various custom robotic manipulators have been investigated which show potential to be applied to ultrasound robotics [4], [5]. To the best of the author’s knowledge, passive compliant systems have not yet been applied to robot-guided ultrasound systems. Methods.

Figure 4: Soft robotic end-effector attached to UR3 robot Figure 2: Concept of end-effector

The parallel platform design, as presented attached to a Universal Robot UR3 manipulator in figure 1 and as a conceptual drawing in figure 2, is based on linear compliant actuators which have been developed according to a principle investigated in previous studies [10]. The actuators can be driven by inducing hydraulic fluid into internal chambers to cause longitudinal extension. By inflating the actuators the platform can be extended longitudinally and tilted about its transversal axes to allow for rotations of the ultrasound probe. The platform comprises of a diameter of 11cm to incorporate commercially available ultrasound probes. The steerable top platform is linked through three compliant actuators to a base plate. The actuators are tilted at an angle of 15° to allow for sufficient space for a tip rotation of a maximum of 15°, which is used as a clinical estimate for fine-tuning the ultrasound transducer after positioning it in the vicinity of the imaging region.

95 Results. It is shown that the platform comprises relatively high longitudinal and transversal stiffnesses, while still being compliant enough to ensure the safety of the patient. The shear stiffness of the platform is 4.2 times greater than the shear stiffness of an individual actuator. The platform is capable of applying loadings of 10N along its longitudinal axis, which makes it suitable for general ultrasound interventions. Discussion. In future investigations the experimentally evaluated stiffness will be incorporated in a model to define the kinematics and dynamics of the system based on a structural mechanical approach. Furthermore, the variation of the hydrostatic pressure with regards to external forces applied to the system will be evaluated to incorporate intrinsic force sensing capabilities into the ultrasound platform. The latter will then be used for closed loop force control to regulate the pressure of the probe during the ultrasound intervention. Conclusion. It has been shown that the platform comprises of a high longitudinal stiffness and lower transversal stiffness. This is helpful in averting the probe in case of malfunction or discomfort in the patient, which could be achieved by pushing the probe away laterally, while a stiff longitudinal axis allows for sufficient pressure induced by the ultrasound transducer onto the patient for image acquisition. Key references [1] K. Mathiassen, J. E. Fjellin, K. Glette, P. K. Hol, and O. J. Elle, “An Ultrasound Robotic System Using the Commercial Robot UR5,” Front. Robot. AI, vol. 3, no. February, pp. 1–16, 2016. [2] H. T. Şen, A. Cheng, K. Ding, E. Boctor, J. Wong, I. Iordachita, and P. Kazanzides, “Cooperative Control with Ultrasound Guidance for Radiation Therapy,” Front. Robot. AI, vol. 3, no. August, pp. 1–12, 2016. [3] L. Santos and R. Cortesão, “A dynamically consistent hierarchical control architecture for robotic- assisted tele-echography with motion and contact dynamics driven by a 3D time-of-flight camera and a force sensor,” Proc. IEEE Int. Conf. Robot. Autom., pp. 2931–2937, 2015. [4] A. S. Bin Mustafa, T. Ishii, Y. Matsunaga, R. Nakadate, H. Ishii, K. Ogawa, A. Saito, M. Sugawara, K. Niki, and A. Takanishi, “Development of robotic system for autonomous liver screening using ultrasound scanning device,” 2013 IEEE Int. Conf. Robot. Biomimetics, ROBIO 2013, no. December, pp. 804–809, 2013. [5] R. P. Goldberg, M. Dumitru, R. H. Taylor, and D. Stoianovici, “A modular robotic system for ultrasound image acquisition,” Lect. Notes Comput. Sci. (including Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinformatics), vol. 2208, pp. 1430–1432, 2001. [6] D. Trivedi, C. D. Rahn, W. M. Kier, and I. D. Walker, “Soft robotics: Biological inspiration, state of the art, and future research,” Appl. Bionics Biomech., vol. 5, no. 3, pp. 99–117, Dec. 2008. [7] D. Rus and M. T. Tolley, “Design, fabrication and control of soft robots,” Nature, vol. 521, no. 7553, pp. 467–475, 2015. [8] M. Cianchetti, T. Ranzani, G. Gerboni, T. Nanayakkara, K. Althoefer, P. Dasgupta, and A. Menciassi, “Soft Robotics Technologies to Address Shortcomings in Today’s Minimally Invasive Surgery: The STIFF-FLOP Approach,” Soft Robot., vol. 1, no. 2, pp. 122–131, 2014. [9] M. T. Perri, A. L. Trejos, M. D. Naish, R. V Patel, R. A. Malthaner, and A. P. Art, “Short Papers System for Minimally Invasive Tumor Localization,” vol. 15, no. 6, pp. 925–931, 2010. [10] J. Fras, J. Czarnowski, M. Macias, J. Glowka, M. Cianchetti, and A. Menciassi, “New STIFF- FLOP module construction idea for improved actuation and sensing,” Robot. Autom. (ICRA), 2015 IEEE Int. Conf., pp. 2901–2906, 2015.

96 Thursday 14th September 2017, 14.00 – 15.00 Radiotherapy Dosimetry I

Determination of absorbed dose from filtered and FFF photon beam fields using an extrapolation chamber 1 Moore AR, 2 Ionescu A, 3 Zatout I 1Medical Physics, Royal Marsden Hospital, SW3 6JJ, UK. 2Department of Physics, University of Hertfordshire, UK. 3GE Healthcare Ultrasound Field Engineering, UK.

Background The aim of this project was to test a parallel-plate Extrapolation Chamber (EC) developed by S C Klevenhagen[1]. This is a type of ion chamber which under appropriate conditions can be used for absolute dosimetry via ionisation. EC are plane-parallel chambers of which the distance between the electrodes can be varied and accurately set. This allows one to derive the gradient of the ionisation with respect to the volume (proportional to the plate separation) and thus to avoid the need of calibrating the chamber as a means to determine its volume. EC are the basis of standards of absorbed dose in low and medium-energy x-rays as well as for beta emitting sources. In the first step of this work, the EC was for its reproducibility, stability, and linearity; then the EC was used to measure the absorbed dose of different therapy sources inside a solid water phantom. The results were compared with conventional traceable chain dosimetry [2], in order to affirm and quantify the uncertainties influencing the output measurement of an accelerator. The work was carried out by students at undergraduate and postgraduate level over the last few years. In this way, developments in technological beam delivery and dose calculation implemented in different classes of algorithms in a well-known treatment planning system have been investigated.

Methods. Klevenhagen’s EC has a variable air-volume cavity and is constructed from water-similar materials. It can be used in a water phantom directly in a beam of a therapy megavoltage machine under clinical conditions. The chamber allows absorbed dose to be determined from first principles i.e. absolute dosimeter. The material used in construction is Perspex (Lucite) which offers good mechanical stability and electron scattering similar to water. Its plane-parallel geometry is well suited for electron dosimetry and its elongated shape ensures that any metal parts such as the stepping motor and driving shaft are kept far enough from the beam to avoid any significant perturbation of the electron fluence. A cross section of the chamber is shown in figure below [1]. Figure 5 Schematic diagram of the chamber (not to scale). 1: guard ring. 2: collecting electrode. 3: sensitive volume. 4: entrance window (12 μm thick polyester film). 5: polarizing voltage lead. 6: external chamber wall (Perspex). 7: mobile cylinder (Perspex) carrying the collector and guard ring. 8: signal lead. 9: screw driving the piston. 10: stepping motor for moving the collector. 11: connector box. [1]

Results.

97 Table 1 EC measured 6MV TMR dose data versus tabulated RMH data 10x10 field Expected EC derived (RMH tables) Depth Dose Dose (mm) (cGy/MU) (cGy/MU) 200 0.532 0.536 150 0.652 0.658 100 0.800 0.800 50 0.958 0.950

15 1.027 1.030 By way of examplars, the above data graphed and converted into dose from raw charge gradient (via Bragg-Gray theory and stopping power ratios) are displayed as TMR. For comparison, RMH standard data are included.

Discussion. Other results non-reference fields, wedge fields, IMRT and FFF beams are in reasonable agreement.

Conclusion. Absorbed doses estimated by EC were in good agreement with conventional calibrations within 2% for photons beams. The present results suggest that the Klevenhagen extrapolation chamber can be considered as an independent measuring system for absorbed dose in addition to Farmer type ion chambers. This work was enabled by Prof. S.C.Klevenhagen (d. 1998) and team of physicists and biomedical- bioengineering group at the Royal London Hospital (and St Bartholomew’s Hospital, London) are acknowledged and thanked for the opportunity to use this EC.

Key references. [1] S C Klevenhagen, “Determination of absorbed dose in high-energy electron and photon radiation by means of an uncalibrated ionization chamber” Phys. Med. Biol., 1991, Vol. 36, No 2, 239-253 [2]. INSTITUTE OF PHYSICAL SCIENCES IN MEDICINE, “Code of practice for high energy photon therapy dosimetry based on the NPL absorbed dose calibration service”, Phys. Med. Biol. 35 (1990) 1355–1360

DECLARATION: IPEM Innovation Award recipient declares IPEM funding spent on materials in this project. Support in kind received from Phoenix Dosimetry Ltd (Mr Mike Sweeney) and Pyramid Technical Consultants (Dr John Gordon). (Rollo Moore 29 June 2017)

Assessing the dose significance of unplanned small air cavities in MR Guided Radiotherapy Jane Shortall (1) Robert Chuter (2) Alan McWilliam (1,2) Marcel Van Herk (1,2) The University of Manchester, Manchester, UK (1) The Christie NHS Foundation Trust, Manchester, UK (2)

Background. It has been proposed that opposing beam configurations can be used to compensate for the Electron Return Effect (ERE) around air cavities during Magnetic Resonance (MR) guided radiotherapy. However, this is not always achievable for organs at risk (OAR) that lie within the path of a single radiation beam. This work evaluates the effect of unplanned slab and spherical air cavities that may cause potential overdosing of nearby OARs due to the ERE [1-5]. Methods.

98 Monaco 5.19.02 (Elekta, AB, Stockholm, Sweden) was used to produce a Monte Carlo simulation of a single radiation beam under the influence of a 1.5 T transverse magnetic field, projected through a water–air–water phantom. A series of air slabs of varying thickness between 0.1 cm and 2.0 cm, and air spheres with varying diameters between 0.5 cm and 4.0 cm were inserted to test their impact on the calculated dose. The maximum dose increase and distance of dose influence were measured for each simulation compared to a reference model without air cavities. The distance of dose influence was defined by measuring the distance from the air boundary to the furthest isodose at half of the maximum dose increase. Results. A position of overdosing is found downstream from the air cavities, illustrated in the dose profiles of the reference beam and a 3.0 cm spherical air cavity in Figure 6. The relative dose increases and distance of dose influence for both geometries are presented in Figure 2 and 3 respectively.

Figure 6 - Dose profiles of reference beam and Figure 2 - % dose increase for varying sizes of air Figure 3 - Build-up distance for varying sizes of air 3.0 cm spherical air cavity. gap under the influence of a 1.5 T magnetic field. gap under the influence of a 1.5 T magnetic field. Discussion. The downstream dose was seen to increase for all air cavities. A dose increase of approximately 35 % is observed in air slabs of 0.6 cm and thicker, which agrees with the typical electron radius of curvature of 6 mm in 1.5T for this beam energy. A much larger dose increase of approximately 87 % is observed in spherical geometries over 3.5 cm in diameter. Whereas the distance of dose influence was seen to step for shift for slab geometries larger than 0.5 cm, it seems to behave somewhat erratically for curved tissue-air boundaries. Conclusion. Small air pockets in or downstream to OARs cause significant overdosing when exposed to a single radiation beam under the influence of a 1.5 T magnetic field. This effect is substantially larger in spherical geometries akin to the likely geometry of biological air pockets. A possible explanation is that in spherical air pockets electrons can circle longer resulting in more focused ERE. Key references. [1] A. J. E. Raaijmakers et al., Phys. Med. Biol., vol. 50, no. 7, pp. 1363–76, 2005. [2] A. J. E. Raaijmakers et al., Phys. Med. Biol, vol. 52, pp. 929–939, 2007. [3] B. W. Raaymakers et al., Phys. Med. Biol., vol. 54, no. 12, pp. N229–N237, 2009. [4] G. H. Bol et al., Phys. Med. Biol., vol. 60, no. 2, pp. 755–68, 2015. [5] S. Uilkema et al., Med. Phys., vol. 42, no. 12, p. 7182, 2015..

99 Thursday 14th September 2017, 14.00 – 15.00 Innovative and Sustainable Healthcare Technology

Health Technology Assessment of Medical Devices: assessing the value of innovation 1,2Chalkidou A 1King’s Technology Evaluation Centre. 2School of Biomedical Engineering & Imaging Sciences, King’s College London, UK.

The NHS 5-year Forward View has placed the adoption of medical technology (MedTech) innovation as a priority. To assess the value of innovation, and to decide which medical technologies should be adopted by the NHS requires evidence-based guidance. The National Institute for Care and Health Excellence (NICE) has established evaluation programmes for producing evidence-based guidance of medical technologies [1]. One of the key methodologies incorporated by these programmes is Health Technology Assessment (HTA). Contrary to pharmaceuticals, HTA for MedTech poses many challenges, from new definitions of what constitutes value to the scarcity of available evidence. HTA of MedTech also requires specialist technical background to understand the impact of upgrades, training and maintenance on safety and efficacy. To make things more complicated, the field of MedTech is highly diverse with more than 500,000 different types of medical devices being produced globally (not including the plethora of digital medical technologies produced) compared to only 20.000 medicinal products. This diversity is reflected in the many ways that they can bring value to the patients and to the system. In recent years, the definition of value is becoming increasingly complex. There are increasing numbers of stakeholders involved in the decision-making process and priorities are shifting towards patient-centred outcomes. This presentation will provide an overview of the different definitions of value for innovative medical technologies. It will also explore the specific challenges posed when assessing the value of innovation for medical technologies. It will also provide an overview 1. The National Institute for Health and Care Excellence. https://www.nice.org.uk/about/what-we- do/our-programmes/nice-guidance/nice-medical-technologies-evaluation-programme. Accessed 31st August 2017.

Health-economics modelling of screening for liver metastases using the Liverpool Uveal Melanoma Prognosticator Online (LUMPO) Azzam Taktak (1) Antonio Eleuteri (1) Helen Kalirai (2) Heinrich Heimann (1) Bertil Damato (3) Sarah Coupland (2) Royal Liverpool University Hospital, Liverpool, UK (1) University of Liverpool, Liverpool, UK (2) University of California San Francisco, San Francisco, USA (3)

Background. Uveal melanoma is the most common primary intraocular malignancy. The Liverpool Uveal Melanoma Prognosticator Online tool (LUMPO) has been thoroughly validated as a reliable tool for predicting survival following treatment of uveal melanoma [1,2,3]. It uses mathematical models based on machine learning algorithms. Almost 50% of patients with the disease develop metastases, which are always fatal. Current practice is to screen all patients every 6 months for liver metastases; however, this can be costly both to the health provider and to the patient. Conversely, screening can miss events in patients with a very poor prognosis. We performed a cost-benefit analysis of using LUMPO to optimise screening frequency according to genetic data available and estimated survival time. Methods. Follow-up data from a cohort of patients treated at the Royal Liverpool University Hospital for uveal melanoma between 2006 – 2013 was obtained. Number of scans carried out using the 6-monthly regime was calculated and compared to the number of scans that would have been performed using LUMPO to estimate metastatic risk. The cost-benefit function was defined as number of metastases missed per unnecessary scans avoided using the tool. Several assumptions were made, such as the time from development of metastases to death, how long in

100 advance metastases could be detected before symptoms, and number needed to screen (NNS). Results. Five-year follow-up data was available on 640 subjects, for whom clinical and cytogenetic data were available as well as outcome. Total number of screening exams conducted using a 6-monthly regime was 3338 and the total number of metastases detected was 91. Using LUMPO, for NNS = 20, the total number of screens dropped to 1550 with 18 (20%) metastases missed. For NNS = 10, the number of screens dropped further to 986 with 31 (34%) metastases missed. If LUMPO was used with clinical data only (i.e., no cytogenetics), 1698 and 1033 screens missed 27 (30%) and 47 (52%) metastases for NNS = 20 and 10 respectively. Discussion. Routine screening for metastases is time consuming and costly to patients and healthcare providers. Adverse effects not investigated in this study were anxiety resulting from false positives and complications of intraocular tumour biopsy. Conclusion. LUMPO enhances the cost-effectiveness of service delivery by reducing the number of screening exams by approximately 49% if only clinical data are known and by 54% if cytogenetic data are also entered into the model with a NNS = 20. Key references. [1] B.E. Damato, A. Eleuteri, A.F.G. Taktak, S.E. Coupland. Estimating prognosis for survival after treatment of choroidal melanoma. Prog Retin Eye Res, 30,5, 285-295 (2011). [2] S. W. DeParis, A. Taktak, A. Eleuteri, W. Enanoria, H. Heimann, S. E. Coupland, B. E. External validation of the Liverpool uveal melanoma progonsticator online, Invest Ophthalmol Vis Sci, 57 6116-6122 (2016). [3] A. Eleuteri, B.E. Damato, S.E. Coupland, A.F.G. Taktak. Enhancing survival prognostication in patients with choroidal melanoma by integrating pathologic, clinical and genetic predictors of metastasis. Int J Biomedical Engineering and Technology, 8,1, 18-35 (2012).

101

Real-time Signal Quality Diagnostics for Low-Cost ECG Telehealth in the Developing World Tingting Zhu (1) Nathaniel D. Cruz (2) Kristine Mae P. Magtubo (2) David A. Clifton (1) University of oxford, Oxford, UK (1) University of the Philippines, Manila, The Philippines (2)

Background The RxBox is an Android-based, low-cost telemedicine device that provides real-time cardiovascular decision support for resource-constrained areas. Currently, 180 RxBox devices have been deployed to geographically-isolated and disadvantaged areas in the Philippines, and have been adopted for use by healthcare workers. However, the usability of the system is limited due to poor-quality data obtained from low-cost electrocardiogram (ECG) sensors, hampering proper diagnosis. Although the “suction cup” electrodes are not commonly used in most clinical settings due to their poor conductivity, they are often the only viable option for resource- constrained areas (such as the Philippines) because they are affordable and reusable. Furthermore, while healthcare workers were trained to use the ECG device, they were often unable to identify the correct position for placing the ECG electrodes, resulting in poor-quality data and unnecessary delay in patient referral. The purpose of our study was to investigate the feasibility of implementing real-time user feedback for ECG lead placements and signal quality enhancement on a low-cost Android-based medical device. Methods Baseline ECG studies were conducted by primary care physicians at a partner health facility, where ECG traces were analysed by a clinical cardiologist, and were deemed poor in quality. These ECGs were then retrospectively analysed with in-house algorithms to remove artefacts. To further enhance the diagnostic interpretability of ECGs, the following signal quality indices (SQIs) were implemented: (1) the “fSQI”: the proportion of the signal which is not a flat line, which is useful in detecting incorrect sensor placement. (2) The “bSQI”: the percentage of beats on which two different QRS segmentation algorithms agree in terms of the R-peak position, as an indication of the presence of noise. (a) Figure 2 (left): Our user- (a) guided ECG lead placement demonstrating (a) RA-RL lead reversal and (b) the correct position. (b) (b) Figure 3 (right): Example of a real-time bSQI indication of 12-lead ECGs for (a) good signal quality, as shown in a green colour-bar and (b) unacceptable quality, as shown in a red colour-bar. Results and Discussion A real-time fSQI algorithm was implemented in the RxBox device to inform the user of the electrode placements while ECGs were taken so that the user can make corrections (see Figure 2). This allowed for accurate ECGs being taken in the primary care facility that improved the interpretation and hence better diagnosis. After correcting ECG electrode placement, real-time bSQI values were used to identify signal quality before recording ECG signals. The user interface in the RxBox was configured so that bSQI values for 12-lead ECGs were easily interpretable by a healthcare worker: a green colour-bar indicates good quality of ECGs, while a red colour-bar indicates poor quality of ECGs. Noting that bSQI was considered from all 12-lead ECG signals. Conclusion This study was able to demonstrate a prototype of an ECG device that provides real-time high quality ECG traces obtained from a low-cost ECG sensor using reusable electrodes (i.e., suction cups), suitable for resource-constrained settings. It also shows the possibility of real-time signal quality indication for further enhancing the interpretability of ECG signals, which potentially speeds up the patient referral process in the Philippines. In so doing, this study was able to set the groundwork for future automatic ECG diagnosis and early warning provision of deterioration using data acquired from low-cost sensors.

102 Thursday 14th September 2017, 14.00 – 15.00 Engineering Design

Biomedical Engineering: Developing Innovative Solutions for the Healthcare sector Dr Helen Meese CEng MIMechE, Head of Healthcare, Institution of Mechanical Engineers

With a growing global population and a universal expectation of longer, more active lives, the technology that promotes health and wellbeing has become ubiquitous. Healthcare is increasingly dependent on technology, and the safe, effective advancement of that technology particularly in the NHS, depends on the work of biomedical engineers and scientists. Yet the engineering and scientific communities continue to face considerable barriers to delivering medical technology into the healthcare sector.

In 2016 the Institution of Mechanical Engineers began a campaign to raise the profile of engineers in healthcare and highlight the importance to society of developing innovative solutions to strategic clinical challenges facing the NHS.

Dr Helen Meese, Head of Healthcare, will discuss today’s med tech sector, some of the barriers facing innovators in getting product to market, the need for strong clinical/technical relationships and some of the routes available to successfully delivering innovative solutions.

Engineering Better Patient Care: Re-design of a Paediatric Anal Dilator for Lower Cost and Higher Volume Production Thomas Stone (1) Sonya Sireau (1) Daniel Marsden (1) Andrea Jordan (1) Cambridge University Hospitals NHS Foundation Trust, Cambridge, UK (1)

Background: Paediatric patients may require anal dilatation as a treatment for anal stenosis such as following ano-plasty surgery or after pull-through operations for Hirschsprung’s disease. Anal dilators are used by medical and nursing staff and by parents to prevent or treat anal stenosis. Dilators which are currently commercially available on the market have significant drawbacks: they are aesthetically disturbing, predominately designed for use in surgery, sold as sets (even though often only a few sizes are required), re-usable between patients therefore requiring de- contamination and they are expensive. Cambridge University Hospitals NHS Foundation Trust (CUH) Paediatric Surgery Lead Clinical Nurse Specialist, Andrea Jordan, worked with Clinical Engineering to develop an improved solution – the Cambridge Paediatric Anal Dilator. The dilator is available in a range of sizes for sale individually at a reduced cost and improved patient and carer experience. The Cambridge Paediatric Anal Dilator was subsequently CE marked through CUH 13485 quality system in order to make it available for patients outside CUH. Thanks to more recent funding from Addenbrooke’s Charitable Trust (ACT) Innovation for Patient Benefit grant, further re- design and iteration work has been carried out, with the aim of further improving the device, increasing the number of devices manufactured and reducing the cost so more patients could benefit. Methods: A Healthcare Design Toolkit was used to look at the different options for re-design of the paediatric anal dilators. Various alternative manufacturing methods were investigated with the aim of reducing costs for higher volume and meeting further design requirements. Design changes were incorporated, based on post market surveillance feedback from the use of the handmade version and outputs from the Healthcare Design Toolkit iterations. Promising designs were prototyped and tested and one design was selected for validation in simulation and in materials testing labs. Results: A less expensive version of the dilator was CNC milled externally to CUH then heat- formed in-house, instead of being hand-made on site. This remained within the existing technical file and CE marking yet enabled the devices to be made at a lower cost. A subsequent version has also been designed which is to be manufactured in resin using additive manufacture which it is

103 hoped will reduce the costs further. This will bring manufacture back in-house, but will be able to be dispatched in collaboration with our 3D printing service based in our Media Studio. Conclusion: Engineers embedded in the NHS have a unique opportunity to improve healthcare through innovative use of technology. One way of doing this can be through championing high patient benefit projects which would not necessarily attract external investment. These ‘orphan’ projects may not be huge commercial opportunity for an investor, since they may not have clearly exploitable intellectual property. One such project was the paediatric anal dilator. Re-designing the dilator to make use of latest developments in manufacturing technology has enabled patient experience to be improved and costs to be lowered.

The LOCPIN Femoral Plate Location Pins: Use of the Healthcare Design Toolkit for Rapid Delivery of In-house Manufactured Medical Devices Sonya Sireau (1) Daniel Marsden (1) David Jackson (1) Matija Krkovic (1) Cambridge University Hospitals NHS Foundation Trust, Cambridge, UK (1)

Background: Femoral plates, such as the Stryker AxSOS 3 Plating System, are used to support bone following distal femoral fracture to hold fragments in place to facilitate healing (Figure 1). Orthopaedic surgeons, at Cambridge University Hospitals NHS Foundation Trust (CUH), identified that after initial reduction and fixation of the fracture and a period of recovery, some plates can become weak. As a precautionary measure, surgeons may replace these plates, rather than allow them to fail in vivo. Replacement of plates can be complicated, since the plate fixation site can be lost. During the replacement procedure, it is especially important that the new plate is placed in exactly the same position as the original, using the same holes in the distal femur to minimise bone loss from drilling. A need was therefore identified for a low cost device to be designed and manufactured to facilitate plate replacement surgery. Methods: The Clinical Engineering Innovation team at CUH, has been working with Cambridge University Engineering Design Centre to implement the Healthcare Design Toolkit. This Toolkit was used to design and manufacture a suitable solution: The LOCPIN femoral plate location pins. The Toolkit facilitated the creation of a technical file for the LOCPIN to be used in-house. Prospective hazard analysis methods were used to identify potential design requirements and to assess the proposed design. A form prototype was printed using additive manufacture in order to aid visualisation and clarify understanding of requirements. Results and Discussion: The Clinical Engineering Innovation team at CUH used the Healthcare Design Toolkit to design and manufacture the LOCPIN location pins to facilitate plate replacement procedure. The Toolkit enabled rapid creation of a technical file to allow in-house manufacture. The LOCPIN comprises surgical steel dowel pins, with a loop attached for retrieval. The LOCPIN was manufactured by CUH Clinical Engineering within a 13485 registered quality management system. To date, the LOCPINs have successfully been manufactured and used in one procedure: It is estimated this operation took 30 minutes to 1 hour less to complete, meaning that the patient benefited from a shorter period of general anaesthetic and suggests potential for releasing costly theatre time. Conclusion: The LOCPIN presents potential for a more efficient method for replacement of femoral plates. It is hoped that further use of the LOCPINs in CUH can produce additional data to form part of a wider study. This would enable evaluation and dissemination of any evidence of the benefits. The design and manufacture of the LOCPIN is an example of how engineers embedded in healthcare can facilitate safe innovation for patient benefit.

104

Figure 1 – Images of distal femoral fracture fixated using femoral plating system

105 Thursday 14th September 2017, 15.30 – 17.00 Radiotherapy Dosimetry II

The Streamlining of Routine Linac QC Through the Use of EPIDs and MATLAB Software to Replace Film Christina Agnew (1) Prakash Jeevanandam (1) Mark Grattan (1) Belfast Health and Social Care Trust, Belfast, UK (1)

Background. The gold standard for a number of linac Quality Control (QC) tests, such as radiation field size, asymmetric junction, MLC transmission, MLC/collimator radiation isocentre and gantry radiation isocentre remains radiographic or radiochromic film [3, 4]. However, as demand for film processors declined, it is known that infrequent use deteriorates film quality and maintenance of such equipment is not cost effective. Furthermore, in the current cost-driven environment, both radiographic and radiochromic films are expensive in terms of consumables and physicist time, both on the linac and afterwards in processing, scanning and analysis of the films. This work describes software, written within a quality management system, to assess radiation field size, asymmetric junction, MLC transmission and MLC/collimator radiation isocentre using images acquired on an EPID. The validation and commissioning of the software for use in our centre is presented, along with results of data captured on 10 linacs (Varian 600CD, 2100CD, TrueBeam and TrueBeam STx linacs) over a minimum of 3 years. The trends in these data have been assessed to inform the frequency of measurement required to detect errors and perform preventative maintenance. Finally the time and cost benefit of switching to digital imaging are also assessed. Methods. Digital images were acquired on 10 Varian linacs using the EPIDs with no additional build-up. Radiation field size, asymmetric junction and MLC/collimator radiation isocentre images were acquired using 6MV beam energy, 400MU/min dose rate and high quality mode. MLC transmission was acquired in dosimetry (integrated) mode. Commissioning of the software was performed by comparison of EPID results to identical measurements using EBT3 film. Following commissioning the EPID was used to perform these tests for a minimum of 3 years. Results. Given the physical distance between the top of the EPID cover and the sensitive layer of the imager is known, measuring the FSD to the cover allows an accurate indicator of the distance to the sensitive layer of the EPID. Use of the effective depth of measurement of the EPID (due to its non-water equivalence), reduced any differences between the known field size measured with both EPID and film measurements to 0.0 ± 0.6 mm (Figure a). The EPID was found to overestimate the junction dose by a factor of 1.2 ± 0.1 (Figure b). EPID analysis of MLC/collimator radiation isocentre (Figure c) and MLC transmission achieved comparable results to film.

14 12

100 1

10 LA1 y = 22.367x 0.8 8 50 LA2 6 0.6 y = 18.63x Frequency 4 0 0.4 LA3 2 -5 0 5 0.2 LA4 0 -50 F… 0

Junction Dose (%) Dose Junction E… LA5

Axis of Rotation (mm) Rotation of Axis

1.30 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 1.10 1.20 1.40

0.00 2010 2012 2014 2016

-1.00 -1.30 -1.20 -1.10 -0.90 -0.80 -0.70 -0.60 -0.50 -0.40 -0.30 -0.20 -0.10 -1.40 -100 LA6 (a) EPID-Film (mm) (b) Gap/Overlap (mm) (c) Year

Discussion. Use of the EPID to measure radiation field size, asymmetric junction, MLC transmission and MLC/collimator radiation isocentre reduced the time for delivery and analysis for each of these tests to < 2 minutes, with comparable accuracy and improved precision and reproducibility compared to film. The over-estimation of the EPID junction dose is comparable to previous literature (1,2). The use of the EPID to assess MLC transmission, MLC/collimator radiation isocentre or Varian linac field sizes has not previously been reported. Trends from the routine EPID results were used to inform planned preventative maintenance. Conclusion. The EPID can be used to replace film measurements for routine QC checks, saving time and money.

106 References. 1) Clew L & Greer P Med Phys., 36(12), 2009. 2) Hernandez V, et al. Med. Phys; 40(2), 2013. 3) Klein E, et al. TG142 report, Med Phys; 36(9), 2009. 4) Mayles W, et al. IPEM Report 81, 1999.

Commissioning of a 3D Scanning Water Phantom for Radiotherapy Dosimetry Andrew Blackmore (1) University Hospital Southampton, Southampton, UK (1)

Background. A new generation of 3D Scanning Water Phantoms for Radiotherapy Dosimetry have become available on the market. These phantoms make use of a number of advances not previously seen in scanning phantoms, namely: automatic tank setup routines, high-precision stepper motor and ballscrew drives, water level sensors, wireless operation, and complex software coordinate corrections [1]. Previous commissioning methods may therefore not be appropriate [2]. This study aims to determine if a new scanning phantom (PTW Beamscan) is acceptable for clinical use. To achieve this, we present a modernised methodology to determine mechanical and scanning accuracy to a high precision. We also verify the automatic setup routines, and statistically determine optimal scanning parameters for common detectors based upon acceptable error thresholds. Our results are then compared against similar studies and manufacturer specifications.

Methods. Commissioning focused on three areas: Mechanics, Scanning, and Software. Mechanical tests, using calibrated digital Vernier callipers and inclinometers, measured the positioning accuracy and reproducibility of the phantom in each axis. We verified that the water- filling level was reproducible, and water-level detection was accurate. Quality assurance baselines were established by referencing external landmarks against the carriage at its mechanical limits. Scanning tests examined the phantom’s automatic setup process in detail, utilising scans before and after to manually determine the deviations and examine if the corrections applied were successful. We investigated the phantom’s repeatability and radiological hysteresis in each axis, and by selecting an acceptable error, we found the optimal scan settings for common detectors. Software tests compared the calculations and indications displayed against manual calculations. We verified the smoothing, centering, symmetrisation, normalisation, and electron depth-ionisation to depth-dose conversion [3]. The Dmax, R50,d, Rp, beam flatness, and beam symmetry indications were also checked for a range of scans.

Results and Discussion. The phantom was mechanically within tolerance, providing positioning accuracy of ±0.1mm and repeatability within ±0.05mm. However, hysteresis of up to ±0.5mm was observed whilst scanning, which was larger than the apparent mechanical backlash. The auto-setup routines accurately level (±0.2°), centre (±0.2mm), and rotate (±0.3°) the measurement axis to coincide with the treatment device. These corrections are all software based. We found small volume ionisation chambers (e.g. Semiflex 3D) required point integration times of ~0.25s to achieve measurements throughout the scan within 0.2% of the true value (95% CI). Diodes required larger integration times (e.g. ~0.5s) to achieve similar results. All software calculations were as expected; however, the software only has a limited feature-set.

Conclusion. This study concluded that the phantom was suitable for clinical use. We determined scan parameters for common detectors based upon acceptable errors, and verified that the automatic setup produced acceptable results. Our results are comparable to the manufacturer’s specifications for the device, and similar to those observed in other studies [1].

Key References. [1] Akino, Y., Gibbons, J.P., Neck, D.W., Chu, C. and Das, I.J., 2014. Intra-and intervariability in beam data commissioning among water phantom scanning systems. Journal of Applied Clinical Medical Physics, 15(4).

107 [2] Mellenberg, D.E., Dahl, R.A. and Blackwell, C.R., 1990. Acceptance testing of an automated scanning water phantom. Medical physics, 17(2), pp.311-314. [3] Party, I.W., Thwaites, D.I., DuSautoy, A.R., et. al. 2003. The IPEM code of practice for electron dosimetry for radiotherapy beams of initial energy from 4 to 25 MeV based on an absorbed dose to water calibration. Physics in medicine and biology, 48(18), p.2929.

The Grandfather Effect- Personalising radiation protection advice for patients receiving prostate seed brachytherapy treatment for prostate cancer Margaret Moore (1,2) Michael O'Neill (2) University Hospital Galway, Galway, Ireland (1) National University of Ireland Galway, Galway, Ireland (2) Background: The aim is to provide suitable radiation protection advice for prostate cancer patients receiving prostate seed brachytherapy. Patients are radioactive following implantation and precautions are given to them concerning close contact with family members and children for 12 months afterwards. Precautions include minimising the time a child spends on the patient’s lap. However, the implanted radiation and the physical size and shape of patients vary widely. Hence precautions vary widely also. This study aims to adopt published research1-5 to define the time period for enacting contact precautions personalised to each patient’s individual treatment. Methods. Sixteen patients were chosen to verify the methodology by resolving measured data with calculated data. The radiation exposure from anterior and lateral surfaces was measured using a survey meter after implant. The depth from prostate to skin surface and the total activity implanted was obtained. The appropriate period to enact the contact precautions for each individual was calculated using this data and the algorithm proposed by Dauer1. Retrospective calculations were done for a further fifty patients by developing a python based calculation programme. The occupancy factor (O = hrs/24) for contact was varied from 1 hour to 8 hours per day and exposure limits of 1mSv and 0.5mSv were tested. Results. The ability of the calculation programme to predict levels of exposure from implanted patients was validated by patient measurement (Fig 1). The programme’s versatility to predict the duration of the precautionary period post implant for individual patients prior to their implantation was also proven. Inverse correlation between anterior dose-rates and depth to prostate was shown. The results compare the 12 month advisory period against individualised precaution periods across the cohort of patients receiving either mono-therapy or boost brachytherapy.

Fig1: Comparing measured to calculated dose-rates. Fig2:Time to achieve 1mSv exposure. Discussion. Patients receiving a boost treatment required shorter precautionary periods due to lower implanted activity. The 12 month period can be reduced to between 7 and 10 months based on patient parameters and significantly reduced further for boost patients. The retrospective study of fifty patients showed very wide variability among patients supporting an individualised patient specific approach for radiation protection advice. The calculation programme provides this function. Conclusion. The calculation tool developed allows customisation of the period for restricted contact for seed implant patients with family members particularly children and grandchildren. This can make prostate seed brachytherapy more appealing as a treatment option for men with prostate cancer particularly grandfathers. Key references. 1Dauer L et al “Less-restrictive patient specific radiation safety precautions can be safely prescribed after permanent seed implantation.”, Brachytherapy, (9) 2010. 2Dauer et al. “Assessment of radiation safety instructions to patients based on measured dose

108 rates following prostate brachytherapy.” Brachytherapy 3 (2004) 1-6. 3Hanada et al. “Assessing protection against radiation exposure after prostate 125I brachytherapy. “ Brachytherapy 13 (2014) 311-318. 4Hanada et al. “Prospective study of direct radiation exposure measurements for family members living with patients with prostate 125I seed implantation: Evidence of radiation safety” Brachytherapy 15 (2016) 412-419 5IPEM, Report 106. “UK Guidance on radiation protection issues following permanent iodine-125 seed prostate brachytherapy.” IPEM, York, UK (2012).

109 Thursday 14th September 2017, 15.30 – 17.00 Professional Session

Delivering Imaging Research in the NHS: the NIHR Clincal Research Network Imaging Research Strategy 1,2Keevil SF 1Department of Medical Physics, Guy’s and St Thomas’ NHS Foundation Trust, UK. 2Department of Biomedical Engineering, King’s College London, UK.

Over the past 10 years, the National Institute for Health Research (NIHR) has become one of the most respected funders of health-related research in the world, which according to an independent think-tank has ‘transformed Research & Development in and for the NHS and the patients it serves. [1]. The NIHR’s numerous funding streams and programmes include the Clinical Research Network (CRN), which provides infrastructure funding to support recruitment of patients into ‘portfolio’ studies, that is to say research that aligns with NIHR objectives and quality standards. Imaging has become an increasingly important element of medical research, whether as a diagnostic tool, to plan and guide therapy or to monitor treatment response. In many organisations, this expansion in research activity is placing strain on clinical imaging capacity, and conversely the amount of imaging research that can be delivered using finite equipment and workforce resources is constrained by growing clinical demand for imaging. In 2016 the NIHR CRN established an Imaging Strategy Group, chaired by Professor Stephen Smye OBE and made up of radiologists, radiographers and physicists, to explore ways of providing more effective and efficient support for imaging research in the NHS. A broad view of ‘imaging research’ was taken, including not just image acquisition technology but also clinical applications, image analysis and emerging ‘big data’ and machine learning applications that combine imaging with genomics and other sources of clinical data. In February 2017, the Group organised a workshop entitled ‘Delivering Imaging Research in the NHS’, with a multidisciplinary invited audience, a series of authoritative presentations on the future of imaging research, and plenty of opportunity for group discussion. Two particular challenges that emerged were developing the workforce needed to deliver imaging research and exploiting the opportunities provided by big data. The workshop was summarised in a report published online in August [2]. Further discussions following the workshop led to development of an ‘Imaging Research Roadmap’, available online alongside the report. In brief, the Roadmap comprises a series of tasks grouped into five workstreams: workforce, science, infrastructure, industry, and engagement. Members of the Strategy Group have been identified to lead each of the workstreams, working with colleagues to deliver the agreed actions over the next few months ahead of a follow-up workshop in 2018. This presentation will give an overview of the five workstreams, including their relationship with current and possible future NIHR funding streams and wider activities that have the potential to bring about transformational change in healthcare over the next 20 years. The Imaging Research Roadmap offers opportunities to all of the professions involved in medical imaging, and indeed multidisciplinary collaboration is going to be vital to its success. To keep in touch with developments, it is possible to join the emerging NIHR CRN Imaging Network at https://goo.gl/forms/OT6wr5pBEfbfHBSp2.

1. The National Institute for Health Research at Ten Years. RAND Corporation. https://www.rand.org/randeurope/research/projects/nihr-ten-years.html. Accessed 31st August 2017. 2. Delivering Imaging Research in the NHS: a Network Approach. https://www.nihr.ac.uk/life- sciences-industry/useful-info/brochures-and-resources/. Accessed 31st August 2017.

110 International impact of the IOMP Journal MEDICAL PHYSICS INTERNATIONAL Slavik Tabakov (1) King's College London, London, UK (1) International Organization for Medical Physics (IOMP), York, UK (2)

Background The study presents the statistics and impact of the IOMP Journal Medical Physics International (MPI), launched in 2013 as a bi-annual online free-access Journal, aiming to support the professional and educational development in the profession. Methods The study uses the statistical data from the MPI Journal server [1] in the period 2013-2017. It includes data about the visits per month, download of specific papers, geographical placement of readers, main access platforms and overall number of readers during the period. Results The first issue of MPI Journal was launched in April 2013 and during its first month attracted c.4000 readers [2]. Since 2015 the Journal has more than 7000 readers per month. This is a significant number for a profession with c.24,000 members. To this moment MPI has produced 9 issues (2 per year). The overall number of pages with professional and educational papers is close to 600. Additionally MPI published the abstracts of 3 International Conferences – total c.1200 pages. The overall visits of the MPI web site (April/2013 to April/2017) is c.280,000 (of which c.100,000 in the last year alone). The journal has published so far 75 professional and educational papers. Additionally, the papers related to technological innovations for educational purposes (most of these being contributed from the industry) attracted c.60,000 downloads. The stats for the global use of the Journal (in a typical month- see figure below) shows c.43% use in Europe; 30% use in Asia; 13% use in N America, 7% use in S America; 7% use in Africa and others. Readers feedback from 30+ countries (at ICTP College) to the MPI Co-Editors revealed that many are using the MPI educational materials in their teaching classes.

Visitors of MPI Journal per day during the busy March 2017, compared with March 2016 (dotted line) Discussion The immediate success of the MPI Journal showed the need of a medical physics professional and educational journal. Its specific impact is seen in the low-and-middle-income countries, where MPI is supporting the educational process. This is also seen in the developed countries, where MPI is a steady platform for exchange of information, especially for the short lived e-learning materials. Recently MPI began publishing full sets of teaching materials and colleagues are encouraged to contribute to this sharing of educational resources. Future MPI developments include publishing the volumes of the new IOMP project “History of Medical Physics”. Conclusion The IOMP Journal Medical Physics International is making a steady impact in the profession, encouraging a number of new professional activities and enriching the educational process. Key references 1. MPI web site: www.mpijpournal.org 2. Tabakov, S, P Sprawls, (2015) IOMP New Professional Journal Medical Physics International Initial Results, Medical Physics, vol.43, No.6, p.3367

111 UKAS Accreditation Programme for Medical Physics and Clinical Engineering Services (MPACE) – Strategy and Action Plan 1Compton D, 2Jarritt P 1United Kingdom Accreditation Services (UKAS), UK. 2Cambridge.

The drive to improve the quality of care for patients, whilst delivering efficiency and productivity, is a key objective for NHS England and most healthcare services. Through diagnostic and therapeutic procedures, Medical Physics and Clinical Engineering (MPACE) directly impacts the majority of patients and the care and treatment they receive. It is therefore essential that these services consistently deliver high quality care by maintaining the safety and the accuracy of equipment used and ensuring that procedures are evidence based, safe and effective.

As the UK’s National Accreditation Body, UKAS has been appointed to manage and deliver the accreditation of MPACE services. UKAS is the only body recognised by Government, to assess and declare the competence of organisations against internationally recognised standards. Accreditation for MPACE services will be against the requirements of BS 70000.

UKAS, the Institute for Physics and Engineering in Medicine (IPEM) and NHS England will work together to build on the previous work undertaken in the ‘iCEPSS’ project and looked to develop an appropriate accreditation scheme in a number of MPACE specialisms.

UKAS has established a Technical Advisory Committee (TAC) in collaboration with NHSE and IPEM to support the development of accreditation for MPACE. It consists of representatives from relevant stakeholders and technical experts and will provide advice and guidance on all aspects of the accreditation of MPACE, both initially and on a continuing basis.

In July 2017, a UKAS pilot assessment programme was announced. The pilot will use BS 70000 as the criteria for assessment and accreditation and will include two service areas. For areas outside the scope of the initial pilot areas, UKAS will continue to work with the relevant experts in the different fields to develop the accreditation scheme in readiness for their inclusion after the pilot activities have been completed.

This presentation will explore the development of the UKAS accreditation programme including iCEPSS and BS70000. It will consider the links between certification standards and the process of accreditation and the implications for Medical Physics and Clinical Engineering services. It will outline the current thinking on the potential scopes for accreditation and provide an update on the status of the project and timelines for accreditation assessments together with the requirements for recruitment of peer assessors. It will also provide guidance on what Trusts need to consider on the road to gaining accreditation.

An Update from the National School of Healthcare Science (NSHCS) Claire Hardiman Professional Lead for Medical Physics and Clinical Pharmaceutical Sciences, National School of Healthcare Science (NSHCS)

This presentation will provide an update from the National School of Healthcare Science (NSHCS) on its programmes; Higher Specialist Scientific Training (HSST), Scientific Training Programme (STP) and Practitioner Training Programme (PTP).

112 Thursday 14th September 2017, 15.30 – 17.00 Clinical Engineering

Engineering and software solutions for making the curriculum accessible in schools Carole Thornett (1) Independent Consultant, Exeter, Devon, UK (1)

Background. This study looks at the difficulties facing the engineer in making the mainstream curriculum accessible through the use of Information and Computing Technologies for children with complex and often multiple disabilities. Case studies spanning a wide age range are examined. While a Year 3 child (age 8) may need to be practising their spellings through an eye gaze system optimised for their use, or taking part in a treasure hunt using their special wheelchair controls, another child (Year 9) without speech may need to participate in as much of GCSE French as possible to allow them to obtain a qualification which reflects their understanding of the subject. Children are willing to make a great effort to participate and be included [2]. It is recognise that the appropriate support is not always available [5]. Children with cerebral palsy may make more errors and take longer to complete a task with eye gaze and it was felt [1] that usability with this group needed to be analysed. Nevertheless it was felt that children with severe physical difficulties could improve their performance with eye gaze [3]. The benefits of independent powered mobility have been highlighted by a number of researchers including [4]. Methods. Knowledge of the mainstream curriculum is essential in determining the best approach available for a pupil with complex physical disabilities. Active participation includes spoken and written communication and independent powered mobility. A comprehensive approach must address all these areas. This may involve programming communication devices so that the software is optimised for eye gaze, alternative access to mainstream software, and designing special powered wheelchair control systems. Solutions are required to overcome problems as they arise in school. Results. The test is day to day participation in the curriculum which is as independent as possible. This should be participation in a range of subjects and activities such as play throughout the school day. A number of case studies illustrate how these problems are tackled. Discussion. Eye gaze has been used since the 1980s but it is only in the last ten years that it has become cheap enough to be used routinely. Access to ICT remains difficult for pupils severely affected by cerebral palsy. Eye gaze is much more successful if the user can maintain a good symmetrical posture with the head in mid-line which can be difficult for a child with cerebral palsy to sustain. The speaking rate for a person talking normally can be at least twelve times faster than for a user communicating through a communication device. Conclusion. An individual approach is necessary in providing access to all relevant parts of the curriculum for a specific pupil with disabilities. Nevertheless, many solutions reflect the technology of the day, and can be adapted for other pupils with similar needs. It is also necessary to address all areas including spoken and written communication and independent mobility. Key references. [1] Amantis, R., Corradi, F., Molteni A., et al.: ‘Eye–tracking assistive technology: is this effective for the developmental age? Evaluation of eye tracking systems for children and adolescents with cerebral palsy’, Assistive Technology Research Series, 2011, 29, pp. 489-496 [2] Asbjørnslett, M., Engelsrud, G. Helseth, S. ’Inclusion and participation in everyday school life: experiences of children with physical (dis)abilities’, International Journal of Inclusive Education, 22 May 2014, pp.1-14 [3] Borgestig, M., Sandqvist, J., Parsons, P., et al.: ‘Eye gaze performance for children with severe physical impairments using gaze-based assistive technology – A longitudinal study’, Assistive Technology, 2016, 28, (2), pp. 93-102 [4] Hardy, P.: ‘Powered wheelchair mobility: An occupational performance evaluation perspective’ Aust. Occup. Ther. J., 2004, 51, pp. 34-42 [5] Ofsted: ‘High needs learners are not always receiving the support they need to reach their potential’ https://www.gov.uk/government/news/high-needs-learners-are-not-always-receiving-the- support-they-need-to-reach-their-potential, 22 March 2016, accessed 1 February 2017

113

Hear Glue Ear: Horizontal Innovation to provide hearing for children with Glue Ear Cara Cooper (1) Sonya Sireau (1) Tamsin Brown (2) Emily Boyce (3) Daniel Marsden (1) Cambridge University Hospitals NHS Foundation Trust, Cambridge, UK (1) Cambridgeshire Community Services NHS Trust, Cambridge, UK (2) Health Enterprise East, Cambridge, UK (3) Background: By the age of 10 years, 80% of children will have had at least one episode of Otitis Media with Effusion (Glue Ear). Glue ear is a condition in which the middle ear becomes filled with fluid, preventing the ossicles to amplify sound for the inner ear. The condition is the most common cause of hearing loss [1] and is known to cause fluctuating mild to moderate hearing loss. However, due to spontaneous resolution in 95% of cases [2], treatments are only offered after extended periods of “watchful waiting” after diagnosis is confirmed through an audiology test. During this time, which can last up to 18months, hearing aid is rarely provided to the child and this in combination with the condition can result in speech and language problems. We aim to transfer consumer bone conduction technology into a medical product which can provide hearing to the child at a low cost. Bone conduction technology by-passes the middle ear and provides sound to the cochlear directly, improving hearing for those with a conductive hearing loss, such as in Glue Ear. This project is targeting delivering a number of complete products which can be used in clinical evaluation as well as developing a technical file sufficient for regulatory approval. Methods: To facilitate the design process the Hear Glue Ear product has been developed using a Healthcare Design Toolkit. This process seeks to explore needs, create concepts and then evaluate in a rapid and iterative cycle. An existing consumer technology had been identified by the clinician at the start of the project. Needs were explored with key stakeholders through design workshops, stakeholder meetings and regulatory analysis. These needs analysis sessions lead to the creation of a requirements specification covering a range of “must have” elements essential for the product and “nice to have” items which could be developed in future products. Early risk analysis of the current consumer technology, using user failure mode effects analysis (uFMEA), has identified a number of key risks in the use of the products in a medical context. These risks have driven the design and concept workshops to innovate additions to the consumer product to provide safety features. Prototypes will then be made to establish viability before further development to produce manufacturable accessories to the consumer product. These accessories, along with an instruction for use, will be packaged with the consumer technology to create the medical device pack. It is expected that this pack will then be made available through audiology as well as direct to consumer, due to an anticipated low cost. Results: The healthcare design toolkit, along with risk management techniques have been effective for capturing user needs, developing specifications and understanding the risk profile of this class IIa medical device. This is leading to the development of concepts to extend the currently available consumer product to a medical use for paediatric patients with Glue Ear. The modifications are being developed using CAD and rapid prototyping techniques. The complete medical product will then go into a clinical investigation which will support the development of a technical file for regulatory submission. Conclusion: A healthcare design toolkit and risk analysis tools have been used to facilitate in- house horizontal innovation in an NHS trust to provide a cheap hearing tool for children with Glue Ear diagnosis during the “watchful waiting” period.

Figure 1: Child hearing using bone conduction with microphone support at story time

References:

114 [1] Haggard M, Hughes G. Screening children’s hearing. A review of the literature and the implications of otitis media. London: HMSO, 1991 [2] Zielhuis GA, Rach GH, Broek PV. Screening for otitis media with effusion in preschool children. Lancet 1989; 1: 311-4

Modelling Medical Technology Sustainability Using Discrete Event Methods 1,2Newman B, 2Akinluyi E 1Department of Medical Engineering & Physics, King’s College London, UK 2Clinical Engineering, Guy’s & St Thomas’ Foundation NHS Trust, UK Background. Health and care activity has increase by 18% since 2007 (1). During that time the Trust’s medical device inventory has doubled. The medical technology utilised to enable this activity cause’s inevitable impact in many areas. These include logistical implications which ensure the continuous provision of consumables and electricity. The production, consumption and resultant refuse associated with this provision negatively impacts the environment. The Climate Change Act 2008 requires the health and care sector to reduce carbon emissions by 34% by 2020 and 80% by 2050; it is not on target to achieve this. Electricity consumption is the largest contributor to Green House Gas (GHG) emissions at 64% (2). Considering this and the legislation, there is a notable lack of evidence regarding the contribution of medical devices to carbon emissions, beyond the acknowledgement their contribution is significant (2)(3)(4).The situation is similar for refuse. Whereas previous studies, including (5) have created models to ensure an optimised and continuous electricity supply, an attempt to model the consumption of electricity and consumables from the side of the health care provider, to better understand sustainability, is yet to be considered. The study attempts to utilise a novel approach of combining coded procedural data, used to obtain financial reimbursement via the Payment by Result Scheme (6), with event-based models of procedural impact, to represent the impact of medical technologies more holistically and explicitly. Methods. The electricity consumption of each medical device was measured using a custom made mains power extension cable. Around the ‘live’ conductor, a Hall-effect current clamp, connected to an oscilloscope, logged electrical current values to Excel. Current (Amperes) data was converted into Power (Watts) and input into MATLAB to plot the Kernel Density Estimate; the peak of which, provided a single value consumption rate per discrete functional status (i.e. idle or in-use) for each device. Where applicable, the total energy (Joules) per procedure was determined by calculating the area under the power-time curve using trapezoidal numerical integration. The stages involved, and the medical devices and consumables required for each procedure, were obtained through interviews with senior clinical staff. The purchase cost of each consumable was obtained from the Trust’s procurement database, the constituent materials (including packing) was taken from packaging or obtained from the manufacturer, the weight-per-material was measured using laboratory scales and the refuse stream obtained from waste management. Clinical coding data provided the hospital reimbursement figure per procedure as well as patient throughput numbers; these were cross referenced with departmental records where necessary. Converting into units of kWh the UK Government conversion factors for GHG reporting (7) calculated the Global Warming Potentials (kgCO2e). The final model was able to represent each discrete procedural event in terms of financial cost (pounds sterling) and environmental impact (tCO2e) per year. Results. Patient Throughput Values per Procedure % of Values per Year Procedure Tariff Per month/day Per year £ (kgCO2e) Tariff £ tCO2e Chemo (via Cannula) 194/384/578 74 (per day) 19,240 7.46 0.46 ≤4 7163 8.8504 Chemo (via PICC) 194/384/578 74 (per day) 19,240 14.63 0.86 ≤8 8962 16.5464 Photopheresis (ECP) 1359 299 (per month) 3,588 627.21 8.0 46 2,250,440 28.704 Photodynamic Therapy 1002 10 (per month) 120 200.00 0.075 20 24,000 9 Discussion. In all cases, the procedural financial costs and environment impact, corresponding to consumable use was significantly greater than for electricity consumption. It was notable the

115 expenses for the more specialised ECP and Photodynamic Therapy (PDT) procedures required a significant proportion of the reimbursement value at 46% and 20% respectively; compared with around 4% with Chemotherapy. A single ECP procedure produces nearly as much equivalent CO2 as a years’ worth of PDT procedures. ECP could be an area to achieve possible reductions. Conclusion. The methodology explored within this study demonstrates one possible approach which could help identify areas where significant CO2 reductions could be achieved within a complex acute health and social care system and help inform decisions across many areas including procurement and strategic infrastructure planning. Key references.

1. NHS England. NHS and health and care sector hit first target on climate change - but more testing goals are to come. [Online] 5 May 2016. https://www.england.nhs.uk/2016/01/climate-change/. 2. Sustainable Development Unit. Sustainable, Resilient, Healthy People and Places – A Sustainable Development Strategy for the NHS, Public Health and Social Care system. [Online] 2014. [Cited: 11 11 2016.] http://www.sduhealth.org.uk/documents/publications/2014%20strategy%20and%20modulesNewFolder/Strategy_FINAL_ Jan2014.pdf. 3. Christiansen, N., et al. Electricity consumption of medical plug loads in hospital laboratories: Identification, evaluation, prediction and verification. Energy and Buildings. 2015, Vol. 107, Pages 392-406. 4. Jensen, A.H and Petersen, P.M. Energy efficiency in hospitals and laboratories. [Online] 2011. [Cited: 3 June 2016.] http://proceedings.eceee.org/papers/proceedings2011/6- 337_Hjorth_Jensen.pdf?returnurl=http%3A%2F%2Fproceedings.eceee.org%2Fvisabstrakt.php%3Fevent%3D1%26doc %3D6-337-11. 5. Bagnasco, A., et al. Electrical consumption forecasting in hospital facilities: An application case. [Online] 2015. [Cited: 19 7 2016.] http://www.sciencedirect.com/science/article/pii/S0378778815004235. 6. Department of Health. Reforming NHS Financial Flows. [Online] 2002. [Cited: 1 3 2016.] http://webarchive.nationalarchives.gov.uk/20130107105354/http://www.dh.gov.uk/prod_consum_dh/groups/dh_digitalass ets/@dh/@en/documents/digitalasset/dh_4018704.pdf. 7. UK Gov, Department for Business, Energy & Industrial Strategy. Government emission conversion factors for greenhouse gas company reporting. [Online] 2016. [Cited: 4 March 2017.] https://www.gov.uk/government/collections/government-conversion-factors-for-company-reporting#conversion-factors- 2016.

An Assessment of the Effects of Passive RFID Transmissions on Medical Equipment Performance Stephen Hunt (1) Stephen Gardner (1) University Hospitals of Leicester NHS Trust, Leicester, UK (1)

The efficient management of medical devices in hospitals requires accurate knowledge of the existing inventory. A significant challenge is that many items are frequently transported between wards and equipment library. Using conventional methods, audit speed and accuracy can only be achieved by highly diligent additional staff, together with support from equipment users. An RFID system offers the potential to dramatically improve the speed of audits without such demands on resources. In a hospital setting, the RFID system must not interfere with existing healthcare systems particularly within intensive care units. This study examined the effect of a 860MHz UHF RFID interrogator on a variety of medical devices commonly found in the intensive care wards of the University Hospitals of Leicester NHS Trust. The tests were conducted in a former intensive care ward where radio frequency interference was not isolated. Devices were switched on, verified to function as expected then irradiated by the interrogator set to 1 Watt, delivering 2.7 Watts (ERP) from the antenna for 1 minute at a distance of 200mm. Any effects on function were investigated. Fifteen devices were tested; only one had any observed interference effects, which was to make the alarm buzzer sound, without an alarm state or any functional change. There was no effect on any device performance. These results indicate that this RFID system is safe for use within the areas of hospital to be visited for medical equipment audits and that R.F. shielding standards for these devices provide adequate electromagnetic protection against these transmissions.

116 Biomechanical Analysis of Balloon Eustachian Tuboplasty Matthew Smith (2) Anna Weir (1) Daisy Prior (1) James Tysome (2) Michael Sutcliffe (1) University of Cambridge Engineering Department, Cambridge, UK (1) Department of ENT Surgery, Cambridge University Hospitals NHS Foundation Trust, Cambridge, UK (2) Background. The Eustachian tube (ET) is a narrow, tapering structure that allows ambient pressure equalisation of the middle ear. In obstructive Eustachian tube dysfunction (OETD) opening of the ET is reduced, resulting in glue ear, pain or destructive processes which cause hearing loss (1). OETD affects up to 5% of adults (2). Balloon Eustachian tuboplasty (BET) is a new treatment for OETD, widening the ET with a high pressure balloon inflated within the tube. Promising early results are reported with two commercially available devices of differing size balloon; 3x20mm and 6x16mm (3). Although crushing and cracking of mucosa and cartilage are suspected (4, 5), the mechanism of BET is unknown, and no development work has been published regarding the device sizes. Small balloon sizes have been proposed as potentially safer. This study aims to investigate if BET has a mechanical effect on the ET, the role of balloon size on this effect, and if multiple repeats of the procedure are beneficial. Methods. Ethical approval was obtained. A bespoke high pressure syringe driver was used to inflate BET and angioplasty catheters with water to the 10 bar clinical operating pressure, before controlled deflation. Balloons sizes were 3x10, 3x15, 3x20, 5x12 and 6x16mm. Balloons were inflated/deflated unconstrained, and then in fresh (defrosted) human cadavers, inserted into the ET using the standard surgical technique (run 1). Inflation/deflation was repeated a further 2 times without removing the balloon (run 2,3). Digital pressure recordings were plotted against balloon inflation volume to generate a loop. Results. Balloon dilation data were collected from 45 ETs (23 cadavers). In every 3x15, 3x20, 5x12, 6x16mm trial, within-ET inflation/deflation loops varied from the unconstrained loops. Higher pressures were seen during inflation in run 1 compared to 2 and 3, which typically were identical. Differences between run 1 and 2 were greatest in the commercial balloon sizes. 3x10mm within-ET runs did not differ from unconstrained inflation.

Example inflation- deflation loop from a single cadaver ET using a 3x20mm BET balloon.

Discussion. The difference in results between run 1 and subsequent runs indicates plastic deformation of the ET tissues for the clinically used balloon sizes. This could correlate with tearing of the mucosal lining or cracking of the cartilaginous structure. The repeatability of subsequent inflations indicates only an elastic deformation on repeat inflation. Conclusion. BET using commercial balloons causes a sustained change in the ET mechanics in cadaver specimens, reducing resistance to opening by inflation, but this was limited to the first inflation. Balloons smaller than those commercially available have little or no mechanical effect. The clinical effects of BET may be due to a reduction in ET stiffness enhancing natural opening. Key references: 1. Bluestone CD. Eustachian tube Structure, function, role in otitis media. Hamilton/London: BC Decker; 2005. 2. Browning GG, Gatehouse S. Clin Otolaryngol Allied Sci. 1992;17(4):317-21. 3. Randrup TS, Ovesen T. Otolaryngol Head Neck Surg. 2015 Mar;152(3):383-92. 4. McCoul ED, Anand VK. Int Forum Allergy Rhinol. 2012;2(3):191-8. 5. Poe DS, Hanna BM. Am J Otolaryngol. 2011;32(2):115-23.

117

Thursday 14th September 2017, 15.30 – 17.00 Magnetic Resonance

Tissue Biomechanics: Fibrosis, Tumours & Impact of Vasculature Ralph Sinkus, King’s College London

The cancer invasion process is regulated by chemical and mechanical clues. Recent insight into the metastatic process suggests an intricate and active interplay between primary tumour and its habitat. In the context of breast-cancer, it was found that cancer-associated fibroblasts initially assembled a strained, viscous, and unfolded Fibronectin matrix. Furthermore, the malignant phenotype has been shown to have higher contractility and this elevated contractility was actively promoted by the stiffness of the extracellular matrix and the local fibre realignment. Local cell invasion was found predominantly to be oriented along certain aligned collagen fibres, suggesting that radial alignment of collagen fibres relative to tumours facilitates invasion. Finally, recent data from invasive cancer cells indicate that firstly there is a sweetspots regarding the matrix stiffness and viscosity that enable metastatic spread.

Non-invasive elasticity imaging via MRI (MR-Elastography) enables the precise mapping of elastic and viscous properties at the millimetre resolution. Hence, MRE has the potential to add valuable diagnostic information in the context of staging and gauging metastatic propensity. For breast cancer for instance, a clear correlation between cancer grade and enhanced viscosity is found. Equally, malignant liver lesions are found to be more viscous than benign. Furthermore, the value of MRE for evaluating response to therapy will be discussed as preclinical data suggest an enormous potential due to high sensitivity to vascular density. Clinical data show the ability to distinguish fibrosis from inflammation hence providing a motivation to investigate immune-therapies and potentially the ability to detect a response at early time points. Preliminary data in mice will be discussed. The Fibroscan (US-based elasticity imaging) has shown to predict the incidence of HCC using liver stiffness measurements. This opens an exciting opportunity of new possibilities in the context of patient management. Finally, we ask the following question: If the metastatic process requires a sophisticated interplay between tumour and habitat, could we perturb that interplay actively via focussed shear waves? Initial in-vitro results will be presented and the potential in the context of therapy discussed.

Development and validation of an electronically controlled algometer for use in fMRI studies to investigate brain activation in response to pressure pain stimuli Francis Hegarty (1,2) Tim Foran (2) Arun Bodke (3) Hari Gopal (2,3) Conail McCrory (2) New Children's Hospital Ireland, Dublin, Ireland (1) St James's Hospital, Dublin, Ireland (2) Trinity College, Dublin, Ireland (3) Pain is one of the most common reasons for people to attend medical clinics however objective assessment of pain remains elusive and traditional radiological imaging is limited by lack of correlation between radiological findings and clinical picture. The authors describes a new method they developed of delivering a controlled pressure pain stimulus to subjects undergoing fMRI studies. The equipment developed allows clinicians specializing in pain medicine to investigate the cortical and subcortical brain activation of subjects in response to pressure pain stimuli.

Existing algometers, used to measure pain induced by applying pressure to a body part, have limitations. Electronic algometers can not be used in the MRI scanning environment. Manual algometers are limited in how fast and precisely the controlled pain stimulus can be delivered. A sequence of pain stimuli from manual algometers can be difficult to deliver since they are subject to operator variability. Furthermore precise synchronize of the operation of manual algometers with MRI scanners is impossible.

Electronically controlled equipment was developed by the authors which delivered a time varying pressure pain stimulus, synchronized with the MRI scanner and which could be used in an MRI

118 imaging environment. The equipment consisted of two parts. The first is an all plastic pneumatic applicator placed on the subjects thumb during the scan. This was connected to the second part an electronically controlled variable gas pressure source positioned in the MRI control room and connected to the applicator via plastic pneumatic tubing.

The applicator contains a plastic pneumatically operated piston. When pressurized the plunger is displaced towards the patients thumb nail. The thumb can not move so the pressure applied to the piston results in the transfer of a mechanical force to the patients thumb nail, inducing a pain stimulus. When the pressure in the piston is allowed to return to atmospheric pressure, the mechanical force on the patients nail bed ceases. By applying different pressures at different times to the piston it is possible to apply an alternating force to the patients nail bed and therefore illicit a varying pain response.

The pneumatic drive unit consists of three monostats connected to a compressed gas source. Each monostat is preset to regulate to a different pressure and is connected to the applicator through a electronically controlled solenoid. By switching these solenoids in time, a series of pressures can be applied to the applicator, and consequently predetermined forces applied to the patients nail bed. An Arduino micro controller was programmed to control the switching of the solenoids to deliver pressure pain stimuli time sequences and the programme it runs is synchronized to the MRI scanner.

This paper will describe the design, operating principles, construction, and testing of the equipment undertaken by the authors. This will include descriptions of the methodology and results of the validation processes and discussion of features of the design that ensure subject safety and the ability to terminate the pain sequence in the event of system failure or at the subject’s request.

We describe the successful use of the equipment in an initial study where the equipment was used with a 3 T MRI scanner to undertake Echo-planar imaging of twenty normal volunteers (ethical approval obtained).

This new method of clinical assessment facilitates objective and dynamic assessment of pain in vivo.

Developing a program of tractography for neurosurgical planning: where best to focus your efforts Jonathan Ashmore (1) Gideon Oluniran (2) Hugh Pemberton (2) Bill Crum (3) Marcelo Lemos (2) Jozef Jarosz (1) Gareth Barker (2) King College Hospital, London, UK (1) Kings College London, London, UK (2) Imperial College London, London, UK (3)

Background. Regulatory approved neurosurgical planning systems which include tractography are becoming increasingly used in the clinical setting. There are however well publicised limitations to the tractography process having serious implications to surgical outcome [1]. The goal of the work presented here was to investigate in a volunteer cohort the accuracy of various diffusion acquisition and post processing methods for the reproduction of white matter pathways which are of common interest for neurosurgical planning. We evaluated a commercial neurosurgical planning system against freely available research software, comparing the tractography output results to cytoarchitectonic maps which act as a gold standard for the true tract location [2]. We refined our findings to a patient population through registration of post-operative to preoperative images, allowing us to evaluate if the pre-operative tractography intersected the post-operative resection cavity. This result was then considered in the context of the patient outcome and the presence of deficits associated with surgery. Methods. Diffusion weighted images were obtained for 10 volunteers at 1.5T: TR=17s, TE=101ms, FOV=32cm, slices=52, 32 diffusion directions, b-value=1500s/cm2, 4xb0 images, isotropic voxels (2.5mm3). This “base” acquisition (labelled 32Dir) was repeated with the following modifications applied separately: (i) 64 diffusion directions, 7xB0 images (64Dir) (ii) peripheral gating to the cardiac cycle (32Dir PG). All diffusion data sets were post processed using the following methods

119 (i) StealthViz neurosurgical planning software utilising a DTI reconstruction with deterministic tractography (labelled STEALTH) (ii) MRtrix (http://www.brain.org.au/software/) utilising constrained spherical deconvolution (CSD) reconstruction with deterministic tractography (CSD- stream). (iii) MRtrix utilising CSD reconstruction with probabilistic tractography for probabilistic thresholds of 0.005 (Prob 0.005), 0.01 (Prob 0.01), 0.025 (Prob 0.025), 0.05 (Prob 0.05), 0.075 (Prob 0.075). Each subject’s tractography images were warped into the standard space of the cytoarchitectonic maps and summed across all subjects to create frequency maps. The degree of overlap between the volunteer subject frequency map and the cytoarchitectonic maps was assessed using a fuzzy-DICE overlap metric [3]. In 7 patients we used the tool PORTR to align post-surgical to pre surgical images [4]. For each patient we inspected if the tractography for each post processing method was able to predict the presence (1 patient) or absence (6 patients) of post-surgical deficit (i.e. we considered if the tracts were “cut” during the surgery). Results. Figure 1 shows the quantitative fuzzy-DICE overlap results. We found that all methods were able to predict the post-surgical outcome in our patient cohort

Figure 1. fuzzy Dice Overlap results between tracts and cytoarchitectonic maps for (a) the cortico- spinal tract and (b) the optic radiations Conclusion. All methods investigated were limited in entirely depicting the CST and OR compared to the cytoarchitectonic maps, having a fuzzy-Dice overlap score of at best 50%. We found the acquisition protocol had limited impact on the accuracy of tractography whilst processing method had the greatest impact. CSD based probabilistic tractography with thresholds of 0.025 and 0.05 for the CST and OR respectively was found to be the optimal technique, a result in agreement with published studies [5]. This tractography technique is not available in any regulatory approved neurosurgical packages and hence the results from this study may provide justification for the clinical use of non-CE marked software within the EU. Key references. [1] Kinoshita M. et.al. NeuroImage 25:424 (2005) [2] Burgel U. et.al. NeuroImage 29:1092 (2006) [3] Crum, W. et.al. IEEE Trans. Med. Im. 25:1451 (2006) [4] Kwon, D. et.al. IEEE Trans. Med. Im. ,33: 651 (2014) [5] Farquharson S. et.al. J Neurosurg 18:1367 (2013)

Complete MR-elastography package for breast lesion characterization Sweta Sethi (2,1) Stefan-Heinz Hoelzl (1) Christian Guenthner (3) Daniel Fovargue (1) Jurgen Runge (1) Sebastian Kozerke (3) Jelizaveta Sudakova (1) Ayse Sila Dokumaci (1) David Nordsletten (1) Arnie Purushotham (1,2) Ralph Sinkus (1) King's College London, London, UK (1) Guy's and St.Thomas' NHS Foundation Trust, London, UK (2) Eidgenössische Technische Hochschule Zürich, Zurich, Switzerland (3)

Background MR-Elastography of the breast has in recent years been considered a promising novel imaging modality complementary to MRI with the ability to visualize and quantify the viscoelastic properties of breast tumors and the characterization of the surrounding tissues (6,9,11). Recent approaches in the field have focused on simple hardware solutions (8), yet are unable to achieve complete wave penetration throughout the entire breast as well as encompassing the lymph nodes which are important in view of metastatic propensity in the breast cancer setting. The lack of the combination of acquisition software and easy to use reconstruction alongside the hardware has been another limitation in the application of MR-Elastography to the breast within the clinical world. This study is introducing a complete MRE setup including a novel MRE transducer approach which yields inherently constant vibration amplitude with frequency, alongside Hadamard wave encoding (3) and SMS acquisition strategies (5) which enable a significant increase in SNR and reduction in

120 scan time without compromising on total scan duration or resolution, as well as a robust MR elastography stiffness quantification using a localized divergence free finite element reconstruction method (4).

Methods MRE was performed on 20 healthy volunteers on a 3T MR scanner (Achieva, Phillips Healthcare). Volunteers are positioned prone using the standard Philips 7 channel Breast coil. A dedicated MRE transducer setup was integrated into this product breast coil consisting of adjustable paddles that gently compress each breast individually, connected via cantilevers to an MR-compatible eccentric rotor that generated the monochromatic vibrations necessary to perform MRE (1). MR- Elastography was performed with at an excitation frequency of fVIB =36Hz using a Cartesian MR readout with two shots per mechanical wave cycle at 2mm isotropic image resolution. Recorded displacement data were processed using the divergence free finite element approach (4) to reconstruct viscoelastic properties.

Fig. 2 Details of the breast MRE set-up. (A) Adapted gravitational transducer. (B) Philips dStream breast 16ch coil (a) Pair of C-shaped pistons mounted on (b) PEEK lamellas

Results Fig. 1A depicts excellent correlation to the anatomy (fat/parenchyma) and enhanced stiffness of the magnitude of the shear model in Gabs (kPa) in Fig 1B. The location of the cyst is clearly visible in Fig 1A and lack thereof in Fig 1B due to the fluid-filled properties that define breast cysts (red ROI in Fig. 1A and in Fig.1B). Fig 1C shows the perfect correlation to the parenchyma in the anatomy scan which is also highlighting the lymph node depicted with elevated values in Gabs (kPa) in the healthy volunteer (red ROI in the magnitude image in Fig 1D). Fig 1F depicts increased stiffness scattered around the breast which has not been observed in the correlating T2 image in Fig. 1E. This illustrates the ability of the technique to highlight the stiffness variation within an otherwise homogenous fatty breast.

121

Fig. 1 The stiffness Gaps is illustrated in Fig 1 (B, D) correlating to the corresponding anatomy scan (A,C). Fig 1(F) depicts the wave length in lambda demonstrating an area of increased wavelengths not seen in the T2 image in Fig. 1(E)

Conclusion An entire MRE package for the characterization of breast lesions has been developed and successfully tested in 20 healthy volunteers demonstrating the feasibility of this integrated MRE package within the clinical breast cancer environment. Very good correlation of anatomy and changes in viscosity have been observed. Changes in parenchyma in terms of elasticity and lymph node detection has been noted. Next steps will involve utilizing this MRE package in the neoadjuvant breast cancer setting.

Key references

1. Abstract, A new gravitational transducer concept for MR Elastography based on a rotational eccentric mass, Runge JH, Hoelzl SH, Sudakova J, Dokumaci AS, Nelissen JL, Lee J,

122 Stoker J, Nederveen AJ, Nordsletten D, Sinkus R, submitted to ISMRM 2017

2. Abstract, Enabling high-resolution MR-Elastography of the breast, Hoelzl SH, Sethi S, Sudakova J, Dokumaci AS, Runge JH, Ng T, Purushotham A, Sinkus R

3. Abstract, Hadamard Encoding for Magnetic Resonance Elastography, Günthner C, Runge JH, Sinkus R, Kozerke S, submitted to ISMRM 2017

4. Abstract, Robust MR elastography stiffness quantification using a localized divergence free finite element reconstruction, Fovargue D, Sinkus R, Nordsletten D, submitted to ISMRM 2017

5. Abstract, Simultaneous Multislice Acquisition for Magnetic Resonance Elastography, Gunther C, Runge JH, Sinkus R, Kozerke S, submitted to ISMRM 2017

6. Chen J, Kugel J, Yin M, Venkatesh SK. Magnetic Resonance Elastrography of Other Organs. In: Venkatesh SK, Ehman RL, eds. Magnetic Resonance Elastography. New York, NY: Springer New York; 2014:119-133.

7. Garteiser, P. et al. Rapid acquisition of multifrequency, multislice and multidirectional MR elastography data with a fractionally encoded gradient echo sequence. NMR Biomed. 26, 1326–1335 (2013)

8. Hawley JR, Kalra P, Mo X, Raterman B, BSRT, Yee LD, MD,2 Kolipaka Quantification of Breast Stiffness Using MR Elastography at 3 Tesla with a Soft Sternal Driver: A Reproducibility Study International Society for Magnetic Resonance in Medicine DOI: 10.1002/jmri.25511

9. Lorenzen J, Sinkus R, Lorenzen M, et al. MR elastography of the breast:preliminary clinical results. Fortschr Röntgenstr. 2002;174:830-834.

10. Rump, J., Klatt, D., Braun, J., Warmuth, C. & Sack, I. Fractional encoding of harmonic motions in MR elastography. Magn. Reson. Med. 57, 388–395 (2007).]

11. Sinkus R, Lorenzen J, Schrader D, Lorenzen M, Dargatz M, Holz D. High-resolution tensor MR elastography for breast tumour detection. Phys Med Biol. 2000;45:1649-1664.

123 POSTERS

Understanding Aortic Blood Pressure by Chasing Reflected Waves Travelling to the Heart Shima Abdullateef (1) Jordi Alastruey (2) Ashraf Khir (1) Brunel University Londonn, Uxbridge, UK (1) King's College London, London, UK (2)

Background. Forward blood pressure waves generated by heart contraction encounter many bifurcations and changes in the structure of the vessels, creating backward waves travelling towards the heart. The existence of backwards waves, their magnitude, and their effect on increasing the pressure measured in aortic root has been recently revisited with an ongoing debate, particularly for the outstanding question about the origin of the reflected wave present in the aortic root, and how far the reflected waves could travel in the arterial tree. The aim of this study is to investigate how the generation of bifurcations affect the magnitude of reflected waves, and how far reflected waves can travel in the backward direction with a discernible magnitude. Methods. The non-linear one-dimensional (1D) equations of blood flow in elastic tubes have been used to develop a computational model to trace the waves as they traverse a simplified branching system as a surrogate of the arterial system. In addition, pressure waves were separated into forward and backward travelling components. A single Gaussian shaped pulse is prescribed at the root of sets of successive bifurcations. All of the terminal points of the bifurcations are prescribed to be absorbent except one point in the 5th generation of bifurcations (figure1.a), in order to analyse the behaviour of the reflected pressure wave originating from that point. The length of each segment is 5 meters in order to magnify the effect of each individual wave.

Results. A significant decay in “reflection” magnitude is distinguishable. There is a threefold decrease in the magnitude of the reflection waves originating from the first generation of bifurcation compared to the magnitude of the waves originated in the second generation of bifurcation. Based on the calculations (t =2L/c), where t is the arrival time, L is distance and c is the wave speed, the timing of the arrival of reflected wave from the blocked end at the 5th bifurcation should be evident after 14.33 seconds. In figure 1.b, The size of the wave at 14.33 is 0.04756 Pa which is 0.1 % of the initial wave. Discussion. We simplified the behaviour of the forward waves in the arterial tree with a numerical 1D simulation by sending a pulse wave from the first point of the sets of bifurcations and absorbing the wave at the end of all terminal points except one blocked end in the final bifurcation. The results show significant decay in the amplitude of the reflection waves reaching the root of the first bifurcation, which may indicate that more than half of the waves are re-reflected and trapped between the bifurcations; which is in agreements with findings of Davis et al (2012). However, the absence of reflected waves generated from other terminal points might be an over-simplification of the in-vivo environment. Conclusion. This simulation shows a significant decrease in the amplitude of the reflected wave reaching to the root of bifurcations. With Increased number of bifurcations, as is the case in-vivo, single reflections originating from the periphery may not be discernible at the aortic root. 1 .Alastruey J, Parker KH, Peiro J, Sherwin SJ. Analysing the pattern of pulse waves in arterial networks: a time-domain study. J eng Math. 2009;64:331-351 2.Davies JE, Alastruey J, Francis DP, et al. Attenuation of Wave Reflection by Wave Entrapment Creates a “Horizon Effect” in the

124 human aorta.Hypertension.2012;60:778-785 3.Khir AW, Parker KH. Measurement of wave speed and reflected waves in elastic tubes and bifurcations. Journal of biomechanics.2002;35:775-783 4. Khir AW, Zambanini A., Parker KH. Local and regional wave speed in the aorta: effects of arterial occlusion. Medical engineering & physics.2004; 26:23-29 5.Li Y, Gu H, Alastruey J, Chowienczyk P. Forward and Backward Pressure Waveform Morphology in Hypertension. Hypertension.2017;69:375-381

E-skin development and virtual prototyping: computational models of tactile sensing Fahad Al Shaibani (1) Jeroen H.M. Bergmann (1) Mark S Thompson (1) University of Oxford, Institute of Biomedical Engineering, Oxford, UK (1)

Background: Modelling tactile sensing is of crucial relevance in driving the innovation of second-generation prosthetics that can relay tactile information to patients using so-called e-skin [1, 2]. Computational models can encapsulate current physiological knowledge, determine parameter sensitivities and may be used to inform the development of biomimetic devices. Experimental data on neuronal responses to touch stimulation have long been available [3], and microstructural models and data fitting approaches have achieved some success. However, models for e-skin development should encompass the entire system of sensing incorporating aspects of both approaches. The aim of this study is to evaluate the current state-of-the-art of computational models that are able to predict neuronal responses in response to a range of external tactile stimuli. In particular, the research question is to determine which computational models have been published and which model meets most of the required criteria to support virtual prototyping and development of e-skin. Methods: A literature search was conducted using the following keywords in PubMed, IEEE Xplore and Google Scholar: (mechanoreceptor, Merkel, Pacinian, touch sensing, tactile receptors, slowly adapting afferent, rapidly adapting afferent) and (computational models, mathematical models, simulation, predictive model); the keywords in the each bracket were explored using OR, while AND was used for combining the two brackets. The resulting papers were selected according to their applicability to the proposed research question. The models were then comparatively evaluated based on 11 criteria’s categorized as computational (2) or physiological (9). Results: Preliminary results of the literature search identified three models [4, 5, 6], [4] is a mechanics of solid approach focusing on what triggers receptors; [5] is a signal processing approach focusing on transforming the input to the output; [6] is a mix between the two approaches with a systems information method. Criterion evaluation showed that [5] and [6] fulfilled 7 out of the 11 criteria, with both models being able to take displacement as input variable and outputting temporal patterns of neuronal output; while [4] only fulfilled 4 out of 11. If the approaches taken by [5] and [6] were combined, a model fulfilling a total of 10 out of 11 would be generated. Discussion: This study found only three models and the number of fulfilled criteria helped in evaluating whether or not a given model can support the development of e-skin. Specifically, [4] was the only model that did not predict temporal patterns of neuronal responses and it did not easily convey hardware implementation. On the other hand, [5] was the only model to incorporate the saturation of response in its prediction, but it had the highest number of free parameters. Additionally, none of the models could predict the response for a variety of different stimuli. Interestingly, [6] was able to discriminate between physical shapes, while [5] could not, however, [5] could predict for an entire population of different afferents while [6] did not have this feature. Models [5] and [6] fulfil criteria that complement each other, hence combining aspects from each would be promising. Conclusion: Integrative models of tactile sensing are lacking and for the virtual prototyping and development of e-skin no single model is optimal. A combination of signal processing approaches [5] plus MOS and transformation approaches [6] is suggested to be a promising direction. Key references:

125 [1] Saal et al, Neuropsychologia, 79, 344–353, 2015 [2] Tan et al, Sci. Transl. Med., 6, 257, 257ra138-257ra138, 2014 [3] Phillips et al, J. Neurophysiol., 46, 1192–1203, 1981 [4] Kumar et al, J. Biomech. Eng., 137, June,1–10, 2015 [5] Dong et al, J Neurophysiol, 109, 1350–1359, 2013 [6] Gerling et al, IEEE Trans. Haptics, 7, 2,216–228, 2014

Distinguishing effects of WSS magnitude and direction on endothelial cells by swirling a modified cell culture plate on an orbital shaker Mehwish Arshad (1) Mean Ghim (1) Spencer J. Sherwin (1) Maarten Van Reeuwijk (1) Peter D. Weinberg (1) Imperial College London, London, UK (1)

Background. The non-uniform distribution of atherosclerosis within the arterial tree[1] has been attributed to low wall shear stress (WSS), to highly multidirectional flow (assessed by the oscillatory shear index - OSI) and to WSS acting at right angles to the mean shear direction (transverse WSS - transWSS)[2,3]. Effects of shear on cultured endothelial cells (which line all blood vessels) are being investigated by swirling culture dishes on an orbital shaker. Computational fluid dynamics can be used to simulate flows in the wells; changes in shear metrics are plotted as a function of radial distance from the centre of the well and compared to cell properties [4,5]. However, the metrics often change together, making it difficult to distinguish effects of WSS magnitude and direction. The aim of this study is to introduce more variation by placing obstacles within the flow and to measure the effect on nuclear orientation. Methods. Cylinders of radius 4mm were prepared using PDMS and placed upright at the centre of 11.05 mm radius wells. Pig aortic endothelial cells were seeded in control wells or wells with the cylinder. The plates were swirled, or kept stationary as a control, for 7 days under 5% CO2 at 37°C, following which cells were fixed and stained with DRAQ5 (nuclei) and imaged en face using a confocal microscope. MATLAB was used for post-processing to extract the orientation of the nuclei. Flow in wells with and without the cylinders was simulated using a volume of fluid method in STAR-CCM+ software (CD-Adapco) and the WSS metrics extracted and plotted across the radius of the well. Results. For static wells with and without the cylinders, the cell nuclei had no net alignment at any location along the radius of the well. For sheared wells without the cylinders, nuclei at 4 mm from the centre aligned at -45° to a line connecting them to the centre of the well, whereas in sheared wells with cylinders, nuclei at 4 mm aligned at -25°. TransWSS peaks in the vicinity of the cylinder and gradually drops to zero at the well wall. Discussion. The decrease in alignment in the vicinity of the cylinder could be attributed to the cylinder wall which distorts and produces more disturbed flow. TransWSS is maximum at the cylinder wall but drops rapidly with radial distance, showing a sharp transition from purely multidirectional to predominantly uniaxial flow; this may explain the change in nuclear orientation. Conclusion. Adding a cylinder to the centre of the culture well modified the pattern of WSS metrics and altered the orientation of the cells. To further modify the WSS patterns, the cylinder could be moved off centre to disturb the symmetry of the flow. This work was funded by the Imperial College BHF Centre of Research Excellence. Key references. [1]. Lusis, A. J., 2000. Atherosclerosis. Nature, pp. 233-241. [2]. Mohamied, Y., Rowland, E. M., Bailey, E. L., Sherwin, S. J., Schwartz, M. A., & Weinberg, P. D. (2015). Change of Direction in the Biomechanics of Atherosclerosis. Annals of Biomedical Engineering. Annals of Biomedical Engineering, 43, 16-25.

126 [3]. Mohamied, Y., Sherwin, S. J., & Weinberg, P. D. (2017). Understanding the fluid mechanics behind transverse wall shear stress. Journal of Biomechanics , 102-109. [4]. Warboys, C. M., Berson, R. E., Mann, G. E., Pearson, J. D., & Weinberg, P. D. (2010). Acute and chronic exposure to shear stress have opposite effects on endothelial permeability to macromolecules. Am J Physiol Heart Circ Physiol, 1850–1856. [5] Potter, C. M., S. Schobesberger, M. H. Lundberg, P. D. Weinberg, J. A. Mitchell, and J. Gorelik. Shape and compliance of endothelial cells after shear stress in vitro or from different aortic regions: scanning ion conductance microscopy study. PLoS ONE 7:e31228, 2012

Generic scanning policies for patients with common implants in clinical and reserach MRI Jonathan Ashmore (1,3) Gareth Barker (2) King College Hospital, London, UK (1) King College London, London, UK (2) NHS Highland, Inverness, UK (3)

Background. Most MRI departments apply a risk averse policy for scanning patients with implants following the advice from bodies such as the MHRA: “Users should refer to implanting clinicians and the manufacturers for advice on the MR safety of all implants” and “When in doubt the user should assume the device is MR UNSAFE” [1]. This guidance is typically implemented in most departmental local rules as an “identify make and model prior to scanning” policy. However personal experience indicated that for practical and logistical reasons these policies were not always followed within our own institution and certain common implants were scanned without full prior assurance of safety. In the work presented here I discuss the results of a national audit highlighting the prevalence of this “blanket scanning” approach. I also present results demonstrating the frequency with which patients present with common implants in a clinical setting. In response to these findings I present our development of generic scanning policies for certain implant types. An approach which we believe better facilitates safe and efficient scanning in MRI units where resources are already overstretched.

Methods. Radiographers from 21 separate clinical departments were surveyed regarding practice for Figure 1 scanning patients with common implants (see figure 1 x-axis for list of implants). When asked about each implant the radiographer provided one of the following responses (for both 1.5T and 3T): (a) Always determine if safe through Identifying make and model (b) Assume it is safe if implanted at your trust (c) Just scan them - always assume it is safe with no need to identify make and model. In a second survey, undertaken at a single institution implant details were (prospectively) recorded for all patient’s entering the scanner (total 596 patients). The risk profile of an implant type was assessed through identifying evidence from (a) mrisafety.com, (b) peer reviewed scientific literature (c) email lists where implants are commonly discussed. These risks were documented and generic implant scanning policies were created based on the risk profiles which were signed off by the clinical and research leads. Results. Figure 1 shows the results for survey 1, highlighting that many implants are scanned without make and model at 1.5T, although this practice is less likely at 3T. In survey 2 it was found that 21% of patients presented with an implant of some sort (figure 2). Thus far we have created generic scanning policies for coronary stents, sternal wires, aneurysm coils, orthopaedic implants, IUD’s, heart valves and general abdominal surgical clips, where certain scanning conditions apply

127 Conclusion. Our survey showed that the Figure 2 majority of hospitals appear to have blanket practices for scanning implants without assurances of safety. This is most likely due to the high number of patients who present with implants and the difficulties in finding the implant information from various clinical systems. In response, to facilitate safe and efficient scanning we have created generic scanning policies for implants which demonstrating an extremely low risk profile. All other implants retain the “identify make and model” approach.

Key references. [1] MHRA: “Safety Guidelines for Magnetic Resonance Imaging Equipment in Clinical Use” (2014),

Three Dimensional Contact Force Estimation for Ablation Catheter using Bi-point Tracking Junghwan Back (1) Lukas Lindenroth (1) Kawal Rhode (2) Hongbin Liu (0) Department of Informatics, King's College London, London, UK (1) Biomedical Engineering, King's College London, London, UK (2)

The heart comprises a conduction system including the sinus and atrioventricular nodes to regulate the heart rhythm as a natural pacemaker. In an arrhythmic heart, the regulated heart rhythm is destroyed by chaotic electrical impulses from ectopic areas and additional electrical pathway between atrium to ventricle. To treat these abnormal heart rhythm problems, steerable catheters are navigated to the targeted area in the heart to block the arrhythmic electrical impulses and the additional electrical pathway using radio frequency tissue ablation. A surgeon relies on contact force feedback to ablate a target area with right amount of the radio frequency energy. Therefore, contact force play a significant role on success of the cardiac ablation.

Available force sensors for the thin and flexible catheter are in use. However, causes of that the catheter is disposable tool after single use and the millimetre-scale catheter, usage and development of the current catheter force sensors are economically inefficient and still technically challenge. Therefore, an intrinsic force sensing based on a shape of the catheter for cardiac ablation catheter had been proposed in our previous work [1]. However the shape based intrinsic force sensing method could be limited by that the shape must be available always in fluoroscopy image. In addition, it is still challenging to estimate contact force when catheter is under large bending and multiple contacts.

Therefore, this work develops a new multi-section static model of tendon driven catheters for both real-time intrinsic force sensing and interaction control [2]. The model allows the external force to be applied at arbitrary location on the catheter and can also cope with multiple contacts. In this study, we validated the model using a robotic platform, which steers a catheter consisting of 4 tendons with tension feedback. The experimental results show that the model can accurately predict the catheter shape with the effect of internal friction and large deflection. The position difference between measured and estimated was 2.5mm. Based on the catheter model, we developed an algorithm to estimate the contact force based on the catheter tip tracking and tension feedback. The validation results show that 3-dimensional contact forces can be estimated accurately using the proposed method. The magnitude of contact force error was 0.0117N with 400Hz update rate.

References [1] J. Back., et al. Catheter Contact Force Estimation from Shape Detection using a Real-Time Cosserat Rod Model, IEEE/RSJ International Conference on Intelligent Robots and System(IROS), pp.2037-2042, 2015

128 [2] J. Back., et al. New Kinematic Multi-Section Model for Catheter Contact Force Estimation and Steering, IEEE/RSJ International Conference on Intelligent Robots and System(IROS), pp.2122- 2127, 2016

A comparison of forward- and back-projection in vivo EPID dosimetry for VMAT treatment of the prostate James Bedford (1) Ian Hanson (1) Vibeke N Hansen (1) The Institute of Cancer Research and The Royal Marsden NHS Foundation Trust, London, UK (1) Background. Both forward-projection (FWD) [1, 5] and back-projection (BCK) [3, 4] methods have been described for carrying out in vivo dosimetry of volumetric modulated arc therapy (VMAT) using an electronic portal imaging device (EPID). In FWD, the treatment planning system produces a prediction of the integrated image which is accumulated during treatment. The actual measured image is then compared with this. In BCK, the image intensity measured at each gantry angle is projected through the patient CT scan and the sum of all such projections is compared to the planned dose distribution. This study reports on the practical agreement between the two methods in VMAT treatment of the prostate. Methods. Treatment plans consisted of either a 220° anterior arc or a 358° whole arc irradiating three planning target volumes similar to the CHHiP trial [2]. AutoBeam (v5.7) was used to produce forward-projection images [1] and iViewDose (Elekta AB) was used to carry out BCK [3, 4]. A total of 46 sets of gantry-resolved portal images were acquired in 13 patients using an iViewGT portal imager v3.4.1. A gamma index for 3% and 3 mm was used for both methods to compare planned and measured intensity (FWD) or dose (BCK). A threshold of 10% intensity or 50% dose was used. Results. Figure 1 shows the agreement between the two different types of EPID dosimetry. All methods show good agreement, with the closest agreement seen in the comparison of isocentric fluence or dose. Consistency within patients is generally higher than between patients.

Figure 1. Scatter plots of forward- and back-projection EPID dosimetry results. Each patient is shown in a different colour. Discussion. The differences between the methods can be largely accounted for by differences in dose calculation and EPID dosimetry projection model. This is most noticeable between patients. Variability within patients is mainly due to differences in rectal filling. FWD has the advantage of simplicity, but BCK provides a dose distribution within the patient, which is easier to interpret. Conclusion. Both the forward- and back-projection EPID dosimetry methods agree well in this study of prostate VMAT patients, so that both methods can be used reliably for in vivo dose verification. Key references. 1. Bedford J L, Hanson I M and Nordmark Hansen V 2014 Portal dosimetry for VMAT using integrated images obtained during treatment Med. Phys. 41 021725 2. Dearnaley D, et al. 2012 Conventional versus hypofractionated high-dose intensity-modulated radiotherapy for prostate cancer: preliminary safety results from the CHHiP randomised controlled trial Lancet Oncol. 13 43-54 (supplementary appendix) 3. Hanson I M, Hansen V N, Olaciregui-Ruiz I and van Herk M 2014 Clinical implementation and rapid commissioning of an EPID based in-vivo dosimetry system Phys. Med. Biol. 59 N171-9 4. Mans A, et al. 2010 3D Dosimetric verification of volumetric-modulated arc therapy by portal dosimetry Radiother. Oncol. 94

129 181-187 5. Schyns L E J R, Persoon L C G G, Podesta M, van Elmpt W J C and Verhaegen F 2016 Time-resolved versus time-integrated portal dosimetry: the role of an object’s position with respect to the isocenter in volumetric modulated arc therapy Phys. Med. Biol. 61 3969-84

Tool for Electro-Diagnostics Statistics (TEDS): Diagnosing Excessive Waiting Times 1Chiu J, 1Lake S, 1Kennedy V 1Medical Physics and Clinical Engineering Department, Royal Liverpool Hospital, UK. Background: NHS England defines waiting time as the time elapsed from the initial referral letter being received to the start of initial treatment3. Waiting times function as a primary rationing mechanism in the NHS2, but do not diagnose causes of changes in waiting times. Lead time plays an important role in the lean thinking process and is calculated using 퐿푒푎푑 푡푖푚푒 = 푊표푟푘 푖푛 푃푟표푔푟푒푠푠1. Work in progress (WIP) refers to the number of items in the system, and clearance 퐶푙푒푎푟푎푛푐푒 푟푎푡푒 rate is the rate at which items are passing through the system. Application of lean principles in healthcare has resulted in improved processes, and quality of patient care5. This work introduced the use of lead times into electro-diagnostic services through the development of a tool which provides lead time data and statistics. Methods: A tool was developed using MATLAB, a mathematical processing system, and a JavaScript web page. The web page extracts data from the local database which is subsequently posted to MATLAB to compute and plot waiting and lead time values. The resultant tool is called Tools for Electro-diagnostic Statistics (TEDS). Results: Table 2. Waiting and Lead time results taken from TEDS for Diagnostic Audiology in 2010 (baseline) and 2014-16 Median Waiting time Lead time Clearance rate Year WIP (days) (days) (per week) 2010 26 41 35 5.7 2014 78 106 88 6 2015 79 97 71 4.8 2016 95 134 101 5.5 Discussion: Waiting times should meet with the NHS national standard for diagnostic tests of 42 days4. From Table 1, the results show that waiting days from 2014-16 are consistently higher than the national standard, however, the cause of this problem cannot be determined by the waiting times. The lead time data can, however, attribute the possible cause of the increased waiting times. In 2010, there were two members of staff in this service, and thus additional resource, and a lower number of referrals (WIP). In the latter years, there was only one member of staff carrying out the service, who was, interestingly, carrying out work at the same rate, however increased referrals which resulted in longer lead times. Conclusion: Waiting times provide a good measure of the performance of a service, however, does not explain possible causes. Lead time diagnoses these excessive waiting times, i.e. is it demand, represented by WIP, or resources, represented by clearance rate. Key references 1George, Michael L., & Mike George. (2003) Lean six sigma for service. New York, NY: McGraw- Hill 2Lindsay, C. M., & Feigenbaum, B. (1984). Rationing by waiting lists. The American Economic Review, 74(3), 404-417. 3NHS England. Recording and Reporting Referral-to-Treatment Guidance. (2013). [ebook] Available at: http://www.england.nhs.uk/statistics/. Accessed 17 Jun. 2017. 4NHS England (2017). https://www.england.nhs.uk/statistics/ [online] Available at: https://www.england.nhs.uk/statistics/ [Accessed 26 Jun. 2017]. 5Yeh, H. L., Lin, C. S., Su, C. T., & Wang, P. C. (2011). Applying lean six sigma to improve healthcare: An empirical study. African Journal of Business Management, 5(31), 12356.

130 A Comparison of Conventional 10MV, 10MV IMRT and 10FFF IMRT Breast Planning David Church (1) Suzanne Smith (1) Beatson West of Scotland Cancer Centre, Glasgow, UK (1)

Background The purpose of this study was to evaluate if there are potential benefits in using intensity modulated radiotherapy (IMRT) with either flattened 10MV or flattening filter free (FFF) 10MV to produce improved treatment plans, in terms of reduction in dose to organs at risk and/or improved planning target volume (PTV) coverage, over conventional forward planned 10MV planning technique. In our clinic, conventional breast planning is currently being replaced with new IMRT and Deep inspiration breath-hold (DIBH) techniques. Patients receiving DIBH treatment are treated using FFF and Flattened MV energies however not all breast cancer patients will be treated with DIBH. Commissioned photon energies include IMRT (6MV, 6FFF and 10FFF) but 10MV IMRT and 16MV IMRT are not currently available. Methods A retrospective planning study was performed for ten randomly selected 10MV breast patients. Each patient treatment had been calculated within the Eclipse Treatment Planning System [Varian Medical Systems]. New IMRT plans were generated for each of the ten patients using 10MV IMRT and 10FFF IMRT. Planning was performed using current planning objectives and dose constraints. Differences in dose-volume histogram parameters were calculated to assess plan quality. The plans were evaluated based on the PTV coverage, the percentage volume of heart that received 2Gy (V2Gy), volume of lung that received 12Gy (V12Gy) and the Monitor Units (MU). The mean patients PTV size was 2000cc (range 740cc – 3585cc). Differences in dose volume histogram parameters were calculated to assess plan quality. The PTV Homogeneity Index (HI) was calculated:

HI = D 5 /D 95; where:

D 5 = dose to 5% of target volume,

D 95 = dose to the 95% of target volume, Significance was assessed by two-tailed t-test (p<0.01). Results When comparing the HI results for the PTV coverage there was a significant difference between the conventional 10MV plans (mean HI 1.121, range 1.102-1.149) and both the 10MV flattened (mean HI 1.094, range 1.079-1.132) and 10MV FFF plans (mean HI 1.099, range 1.075-1.135). However, the difference in PTV HI between the 10MV FFF and 10MV flattened plans was not statistically significant. The statistical comparison of organs at risk (OAR) showed no significant variation between the 10MV plans and either of the IMRT plans observed but a significant t-test result showing significance between each set of IMRT results for heart V2Gy with the FFF IMRT (mean 6.5%, range 0%-15.6%) demonstrating a reduction in dose when compared with the flattened IMRT plans (mean 8.7%, range 0%-19.8%). When comparing the Monitor Units there was a significant difference between the conventional 10MV plans (mean MU 354, range 306-573) and either 10MV flattened (mean MU 643, range 464- 955) or FFF plans (mean MU 1326, range 872-2180) Discussion IMRT 10MV Flattened and 10MV FFF have shown increased MU with large variations observed in 10FFF. In our centre 10MV flattened IMRT will be fully commissioned for use in Breast IMRT treatments and if there is no difference between 10MV and 10FFF, DIBH will be treated using FFF.

131 MRI Acoustic Noise: Are Headphones Sufficient? Cluny L, Cavin I. Medical Physics, NHS Lothian Background Acoustic noise generated by MRI scanners is a concern for patients, carers and staff in terms of safety and comfort. During scanning the sound levels are capable of causing transient or permanent hearing damage to persons inside the scanner room. The sound arises as a result of the Lorentz forces generated while time-varying electric current pass through the gradient coils. The Medicines and Healthcare Product Regulations Agency (MHRA) MRI Guidelines and Health and Safety legislation recognise acoustic noise as a hazard and indicate that hearing protection be provided to reduce the noise levels to below 85dB. MRI facilities may use a combination of headphones and earplugs to achieve this, however measurements of acoustic noise and sound reduction are not performed routinely. This project aims to evaluate the noise levels experienced by patients, staff or carers, present in the scanning room, as well as evaluate the effectiveness of the headphones provided. Methods Acoustic noise measurements were taken using an OptiSLM MR Conditional sound level meter during a diffusion weighted imaging (DWI) scanning sequence; one of the loudest MRI sequences in clinical use. Measurements were obtained for two Siemens systems, a 1.5T Aera and a 3T Prisma, and for a GE 1.5T Optima 450 across 3 different sites. The sound meter transducer was attached to a spherical head phantom and measurements were taken during the scanning sequence at 20cm intervals along the patient couch with and without the manufacturer supplied patient headphones. For each measurement, the peak A-weighted sound level was recorded. Further sound measurements were recorded at locations in the scanning room, representative of positions occupied by parents, carers or intraoperative MRI staff. These measurements were taken with the transducer secured inside the staff headphones whilst being worn by a staff member, and then positioned outside the headphones. Results At positions along the patient couch from isocentre to 150cm, the peak sound levels measured ranged from 81.0dB – 100.4dB without patient headphones and from 75.0dB – 91.2dB with patient headphones. Sound levels from locations around the scanning room ranged from 88.3dB – 100.4dB without staff headphones and 53.0.dB – 66.8dB with staff headphones. Discussion The results obtained suggest that whilst hearing protection is required by staff and visitors present during MRI scanning, the current headphones provided are adequate; peak noise levels were reduced to below 85dB at all measured positions on all systems when staff headphones were used. However, the use of headphones provided to patients failed to reduce the noise levels to below 85dB at many positions along the patient couch. Further, the patient headphones were found at several positions to provide significantly less attenuation than the figure provided by the manufacturer, with the lowest levels of attenuation seen at the isocentre. With some MR sites using a combination of headphones and earplugs, further investigation is required to assess whether sufficient attenuation is achieved with these measures. Conclusion Multiple acoustic noise measurements were taken in order to evaluate the noise levels experienced by both patients, staff and visitors during MRI scanning. It was found that whilst adequate headphone protection is currently provided for staff and visitors, additional measures may be required to sufficiently reduce the noise levels experienced by patients. Acknowledgments Thanks to Julie Childs (East Kent Hospitals University NHS Foundation Trust), Nicola Tyler (Royal Surrey County Hospital NHS Foundation Trust) and Malcolm Philips (NHS Lothian) for the loan of the OptiSLM MRI Compatible Sound Meter and description of acoustic noise measurement techniques.

132 Wheelchair robotic arm mount for automated mobile communication aid positioning Abdul Daya (1) Jeffrey Morris (2) Electronic Assistive Technology, ALAC - Wrexham Maelor Hospital, Wrexham, UK (1) National Centre for Electronic Assistive Technology, Rookwood Hospital, Cardiff, UK (2)

Background. Wheelchair arm mounts are widely used to mount communication aids for the disabled wheelchair user. The arm mounts are fixed devices which require proper alignment and positioning during installation, for the user to communicate effectively1,2. As many of the communication aids now incorporate some sort of eye tracking device, correct positioning of the device is essential for the user to communicate effectively. As most persons tend to move their head position naturally, the optimum positioning of the communication aid that was setup during installation can change after installation, making it difficult for the disabled user to communicate effectively. The purpose of this project was to develop an automated robotic arm mount that will re align the communication aid with changes in head movement, so that the communication aid is always in the optimum position for the user. Methods. Facial tracking software was developed using C# and EMGU to track head position using a webcam. Facial tracking is widely used in the surveillance and photographic industry but its use in medical robotics is fairly limited. As the facial tracking software will be converted to an “app” for installation on tablet devices, it should be able to support USB communication. The control unit for the robotic arm/mount is designed around a PIC 18f4550 which has been successfully configured to use the USB port as a virtual COM port for data communication. The control unit automatically loads the necessary device drivers for the relevant operating system when plugged into the USB port and can support a variety of operating systems including; Windows, Max OSX (v3.1), Linux and Android 4.2, with full support for USB2 protocol. Programmable servo motors will be used to respond to the facial tracking data. They can be mounted in a dedicated robotic arm or a pan/tilt bracket which is specifically designed to be supplied as an add on unit to existing wheelchair arm mounts. Depending on the weight of the tablet device being mounted, industrial grade servo motors which can support up to 40Kgf can be incorporated to the robotic arm with the necessary motor drives. Results. The image processing libraries included with EMGU are easily incorporated in C# and are used for facial and eye tracking. It has been shown that we can reliably track head movement using a webcam and transfer the real time data of the position of the head to the control unit of the robotic arm via the USB port. The control unit can then be used to control the servo motors on the robotic arm and reposition the communication aid to the optimum position for the user. Discussion. Unlike existing devices which are controlled using a joystick or push buttons to reposition the communication aid in and out of the line of sight of the wheelchair user, our device will be fully automated and will operate dynamically in real time with changes in head movement. The facial tracking software that has been developed can be readily transferred to tablet devices with the Windows operating system or be converted to an “app” suitable for Android and iOS using software utilities such as Xamarin. It is anticipated that future developments in the project will enable data transmission to the robotic device using Bluetooth technology. Conclusion. Most facial tracking software requires the head be close to the vertical position for it to be tracked reliably. As some of our patients cannot maintain good head position due to degenerative motor neurone disease (MND) and lack of motor control, it is imperative that we are able to track head drop/tilt reliably. The uniqueness of our facial tracking software is its ability to track head movements and cater for the various head tilt angles which is necessary for the positioning of the robotic arm/mount. The benefits of having a fully automated and dynamic robotic arm will be beneficial to most disabled users especially those with limited or no motor control. Key references. 1. Barrett G, et al. Wheelchair-mounted robotic arm to move and hold a communication device- final design. Biomed Sci Instrum. 2015. 2. Hurley K, et al. Wheelchair-mounted robotic arm to hold and move communication device. Biomed Sci Instrum. 2014;50:115-8.

133 MR Imaging of plasma macromolecule uptake by the arterial wall Marta Dazzi (1) Rene Botnar (2) Peter Weinberg (1) Imperial College London, London, UK (1) King's College London, London, UK (2)

Background The focal distribution of atherosclerosis has been attributed to local haemodynamic influences on the endothelium, which are thought to increase endothelial permeability to macromolecules [1]. Imaging their uptake by the artery wall can provide an insight into disease development, but existing methods are limited to animal models [2]. A recently developed non-invasive, real-time technique based on MR imaging of an albumin-binding contrast agent (gadofosveset trisodium) holds promise for human studies; however, only 75% of the contrast agent is bound to albumin at the clinical dose, adversely affecting measurements [3,4]. A mathematical model can quantitate the concentration of bound contrast agent within the artery wall provided the bound and unbound relaxivities, r1 and r2 are known [5]. Results have also not been validated against existing methods.

Methods Relaxivities were calculated from T1 and T2 maps of PBS-based solutions and tissue- mimicking collagen gels at 21oC and 37oC with a 9.4 T Bruker MR scanner. Preliminary in vivo imaging was performed on male Sprague Dawley rats intravenously injected with 0.03 mmol/kg of gadofosveset trisodium. Cross-sectional images of the brachiocephalic artery were acquired before and after 30 mins of administration.

Results r1 remained unchanged upon binding to albumin as expected at 9.4 T. Conversely, r2 exhibited a 5-fold increase at 21oC upon binding to albumin in PBS and collagen gels. At 37oC, a 23-fold and 11-fold increment was observed for PBS and collagen gels, respectively. In vivo scans revealed a brighter signal along the vessels walls after administration of contrast agent.

Discussion Relaxivity was shown to be dependent on environmental conditions. Therefore, r1/r2 of collagen gels at 37oC should adequately mimic in vivo relaxivity which should reliably quantitate the concentration of bound contrast agent in the arterial wall. Furthermore, in vivo scans confirmed that scanner resolution is sufficient to visualise the brachiocephalic artery in rats. Conclusion Bound and unbound relaxivities for gadofosveset trisodium to human serum albumin were determined at 9.4 T. Future work will focus on directly validating MRI data with current techniques that require imaging of excised vessels in animals. Once confirmed, this will allow a method to correlate lesion maps, wall shear stress and permeability in humans for the first time. This study was funded by the EPSRC and the BHF. Key references [1] Staughton TJ, et al., Effect of altered flow on the pattern of permeability around rabbit aortic arches, Am J Physiol Heart Circ Physiol, 281: H53–H59, 2001. [2] Bailey EL, et al., Mass Transport Properties of the Rabbit Aortic Wall, PLoS One, 10: 1–20, 2015. [3] Phinikaridou A, et al., Noninvasive Magnetic Resonance Imaging Evaluation of Endothelial Permeability in Murine Atherosclerosis Using an Albumin-Binding Contrast Agent, Circulation, 126(6):707-19, 2012. [4] Caravan P., et. al., The Interaction of MS-325 with Human Serum Albumin and Its Effects on Proton Relaxation Rates, J. Am. Chem. Soc., 124: 3152- 3162, 2002. [5] Richardson OC, et al., Gadofosveset-Based Biomarker of Tissue Albumin Concentration: Technical Validation In Vitro and Feasibility In Vivo, Magn. Reson. Med, 73:244–253, 2015.

134 Method for Exposing Cultured Endothelial Cells to Specific Shear Stress Patterns Combined with Spatially-Resolved Permeability Measurements Mean Ghim (1) Kuin Pang (1) Peter Weinberg (1) Imperial College London, London, UK (1)

Background. Endothelial permeability to circulating macromolecules and patterns of haemodynamic wall shear stress vary from site to site within the arterial system. Influences of local blood flow on macromolecule transport across endothelium may account for the patchy nature of atherosclerosis [1][2]. To investigate this relation, chronic shear stresses can be imposed on cultured endothelial cells grown in multiwell plates by swirling the plates on an orbital shaker [3]. Computational modelling of the flow reveals that shear stress vectors experienced by the cells vary depending on location within the well [4]. Combined with the use of fluorescent tracers based on avidin that bind to a biotinylated substrate under the cells, permeability can be mapped and compared to the pattern of shear stress [5]. However, cells in one region of the well may release soluble mediators into the medium, thus affecting the behaviour of all the cells. We therefore developed a method that allows culture of endothelial cells only at specific locations in a well, prohibiting the cells from spreading to other areas, whilst still permitting the spatially-resolved permeability measurements. Methods. Fibronectin was biotinylated using a commercial kit and binding of FITC-conjugated avidin (FITC-A) to it was optimised. Region-specific coating of well-plates with biotinylated fibronectin (BF) was carried out using either PDMS masks or custom machined gaskets to confine the BF solution to specific areas within the wells. The mask was then removed and the uncoated surface was passivated with a pluronic solution. Porcine aortic endothelial cells (PAECs) were cultured until confluent in the wells. FITC-A was applied above the monolayer for a short duration and accumulation of this tracer was observed using confocal microscopy. The system was tested by culturing PAECs at the centre or edge of the well, exposing the cells to shear stress for 48 hours and comparing the tracer accumulation under the monolayers. Results. Prolonged culture of the cells did not result in the expansion of the monolayer onto the passivated surface. Images of tracer accumulation at cellular resolution within the BF-coated regions were obtained. Cells at the edge of the well exhibited an approximately 10% decrease in tracer accumulation compared to cells at the centre of the well after exposure to shear stress. Discussion. Permeability measurements were made in PAEC monolayers cultured only in specific regions of the wlll. Sheared monolayers exhibited different permeability depending on location, highlighting the different shear stress characteristics present in the wells. Conclusion. Measuring spatially-resolved permeability and allowing region-specific culture of cells within wells will allow for the investigation of cellular responses to specific shear stress characteristics whilst avoiding interference from cells in other regions of the well, exposed to different shear stress patterns, via unknown and possibly competing effects of soluble mediators released from those cells. This study was funded by the British Heart Foundation and by the A- STAR foundation. Key references. [1] Caro, C.G. et al., 1971. Atheroma and arterial wall shear. Observation, correlation and proposal of a shear dependent mass transfer mechanism for atherogenesis. Proceedings of the Royal Society of London. Series B, Biological 170 sciences, 177(1046), pp.109–59. [2] Giddens, D.P. et al., 1993. The Role of Fluid Mechanics in the Localization and Detection of Atherosclerosis. Journal of Biomechanical Engineering, 115(4B), p.588. [3] Warboys, C.M. et al., 2010. Acute and chronic exposure to shear stress have opposite effects on endothelial permeability to macromolecules. American journal of physiology. Heart 189 and circulatory physiology, 298(6), pp.H1850–6. [4] Salek, M. M. et al., 2012. Analysis of fluid flow and wall shear stress patterns inside partially filled agitated culture well plates. Ann. Biomed. Eng. 40, 707–728 [5] Dubrovskyi, O. et al., 2013. Measurement of local permeability at subcellular level in cell models of agonist- and ventilator-induced lung injury. Laboratory investigation; a journal of technical methods and pathology, 93(2), pp.254– 63.

135 Machine Learning for detection of COPD Exacerbations Peter Gyring (1) Carmelo Velardo (1) David Clifton (1) University of Oxford, Oxford, UK (1)

Background. Chronic Obstructive Pulmonary Disease (COPD) is a progressive and life-long condition characterised by chronically poor airflow, with key symptoms being shortness of breath, coughing, and sputum production. Severity is increased by exacerbations, which involve an acute worsening of symptoms, often leading to hospitalisation [4]. Early treatment of exacerbations is associated with better outcomes [5], but is challenging due to difficulties in predicting such episodes. The advent of mobile-health systems has opened new possibilities for managing COPD exacerbations, such as hospital-at-home schemes, but no systems are yet sufficiently robust for deployment at scale [2]. Here we present several multivariate approaches to detecting exacerbations, using data from a recent randomised clinical trial on COPD home monitoring [1]. Methods. Daily vital signs, symptom and medication data were collected from 107 patients, with a median (interquartile range) of 287 (226-325) days collected per patient. Heart rate (HR) and peripheral blood oxygen saturation (SpO2) were recorded with a pulse oxymeter, and a tablet computer was used to collect self-reported symptoms and medication usage. Symptoms included coughing, sputum and breathlessness, and were aggregated into a symptom score (SS). Medication usage was aggregated into a medication score (MS) and used to define exacerbation occurrence, while HR, SpO2 and SS were used as predictors. Population-based algorithms, including logistic regression (LR) and support vector machines (SVM), were trained with cross- validation on data from the population of patients. Patient-specific models, including kernel density estimation (KDE) and one-class SVMs, were trained on the initial 40 days of data from any given patient, without using information from other patients. All models assumed iid predictors. Testing was done on a day-by-day basis, and used to construct a receiver operating characteristics (ROC) curve. The Area Under the ROC curve (AUROC) was used to compare algorithms. Results. For population-based models, both LR and SVM achieved an AUROC of 0.74, when using the full set of predictors. Correspondingly, for patient-specific models, the KDE achieved an AUROC of 0.68, and the one-class SVM achieved 0.67. Multivariate algorithms were generally better than univariate threshold algorithms. The strongest predictor was SS, and all algorithms fell below an AUROC of 0.65 without this predictor. Discussion. While previous work has reported a significantly higher performance of patient- specific models [3], this was based on biased performance metrics. The current results demonstrate that both population-based and patient-specific models suffer from significant limitations, especially if one wishes to base algorithms on vital signs alone. Conclusion. The results highlight the need for hierarchical algorithms that integrate information from the population- and patient-level, or algorithms that explicitly model time series. We propose specific model formulations based on hierarchical linear models and gaussian processes. Key references. [1] Farmer, A.; Toms, C.; Hardinge, M.; Williams, V.; Rutter, H. & Tarassenko, L. Self-management support using an Internet-linked tablet computer (the EDGE platform)-based intervention in chronic obstructive pulmonary disease: protocol for the EDGE-COPD randomised controlled trial. BMJ Open, 2014, 4, e004437-004437 [2] Sanchez-Morillo, D., Fernandez-Granero, M. A., & Leon-Jimenez, A. Use of predictive algorithms in-home monitoring of chronic obstructive pulmonary disease and asthma: A systematic review. Chronic Respiratory Disease, 2016, 13(3), 264–283. [3] Shah, S. A., Velardo, C., Gibson, O. J., Rutter, H., Farmer, A., & Tarassenko, L. Personalized alerts for patients with COPD using pulse oximetry and symptom scores. Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2014, 3164–3167. [4] Vestbo, J. (Ed.). Global Strategy for the Diagnosis, Management, and Prevention of Chronic Obstructive Pulmonary Disease. Global Initiative for Chronic Obstructive Lung Disease, 2013. [5] Wilkinson, T. M. A.; Donaldson, G. C.; Hurst, J. R.; Seemungal, T. A. R. & Wedzicha, J. A. Early Therapy Improves Outcomes of Exacerbations of Chronic Obstructive Pulmonary Disease American Journal of Respiratory and Critical Care Medicine, 2004, 169, 1298-1303.

136 What is the impact of iodinated contrast media on the radiation doses from diagnostic x-ray examinations? Richard Harbron (1,2) Elizabeth Ainsbury (3) Mark Pearce (1,2) Simon Bouffler (3) Rick Tanner (3) Jonathan Eakins (3) Newcastle University, Institute of Health and Society, Newcastle-upon-Tyne, UK (1) NIHR Health Protection Research Unit in Chemical and Radiation Threats and Hazards, Newcastle-upon-Tyne, UK (2) Public Health England, Centre for Radiation, Chemical and Environmental Hazards, Chilton, Didcott, UK (3) Background: Iodinated contrast media (ICM) are frequently used in x-ray imaging to improve visualization of blood vessels and other structures. Although it may be expected that ICM modify absorbed doses for a given x-ray output, most current dose estimates do not take ICM enhancement into account. A number of studies have reported increases in absorbed dose to the blood and damage to circulating lymphocytes by several hundred percent for ICM-enhanced clinical x-ray procedures, compared to unenhanced equivalents [1-5]. The implications of these findings are unclear, however. The impact on dose and damage to cells outside of blood itself, including those cells prone to malignant transformation, is not known. Methods: A microdosimetric evaluation of the impact of ICM was carried out using the general purpose Monte Carlo code MCNP 6.1. Models representing arteries and capillaries were constructed using ICRP tissue composition specifications. Each blood vessel model was embedded into a larger soft tissue cylinder to simulate beam hardening and electron build-up. Iodine concentrations of 0, 5, 10, 50 and 100 mg/ml were added to the blood. These models were irradiated with four x-ray beam spectra typical of clinical imaging (tube potential ranging from 70- 120 kV, with 0, 0.2 and 0.6 mm copper filtration). Results: For the artery model, the energy imparted (ε) to the blood itself was increased by 66%, 134%, 712% and 1500% for iodine concentrations of 5, 10, 50 and 100 mg/ml respectively, compared to no iodine (120 kV beam spectrum). The respective enhancement effect was reduced to 26%, 51%, 238% and 445% in the region 0-1 μm outside the vessel lumen (equivalent to the endothelium) and falling to close to zero beyond around 50 μm. For capillaries (radius: 5 μm), the increase in ε was smaller and the gradient away from the vessel steeper. Compared to iodine concentration, beam spectrum had relatively little impact. Discussion: Iodinated contrast media may increase energy deposition by several hundred percent. This dose enhancement effect is extremely localised, however. The increase in dose to cells outside blood vessels may be very small. This suggests the impact on cancer risks is also small, certainly in comparison with the apparently large increase in blood cell damage. As the dose to the endothelium of blood vessels may be much higher when ICM is present in the blood, the potential for radiation-induced cardiovascular disease may be underestimated by dose models in which contrast enhancement is not accounted for. Conclusion: While ICM should be taken into account in dose estimates, the overall impact on the radiation doses, and associated cancer risks from diagnostic x-ray exposures is likely to be much smaller than suggested by assays of blood cells or estimates of blood dose alone. Key references. 1. Jost G, Golfier S, Pietsch H, Lengsfeld P, Voth M, Schmid TE, et al. The influence of x-ray contrast agents in computed tomography on the induction of dicentrics and gamma-H2AX foci in lymphocytes of human blood samples. Physics in medicine and biology. 2009;54(20):6029-39. 2. Piechowiak EI, Peter JF, Kleb B, Klose KJ, Heverhagen JT. Intravenous Iodinated Contrast Agents Amplify DNA Radiation Damage at CT. Radiology. 2015;275(3):692-7. 3. Pathe C, Eble K, Schmitz-Beuting D, Keil B, Kaestner B, Voelker M, et al. The presence of iodinated contrast agents amplifies DNA radiation damage in computed tomography. Contrast Media Mol Imaging. 2011;6(6):507-13. 4. Grudzenski S, Kuefner MA, Heckmann MB, Uder M, Lobrich M. Contrast medium-enhanced radiation damage caused by CT examinations. Radiology. 2009;253(3):706-14. 5. Callisen HH, Norman A, Adams FH. Absorbed dose in the presence of contrast agents during pediatric cardiac catheterization. Medical physics. 1979;6(6):504-9.

137 Image quality benchmarking and quality assurance in digital radiography using threshold contrast detail test objects Anna Janeczko (1) Elli-Noora Salo (1) Avril Weir (1) Nick Weir (1) Department of Medical Physics, Royal Infirmary of Edinburgh, NHS Lothian, Edinburgh, UK (1)

Background. In digital radiography (DR) quality assurance programmes, detector performance is typically assessed separately from image processing, using raw images analysed off-site. However, immediate feedback on overall system performance at site and evaluation of the full image processing chain also provide valuable information in managing and optimising equipment. NHS Lothian has introduced the use of two contrast detail test objects as complementary aspects of our DR testing programme. CDRAD[1] was introduced for commissioning testing and optimisation using clinical image processing, whereas test object TO20[2] was introduced for benchmarking and routine surveys using raw images. A previous study at our institution showed that under these conditions the test objects provide results which are sensitive to changes in exposure conditions with good inter- and intra- observer repeatability [3]. Methods. The CDRAD was imaged in a scatter environment, positioned in the middle of 15 cm of Perspex for the table Bucky, 10 cm for the wall Bucky and 2.5 cm of Perspex centrally on the top of CDRAD for out of Bucky exposures (simulating an abdomen, chest and wrist respectively). The images were acquired using default AEC settings for table and wall Bucky, and default manual settings for extremity work for out of Bucky. The TO20 was imaged in scatter free environment, positioned directly on the detector and the images were acquired at 80kV with a fixed dose of 3 µGy to the detector surface. CDRAD images were scored using the dedicated CDRAD Analyser software whereas the TO20 images were scored manually by two experienced observers. Results. The TO20 threshold contrast curves are presented for 18 DR detectors (Fig. 1). CDRAD contrast detail curves are compared for a range of manufacturers and examination types.

100

10 Threshold Threshold Detection Index HT(A) 1 1/2 0.1 Square Root Detail1 Area, A mm 10 Fig.2. Contrast detail detection curves of 5 Fig.1. Threshold contrast detail detection curves derived from repeat images derived from CDRAD Analyser TO20 scores for 18 different DR detectors. for one DR table Bucky system. Discussion. The acquired image of the TO20 is easy to interpret and can give an instant indication of the detector performance by counting the number of detectable discs using defined criteria. The CDRAD results offer a more objective analysis and encompasses the performance of DR detector, image processing and x-ray beam quality. Scores for DR detectors so far tested indicate generally consistent performance across the installed base of equipment. Conclusion. Both test objects have been introduced into our testing programme successfully. We have found that direct on-site verification of image quality using TO20 is particularly useful before handing the equipment back into clinical use. Both TO20 and CDRAD provide a sensitive image quality assessment, with the data collected providing a benchmark which will be particularly useful for ongoing optimisation of clinical imaging. Key references. [1] Manual CDRAD 2.0 Phantom and Analyser software version 2.1, Artinis Medical Systems, The Netherlands. [2] Leeds Test Objects LTD, United Kingdom. [3] Weir A., Digital radiography test objects comparison: CDRAD vs TO20, Internal Report, 2016.

138 Assessing the impact of the interplay effect on 10FFF lung SABR treatment dosimetry Robert Julian (1) Mohammad Hussein (1) Veni Ezhil (1) Royal Surrey County Hospital, Guildford, UK (1)

Background. Lung stereotactic ablative body radiotherapy (SABR) treatments are currently planned locally using 6MV. Changing to 10MV Flattening Filter Free (FFF) could reduce treatment times by >50%1, allowing for better efficiency of service and improved patient experience and stability2. However, while interplay effect has been shown to be negligible with flattened beams3,4, the faster treatment offered by FFF potentially exacerbates the effect5,6. The aim of this study was to assess the interplay effect on both 6MV and 10FFF lung SABR treatment plans in order to determine whether it was clinically significant. Methods. 4DCT scans were taken of the CIRS dynamic thorax phantom with spherical tumour inserts with 1cm and 3cm diameter, and varying breathing amplitudes and periods. 55Gy / 5# SABR plans were produced for both 6MV and 10FFF. Dose measurements within the tumour inserts were taken using a pinpoint ionisation chamber. For a range of plans, measurements were repeated commencing the delivery at different stages of the breathing cycle; these values were then averaged, to simulate the blurring out of interplay effects over multiple fractions. Film measurements were also taken, and gamma analysis used to compare the static and moving films. A 10FFF treatment plan was then produced on a patient scan. The plan was delivered to the phantom with the 3cm tumour insert, 2cm amplitude, 7s period. Measurements were repeated with the insert at various starting positions. The dose delivered from each combination of breathing phases was calculated - 16 combinations in total. In order to simulate the blurring effect caused over multiple fractions, every possible combination of those 16 doses delivered over 5# was calculated (1,048,576 combinations in total), as was the ratio to static for each. Results. Maximum difference between static and motion was seen for 10FFF, 1cm insert, 2cm amplitude (-7.3%). The 10FFF average difference was 0.1% (stdev ±6.7%), while the 6MV average difference was 0.6% (stdev ±2.9%). Averaging over breathing phases improved consistency, giving a 10FFF average difference of -0.7% (±1.5%), and a 6MV average difference of 0.6% (±2.2%). Gamma analysis between static and moving films saw an agreement of >98.5% (3% 3mm) for both 6MV and 10FFF with 1cm insert, 1cm amplitude, and of >95% (5% 5mm) for 1cm and 2cm insert, 2cm amplitude. Most failures were seen in the shoulder regions of the field, as expected. For the patient case, of all combinations of starting phases, the mean difference between static and moving measurements was -0.2%, range -4.9 to 5.4%, stdev 2.9%. Dose difference vs static of all possible combinations delivered over 5# gives a normal distribution, centred about a mean of -0.2%. In total, 98.0% of combinations agree within 3%, and 100% agree within 5%. Discussion. Large differences were seen between moving and static measurements, particularly with 10FFF, with the larger amplitude (2cm) seeing the largest differences. However, these differences were significantly reduced when taking average values over breathing phases. Film measurements saw acceptable agreement (>95%, 5% 5mm) in the high dose region in all cases. The patient plan corroborated these findings – while disagreement is seen on an arc-by-arc basis, averaged over 5# the risk for significant disagreement becomes negligible. Conclusion. Treating SABR lung with 10FFF has been shown to suffer from greater interplay effect than 6MV, but when averaged over several fractions these effects become insignificant. References. 1. Tambe, N. S. et al, 2016. Biomed Phys & Eng Exp, 2(6). [online] goo.gl/1vw4wE [Acc. 12/6/17]. 2. Stieb, S. et al, 2015. Radiation Oncology, 10(27). [online] https://goo.gl/05MfYV [Acc. 12/6/17]. 3. Rao, M. et al, 2012. Int J Rad Onc, Bio, Phys, 83(2):251-256. 4. Ong, C. et al, 2011. Int J Rad Onc, Bio, Phys, 79(1):305-311. 5. Ong, C. et al, 2013. Int J Rad Onc, Bio, Phys, 86(4):743-748. 6. Budgell, G. et al, 2016. Physics in Medicine & Biology, 61(23):8360-8394.

139 Designing affordable and modern smart incubator for easy monitoring and management of premature babies Edwin Khundi (1) Malawi University of science and Technology, Blantyre, Malawi (1)

Background. Many countries experience shortage of medical devices due to high cost of such devices. As such many developing countries depend on donated equipment which usually end up being piled up in storage rooms due to lack of spare parts or expertise [1]. Some of the medical devices include incubators, oxygen concentrating machines and patient monitoring machines. Another problem affecting health care systems especially in low income countries is the shortage of human resource such as nurses, doctors and laboratory scientists [5] [4]. This leads to poor management of many patients, expectant mothers and premature babies that are usually placed in incubators or Kangaroo care system. Many premature babies die due to lack of incubators and also due to lack of proper monitoring systems for the premature babies [2] [3]. Therefore the aim of this study was to design a smart incubator with systems for monitoring and managing premature babies to ensure their highest chance of survival. Methods.The method used to design the device was using computer drawing and modelling software such as Solid Works. The incubator hardware was made from wood, flex form and mild steel. The electronic circuit was made using soldering and simple fastening using screws. The cameras videos were being streamed using device Viewer software. The testing of the devices was conducted in a workshop and not on premature babies. Results. The smart incubator was working properly and images were being streamed successfully to a computer. As shown in figure 1 below a device viewer software has capability to display different views from different cameras. In figure 2 the device viewer is displaying images from two different cameras.

Figure 1 and 7: device viewer connected to two cameras

Discussion. The smart incubator was manufactured using locally available materials. The cameras could remotely capture images that were being displayed by the image viewer. Similarly temperature sensor and humidity sensor could also remotely measure temperature and humidity respectively. This is an innovative way that can be implemented to monitor babies premature and identify those babies that requires urgent medical attention. Thus this system could minimize manual monitoring of babies who is tiresome and limited since a nurse can only monitor one baby at time. However this system can allow the nurse to monitor many babies. Alarm and vital signs measuring tools could assist the nurse to make quick decisions on which babies to attend first. Conclusion. The design project was successful. Many babies in different incubators could be monitored easily using this method. There will be need to find better IP cameras that are small and easy to attach to incubators. The remote monitoring of babies through use of camera and remote sensing can read to non-tiring way of monitoring babies and can also lead to a quick response to help babies whose vital signs are changing or if their breathing is not normal. This can also lead to monitoring many babies and attending to those that require urgent help. Key references. In alphabetical order, numbered.

140 1. A. Jones, “scientificamerican,” 6 5 2013. [Online]. Available: https://www.scientificamerican.com/article/medical-equipment-donated-developing-nations-junk- heap/. [Accessed 2 2 2017]. 2. ARIJ, “ARIJ,” 2017. [Online]. Available: http://en.arij.net/report/incubator-shortage-in-egypt- hospitals-kills-100000-infants-a-year/. [Accessed 17 03 2017]. 3. G. Olawale, 2016. [Online]. Available: http://www.vanguardngr.com/2016/09/lack- incubators-tragedy-nigerias-premature-babies/. [Accessed 20 03 2017]. 4. Oxfam. [Online]. Available: https://www.oxfam.org/sites/www.oxfam.org/files/hcic.pdf. [Accessed 02 02 2017]. 5. S. Naicker, J. Plange-Rhul, R. C. Tutt and J. B. Eastwood, “Shortage of Healthcare worker in Developing Countries - Africa,” Ethnicity & Disease, vol. Volume 19, no. http://forms.ishib.org/journal/19-1s1/ethn-19-01s1-60.pdf, 2009.

OPT Image Quality Phantom for Dose Optimisation and establishing QA programme Nadia Latif (1) Stephen McCallum (1) NHS Grampian, Aberdeen, Scotland, UK (1) Background. Following an audit of Orthopantomography (OPT) image quality, it was concluded that a test protocol for commissioning, routine and user quality assurance (QA) tests should be developed. The protocol should include tests for monitoring resolution, contrast and changes in image magnification during life time of equipment caused by wear of moving components. The available image quality phantoms designed as per IEC 61223-3-4 standard had limited number of resolution bar patterns up to 3.0 lp/mm. Moreover, only four low contrast objects were available in these phantoms. These phantoms were not sensitive to changes in exposure factors. Due to the limitations of these phantoms there was clearly a need for a new Image Quality (IQ) phantom that has better spatial resolution object with smaller gradation between line pair groups, low contrast objects of varying thickness and test object for monitoring geometry changes. Methods. We fabricated two phantoms to ensure that the results of tests do not blur each other out. IQ-A phantom has embedded tests objects for 7 low contrast objects of varying thickness, 7 steps of step wedge object, 4 low contrast holes of different sizes and ball bearings for geometry tests. IQ-B phantom has embedded spatial resolution test object (Leeds test object, Type 18UL) that has smaller gradation between line pair groups. A 0.8mm Cu filter was attached at primary collimator as a head equivalent. A jig was used to overcome the problem of positional reproducibility. Results.

Figure 1:IQ-A Figure2: IQ-B Figure 3: Low contrast Figure 4: Spatial resolution The images were obtained on Instrumentarium OP100D at a range of available exposure settings. The preliminary results (figure 3 and figure 4) show that these phantoms are sensitive to changes to tube voltage (kvP) and current (mA). It was observed that the spatial resolution and low contrast can be improved to a certain extent by increasing the dose which will reduce noise as expected. Due to the limitations of the detector further increase in dose will not result in improvement of image quality as it also depends on the detector resolution.

141 Discussion. There is a potential for dose optimisation in direct digital radiography (DDR) systems by using these phantoms as a benchmark for image quality. IQ phantoms will be useful for acceptance tests to establish baselines and optimise patient doses and to monitor any changes during routine tests. It will enable dental practices to set up an efficient QA program to monitor changes in performance. A dedicated software can potentially be developed to eliminate viewer bias and automate the image quality tests in a quick and an objective way. Conclusion. An effective QA program can be established to better meet the legal requirement. The QA results with the new phantom will be more sensitive to the variation in image quality and to changes in image magnification due to wear of moving components. It can be used to optimise the OPT units settings to keep doses as low as reasonably practicable.

Assessing the feasibility of adaptive planning for prostate radiotherapy using SmartAdapt deformable image registration Virginia Marin Anaya (1) Jamie Fairfoul (2) Cambridge University Hospitals NHS Foundation Trust, Addenbrooke's Hospital, Cambridge, UK (1) Peterborough and Stamford Hospitals NHS Foundation Trust, Peterborough City Hospital, Peterborough, UK (2)

Background: Prostate cancer presents a number of challenges such as inter-and intra-fractional variations in patient set-up and anatomy. The purpose of this study is to assess the feasibility of adaptive planning for prostate RapidArc using the Varian SmartAdapt Version 11 deformable image registration tool, in an effort to reduce the effect of patient-specific variations. Eiland et al [1] suggested that the deformed dataset produced by SmartAdapt could be used as a decision tool to establish whether the patient would need to be rescanned. In order to implement this clinically, the deformed dataset should overestimate the need for re-planning. Methods: 18 prostate patients treated with RapidArc (74 Gy in 37 fractions) were selected. Each patient had a planning CT and CBCT images acquired at fractions 1, 6, 11, 16, 21, 26 and 31. For the first fraction, the planning CT was registered to the CBCT image dataset, using SmartAdapt. A rigid registration was performed between the deformed image dataset and the original planning CT. The deformed volumes/structures were then propagated to the planning CT. In Eclipse, the original treatment plan was copied and pasted, assigning the new structure set that contains the original, deformed and the reference contours. The manual contours on the CBCT images by the same clinician were used as reference. The new treatment plan was re-calculated with the same settings as in the original plan. DVH points and the Conformity Index were used for the analysis. Results: The difference in DVH points for the original planning CT and the deformed volumes/structures relative to the reference were determined.

The median and range CI values obtained from the deformed PTV2 were very similar to those from the reference PTV2 and higher than one, indicating that healthy tissue was being included in the 95% isodose.

Discussion: The difference in DVH points between the deformed and the reference CTV2 and

142 PTV2 was worse than the difference between the planning CT and the reference, overestimating the need for re-planning. However, there was a tendency for the deformed Rectum to underestimate the need for re-planning, when compared to the reference Rectum (manually-drawn contours on the CBCT image dataset). The CI values obtained from the deformed PTV2 are very similar to those from the reference PTV2. Thus, in combination with the CI from the deformed PTV2, the deformed CTV2 and PTV2 could be used to evaluate whether a re-scan is needed when disregarding the deformed Rectum. Conclusion: The aim of this study was to determine whether SmartAdapt allows us to make judgements on the impact of anatomical changes that we would make by manually re-outlining the volumes and structures. There is a tendency for the deformed CTV2 to overestimate the need for re-planning, when compared with the reference CTV2. After cautious analysis of the results obtained in this study, it has been decided not to implement SmartAdapt clinically to trigger adaptive decisions for prostate cancer in our department. Further work is needed to evaluate the performance of SmartAdapt for other sites such as head and neck. Key references: [1] Eiland, R. B., Maare, C., SjöStröm, D., Samsøe, E. and Behrens, C. F. (2014) ‘Dosimetric and geometric evaluation of the use of deformable image registration in adaptive intensity-modulated radiotherapy for head-and-neck cancer’, Journal of Radiation Research, 55(5), pp. 1002–1008. doi: 10.1093/jrr/rru044.

Gait-Stability in simulated maritime conditions Michael McLellan (1) University of Strathclyde, Glasgow, UK (1)

Background. Gait analysis in maritime conditions has seen very little research. Human locomotion is now well understood and documented through assigning mathematical and physical principles to quantitative measurements systems like motion capture and force plate technology. The aim of this study is to adopt biomechanical measurement methods for the gait analysis of walking in maritime conditions. The investigations objectives include deriving and simulating various maritime conditions for output on the Computer Aided Rehabilitation ENvironment (CAREN) platform, defining and visualising gait events and the detection of statistical differences between control groups. This is the first experiment of its kind and aims to establish several objectives; 1) investigate differences in performance between participants with maritime experience and those who do not 2) establish possible reasons for difference in performance between group through biomechanical gait analysis techniques 3) define or derive possible mathematical relationship hypotheses between hydrodynamics, naval architecture and biomechanical gait analysis 4) finally, relate findings to further research and possible applications. Methods. The participant was exposed an open sea simulated environment where they walked in- sim across the deck of a ship in increasing stormy conditions. They performed a 50m walk to complete each trail. They walked the same 50m distance at 5 different levels of increasing intensity and each stage was repeated 3 times. Results. Results are recorded in several forms; time to finish, walking speed, step parameters (width and length), and margins of stability (MoS). MoS being a measure between centre of mass (CoM) and the base of support (BoS) and giving an indication of balance control. This developed method by Hof et al, (A. L. Hof,Gazendam and Sinke, 2005, At L. Hof et al., 2007) has been very successfully utilised by in previous studies within the CAREN setting studying balance control by introducing perturbations and examining MoS comparisons (Hak et al., 2012, Hak et al., 2013, Hak et al., 2015). It is from the success of these studies and their similarity to this study that the MoS approach is one of the methods used to assess balance control between groups. Discussion. The experiment will take place in in June 2017. Conclusion. As Above.

Key references. Hak, L. et al. (2012) 'Speeding up or slowing down?: Gait adaptations to preserve gait stability in response to balance perturbations'. Gait Posture, 36 (2), pp. 260-264. Hak, L. et al. (2013) 'Stepping strategies for regulating gait adaptability and stability'. J Biomech, 46 (5), pp. 905-911. Hak, L. et al. (2015) 'Stride frequency and length adjustment in post-stroke individuals: influence

143 on the margins of stability'. J Rehabil Med, 47 (2), pp. 126-132. Hof, A.L., Gazendam, M.G.J. and Sinke, W.E. (2005) 'The condition for dynamic stability'. Journal of Biomechanics, 38 (1), pp. 1-8. Hof, A.L. et al. (2007) 'Control of lateral balance in walking: Experimental findings in normal subjects and above-knee amputees'. Gait & Posture, 25 (2), pp. 250-258. Oddsson, L., I. et al. (2004) 'Recovery from perturbations during paced walking'. Gait Posture, 19 (1), pp. 24-34. Paloski, W.H. et al. (2006) 'Destabilization of human balance control by static and dynamic head tilts'. Gait & Posture, 23 (3), pp. 315-323. Priplata, A. et al. (2002) 'Noise-Enhanced Human Balance Control'. Physical Review Letters, 89 (23), pp. 238101.

Evaluation of local irradiation and magnetic resonance imaging with remote analysis for radiotherapy dosimetry Rollo Moore (1) Tess A'Lee (3) Maria Schmidt (1,2) Eva Kousi (2) Filipa Costa (2) Evangelos Pappas (4) Emma Wells (1) Royal Marsden Hospital, Chelsea, London, UK (1) Institute of Cancer Research, Sutton, Surrey, UK (2) Queen Mary, University of London, Mile End, London, UK (3) RTsafe,P.C., Athens, Greece (4)

Background. The study was initiated as part of a commissioning program for image guided radiation therapy for SRS/SABR. It was a co-funded collaborative project including an undergraduate placement July-August 2016 (SEPnet [2] / Royal Marsden / RTSafe [1]). The aims included verification of precision radiation therapy under image guidance on Cyberknife and TrueBeam delivery systems of cranial radiotherapy. This work aims to assure quality appropriate to high dose and multiple sites. It leads the way for extracranial work. It forms part of the departmental risk mitigation strategy developed from ideas in references [3] to [6]. Methods. In order to assess end-to-end relative dose delivery precision RTSafe create an anthropomorphic phantom from an xray CT scan data set using 3d printing material of bone-like density in place of bone and in an ~2mm layer in place of skin (to seal and provide rigidity). Radiosensitive gel is held in the cranial cavity and a filler gel is used in other space between bones and skin. MRI scanning was performed at 1.5T (Siemens, Aera) using an 18-channel surface coil and a standardized RT positioning for head/neck. A 2D fast spin echo sequence with multiple echo times was employed (TR/TE=2000ms / 36, 436, 835, 1230ms, voxel size = 1.4 x 1.4 x 2 mm3, bandwidth = 780 Hz / px) and T2 relaxation time was calculated on a pixel-by-pixel basis using the workstation provided by the vendor (Siemens). T2 maps were subsequently transferred off- line and R2 (=1/T2) maps were also computed in excel to manually independently check a sample of points. The system-based geometric distortions are routinely assessed and errors are kept to a minimum in the clinically useful volume. The acquired raw images of the irradiated phantom were separately processed by RTsafe for dosimetry. A radiotherapy plan was produced as if to irradiate small, medium and large volume metastatic lesions, reasonably separated in the brain phantom (mimicking a patient like treatment). A calibration experiment was also carried out where a gel-filled cuboid with pmma walls was irradiated with three orthogonal open photon beams to create a distribution of dose. The calibration uncertainty influences include the accuracy of the relevant planning system, linac delivery in addition to the MR readout and dose conversion model. Results: The study indicated adequate PTV coverage and OAR sparing on both platforms.

144 Figure 1 Graphical display from RTSafe indicating CT planning contoured structures PTV, OAR (LHS). Dose profile normalised by equating maxima in TPS and RTSafe gel (top RHS); 1D gamma (5%,2mm) plot along representative profile on greyscale image Conclusions include that this method is worth the financial and resource implications in commissioning and periodic end-to-end risk mitigation in high dose IGRT. [1] A new polymer gel for magnetic resonance imaging (MRI) radiation dosimetry. E Pappas, T Maris, A Angelopoulos, M Paparigopoulou, L Sakelliou, P Sandilos, S Voyiatzi and L Vlachos. 1999 Phys. Med. Biol. 44 2677 (http://iopscience.iop.org/0031-9155/44/10/320) www.rt-safe.com [2] SEPnet: South East Physics Network www.sepnet.ac.uk [3] National Radiotherapy Implementation Group Report; Stereotactic Body Radiotherapy; Guidelines for Commissioners, Providers and Clinicians in England 2011 [4] The Royal College of Radiologists, Society and College of Radiographers, Institute of Physics and Engineering in Medicine, National Patient Safety Agency, British Institute of Radiology. Towards Safer Radiotherapy. London: The Royal College of Radiologists, 2008 [5] On Target: Ensuring Geometric Accuracy In Radiotherapy. A joint report published by the Society and College of Radiographers, the Institute of Physics and Engineering in Medicine and The Royal College of Radiologists. [6] National Radiotherapy Implementation Group Report; Image Guided Radiotherapy (IGRT): Guidance for implementation and use. August 2012

Review of IMRT/VMAT Practice at ICH 1Morris M, 1Harrison L, 1McLauchlan R. 1Department of Radiation Physics and Radiobiology, Imperial College Healthcare NHS Trust (ICH), London. Background. This study presents some of the further work highlighted in the “Review and Rationalisation of IMRT/VMAT Practice at ICT” 2D to 3D volumetric analysis with a view to rationalisation of QA practises at Imperial College Healthcare NHS Trust” which was presented to the IPEM radiotherapy special interest group in April 2016. The current practice for patients with prostate cancer with seminal vesicle involvement (PSV) at ICH is to use a PTW® 729 array for both IMRT and VMAT but in solid water at gantry 0 for each field for IMRT plans and in a 4D-Octavius phantom for VMAT plans with the whole planned dose integrated over all of the arcs. This study investigates the rationalisation of the patient QA process to use only the 4D Octavius. Methods. During the initial VMAT commissioning, a unique cohort of 20 patients with PSV were dual planned in both VMAT and (backup) IMRT. These plans were originally exported to a verification plan to be delivered to a PTW Octavius 729 array in a solid water phantom at gantry 0 and subsequently analysed. The same IMRT plans were then re-exported to a verification plan using the 4D Octavius rotational phantom and subsequently re-analysed by comparison with a single (integrated) acquisition for all of the fields. Both sets of analyses were carried out with PTW Verisoft 6.2. These results were compared against the original (gantry 0) verification method using the PTW 729 array in a solid water phantom with the same gamma index criteria to highlight any difference in the two methods. Results. With the IMRT at gantry angle method with using the PTW Octavius rotational phantom, an average of 92.5 ± 1.6% of voxels for all of the plans in this study fell within the pass criteria, whereas with the original IMRT at gantry 0 and solid water phantom method, an average

145 98.2 ± 2.0% of voxels averaged over all of the plan fields fell within the pass criteria. In this study the gamma index criteria was set so that a pass is recorded for a particular field if 95% of the voxels are found to be within 3%/3mm of the expected (local) dose, with suppression of any voxels  10% of the maximum dose. Discussion. Taking into account the previous work undertaken which identified amongst other things that for IMRT there is no difference between 2D and 3D analysis with planar acquisition, the lower average percentage scores found when plans were delivered to the Octavius rotational phantom could possibly be explained by the more complicated delivery method having an impact on the accuracy of plan delivery. Conclusion. The Octavius rotational phantom method could be adopted for both IMRT and VMAT patient QA verification as it provides a reduced setup time when a mixture of IMRT and VMAT patient plan verifications are to be undertaken during a single QA session. Key references. 1. D. Low et al. A technique for the quantative evaluation of dose distributions. Medical Physics. 5, 1998, Vol. 25, pp. 656-661. 2. Depuydt et al. A quantitative evaluation of IMRT dose distributions: refinement and clinical assessment of gamma evaluation. Radiotherapy and Oncology. 2002, Vol. 62, pp. 309-319.

Clinical commissioning process of Volumetric Arc Therapy(VMAT) using Varian Eclipse treatment planning system and RapiARC on Varian Truebeam machines Munshi M,Colchester Hospital University NHS Foundation Trust Background. VMAT is one of the preferred methods of Intensity Modulated Therapy (IMRT) in the UK and worldwide (1,6) mainly because of ease of planning and quick in execution. At CHUFT, VMAT is clinically in use for Prostates and Prostate node (PPN) plans using Eclipse treatment planning system (TPS) (ver. 11) & Varian TrueBEAM therapy machines. We routinely use Sliding window technique for all other IMRT plans. Methods: For commissioning purpose, Five prostates and four PPN patients already been treated with sliding window technique are chosen for different test plans following CHUFT clinical protocols to maintain same quality of each plan. One of these patients had left hip prosthesis. Original CT data and contours were copied,renamend and reused for the VMAT plans with single full arc rotation for prostates and two opposing full arcs for PPNs.First arc rotates 181.0 to 179.0 degree clockwise having 30 degree collimation and second arc rotates 179.0 to 181.0 degree rotation with collimation 330 degrees. Different plans tested for (A) 6 and 10MV energies(B)with and without Jaw tracking(C)with NTO and with ANTO settings(D)with different upper limits based on prescription is 2Gy per fraction or more to check plan modulation.(E) One of the patients had different collimation rotation to check the effect of different collimator rotation to the recommended settings(2).Different plan parameters like gantry speed, dose rate variations, difference in dose between two arcs in case of two arc plans and effect of increased difference between two arcs MUs are noted for estimation of modulation of each plan. Each plan is independently checked using RadCalc ,Average Leaf pair opening(ALPO) is noted for each plan that also act as a parameter for checking plan modulation. In terms of machine QA specific to VMAT we use a standard VMAT plan delivered on every treatment machine and compare with the standard plan. We check flatness and symmetry variations for constant dose rate, variable dose rate and variable gantry speed as per NCS24 guidelines (3).Q.A.of each plan is done using gantry mounted MapCHECK2, Varian’s Portal dosimetry tool and point dose measurements specially for PPN patients in which seven different locations were tested for point dose measurements using solid water and 0.6cc ion chamber. Results: All the prostate plans passed RadCalc measurements for dose tolerance of +-2%.In case of PPNs, three out of four plans had near tolerance passing or just failed particularly for the second arc. ALPO of each arc is noted and statistics indicate minimum ALPO value of 3.36 and mean of 4.26.All plans passed MapCHECK2 with greater than 95% points passing 2%/2mm global criteria for every individual arc and mean pass rate for all the plans was 98.08%(std.dev 1.23%). Portal

146 Imaging shows all but one plan, passed criteria of more than 95% points passing 3%/2mm criteria with mean pass rate of 98.23%.All PPN patients point dose checks pass in PTV within ±2.0% except in one case and worst result in gradient is -4.94% for Arc2 but for total dose it is - 3.8%.Comparison between Jaw tracking and without Jaw tracking shown better results with former.Also comparison between NTO settings shown better results to that of ANTO settings. Conclusion. Following the commissioning results, a consultation with oncologists and radiographers been done. Some necessary training is given for planning, plan checking and treatment execution with detailed work instructions. This work led to routine use of VMAT for prostates and PPNs and now further QA data are being collected with all the mentioned QA methods to tighten the QA passing criteria and further updates in work instructions. Key references. 1.A Comprehensive comparison of IMRT and VMAT plan quality for prostate cancer treatment.E.M.Quanet.al.;IJROBP,July12,2012,vol.83,Issue4,Pages 1169-1178.2.Analysis of RA optimization strategies using objective function values and DVHs.M.Oliver et.al.;JACMP,Vol 11,No 1(2010).3.Code of practice for Quality assurance and control for volumetric modulated Arc therapy.(NCSReport24).4.Implementing RA in to clinical routine:A comprehensive program from machine QA to TPS validation and Patient QA.A.V.Esch et.al.Med.Phys.38(9),September 2011.5.RapidArc(RA)-Planning techniques,tips and tricks.Presentation by O.Apinorasethkul;AAMD ,FL-June 25,2015 6.VMAT:a review of current literature and clinical use in practice.M.Teoh et.al.;BJR. 2011,Nov.84:967-996.

Creation of HTML based clinical site specific Eclipse Scripts based on the PlanQuality Metrics project Munshi M., Colchester Hospital University NHS Foundation Trust Background. Eclipse scripting API uses Microsoft’s Visual studio (Ver.10 and above) C# language integrated with Eclipse treatment planning system. One such project called PlanQualityMetrics was uploaded on variandeveloper.codeplex.com (Feb 28, 2014) that generates HTML based Plan quality reports for the selected plan or plan sum using QUANTEC style metrics. Methods. The original PlanQualityMetrics project was modified and additional codes were added to customise the project as per local requirements. Unlike original program, new codes give statistics for dose to body for maximum dose and dose above 107% of prescription with objective of 1cc and constraint for 2cc.New program will give dose information for multiple planning Target volumes (PTVs) and it will also give information about the median dose to the PTVs. Based on the local department requirements of reporting the dose constraints in terms of Dose to the volume, additional codes were written to generate reports with Dose to volume, Median dose, Mean dose, body dose, Dose above max tolerance apart from Volume at dose which was already present in the original program. HTML reporting codes were also modified from the original program to give information on Headings displaying hospital information followed by Patient information -Name and hospital Identification, Signature spaces for Planner, Approver and checker with date and time on the top of the report. Each report is generated in HTML format with date and time stamp and also with the initials of the one who has generated the report. It is then followed by the script name and published version, tick boxes for the clinical sites are provided for the planner to tick before printing or publishing on the system database. As taken from the original program, each HTML report shows Eclipse version and statistics for each structure starts with the structure name, its volume in cubic centimetre(cc) followed by quality metrics that shows a table with Name of the evaluation method,type,Evaluation point, Calculated value, plan Objective,Cosntraint and Evaluation in terms of Pass, Fail or Warn. These tabular codes were taken from original program. Modifications were made in terms of adding different Evaluation names,Types,Methods to show Evaluation points in percentage or absolute terms as required by the local department and Evaluation that shows different background colour when PASS(white),Warn(light grey) and FAIL(larger fonts and darker grey).Additional codes in HTML written to add comment section after each structure information and in the end each plan quality report ends with **End of Plan quality report, patient name, ID and Date time stamp**.It is then followed by Special comments section for any additional observations about the plan. Results: Total Six Eclipse script APIs were generated which have outputs in “.esapi.dll” format and can be accessed from any computer that has access to Eclipse treatment planning system or ARIA

147 system including those which have remote access to ARIA. Each of these APIs has on an average three different clinical sites or site with different prescriptions and dose constraints. Each script initially generates an HTML report that can be either printed on paper, in PDF or can be saved as HTML.A PDF report can then be signed electronically and can be uploaded on ARIA database. It can also be sent securely through NHS mail if required. Conclusion. Eclipse scripting API can be used effectively to generate a report indicating plan quality and can also be customised as per departmental needs and can be useful for paperless plan quality reporting in Radiotherapy. Key references. 1. CHHiP clinical protocols; ICR-CTSU 2006. 2. Eclipse scripting API reference guide. Varian Medical systems. March 2014. 3. PARSPORT trial protocol; March 2004. 4. Varian APIs in V11 and V13.Presentation by W.Keranen, 17Jan 2014. 5. Write it the right way-an introduction to scripting, my Varian Webinar April 2015; W.Keranen.

Similitude in Biomechanical Experimentation 1Ochoa R, 1Alonso-Rasgado M. T, 1Davey K. 1Bioengineering Research Group, School of Materials, University of Manchester, UK.

Background. Many obstacles arise when considering biomechanical experimentation: resource limitations, unavailability of materials, manufacturing complexities, time scale difficulties and, ethical compliance issues. As such, a scaling approach resolving all or some of these issues is of some interest to the bio-community for which the most commonly used approach is dimensional analysis1 but there are more approaches within biology2. The focus of this work is the application of a novel scaling approach3 to the design of scaled biomechanical experiments. Similitude exists when the physics on scaled spaces is described by proportional transport equations4 and ultimately relates the balance laws for each space to guarantee the physics remain representative. This work demonstrates how the concept can be applied to biomechanical experimentation using a commercial finite element package and validated by means of image correlation software. Methods. The use of synthetic composite bone in biomechanical experimentation is studied using the scaling approach which leads to the selection materials manufacture by means of FDM (fuse deposition modelling) for the small-scale, trial-space materials. The materials derived from the theory are analysed in finite element models of a simple geometry (cylinder) and in a femur geometry undergoing compression, tension, torsion and bending tests to assess the efficacy of the approach. Furthermore, the physical-trial experimentation using these materials for the compression and bending tests and image correlation software is carried out. Results. The results show similar strain patterns in the surface for the cylinder with a maximum percentage difference of less than 10% for all the tests and for the femur geometry a maximum percentage difference of less than 4% for all the tests. Finally, physical-trial space experimentation using the proposed materials for the cylinder tests provides a mean difference no greater than 5% and a good agreement using the Bland Altman statistical analysis providing good supporting evidence for the practicality of the scaling approach. Discussion. Finite element models analyse a wide variety of subjects such as the study of the role of longitudinal bone curvature in the design of limb bones, the optimization of the position of the acetabulum in a periacetabular osteotomy or the study of stress in the femoral head neck junction after osteochondroplasty. With this in mind, facilitating the production of these models by reducing the cost and time with easily manufactured materials is important and the results shown demonstrate the efficacy of the approach. Conclusion. The use of finite similitude and materials manufactured by means of FDM in biomechanical experimentation is a viable alternative to the synthetic composite bone when validating finite element models by means of image correlation software. The significant reduction in cost and the availability of each specimen result in an efficient alternative that can potentially significantly reduce the time required to produce finite element models in the field of biomedical engineering. Key references.

148 1Barenblatt G. I. Scaling, Self-similarity and intermediate asymptotics. Cambridge University Press, 1996. 2Brown J. H, West G. B. Scaling in Biology. Oxford University Press, 2000. 3Davey K, Darvizeh R, Al-Tamimi A. Finite similitude in metal forming. MATEC Web Conf, 2016. 4Davey K, Darvizeh R. Neglected transport equations: extended Rankine-Hugoniot conditions and J-integrals for fracture. Continuum Mechanics and Thermodynamics, 28(5):1525-1552, March 2016.

Diagnostic Reference Level for Brain and Chest Computed Tomography Scans: An Evaluation of CT Protocol Dose Index in a Diagnostic Facility Abayomi Opadele (1) Micheal Akpochafor (2) Moses Aweda (3) College of Medicine, University of Lagos, Lagos State, Nigeria (1) College of Medicine, University of Lagos, Lagos State, Nigeria (2) College of Medicine, University of Lagos, Lagos State, Nigeria (3)

Recent surveys in large medical centers have found that CT studies now often account for 25% of all examinations and 60 to 70% of the patient dose received from diagnostic radiology (IAEA, 2009). This has contributed a sense of urgency to the impetus for establishing diagnostic reference levels and dose reduction strategies. Hence, the concept of diagnostic reference levels (DRLs) is widely recommended for use with CT (ICRP, 1996). The need to establish local diagnostic reference level for brain and chest computed tomography scans was considered in this study. The quantities estimated in this study were the mean (75th percentile) volumetric computed tomography dose index (CTDIv) value, dose-length product (DLP) and the effective dose (DE) received by the patients; the data and results were compared with that of national and international established work. The study was carried out on a total of 157 patients; 87 patients underwent routine brain computed tomography examinations, while 70 patients underwent chest computed tomography examinations. For brain protocols, the CTDIv and DLP obtained from this study were 51.19mGy and 852.96mGy.cm respectively which is below the reference value by European Commission (60mGy and 1050mGy.cm). The effective dose obtained for male and female patients were 1.90mSv and 1.85mSv respectively while 1.88mSv was obtained for both male and female patients. For chest protocols, the CTDIv and DLP obtained from this study were 16.8mGy and 490.07mGy.cm respectively which comparably higher than the reference value by European Commission (12mGy and 430mGy.cm). When compared with international standards, the great variation in dose distribution was attributed to technical parameters, clinical complexity of the patients and untimely quality control program. Hence, there is a dire need to establish local diagnostic reference levels which are corrective measures that are required in elimination of unnecessary radiation that does not contribute to overall profile of the patients.

Assessment of IMRT and VMAT using RapidPlanTM versus IMRT using a class solution for the treatment of prostate cancer in a 60Gy in 20 fraction dose regimen Kate Sexton (1) Suzanne Smith (1) Beatson West of Scotland Cancer Centre, Glasgow, UK (1)

Background: To compare IMRT and VMAT methods using a RapidPlanTM model to assess if IMRT plans using a model based on VMAT treatment would give clinically acceptable plans for treatment of prostate cancer using a 60Gy in 20 fraction schedule. In addition, to use a class solution for the IMRT method to compare to the RapidPlanTM results for VMAT and IMRT and assess if these methods differ in terms of improved dose to organs at risk and/or planning target volume coverage. Methods: A planning study was performed for 10 prostate patients using the Eclipse Treatment Planning System [Varian Medical Systems] v13.6. These patients were chosen from a larger sample as they had average bladder and rectum volume, thus eliminating potential outliers that may affect performance. Treatments were planned using three different methods: a class solution for 60Gy in 20 fractions to optimise an IMRT plan, a RapidPlanTM model for 60Gy in 20 fractions for the optimisation of an IMRT plan and, using the same RapidPlanTM model, the optimisation of a

149 VMAT plan. The dose and fractionation used for prostate treatment have been adopted following the Phase 3 CHHiP trial results. Constraints for the 60Gy in 20 fraction dose regimen have been set according to this trial. However, this centre uses a tightened bladder constraint set compared to CHHiP trial recommendations. All plans were normalised to ensure that the target mean was 100% and the doses to organs at risk (OARs) and coverage of PTVs were recorded and assessed. Differences in OAR dose volume histogram parameters were calculated to assess plan quality. Significance was assessed by two- tailed t-test (p<0.01). Results: Some differences were observed in comparing the planning methods. The majority of plans generated using each method were clinically acceptable and met local OAR and PTV dosimetric objectives without alteration of the optimisation models for individual cases. Plans generated using RapidPlanTM (both VMAT and IMRT) gave superior results to IMRT plans generated using the class solution for 60cGy in 20 fraction treatment. It was found that there is a statistically significant difference between RapidPlanTM VMAT and RapidPlanTM IMRT in the dose received by 3% of bladder volume with VMAT showing an increase in dose. For all other constraints VMAT gave better OAR sparing than both IMRT planning methods. For six patients VMAT gave the most optimal plan and for four patients RapidPlan with IMRT gave the most optimal plan. VMAT plans have a statistically significant increase in monitor units compared to that of IMRT plans. Discussion: When compared with an IMRT plan using a class solution for plan generation, it was found that the use of the RapidPlanTM model demonstrates improved relationship between balancing OAR objectives and PTV conformity. This indicates that the specific RapidPlanTM model used is superior to the class solution used for prostate treatment in this dose prescription. IMRT using RapidPlanTM outperformed VMAT when delivering locally acceptable dose to 3% of bladder volume. This indicates that it may be useful to propose an alternative IMRT treatment plan using the RapidPlanTM model should the bladder dose in a VMAT treatment plan be unacceptable. Conclusion: A RapidPlanTM model was used to optimise VMAT and IMRT treatment plans for prostate and compared to optimising IMRT plans using a class solution. Plans were assessed and it was found that the methods differ in terms of improved dose to organs at risk, most noticeably in the dose received by bladder when comparing VMAT and IMRT plans optimised using RapidPlanTM. The results obtained may lead to the in-house implementation of IMRT plans for certain patients where bladder dose delivered in a VMAT treatment plan is out of tolerance according to local dose constraints. Key references: 1 3rd ESTRO Forum 2015: A comparative audit of IMRT and VMAT for prostate cancer; 2 The Green Journal: Clinical validation and benchmarking of knowledge-based IMRT and VMAT treatment planning in pelvic anatomy; 3 JMIRO: Dose comparisons for conformal, IMRT and VMAT prostate plans; 4 Lancet: Conventional versus hypofractionated high-dose intensity- modulated radiotherapy for prostate cancer: 5-year outcomes of the randomised, non-inferiority, phase 3 CHHiP trial; 5 Radiation Oncology: Applying a RapidPlan model trained on a technique and orientation to another: a feasibility and dosimetric evaluation.

Heat, Mass and Momentum Transfer in Laminar 2-Phase Flow Lawrence Shaban (1) Kenny Low (1) Swansea University, West Glamorgan, UK (1)

Background: There are many situations in which it is necessary to transfer heat concurrently with membrane gas transfer. There are also applications in which it is desirable to transfer gas and ionic species. Examples in medicine are that of Extracorporeal Membrane Oxygenation (ECMO)1 and Cardiopulmonary Bypass (CPB)2. Traditionally, the need is met by a 2-stage device in which the blood is first heated and then oxygenated. There would be substantial benefits to the patient in combining the two steps into one, whereby the blood is less stressed and a smaller volume of blood is required outside the body2,3. The aim of the research is to determine whether the two stages can be combined by employing a 2-phase flow4,5,6 of, for example, oxygen (for gas transfer) and an aqueous solution (for heat transfer). There are also potential applications in which refrigerated stored transfusion blood can be simultaneously heated and oxygenated for delivery to

150 the patient.7 Methods: Step 1) Identify the 2-phase flow regime arising within the hollow fibres used in such devices. Step 2) Solve the first order differential equations describing the heat and mass transfer in such a flow regime. Step 3) Undertake a Computational Fluid Dynamic (CFD) simulation of a complete hollow fibre membrane device. Step 4) Compare the CFD results to laboratory studies of both single fibre and whole device measurements. Step 5) Parameterisation of Step (3) to identify the best combination of fluid and gas to achieve the most effective performance. Discussion: CFD has proved an effective tool in studying the complex flows arising in these devices and has potential for more detailed studies including inlet and outlet effects. Experimental results on small units suitable for blood transfusion show that the distribution of liquid and gas is non-uniform between fibres in a multi-fibre device. It remains to determine whether this non- uniformity has a significant adverse impact on performance, and whether improved design can achieve a more uniform distribution. Conclusions: It has been determined that, within such small diameter fibres (less than 300 micron), slug flow is the dominant flow regime. It has been further established that this 2-phase slug flow enhances convection and aids the transfer of oxygen and heat, by use of recirculation velocities to enhance convection. Consequently, it should be possible to employ 2-phase flow simultaneously to oxygenate and heat blood with benefits to patient outcomes. However, further work is required to determine (and possibly reduce) the impact of non-uniform flow between fibres in a full-scale multi-fibre device, such as an ECMO that may have tens of thousands of fibres. Key references: [1] Faruk O, Baran A, Ali U. Use of Extracorporeal Membrane Oxygenation in Adults. 2014;10–23. [2] Passaroni AC, Augusto M, Silva DM. Cardiopulmonary bypass : development of John Gibbon’s heart-lung machine. 2015;235–45. [3] Murphy DA, Hockings LE, Andrews RK, Aubron C, Gardiner EE, Pellegrino VA, et al. Extracorporeal Membrane Oxygenation — Hemostatic Complications. Transfus Med Rev [Internet]. 2015;29(2):90–101. Available from: http://dx.doi.org/10.1016/j.tmrv.2014.12.001 [4] V. Talimi, Y.S. Muzychka, S. Kocabiyik, A review on numerical studies of slug flow hydrodynamics and heat transfer in microtubes and microchannels, International Journal of Multiphase Flow, Volume 39, 2012, Pages 88-104, ISSN 0301-9322. [5] Thome JR. Two-Phase Flow Patterns. Eng Data B III. 2007;1–34. [6] Che Z, Wong TN, Nguyen NT. An analytical model for plug flow in microcapillaries with circular cross section. Int J Heat Fluid Flow. 2011;32(5):1005–13. [7]Johns W R, Evans A F and Knight R K. Mass Exchange Apparatus and Methods for the Use Thereof. Patent Application WO2016/120591, January 2015 (International August 2016).

Conceptual Design of a Novel Catheter Ablation Guidance System based on a Soft Stewart Platform Mark Street (1) Yohan Noh (1) Kawal Rhode (1) Kings College London, London, UK (1)

Background This objective is to develop a novel catheter ablation guidance system based on a soft Stewart platform This will be designed to be part of a patient specific guide to aid in ablation in atrial fibrillation. The platform will be able to change the orientation of the guides enabling better ablations to be made. This is hoping to reduce the amount of patients that come back for a second surgery and increase accuracy, preciseness and safety. Moreover the orientation of the Stewart Platform will be changed using water pressure which will mean no electrical actuators will be placed inside the patient making it significantly safer. This will be done using silicon bellows designed to extend a relatively large length for their size. A relationship between the water pressure and elongation of the bellows will be established as well as their maximum extension length. A maximum load for the soft actuator will also be found. In this paper, the design and performance of this Stewart Platform are discussed. Methods Three moulds were used to make the bellows which were designed on Solidworks. The inner mould was filled with wax. This was in the shape of the inner part of the bellows and was then placed in the

151 second mould. Soft silicon was then poured into the second mould around the wax mould to get the outer bellow shape. Once the silicon had solidified the wax inside was then melted out, leaving the bellows hollow. The bellows were then placed in a third mould and hard silicon was poured on the top and bottom, this stopped the ends of the bellows from inflating. A pipe connected to a syringe and an actuator was placed in a hole at the bottom of the bellows to change its shape using water pressure. Six bellows were placed in between a base and a platform. The elongation of these bellows allows for six degrees of freedom. Discussion The soft actuator Stewart Platform will be able to function like a normal Stewart Platform inside the body. This enables surgeons to perform small specific movements without any electrical actuators inside the body which will increase patient safety during surgical procedures. In particular the platform will be able to adjust patient specific guides inside the heart to aid in ablation for atrial fibrillation. By gently pushing the guides into the heart the platform will enable the catheter to carry out better ablations which will reduce the amount of patients needing a second surgery. This will increase patient safety and save money for health institutions such as the NHS. As the actuator is soft the heart will not become bruised whilst it is pumping. Results The platform is expected to operate like a normal Stewart Platform however there may be a slight lack of stiffness when confronted with changing external surroundings such as the pumping of the heart. This is good for this specific situation as there will be no damage to surrounding tissue. However this will be a problem in situations where stiffness is required and therefore platform should not be used when a large load needs to be supported. A mathematical relationship between water pressure and the elongation of the bellows is expected to be found as well as the maximum elongation. A maximum load for the platform will also be established. Conclusions The platform will have many advantages and disadvantages. The main advantage is the size of the platform which could be used in many surgical situations. A surgeon would be able to perform small precise movements in keyhole surgery that they would not be able to perform otherwise. The platform also provides safety to the patient as it is soft and non-electric which maximises patient safety. However, a lack of stiffness would not be ideal in certain medical scenarios and a maximum load will be found to establish the limits of this technology. References 1. Gillinov, AM. January 2007. Advances In Surgical Treatment of Atrial Fibrillation. Stroke, 38(2), pp.618-623. 2. Calkins, H et al. 2012. Expert Consensus Statement on Catheter and Surgical Ablation of Atrial Fibrillation. Heart Rhythm, 9(4).

Adjacent Anatomy-based Registration for Cardiac Resynchronization Therapy Daniel Toth (1,2) Maria Panayiotou (2) Alexander Brost (3) Tanja Kurzendorfer (3) Jonathan M. Behar (2,4) Christopher A. Rinaldi (2,4) Kawal S. Rhode (2) Peter Mountney (5) Siemens Healthineers, Frimley, UK (1) Division of Imaging Sciences and Biomedical Engineering, King's College London, London, UK (2) Siemens Healthineers, Forchheim, Germany (3) Department of Cardiology, Guy's and St. Thomas' Hospitals NHS Foundation Trust, London, UK (4) Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA (5)

Background. Moderate to severe, drug-refractory heart failure is often treated by cardiac resynchronization therapy (CRT). However, the failure rate of CRT interventions is 30-40% [2]. Failure of therapy is associated with inadequate left ventricular (LV) lead placement. Interventional image guidance can be used to optimize LV lead placement. Preoperative MR imaging can be used to identify the ideal position of the LV lead. The preoperative information can be registered to the interventional images manually or by catheters [1,4], but this is highly challenging, since no anatomical landmarks are available. We propose to perform an automatic registration by exploiting the adjacency of anatomical structures, the LV and the coronary veins, see Figure 1.

152

Figure 8. Overview of registration workflow.

Methods. Preoperative short and long axis (SA and LA) cine MR images are automatically segmented to extract a model of the LV [3]. During the CRT intervention, two contrasted X-rays of the coronary veins are acquired. Points of the coronary veins are reconstructed to provide a 3D model of the coronary veins. The reconstructed vessel model is registered to the LV epicardial model segmented from the MR by the globally optimal iterative closest points (Go-ICP) method, to guarantee to find the global optimum of the registration problem [5]. Results. The proposed method was evaluated on simulated data, a phantom dataset and demonstrated on a clinical CRT dataset. First, selected vertices of a segmented LV model were perturbed and reregistered to the LV model. The registration was performed by the ICP and the Go-ICP methods. While ICP failed in 98.9 % of cases, Go-ICP always found the globally optimal alignment. In a second experiment a phantom dataset was created by acquiring a cone beam CT (CBCT) and X-ray images of a 3D printed LV model with metal wires attached to represent coronary veins. The LV was segmented from CBCT and points of the veins were reconstructed from two X-ray images. The LV model was perturbed and reregistered to the vessel model. The mean absolute error was 3.28±1.18 mm. The proposed method was also applied to a clinical CRT case. The resulting overlay showed a good agreement with the shadow of the left ventricle. Discussion. The results show that the proposed approach is capable of registering simulated and phantom data. Performance evaluation on clinical datasets is still outstanding, due to the lack of shared landmarks. Conclusion. This paper proposed a fundamentally different registration method for cardiac MR to X-ray registration for CRT procedures, by exploiting the adjacency of the LV and the coronary veins and applies the Go-ICP method to register the two models. The method was validated on simulated and phantom data and demonstrated on a clinical CRT case. Concepts and information presented are based on research and are not commercially available. Key references. [1] Bourier, F. et al.: Coronary Sinus Extraction for Multimodality Registration to guide Transseptal Puncture. In: 8th Interventional MRI Symposium. pp. 311-313. Leipzig (2010) [2] Claridge, S. et al.: Effects of Epicardial and Endocardial Cardiac Resynchronization Therapy on Coronary Flow: Insights From Wave Intensity Analysis., J. of the AHA, 4(12), pp. 1–12. (2015) [3] Jolly, M.P. et al.: Automatic Segmentation of the Myocardium in Cine MR Images Using Deformable Registration. In: MICCAI-STACOM-I&M Challenges. vol. 8896, pp. 105-113 (2011) [4] Truong, M.V.N. et al.: Preliminary Investigation : 2D-3D Registration of MR and X-ray Cardiac Images Using Catheter Constraints. In: MICCAI – CVII. pp. 1-9. London (2009) [5] Yang, J. et al.: Go-ICP: A Globally Optimal Solution to 3D ICP Point-Set Registration, IEEE transactions on PAMI, PP(99), pp. 1–14. (2015)

Development of an ultrasound phantom for investigating brain tissue pulsations generated by the major cerebral arteries Poppy Turner (1) Jyoti Nath (1) Stefanie Berger (1) Emma Chung (1,2) Georgina de Vries (1,2) James Campbell (1) Imane Bnini (1) Caroline Banahan (1,2) Sara Venturini (1) Evangelia Kaza (1,2) Mark Moehring (3) Asanka Dewaraja (3) Kumar Ramnarine (1,2) University of Leicester, Department of Cardiovascular Sciences, Leicester, UK (1) University Hospitals of Leicester NHS Trust, Medical Physics, Leicester, UK (2) Broadview Laboratories, Seattle, USA (3)

Background Doppler ultrasound can be used to investigate how pulsatile flow through the major cerebral

153 arteries contributes to periodic motion of surrounding brain tissue [1-3]. The aim of this study was to develop an arterial brain tissue pulsation phantom to better understand the impact of the major cerebral arteries on brain tissue motion. Methods To mimic the contribution of major artery pulsation to brain tissue motion, physiologically realistic pulsatile flow was generated within a silicone anatomical replica of the major cerebral arteries using a programmable pump [4]. Resistance tubing was selected based on an electrical circuit analogue approach to achieve a 75:25 split of flow between the anterior and posterior circulations, with a total flow rate of ~430 ml/min and physiological pressure of ~90 mmHg. The vascular replica was surrounded by a polyvinyl-alcohol (PVA) material intended to mimic the elastic and acoustic properties of brain tissue [5]. For comparison with phantom measurements, tissue velocities were also estimated from healthy volunteers with the ultrasound probe held in place using an elasticated headband. Brain tissue velocities were measured in the human and phantom using a Spencer Technologies transcranial Doppler (TCD) ultrasound system equipped with a 2 MHz transducer. Doppler recordings were synchronised to heart rate measured using a 3-lead ECG (Nihon Kohden, Japan). Data were analysed in MatLab to estimate brain tissue displacement over the cardiac cycle; each 8 second recording provides information on tissue displacements in the direction of the beam for 30 overlapping 3 mm Doppler gates spaced 2 mm apart; corresponding to a depth range of 22-80 mm. To compare the path of the ultrasound beam with brain anatomy, volunteers underwent 3T MRI. Results Maximum pulsation amplitudes and displacement waveforms in the phantom were similar to those of healthy volunteers (Forehead-Phantom: 2 to 165 m, Healthy Volunteer: 2 to 183 m; Temporal- Phantom: 3 to 141 m, Healthy Volunteer: 15 to 92 m). Features such as localised changes in pulsation amplitude, phase shifts and asymmetry in pulsation waveforms could be seen in tissue adjacent to vessels. Discussion This arterial brain tissue phantom mimics tissue motion due to vessel pulsation, excluding ventricular flow, tissue perfusion and intracranial pressure. Properties of the tissue mimic were comparable to brain (speed of sound 1540 m.s-1, density 1.06 kg.m-3, Young's Modulus 8 kPa). Further work is underway to improve the phantom to investigate the impact of other parameters such as intracranial pressure (ICP), arterial blood pressure, and cerebral compliance on brain tissue pulsations. Conclusion This study demonstrated the feasibility of developing a suitable phantom for studying arterial contributions to brain tissue motion including features observed in vivo such as localised changes in pulsation amplitude, phase shifts and asymmetry. Key references [1] Kucewicz JC, Dunmire B, Leotta DF, et al. Functional tissue pulsatility imaging of the brain during visual stimulation. Ultrasound Med Biol. 2007 May;33(5):681–90. [2] Kucewicz JC, Dunmire B, Giardino ND, et al. Tissue pulsatility imaging of cerebral vasoreactivity during hyperventilation. Ultrasound Med Biol. 2008 Aug;34(8):1200–8. [3] Wagshul ME, Eide PK, Madsen JR. The pulsating brain: A review of experimental and clinical studies of intracranial pulsatility. Fluids Barriers CNS. 2011 Jan;8(1):5. [4] Ramnarine KV, Anderson T, Hoskins PR. Construction and geometrical stability of physiological flow rate wall-less stenosis phantoms. Ultrasound Med Biol. 2001; 27 (2): 245-250. [5] Cournane S, Cannon L, Browne JE, Fagan AJ. Assessment of the accuracy of an ultrasound elastography liver scanning system using a PVA-cryogel phantom with optimal acoustic and mechanical properties. Phys Med Biol 2010;55:5965–5983.

A Steerable Drill System for Robotic Orthopaedic Surgery Ahmad Nazmi Ahmad Fuad (1) Hariprashanth Elangovan (1) Wei Yao (1) University of Strathclyde, Glasgow, UK (1)

Background: In orthopaedic surgery, Computer-Aided Orthopaedic Surgery (CAOS) has been advancing by using robotic surgical devices and navigation systems [1][2]. However, some areas in

154 surgical operations are still not accessible via rigid surgical tools [3]. Thus, steerable surgical tools for surgery have been widely investigated. Through these steerable tools, surgeons can access problematic zones, such as visualise hidden tissue structures in arthroscopy [4], and avoid lateral collateral ligament (LCL) injury in ACL reconstruction surgery [5]. However, there are not steerable robotic system available that are able to provide sufficient precision and force transformation for bone milling in orthopaedic surgery. Methods: The new robotic system provides an unprecedented and unique capability to perform curved femoral milling under the guidance of a multimodality navigation system. The robotic system consists of three components. Firstly, a flexible drill manipulator in Figure 1 comprises multiple rigid segments. The second part of the robotic system is a hybrid tracking system that consists of an optical tracking system and a position tracking system. Finally, the flexible drill is integrated into a computer-aided navigation system shown in Figure 2. The navigation system provides real time guidance for a surgeon during the procedure. The flexible and steerable drill and its navigation system for femoral milling is evaluated in sawbones.

Figure1.A Steerable Drill.Figure2.NavigationFigure3. Deviation distribution Figure 4. chromatogram Results: It has been found out in the figure 3 that 75.232% of the point cloud data were within ±1 SD and 93.924% of point cloud data were within ±2 SD. This indicates that majority of cloud data from the geometric shape of milled area boundary is within 1.728mm to the pre-planned cut area. Discussion: Although the accuracy of the navigation system is 1.728mm, the difference is small since the defect is on inner side of aim (-1SD) as shown in Figure 4. If the femoral stem implant is a cemented implant, the accuracy of 1.728mm is good since there is space up to nearly 2mm for filling the cement and press fit the implant into the desired position. Conclusion: The new system has demonstrated the ability to perform femoral milling using a unit flexible and steerable drill coupled with a novel tracking and navigation system. Key references. [1] L. Nolte, P. and T. Beutler, “Basic Principles of CAOS”. International Journal of the Care of the Injured, vol. 35, pp. 6-16, 2004. [2] J. Lonner, “Indications for Unicompartmental Knee Arthroplasty and Rationale for Robotic Arm-Assisted Technology,” American Journal of Orthopedics, vol. 38, pp. 3-6, 2009. [3] P. Dario, M. Carrozza, M. Marcacci, S. D’attanasio, M. Bernardo, O. Tonet, and G. Megali, “A Novel Mechatronic Tool for Computer-Assisted Arthroscopy,” IEEE Transactions on Information Technology in Biomedicine, vol. 4, pp. 15-29, 2000. [4] P. Dario, C. Paggetti, N. Troisfontaine, E. Papa, T. Ciucci, M. Carrozza, and M. Marcacci, “A miniature steerable end-effector for application in an integrated system for computer-assisted arthroscopy,” International Conference on Robotics and Automation, Albuquerque, New Mexico. IEEE, 20-25, pp. 1573-1579, April 1997. [5] H. Watanabe, K.. Kanou, Y. Kobayashi, and M. Fujie,”. Development of a "Steerable Drill" for ACL Reconstruction to Create the Arbitrary Trajectory of a Bone Tunnel,” International Conference on Intelligent Robots and Systems, San Francisco, California. IEEE, 25-30, pp. 955- 960. September 2011.

155 A unified dynamic model of movement disorders Nada Yousif (1) Roman Borisyuk (2) Dipankar Nandi (3) Peter Bain (3) University of Hertfordshire, Hatfield, UK (1) University of Plymouth, Plymouth, UK (2) Imperial College London, London, UK (3)

Background. Movement discorders such as Parkinson's disease (PD) and essential tremor (ET) are thought to be caused by abnormal electrical activity of brain cells leading to disabling symptoms such as tremor, an uncontrollable shaking of the limbs, or an inability to move [1]. For example, one popular hypothesis is that essential tremor is caused by synchronous oscillatory activity involving thalamus, cerebellum and the motor cortex [4] and that DBS disturbs pathological synchrony Our recent work has demonstrated that oscillatory activity is supported by the dynamics of the thalamocortical-cerebellar network proposed to be involved in essential tremor [5]. We have also previously shown that the subthalamic–globus pallidus network oscillates at the frequency associated with hypokinetic symptoms of PD [3]. We hypothesized that the dynamics of a single network could support physiological activity and transition to pathological oscillations representing essential tremor, Parkinsonian tremor and bradykinesia and rigidity [2]. Methods. The basal ganglia-thalamocortical circuit was modelled using a population level description, to test the hypothesis that tremor and Parkinsonian symptoms emerge due to pathological synchrony across a network of brain regions. A network model of the relevant brain regions was constructed based on the data from the literature using the Wilson-Cowan mean-field description of neuronal dynamics. Results. Transition between different disorders was controlled by changes in the connectivity, for example in PD the structure of the basal ganglia-thalamocortical network, hypothesized to be abnormally functioning due to the lack of dopamine is relatively well understood. For essential tremor, the thalamocortical cerebellar network is being increasingly identified as the pathological network involved in causing the symptoms of the disorder. Discussion. This study shows that the dynamics of the basal ganglia, thalamocortical cerebellar network support oscillations at both the essential tremor frequency and multiple pathological frequencies associated with Parkinson’s disease. Conclusion. We suggest that our study can be used to better understand the link between the neurophysiological underpinnings of different movement disorders. Key references. [1]. Boraud T, Brown P, Goldberg J, Graybiel A, Magill P: Oscillations in the basal ganglia: the good, the bad, and the unexpected. In The Basal Ganglia VIII. Edited by Bolam J, Ingham C, Magill P.Berlin: Springer; 2005:3-24. [2] Kühn AA, Kupsch A, Schneider GH, Brown P: Reduction in subthalamic 8–35 Hz oscillatory activity correlates with clinical improvement in Parkinson’s disease. Eur J Neurosci 2006, 23(7):1956-1960. [3] Merrison-Hort R, Yousif N, Njap F, Hofmann UG, Burylko O, Borisyuk R. An interactive channel model of the basal ganglia: bifurcation analysis under healthy and parkinsonian conditions. The Journal of Mathematical Neuroscience 3 (1), 14. [4] Raethjen J, Deuschl G. The oscillating central network of Essential tremor. Clin Neurophysiol. 2012; 123(1):61±4. [5] Yousif N, Mace M, Pavese N, Borisyuk R, Nandi D, Bain P. A Network Model of Local Field Potential Activity in Essential Tremor and the Impact of Deep Brain Stimulation. PLOS Computational Biology 13 (1), e1005326

156 3D Full-field Intravascular Flow Reconstruction Using 2D-UIV Measurements and Regularised Radial Basis Function Interpolation Xinhuan Zhou (1) Virginie Papadopoulou (1) Chee Hau Leow (1) Peter Vincent (1) Mengxing Tang (1) Imperial College London, London, UK (1)

Background. Imaging modalities capable of measuring flow in-vivo are of great interest for a wide range of clinical applications. Studies have shown atherosclerotic lesions appear preferentially in regions with particular wall shear stress patterns and velocity fields. Although CFD has been widely used to quantitatively model intravascular blood flow, it depends strongly on the fidelity of the geometry acquired from imaging and the accuracy of initial/boundary conditions. Ultrasound Imaging Velocimetry (UIV), also known as echo-PIV, is a minimally invasive technique, which when combined with High Frame-Rate (HFR) plane wave imaging, is capable of characterising intravascular flow with high accuracy[1]. However the application of UIV and many other velocimetry methods is limited by 2D data acquisition. In this study, a methodology for 3D flow reconstruction from multiple 2D velocity data obtained via plane wave UIV is presented. Methods. A regularised 3D flow interpolation method, constructed through Gaussian Radial Basis Functions (RBF), is proposed to reconstruct a 3D flow velocity field from 2D velocity measurements. This method leverages the fact that the incompressible blood flow velocity field is divergence free to interpolate a 3D velocity from the 2D UIV velocity input at discrete 2D planes. The 3D velocity field reconstruction is an inherently ill-posed problem and is solved by truncated Singular Value Decomposition (SVD). Numerical simulations were performed for comparison, to investigate the accuracy of the technique. Finally, the reconstruction method is tested on both Poiseuille flow and experimental data acquired from a carotid bifurcation model in-vitro. Results. On analytical Poiseuille flow the accuracy of reconstructed 3D flow can be as high as 99%. The reconstructed 3D flow for bifurcation phantom is show in Fig.1.

(a) (b) Fig1. (a) Geometry of phantom. (b) Reconstructed 3D velocity field from UIV Measurements. Discussion. Accuracy of this method can be further improved, and it is feasible to calculate wall shear stress by this method. Conclusion. The method is simple, computationally efficient, mesh-free, and less sensitive to inappropriate geometry/boundary conditions. High accuracy in numerical experiments is achieved with less computational requirement than CFD, and in vitro results demonstrate its applicability to real ultrasound data acquisitions. The methodology could be extended to study in-vivo flow and vessel wall mechanics simulation in future work. Key references. 1.Leow, C.H., et al., Flow velocity mapping using contrast enhanced high-frame-rate plane wave ultrasound and image tracking: Methods and initial in vitro and in vivo evaluation. Ultrasound in medicine & biology, 2015. 41(11): p. 2913-2925.

157

Synthetic Aperture Imaging and Therapy using Random Phased Array Muhammad Zubair (1) R. J. Dickinson (1) Imperial College London, London, Pakistan (1)

Randomized phased arrays have been used for generating and steering single focus and multiple foci with low levels of grating lobes due to the breakage of periodicity of the elements and are considered as useful source of HIFU. However, the reliance of HIFU on MRI for real time visualization of the targeted tissue is a major constraint in its clinical use due to the high cost of MRI and its low temporal resolution. Dual mode ultrasound phased arrays would have the advantage of using the same array for both therapy and imaging due to the inherent registration between imaging and therapeutic frames of reference. Since strong scattering objects in path of HIFU beam are also in path of imaging beam, such objects can be detected in real time and the HIFU beam can be adjusted accordingly. However, the random spherical array would have limited field of view due to the fact that the array is optimized for therapy only and has large, directive elements sparsely positioned on a spherical surface. Nevertheless, images obtained will be useful as they will be aligned with the therapy transducer. In this paper, preliminary results of synthetic aperture imaging with the random phased array are shown. Sub-apertures are being used to image the field of interest, where each sub-aperture contains only those elements which has the pixel in the range of its main lobe. This not only improves the resolution but also significantly reduces the computation time. A MATLAB based pulse-wave simulation model is used to generate Synthetic-Aperture beamforming data for imaging wire-target array. The imaging field of view (IxFOV ) is determined at various depths and lateral resolution of 1.5 mm is achieved near the geometric focus.

Fig: a) Predicted intensity distributions in x-y plane at z = 130 mm for four foci, b) Lateral cross section at depth of 130 mm

Fig: Grayscale image of the wire target array using STA imaging

158 Validation and clinical testing of a wearable knee flexion angle sensor Samantha Micklewright (1,4) presenting Paul Taylor (1,3) Tim Quigg (2) Gareth Jones (2) Duncan Wood (1) Department of Clinical Science and Engineering, Salisbury NHS Foundation Trust, Salisbury, UK (1) University of Bath, Bath, UK (2) Bournemouth University, Bournemouth, UK (3) King's College London, London, UK (4)

Background Wearable sensors can offer a method of providing quantitative evidence-based feedback to aid clinical decision making and contribute to improvements in rehabilitation [4]. Their speed and ease of use, coupled with their ability to take measurements such as activity type and duration, number of falls and muscle activity [2] outside of a laboratory environment, allows us to better understand the activity and participation levels of our patients [1]. A recent ‘Visionary Healthcare’ publication identifies wearable sensors as one of the technologies most likely to impact healthcare by 2025 and describes it as the most established at this time [3]. One example of the use of a wearable sensor in rehabilitation is presented, the study aimed to validate a knee flexion angle sensor and establish the feasibility of evaluating the effect of Functional Electrical Stimulation (FES) on knee motion. Methods The test device was an inertial based sensor mounted above and below the knee to measure knee flexion angle. The first phase of the study involved static and dynamic validation of the sensor using a mechanical rig to replicate the lower limb and testing on volunteers in an optical motion capture laboratory. Following validation, the sensor was trialled on volunteers with unimpaired gait (n=16) to demonstrate the feasibility of using the wearable sensor to measure knee flexion angle during gait. It was then used by established FES users with pathological gait including a dropped foot (n=10) to confirm its feasibility whilst making initial comparisons between walking with and without FES to assess the effect of stimulation on knee flexion angle during swing. Results The results of the static validation found the sensors were precise (SD 0.04°), accurate (RMS error = 0.49°) and reliable (ICC=1). The dynamic validation results were also considered clinically acceptable. The results demonstrated that the use of the sensor was feasible and FES had no significant effect on the knee flexion range achieved during swing in normal or pathological gait. Though data was limited, the FES user results indicated that the use of stimulation may reduce knee flexion range compared to that measured without stimulation. Discussion The study found that it is feasible to use a wearable sensor to measure knee flexion angle in gait and evaluate the effect of FES on knee motion. The level of knee flexion measured with and without the use of FES indicates that the use of stimulation does not have a significant effect on knee flexion and that in some subjects it can reduce knee flexion due to increase dorsiflexion allowing the foot to clear the floor. Further research should be performed in this area using wearable sensor technology outside of the clinical setting. Conclusion This study concludes that wearable sensors can be used in a clinical setting as tool for future evaluation of FES and may lead to a greater understanding of the effect of FES on the motion of the knee and how it may be optimised. It is an appropriate example of how wearable sensors may be used in rehabilitation to guide clinical practise. Key references [1] Aminian, K. and Najafi, B., 2004. Capturing human motion using body-fixed sensors: outdoor measurement and clinical applications. Comp. Anim. Virtual Worlds. 15, pp. 79–94. [2] Cooper, G., Sheret, I., McMillan, L., Siliverdis, K., Sha, N., Hodgins, D., Kenney, L. and Howard, D., 2009. Inertial sensor-based knee flexion/extension angle estimation. J Biomech. 42(16), pp.2678-85. [3] Das, R., 2016. Ten Top Technologies That Will Transform The Healthcare Industry. Forbes [online]. [4] Hodgins, D. and McCarthy, I., 2015. Sensor-based Gait Rehabilitation for Total Hip and Knee Replacement Patients and Those at Risk of Falling. Phys Med Rehabil Int. 2(10), pp. 1073.

159 INDEX OF PRESENTING AUTHORS

A Ghim, Mean ...... 135 Abdullateef, Shima ...... 124 Glatz, Marlies ...... 56 Agnew, Adam ...... 69 Gyring, Peter ...... 137 Agnew, Christina ...... 66, 106 H Al Shaibani, Fahad ...... 125 Harbron, Richard...... 138 Arshad, Mehwish ...... 126 Hardiman, Claire ...... 112 Asgharpour, Zahra ...... 50 Hegarty, Francis...... 118 Ashmore, Jonathan ...... 79, 119, 127 Housden, James ...... 94 Austin, Christopher ...... 86 Hughes, Anna ...... 65 B I Back, Junghwan ...... 128 Ireland, Maighread ...... 62 Barnes, Gareth ...... 72 J Bedford, James ...... 76, 129 Jackson, Steven ...... 37 Bergmann, Jeroen ...... 93 Jamin, Yann ...... 29 Biggans, Thomas ...... 23 Janeczko, Anna ...... 139 Black, Richard ...... 79 Jarritt, Peter ...... 112 Blackmore, Andrew ...... 107 Jassel, Inderbir ...... 22 Blake, Lauren-Emma ...... 19 Julian, Robert ...... 140 Bonnici, Timothy ...... 89 K Brambilla, Marco...... 5 Karakashian, Kevork ...... 9 Bryce-Atkinson, Abigail ...... 83 Käs, Josef ...... A ...... 4 Budyn, Elisa ...... 57 Keevil, Stephen ...... 110 C Khundi, Edwin ...... 141 Carson, Jason ...... 33, 34 Klemt, Christian ...... 46 Chalkidou, Anastasia ...... 100 Koenig, Carola ...... 91 Charlton, Peter ...... 89 Kortesniemi, Mika ...... 7 Charnock, Paul ...... 81 Koutsouveli, Efi ...... 6 Chatterjee, Subhasri ...... 13 L Chiu, Jessica ...... 130 Latif, Nadia...... 20, 142 Choi, Oi-Ching ...... 70 Lewis, Andrew ...... 41, 44 Chughtai, Haroon ...... 64 Lindenroth, Lukas ...... 95 Church, David ...... 131 Liu, Julian ...... 68 Clifton, David A ...... 9 Lynn, Michael ...... 79 Cluny, Laura ...... 132 M Compton, David ...... 112 MacDonald, Conor ...... 40 Constantinou, Loukas ...... 12 Mahmood, Fahd ...... 49, 52 Cooper, Cara ...... 114 Marin Anaya, Virginia ...... 143 Coutts, Glyn ...... 77 Mariscal-Harana, Jorge ...... 35 Curwen, Georgina ...... 61 McLauchlan, Ruth ...... 71 D McLellan, Michael ...... 144 Damilakis, John ...... 5 McMullan, Thomas ...... 58 Daya, Abdul ...... 133 Meese, Helen ...... 103 Dazzi, Marta ...... 134 Micklewright, Samantha ...... 161 de Vries, Georgina ...... 25 Moore, Margaret ...... 108 Di Federico, Erica ...... 47 Moore, Rollo ...... 97, 145 Diallo, Mamadou...... 89 Morris, Max ...... 146 Donoghue, Matthew ...... 81 Munshi, Mayur ...... 147, 148 Dye, Julian ...... 53 Muscat, Carl ...... 48 E N El Haj, Alicia ...... 30 Newman, Ben ...... 115 Esteve, Sergio ...... 67 Nicholls, Rowan ...... 59 G Nordsletten, David ...... 14 Gandecha, Ruchi ...... 39 O Gardner, Stephen ...... 116 Ochoa, Raul ...... 149

160 O'Connor, Stephen ...... 88 Taktak, Azzam ...... 72, 100 Opadele, Abayomi ...... 150 Taylor, Thomas ...... 63 Orozco-Diaz, C. Amnael ...... 45 Tehrani, Marzieh A ...... 55 P Teng, Zhongzhao ...... 24 Panayiotou, Maria ...... 42 Thornett, Carole ...... 113 Peach, Thomas ...... 16 Toth, Daniel ...... 153 Probst, Heidi ...... 86 Totti, Stella ...... 54 Q Tsapaki, Virginia ...... 8 Qassem, Meha ...... 11 Turner, Poppy ...... 154 R V Rashid, Adil ...... 21 Vennin, Samuel ...... 15 Richmond, Ashley ...... 73 Viceconti, Marco ...... 27 Riemer, Kai ...... 90 Vitharana, Kalpani ...... 29 Roberts, David ...... 87 W S Watson, Daniel ...... 18 Sazonov, Igor ...... 32 Weir, Avril ...... 82 Serrani, Marta ...... 14 Wijnen, Stefan ...... 86 Sethi, Sweta ...... 120 Worrall, Mark ...... 84 Sexton, Kate ...... 150 Y Shaban, Lawrence ...... 151 Yao, Wei ...... 155 Shelley, Leila ...... 75 Yousif, Nada ...... 157 Shortall, Jane ...... 98 Z Sireau, Sonya ...... 103, 104 Zar, Areeb ...... 38 Smith, Matthew ...... 117 Zhou, Xinhuan ...... 158 Solomou, Natasa ...... 74 Zhu, Tingting ...... 102 Street, Mark ...... 152 Zubair, Muhammad ...... 159 T Tabakov, Slavik ...... 111

161