Implementation of Silicon Based Dosimeters, the Dose Magnifying

University of Wollongong Research Online

University of Wollongong Thesis Collection University of Wollongong Thesis Collections

2011 Implementation of silicon based dosimeters, the dose magnifying glass and magic plate for the dosimetry of modulated radiation therapy Jeannie Hsiu Ding Wong University of Wollongong

Recommended Citation Wong, Jeannie Hsiu Ding, Implementation of silicon based dosimeters, the dose magnifying glass and magic plate for the dosimetry of modulated radiation therapy, Doctor of Philosophy thesis, Centre for Medical Radiation Physics, Engineering Physics, University of Wollongong, 2011. http://ro.uow.edu.au/theses/3348

Research Online is the open access institutional repository for the University of Wollongong. For further information contact Manager Repository Services: [email protected].

IMPLEMENTATION OF SILICON BASED DOSIMETERS, THE DOSE MAGNIFYING GLASS AND MAGIC PLATE FOR THE DOSIMETRY OF MODULATED RADIATION THERAPY

A Thesis Submitted in Fulﬁlment of the Requirements for the Award of the Degree of

Doctor of Philosophy

from

UNIVERSITY OF WOLLONGONG

Jeannie Hsiu Ding Wong BBiomed. Eng., MMed.Phys.

Centre for Medical Radiation Physics, Engineering Physics Faculty of Engineering

I, Jeannie Hsiu Ding Wong, declare that this thesis, submitted in fulﬁlment of the requirements for the award of Doctor of Philosophy, in the Centre for Medical Radia- tion Physics, Engineering Physics, Faculty of Engineering, University of Wollongong, is wholly my own work unless otherwise referenced or acknowledged. The document has not been submitted for qualiﬁcations at any other academic institution.

(Signature Required) Jeannie Hsiu Ding Wong 30 June 2011 Table of Contents

ListofTables...... vi List of Figures/Illustrations ...... xii

ABSTRACT...... xiii Acknowledgements...... xv Contribution of Collaborators ...... xviii Publication List ...... xx Conferences...... xxi

1 Introduction 1 1.1 Project aim ...... 2

2 Literature review 4 2.1Cancerstatistics...... 4 2.2Radiationtreatmenttrend...... 4 2.3 Intensity Modulated Radiation Therapy ...... 6 2.3.1 StepandshootIMRTdelivery...... 6 2.3.2 Dynamic sliding window IMRT delivery ...... 7 2.4 Volumetric Modulated Arc Therapy ...... 7 2.5StereotacticRadiosurgery/Radiotherapy...... 8 2.6HelicalTomotherapy...... 8 2.7Currentqualityassuranceanddosimetricapproaches...... 9 2.7.1 Ionisationchamber...... 10 2.7.2 Filmdosimetry...... 10 2.7.3 Silicon diode ...... 11 2.7.4 Diamonddetectors...... 14 2.7.5 Geldosimetry...... 14 2.7.6 Electronicportalimagingdevice...... 15 2.8Highspatialresolutiondosimetry...... 16 2.8.1 Concept of silicon strip detector ...... 16 2.8.2 Application of high spatial resolution dosimeters in medical ra- diationtherapy...... 19 2.9 Concept of a two dimensional array detector ...... 20 2.9.1 Pixelateddetector...... 24 2.9.2 MAESTRO project ...... 25

i TABLE OF CONTENTS ii

3 Methodology 26 3.1DoseMagnifyingGlass...... 26 3.1.1 Designandfabrication...... 26 3.1.2 Detectorpackaging...... 28 3.2MagicPlate...... 28 3.2.1 Epitaxialdiodes...... 29 3.3TERAreadoutsystem...... 31 3.3.1 HowTERAworks...... 33 3.3.2 Amplitude and timing ...... 37 3.3.3 Charge collection in silicon strip detector ...... 38 3.4Filmdosimetry...... 39 3.4.1 Radiographic ﬁlm ...... 39 3.4.2 Radiochromicﬁlm...... 40

4 Radiation response and basic characterisation of the Dose Magnifying Glass 43 4.1Introduction...... 43 4.2Materialsandmethods...... 44 4.2.1 Percentdepthdosemeasurement...... 44 4.2.2 Dose per pulse response measurement ...... 45 4.2.3 Stem eﬀect measurement ...... 48 4.2.4 Dose linearity measurement ...... 49 4.2.5 Energyresponsemeasurement...... 49 4.2.6 Angularresponsemeasurement...... 50 4.3Results...... 52 4.3.1 Percentdepthdose...... 52 4.3.2 Dose per pulse response ...... 54 4.3.3 Stem eﬀect ...... 58 4.3.4 Dose linearity ...... 58 4.3.5 Energyresponse...... 59 4.3.6 Angularresponse...... 61 4.4Conclusion...... 63

5 Application of the Dose Magnifying Glass in the dosimetric veriﬁca- tion of an intensity modulated radiation therapy treatment delivery 65 5.1Introduction...... 65 5.2Materialsandmethods...... 66 5.2.1 DoseMagnifyingGlass...... 66 5.2.2 Uniformitymeasurement...... 67 5.2.3 Penumbraresponsemeasurement...... 67 5.2.4 Clinical application in IMRT ﬁelds ...... 68 5.3Resultsanddiscussions...... 69 5.3.1 Uniformity...... 69 5.3.2 Penumbrameasurement...... 69 TABLE OF CONTENTS iii

5.3.3 Clinical application in IMRT ﬁelds ...... 70 5.4Conclusion...... 76

6 Application of the Dose Magnifying Glass in the dosimetric verification of a stereotactic radiosurgery treatment delivery 78 6.1Introduction...... 78 6.2Materialsandmethods...... 82 6.2.1 DoseMagnifyingGlass...... 82 6.2.2 SRSphantom...... 82 6.2.3 Detectorrelativesensitivityfactormeasurement...... 83 6.2.4 Angulardependencecorrection...... 85 6.2.5 Determination of the center of rotation measurement ...... 86 6.2.6 SRSconeprofilesandtotalscatterfactormeasurement..... 87 6.2.7 Clinical stereotactic arc measurement ...... 88 6.3Resultsanddiscussions...... 89 6.3.1 Uniformity...... 89 6.3.2 Determination of center of rotation ...... 89 6.3.3 SRSconeprofilesandpenumbra...... 91 6.3.4 SRSconetotalscatterfactor...... 91 6.3.5 Clinical SRS application ...... 93 6.3.6 On the volume averaging effect of small field dosimetry ..... 94 6.4Conclusion...... 102

7 Application of the Dose Magnifying Glass in the quality assurance of Helical Tomotherapy 105 7.1Introduction...... 105 7.2Methodsandmaterials...... 107 7.2.1 DoseMagnifyingGlass...... 107 7.2.2 Multileaf collimator (MLC) alignment measurement ...... 107 7.2.3 Leafopentimemeasurement...... 109 7.2.4 Leafﬂuenceoutputfactormeasurement...... 111 7.3Resultsanddiscussions...... 113 7.3.1 Multileaf collimator alignment ...... 113 7.3.2 Leafopentimethreshold...... 113 7.3.3 Leafﬂuenceoutputfactor...... 118 7.4Conclusion...... 119

8 Radiation response and basic characterisation of the Magic Plate 121 8.1Introduction...... 121 8.2Materialsandmethods...... 122 8.2.1 PackagingoftheMagicPlate...... 122 8.2.2 Radiation damage studies ...... 123 8.2.3 Dose per pulse response measurement ...... 126 8.2.4 Percentdepthdosemeasurement...... 129 TABLE OF CONTENTS iv

8.2.5 Dose linearity measurement ...... 129 8.2.6 Energydependencemeasurement...... 130 8.2.7 Temperaturedependencemeasurement...... 130 8.2.8 Fieldsizedependencemeasurement...... 132 8.2.9 Angularresponsemeasurement...... 133 8.2.10Beamperturbationmeasurements...... 135 8.3Results...... 137 8.3.1 Radiationdamage...... 137 8.3.2 Dose per pulse response ...... 139 8.3.3 Percentdepthdose...... 146 8.3.4 Dose linearity ...... 146 8.3.5 Energydependence...... 148 8.3.6 Temperaturedependence...... 150 8.3.7 Fieldsizedependence...... 152 8.3.8 Angulardependence...... 153 8.3.9 Beamperturbationstudy...... 154 8.4Conclusion...... 156

9 Investigation of the Magic Plate in clinical application 158 9.1Introduction...... 158 9.2Materialandmethods...... 161 9.2.1 Uniformity and absolute dose calibration ...... 161 9.2.2 Dosemeasurementinsolidwaterphantom...... 164 9.2.3 Fluencemeasurementsintransmissionmode...... 165 9.2.4 Performance index used for the dose distribution comparison . . 166 9.3Resultsanddiscussion...... 166 9.3.1 Uniformitycorrection...... 166 9.3.2 Dose distribution comparison ...... 167 9.4Conclusion...... 174

10 Conclusion 176 10.1DoseMagnifyingGlass...... 176 10.2MagicPlate...... 179

Appendices 183

A SolidWorks drawing 183 A.1 The 2nd generationDoseMagnifyingGlassdetectorholder...... 183 A.2SRSphantom...... 183 A.3MagicPlateY-shapeperspexframe...... 186 A.4MagicPlateholdertobeusedwiththeI’mRTphantom...... 186 TABLE OF CONTENTS v

B Matlab Scripts 189 B.1Uniformitycorrectionscript...... 189 B.2Gammaanalysisscript...... 192

References 219 List of Tables

2.1Speciﬁcationofcommercial2Darrays...... 22

3.1 Set reference values for the CMRP TERA 03 board...... 35

5.1 Comparison of the 80-20% penumbra width measurements for a 6 MV beam at 1.5 cm and 10.0 cm depth between the DMG, Gafchromic EBT ﬁlms, and other published literature. The uncertainty reported represent the 95% CI of the mean for three sets of measurements. . . . 71

6.1 The penumbra and FWHM measurements of the dose profiles of 5 mm to 20 mm cone diameter at SDD 100 cm, comparing the DMG and EBT2measurements...... 92 6.2 Dose calculated based on the analytical model showing the effect of dose averaging effect with different detector sizes...... 98

7.1 Programmed leaf opening conﬁguration for LFOF measurement for LOI = 33. The same delivery sinogram was also used with leaf 47 and 62 as the LOI. The symbol “|”denotesanopenleaf...... 112 7.2 LFOF measurements by DMG and tomo detectors. The uncertainties represent 1 s.d. of the mean...... 119

8.1 Measurement setup for the dose per pulse measurement...... 126 8.2Specificationsofthedosimetersused...... 128 8.3Energydependencemeasurementconfigurations...... 131 8.4 Dose per pulse dependence sections...... 141 8.5 Definition of symbols...... 142 8.6 Difference between the surface doses measured with open fields and a MP fields in percent. The values in the parentheses are the measured surfacedosenormalisedtothemaximumdose...... 155

9.1 MP positioning for the uniformity calibration of the diodes...... 163 9.2 Gamma analysis (3% DD,3 mm DTA) of IMRT dose distributions com- paringTPScalculations,MPandEBT2measurements...... 172

vi List of Figures

2.1 Silicon/water stopping power ratio...... 13 2.2 Schematic of a silicon p-n junction diode (Shi et al., 2003)...... 13 2.3 Operation principle of a silicon strip detector [modiﬁed from (Damerell, 1995)]...... 18

3.1SchematicdiagramoftheDoseMagnifyingGlass(DMG)...... 27 3.2 Prototype version of DMG (left) and the 2nd generation DMG (right). . 28 3.3TheMagicPlate(MP)2Darraydetector...... 29 3.4 The Magic Plate mounted on the linac gantry. The mounting plate was a modified TBI applicator...... 30 3.5Schematicdiagramofanepitaxialdiode(withguardring)...... 32 3.6Schematicdiagramofanepitaxialdiode(withoutguardring)...... 32 3.7 Schematic diagram of the TERA readout [modified from (Mazza et al., 2005)]...... 33 3.8 Charge subtraction waveform (Mazza et al., 2005)...... 35 3.9 Physical structure of Gafchromic (a) EBT and (b) EBT2 film (ISP, 2006, 2009)...... 42

4.1SketchofthesolidwaterencapsulationfortheprototypeDMG..... 45 4.2 The 2nd generation DMG in the solid water holder and the custom made phantom which has a cylindrical body and hemispherical head...... 53 4.3 Definition of polar and azimuthal angle...... 53 4.4 Depth dose curve for a 10 × 10 cm2 field size of a 6 MV photon beam comparing DMG with Farmer ion chamber (NE-2571) measurements in solidwater...... 54 4.5 Dose per pulse response of selected DMGs with different resistivity and preirradiation conditions for the dose per pulse range of 9.45 × 10−5 to 2.72 × 10−4 Gy/pulse. The error bars represent the 1 standard deviationofthemeanofallthechannelsinthestripdetector...... 55 4.6 Dose per pulse response for a low resistivity, preirradiated device for the dose per pulse range of 6.42 × 10−7 to 2.92 × 10−4 Gy/pulse. The dose per pulse responses were normalised to the dose per pulse of 2.2 × 10−4 Gy/pulse...... 57 4.7 Dose per pulse corrected response...... 58

vii LIST OF FIGURES viii

4.8 Dose linearity of the DMG...... 59 4.9 Energy response of the strip detector, normalized to 1 at a 6 MV photon energy. Error bars represent the combined uncertainties of the 95% CI of the mean for three sets of measurements and the 0.5% calibration uncertainty...... 60 4.10 Angular response correction factor generated as the measured dose by DMG/CC13 ion chamber measurements. As the angle of the incident beam increased from perpendicular beam (θ =0◦)toparallelbeam(θ = 90◦), the strip detector under-responded compared to the dose measured by CC13 ion chamber. The maximum angular response 28.1% ± 0.1% was found to be at the parallel incident beam at the gantry angle of 90◦. The diﬀerence between the response of the detector channels at the position of ± 1 cm away from the central axis (mid channels) were <1%...... 62 4.11 Angular response for the 2nd generationDMG...... 63

5.1 Detector response before and after uniformity correction. The error bars represent the 95% CI of the mean for three sets of measurements. 70 5.2 Dose profiles showing the penumbra of a 6 MV beam at 1.5 cm depth for the secondary X-jaw and the rounded leaf ends of a multileaf collimator (MLC) using the strip detector and Gafchromic EBT films. The error bars for the EBT profiles represent the 95% CI of the mean of three sets of measurements while the error bars for the DMG measurements represents the 2% reproducibility uncertainty...... 71 5.3 IMRT dose profiles for each gantry angle and the total summation of all the beams. The error bars represent the reproducibility uncertainty ofthemeasurementsatthelevelof95%CIofthemean...... 72 5.4 Comparison between Pinnacle predicted dose profiles, measurements with Gafchromic EBT film, and DMG. The error bars represent the reproducibility of the measurements which are in the order of ± 2% for DMG and 3% for EBT film, respectively. On average, the difference between the dose measured using the DMG with the dose points from the Pinnacle dose matrix and the dose measured with the EBT film were 1.1%± 1.8% (1 s.d.) and 1.0%± 1.6% (1 s.d.)...... 73 5.5 (a) 3D dose profiles of a single step and shoot IMRT field, depicting the modulation of the dose during each segment, within the width of 2.56 cm. (b) Dose profiles of a full IMRT delivery, each gantry angle and modulation in each IMRT segments can be shown temporally. The measurements were performed with the acquisition pulse width of 0.1 s. (c) The temporal cumulative 3D dose profiles of a full IMRT delivery. (d) Temporal dose rate pattern of an IMRT delivery for channel no. 8 and channel no. 127...... 74 LIST OF FIGURES ix

6.1 SRS phantom mounted on the Radionics SRS couch mount. A plate with flat top and matching concave bottom was used to achieve a flat calibration phantom. Solid water spacers were used to allow the lateral shiftingoftheDMGwithinthephantom...... 83 6.2 Detector response of the DMG before (open circle) and after uniformity correction(closedcircle)...... 89 6.3 Two profiles of a 5 mm diameter cone measured with the DMG showing discrepancy in the center of rotation due to couch rotation from 90◦ - 270◦...... 91 6.4 SRS cones total scatter factor comparing DMG with EBT2, CC04 measurements and Monte Carlo calculation. The measurements were taken within the spatial distance of ± 0.2 mm from the center axis. The CC04 ion chamber was used to measured the Scp for the cones with 15 mm - 40 mm. The Monte Carlo calculation was made with voxel size of 1 × 1 × 2mm3 for cone diameter ≤ 10 mm while voxel size of 2 × 2 × 2 mm3 was used to calculate for the cone diameter >10cm...... 93 6.5 SRS arc (0◦ - 180◦)relativeintensityprofile...... 94 6.6 (a) Pseudo colored image from EBT2 film measurement of a 5 mm diameter beam, normalized to the center of the beam, (b) isodose lines for the 5 mm diameter beam, (c) signal profile alone a horizontal line drawn across the middle of the beam and (d) zoom in to the top beam profile, data points >0.3 were extracted and curve fitted with a 4th degree polynomial function (r2 = 0.998)...... 95 6.7 Schematic drawing of detector area superimposed on the radial beam contour...... 96 6.8 Schematic diagram representing the top view of the beam and the denotation of the r- and the τ-directions. The isodose lines are represented astheconcentriccircles...... 99 1 6.9 Dose gradient in (a) r- and (b) τ-direction for a detector with 2 longitudinal length of b = 0.5 mm and 1 mm. The x-axis is the radial direction extending from the center of the beam (r = 0 mm) to the edge of the beam (r = - 6 mm). The negative gradient at the region r = -0.5 mm to 0 mm is due to the noise in the EBT2 film picked up by the fitted curve...... 101 6.10 Dose map of (a) 2 × 10 mm2 and (b) 1 × 10 mm2...... 102 6.11 Numerical integration of dose for detector of 0.2 × 2mm2 and 1 × 1 mm2 area...... 103 LIST OF FIGURES x

7.1 (a) EDR2 film measured leaf alignment test. The distances from the two outside peaks to the central peak were X1 = 24.94 ± 0.09 mm and X2 = 23.53 ± 0.09 mm. The MLC alignment error was taken as half the difference between these two distances (dX = 0.71 ± 0.09 mm) and (b) DMG measured leaf alignment test. Half the distance between the centre of each profile gives the MLC alignment error (dX = 0.55 ± 0.10 mm)...... 114 7.2 Actual measured leaf open times plotted against programmed leaf open times for (a) 50 ms, (b) 100 ms, (c) 200 ms and (d) 303 ms projection time. A linear curve has been fitted for guidance. The error bars represent 1 s.d. of the mean of the five projections within the same modulation bracket...... 115 7.3 Comparison of the DMG and MVCT measured leaf open time for leaf 32 (a) 50 ms, (b) 100 ms, (c) 200 ms and (d) 303 ms projection time. A linear curve has been fitted for guidance...... 116 7.4 DMG measurement of the leaf opening and closing for a 200 ms projection time. Each row of the ribbon represents a 3 ms acquisition. Red arrowspointtothedirectionofleaftravel...... 118

8.1 MP packaging: (a) MP mounted on the Y-shaped Perspex frame. The readout electronics box is connected to the MP on the left. (b) MP sandwiched between two 5 mm solid water slabs for measurements in theI’mRTphantom...... 123 8.2 Schematic diagram of the MP packaging for the use in (a) transmission modeand(b)inanI’mRTphantom...... 124 8.3 Modified accessory tray to position the Pb attenuator closed to linac gantry...... 127 8.4 Schematic drawing of the ‘face-up’ and ‘face-down’ configurations for theMPdiode...... 130 8.5Experimentsetupforthetemperaturedependencemeasurement.....132 8.6 Sensitivity of the detector as a function of the accumulated radiation dose. The error bars represent the ± 1 standard deviation of the channels located within the central 20-80% region of the 11 × 11 MP detector. The left axis shows the detector sensitivity expressed as counts/1Gy while the right axis shows the detector sensitivity normalised to the first irradiationof1Gy...... 138 8.7 Reproducibility of the MP. The error bars represent ± 1 standard de- viationofthemeanofthreemeasurements...... 138 8.8 Dose per pulse response for the MP and the commercial diode, normalised to the dose per pulse of 2.78 × 10−4 Gy/pulse...... 140 8.9 Division of the dose per pulse measurements into four (A to D) sections. 140 8.10 Minority carrier lifetime as a function of ionising radiation dose rate 16 −3 in a highly doped n-type silicon (NA =10 cm ) [reproduced from (Alexander, 2003)]...... 143 LIST OF FIGURES xi

8.11 Depth dose curve for a 10 × 10 cm2 field size of a 6 MV photon energy measuredwithaCC13ionchamberandtheMP...... 147 8.12 Build up region of the depth dose curve, comparing the MP, CC13 and Attix chamber measurements (courtesy of Dr. Bradley Oborn, ICCC andCMRP)...... 147 8.13 The dose linearity of the MP...... 148 8.14 The energy dependence of the MP, measured with four different setups; comparing ‘face-up’/‘face-down’ configurations. The measurements were made in free air geometry and with a solid water phantom. 150 8.15 The energy response of the MP in ‘face-up’ and ‘face-down’ configurations for 6 and 10 MV photon energies. The detector response was normalised to 1 at the reference setup of SSD 100 cm and dmax of 1.5 cm and 2.0 cm for the 6 and 10 MV photon energies, respectively. . . . 151 8.16 Sensitivity variation with temperature (svwt) of the MP. The detector response was normalized to 1 at 23.32◦C. Error bars represent 1 s.d. of themeanofthreemeasurements...... 151 8.17 The field size dependence of MP at SDD of 101.5 and 58 cm compared with the standard data measured using the Farmer chamber measured at SDD of 101.5 cm. The data was normalised to 1 at the field size of 10 × 10 cm2...... 153 8.18 Mean angular response of the MP for the detectors located at the central column (0 cm) and at ± 5cmoffaxisdistance...... 155 8.19 Dose at the build up region for a 20 × 20 cm2 and 30 × 302 cm at (a) SSD of 90 cm and (b) SSD of 80 cm. The measurement uncertainty was ± 0.5%, representing 1 standard deviation of the mean of three measurements...... 156

9.1 Sketch of the MP coordinate system. The diode at coordinate [0,0] is thecentrediode...... 162 9.2 Mean cross plane and in plane profiles of the MP diodes before and after the uniformity correction. The error bars represent the ± 1 standard deviation of the mean of 11 detectors located within the same row or column...... 167 9.3 Dose measurement for a single IMRT field delivered with gantry angle set to 0◦. (I.a) shows the normalised cross plane dose profile at an off axis distance of 0 mm. (I.b to d) shows the three planar dose distributions of the MP, EBT2 and the Pinnacle TPS. Gamma analysis of the (I.e) TPS versus MP (pass rate = 97.5%), and (I.f) TPS versus EBT2 (pass rate = 92.4%) dose distributions. (II) shows the horizontal and (III)verticaldoseprofilesforthethreedatasets...... 169 LIST OF FIGURES xii

9.4 Dose measurement for a composite IMRT plan delivered with the actual treatment gantry angles. (I.a) shows the normalised cross plane dose profile at an off axis distance of 0 mm. (I.b to d) shows the three planar dose distributions of the MP, EBT2 and the Pinnacle TPS. The MP measurements were corrected for angular dependence. Gamma analysis of the (I.e) TPS versus MP (pass rate = 80.2%), and (I.f) TPS versus EBT2 (pass rate = 82%) dose distributions. (II) shows the horizontal and(III)verticaldoseprofilesforthethreedatasets...... 171

A.1 The 2nd generationDoseMagnifyingGlassdetectorholder...... 184 A.2SRSphantom...... 185 A.3MagicPlateperspexframe...... 187 A.4SolidwaterplatesusedwithanI’mRTphantom...... 188

B.1 Matlab GUI for processing and comparing the three data sets, (i) Magic Plate, (ii) EBT2 ﬁlm and (iii) Pinnacle predicted and (iv) Monte Carlo simulated dose distributions...... 190 Implementation of silicon based dosimeters, the Dose

Magnifying Glass and Magic Plate for the dosimetry of

modulated radiation therapy

Jeannie Hsiu Ding Wong

A Thesis for Doctor of Philosophy

Centre for Medical Radiation Physics, Engineering Physics University of Wollongong

ABSTRACT

New cutting edge radiation therapy techniques such as Intensity Modulated Radiation Therapy (IMRT), Stereotactic Radiosurgery (SRS), Helical TomoTherapy and most recently Volumetric Modulated Arc Therapy (VMAT) produce radiation dose maps with high dose modulation and tight gradients between the high and low dose region. Difficulties in the dosimetric verification of these new complex treatment methods using existing dosimeters have led to the need for a new generation of fast responding real time dosimeters with submillimetre precision. This thesis describes two detector systems based on silicon substrates that were developed to address this need. The first detector system was a silicon strip detector called the “Dose Magnifying Glass” (DMG). It consisted of 128 detectors spaced 0.2 mm apart. It was coupled to a TERA ASIC chip that enabled simultaneous readout of multiple channels at high temporal resolution. The first part of the thesis involved investigation of the basic characteristics of the DMG followed by its application in the dosimetric verification of IMRT, SRS and Helical TomoTherapy treatment deliveries. The high spatial resolution of this device was ideal for the measurement of high dose gradients in IMRT and small fields encountered in SRS. When compared with film dosimetry, DMG measurements showed agreements within 3% for a SRS treatment plan. The DMG was also successfully employed as an independent quality assurance tool for the verification of helical Tomotherapy machine binary MLC leaf parameters. The second detector was a two dimensional array detector so named the “Magic Plate” (MP). The diode was based on epitaxial technology and has a very thin sensitive volume of 50 μm. The MP comprised of 11 × 11 epitaxial diodes mounted on a 0.6 mm thick Kapton substrate. This detector was designed to be used either as a transmission detector or to measure dose distributions in a solid water phantom. Preliminary testing of the MP in a clinical IMRT treatment delivery was carried out. The MP measurements demonstrated good agreement (>80%) with conventional EBT2 film dosimetry and with treatment planning system predicted dose distributions using 3%/3mm gamma criteria.

KEYWORDS: IMRT, small ﬁeld dosimetry, Dose Magnifying Glass, Magic Plate, Tomotherapy, high spatial resolution silicon detector, transmission detector, quality assurance Acknowledgements

On the 31st December 2007, I boarded a flight to Australia, embarked on the journey of a PhD study. For the next 3.5 years, I have had the opportunity to learn so much and met so many wonderful people. It was a wonderful and enriching 3.5 years, indeed. First of all, I would like to thank my three supervisors; Professor Anatoly Rosen- feld, Professor Peter Metcalfe and Dr. Martin Carolan, for without their guidance and advice, this thesis would not be possible. Professor Anatoly Rosenfeld is a wonderful mentor and supervisor. Despite having many other students, he still find time to pull me out of my research portholes. His vast knowledge and experience in physics and solid state dosimetry is inspirational and am humbled by his patience to teach and explain to me the basic every time I ask a question. I also thank him for giving me the opportunities to extend my professional horizon in international research collab- orations and the financial support that goes with it. I would like to thank Professor Peter Metcalfe for his advice and help with my thesis writing. Professor Peter Met- calfe and Dr. Martin Carolan, with their extensive experience in the clinical medical physics provided sound advice to me. I would like to thank Dr. Martin Carolan for the many late nights that he stayed back at the Illawarra Cancer Care Centre (ICCC), supervising and assisting me with the measurements. His generosity and kindness in giving his time and energy towards nurturing young researcher is humbling. It was a priviledge to have worked and learned under these three great people. I would like to thank Dr. Michael Lerch and Dr. Marco Petesecca for their kind

xv assistance and advice in the electronic readout system and explanation of the basic physics of solid state detectors. I thank them, as well as Dean, Sami, Iolanda and Paul for giving me lifts with the equipment to the ICCC during those measurements days. Special thanks to Iolanda and Paul who had worked with me in some of my measurements. In my work, I often have to make special phantoms or holders for the detectors. I was so glad the UoW workshop was there to help me out with this. Thank you to Ron Marshall, Martin, Stuart, Keith, Andrew and Doug for their help in making all the big and small devices that I needed. I would also like to thank Peter Ihnat for the technical support. Special thanks to Terry Braddock for his friendship and ever ready help whenever I come and look for him. I would also like to thank Karen Ford for the administrative assistance and for her amazing organizing skills. I would like to thank Dr. Martin Butson and Dr. Matthew Williams for the useful advice in the use of the EBT and EBT2 ﬁlms. Thank you to Jo and Ab for helping out and staying with us during some of our measurement evenings. I would like to thank Dr. Tony Knittel for the experience working with him in the SRS project. I thank him for his generosity with his time, spending the many weekends working with me at the Prince of Wales hospital, Randwick, even until the late hours. His vast knowledge, experience and hands on approach taught me a lot of things, not only in medical physics, but in conducting research. I will always remember the advice given to me “Change only ONE variable at a time!” To the physicist at POWH: thank you to Soo Ming for bunking me for a night, Dean Inwood for helping with one of the measurement and Carl Chan for the Monte Carlo calculation of the

SRS cone Scp values. During my short trip to the University of Wisconsin-Madison, WI, I had the opportunity to work with Dr. Nick Hardcastle, Professor Wolfgang Tom´e, Ranjini To-

xvi lakanahalli and Dr. Adam Bayliss. I thank all of them for the help and advice given to me, leading to the success of this project. I also thank Dr. Peter Hoban for the tour of TomoTherapy Inc. and the TomoTherapy engineers for the feedback on the MLC leaf open time. To my fellow office mates: Thank you Amir for teaching me Geant4, although I do hope one day I can put it to use. Sianne, Norlaili, Elise, Amir, Ashley and Amy, thanks for the friendship, cakes and banter. To my family: Dad, mom, Ah Ma, Mama, Evelyn, James and Linda, thank you for your encouragement, support, prayers and for just being my family. To my fiancé, Kong Sih Ying, I am ever so glad that I found you. Thank you for your support and for being there for me. I was never alone because I had you. To my friends outside the CMRP, my housemates, the Verbum Dei community, my church choir members, it was a wonderful 3.5 years in Wollongong and you guys have made my life so much richer and memorable. Special thanks to Chris for proof reading part of my thesis.

xvii Contribution of Collaborators

Professor Anatoly Rosenfeld is the inventor of the Dose Magnifying Glass (DMG) and the Magic Plate (MP). He also advised on the experiments design and results discussion. Professor Peter Metcalfe provided advice on the clinical aspects of the experiments and writing of the thesis. Dr. Martin Carolan provided supervision and assistance with the measurements conducted at the Illawarra Cancer Care Centre (ICCC), Wollongong, advised on the experiment designs and the writing of the thesis. Dr. Bradley Oborn provided the build up region depth dose data for a 10 × 10 cm2 ﬁeld of a 6 MV photon beam. Dr. Michael Lerch and Dr. Marco Petasecca provided technical advice and assistance on the electronic readout system. Mr. Sam Khanna wrote the initial Labview software that was used for the data acquisition. Dr. V. Perevertaylo manufactured the DMG and MP detectors. Dr. Dean Cutajar provided advice and assistance on the use of DMG in brachytherapy. Miss Iolanda Fuduli assisted in the electronic readout and the measurements of the Magic Plate work. Mr. Paul Geenty assisted with the Magic Plate measurements. Professor Wolfgang Tom´e, Dr. Nicholas Hardcastle, Dr. Adam Bayliss and Ms Ranjini Tolakanahalli provided assistance with experimental design and data analysis in the Tomotherapy project. Dr. Tony Knittel provided advice and technical assistance with the design and machining of the SRS phantom, assisted in the measurement, and result discussion

xviii in the stereotactic radiosurgery project. Dr. Simon Downes and Dr. Michael Jack- son facilitated the experimental sessions at the Prince of Wales Hospital, Randwick, Sydney.

xix Publications

Wong,J.H.D., Carolan, M., Lerch, M. L. F., Petasecca, M., Khanna, S., Perever- taylo, V. L., Metcalfe, P. and Rosenfeld, A. B. (2010). A silicon strip detector dose magnifying glass for IMRT dosimetry. Medical Physics 37(2): 427-439.

Wong,J.H.D., Hardcastle, N., Tome, W. A., Bayliss, A., Tolakanahalli, R., Lerch, M. L. F., Petasecca, M., Carolan, M., Metcalfe, P. and Rosenfeld, A. B. (2011). Inde- pendent quality assurance of a helical tomotherapy machine using the dose magnifying glass. Medical Physics 38(4): 2256-2264.

Wong,J.H.D., Knittel, T., Downes, S., Carolan, M., Lerch, M. L. F., Petasecca, M., Perevertaylo, V. L., Metcalfe, P., Jackson, M. and Rosenfeld, A. B. (2011). The use of a silicon strip detector dose magnifying glass in stereotactic radiotherapy QA and dosimetry. Medical Physics 38(3): 1226-1238.

Wong,J.H.D., Cutajar, D., Lerch, M. L. F., Petasecca, M., Knittel, T., Carolan, M., Perevertaylo, V. L., Metcalfe, P. and Rosenfeld, A. B. (2011). From HEP to Med- ical Radiation Dosimetry - the silicon strip detector Dose Magnifying Glass. Radiation Measurements. DOI:10.1016/j.radmeas.2011.06.031(Accepted, 14 June 2011).

xx Conferences

Wong,J.H.D., Carolan, M., Lerch, M. L. F., Petasecca, M., Khanna, S., Perever- taylo, V. L., Metcalfe, P. and Rosenfeld, A. B. (2010). A Silicon Strip Detector Dose Magnifying Glass for IMRT Dosimetry, Abstracts in: EPSM-ABEC 2009: Engineering and Physical Sciences in Medicine & the Australian Biomedical Engineering College Conference, 8-12 November 2009, Hotel Realm, Canberra, ACT.” Australas Phys Eng Sci Med. 33(1): 68.

Wong,J.H.D., Carolan, M., Lerch, M. L. F., Petasecca, M., Khanna, S., Perever- taylo, V. L., Metcalfe, P. and Rosenfeld, A. B. (2009). A Silicon Strip Detector Dose Magnifying Glass For IMRT Dosimetry. Poster presented at the AIP Physics in In- dustry Day 2009. CSIRO Materials Science and Engineering, NSW, Australia.

Wong,J.H.D., Cutajar, D., Lerch, M. L. F., Petasecca, M., Knittel, T., Carolan, M., Perevertaylo, V. L., Metcalfe, P. and Rosenfeld, A. B. (2010)Silicon strip detector dose magnifying glass on stereotactic radiotherapy QA and dosimetry. Presented at the Solid State Dosimetry 16th International Conference, Sydney, 19-24 September 2010.

Wong,J.H.D., Hardcastle, N., Tome, W. A., Bayliss, A., Tolakanahalli and Rosen- feld, A. B. (2010). Independent quality assurance of a helical tomotherapy machine using the dose magnifying glass. Presented at the Engineering and Physical Sciences in Medicine and the Australian Biomedical Engineering Conference (EPSM-ABEC) 2010, Melbourne, Australia, 5 - 9 December 2010.

Wong,J.H.D., P Geenty, M Carolan, M L F Lerch, M Petasecca, V Perevertaylo, P Metcalfe, and A B Rosenfeld (2010), A novel 2D array silicon detector ‘Magic Plate” for IMRT dose veriﬁcation. Poster presented at the Engineering and Physical Sciences in Medicine and the Australian Biomedical Engineering Conference (EPSM- ABEC) 2010, Melbourne, Australia, 5 - 9 December 2010.

Wong,J.H.D., M Carolan, I Fuduli, M L F Lerch, M Petasecca, P Metcalfe, and A B Rosenfeld (2010), Development of a 2D array silicon detector Magic Plate for the dosimetric veriﬁcation of IMRT treatment delivery. Abstract accepted for the Engi- neering and Physical Sciences in Medicine and the Australian Biomedical Engineering

xxi Conference (EPSM-ABEC) 2011, Darwin, Australia, 14 - 18 August 2011.

xxii 1 Chapter 1

2 Introduction

3 The advancement of the treatment of cancer patients with radiation therapy is driven

4 by the intention to achieve higher tumour control and fewer side eﬀects to critical

5 organs. Historical studies show a higher dose enhances tumour control and a lesser

6 dose reduces normal tissue complication (Pollack et al., 2000). In order to provide a

7 higher targeted dose to tumour and a lesser dose to critical organs, modern radiation

8 therapy delivery techniques are continually evolving and becoming more complex.

9 Delivery techniques such as intensity modulated radiation therapy (IMRT) challenge

10 the conventional mindset of boxed 3D conformal radiation therapy, while the issues

11 pertaining to the veriﬁcation of small ﬁeld dosimetry are still not completely resolved.

12 Conventional quality assurance (QA) approaches to dosimetry are used with vary-

13 ing degrees of success. Their adequacy in addressing the new dosimetric challenges

14 presented by modulated and small ﬁeld radiation therapy delivery techniques in par-

15 ticular, lend themselves to development of new detector systems that complement the

16 existing detectors. In particular it is hoped these new detector systems will provide

17 higher spatial and temporal resolution than the existing dosimeters. This has led to the

18 development of new purpose-speciﬁc dosimeters to address the various requirements

19 of dosimetric veriﬁcation.

1 1.1. Project aim 2

20 The technological advancement of the medical radiation physics tools has been

21 complemented somewhat by the rapid advancement in semiconductor and computer

22 technology. This enabled the development of device control systems such as the dy-

23 namically driven multi leaf collimators (MLCs). The same technology evolution has

24 also helped push the frontiers of semiconductor detectors.

25 1.1 Project aim

26 This thesis describes the development and characterisation of two new detector systems

27 based on silicon substrates. Chapter 3 describes the design and fabrication of these

28 two systems.

29 The thesis then describes the use of a silicon strip detector as a radiation dosime-

30 ter. The device is referred to in this thesis as the “Dose Magnifying Glass” (DMG).

31 This device consists of a linear array of silicon strips with high spatial resolution. It

32 was coupled to an electronic readout system capable of high temporal resolution ac-

33 quisition. This device has been successfully employed and characterised under simple

34 open medical linear accelerator x-ray ﬁelds (chapter 4). The device was then used to

35 characterise IMRT beams (chapter 5). It was subsequently used to characterise the

36 spatial and temporal properties of extremely small treatment ﬁelds as used in stereo-

37 tactic radiosurgery (SRS), see chapter 6. Then the DMG was used to verify temporal

38 and spatial properties of extremely fast moving binary MLC systems used in a helical

39 tomotherapy system, see (chapter 7).

40 A second detector system which is a two dimensional array detector so named

41 the “Magic Plate” (MP) is then described in subsequent chapters. This detector

42 prototype has been designed to be used as a transmission detector to verify dose from

43 IMRT and volumetric modulated arc therapy (VMAT) treatment delivery. The basic

44 characterisation of this system under simple open ﬁeld linac x-ray beams is described 1.1. Project aim 3

45 in chapter 8. The clinical implementation and testing under more complex clinical

46 IMRT treatment beams is described in chapter 9.

47 Chapter 10 presents the summary of the main ﬁndings of this thesis. The ad-

48 vantages, limitations and the future direction of the development of these two novel

49 detector system prototypes are also discussed. 50 Chapter 2

51 Literature review

52 2.1 Cancer statistics

53 According to a report by the Australian Institute of Health and Welfare & Australasian

54 Association of Cancer Registries (AIHW & AACR, 2010), the highest number of re-

55 ported cancer incidence is prostate cancer (19403 cases per annum), followed by breast

56 cancer (12670 cases) for the Australian male and female population, respectively. This

57 is followed by melanoma of the skin (10340 cases) and lung cancer (9703 cases). One

58 in every ten deaths in Australia is caused by cancer, second only to cardiovascular

59 diseases. However, between 1982 and 2007, rate of deaths decreased for most cancer

60 sites. The fall in the mortality rate is attributed to improved treatment and early

61 detection of cancer.

62 2.2 Radiation treatment trend

63 Radiation therapy has long been recognised as an eﬀective method in the treatment of

64 cancer. Radiation can be delivered to patients in three ways, externally using medical

65 linear accelerators, also called external beam radiation therapy (EBRT), internally via

4 2.2. Radiation treatment trend 5

66 brachytherapy, or systemically using radioisotope therapy.

67 Since the invention of the computed tomography (CT) scanner by Sir Godfrey

68 Hounsﬁeld in 1971, this has been incorporated as an essential part of the treatment

69 planning work chain allowing three dimensional radiation therapy treatment tech-

70 niques (Battista et al., 1980). The use of blocking material to create irregularly shaped

71 ﬁelds and subsequently the use of multileaf collimators (MLC) enabled the pathway

72 to 3D conformal radiation therapy. The idea of 3D conformal radiation therapy is

73 essentially a forward planning method whereby uniform radiation ﬁelds are delivered

74 from multiple angles to the planning treatment volume (PTV), guided by the 3D CT

75 images, providing a homogeneous dose to the target while sparing surrounding nor-

76 mal tissues. This technique was largely successful in most cancer sites and reduced

77 radiation toxicity to the surrounding normal tissues.

78 In 1982, Brahme et al. (1982) published their paper which was later regarded as the

79 conceptual paper that introduced the concept of a treatment technique we now know

80 as intensity modulated radiation therapy (IMRT). They presented the problem of the

81 treatment planning as an inverse problem, as opposed to the forward planning method

82 that the experts of that time were used to. They also proposed that in order to provide

83 a uniform dose to a circular target structure while sparing the critical structure located

84 at the centre of the circular target, one needs to provide a radiation beam which is

85 non uniform. The advent of IMRT treatment technique increased the dose gradient

86 between tumour and close neighbour re-entrant normal structures. This allowed the

87 potential for dose escalation to PTVs as well as produced steep dose gradients between

88 PTV and some adjacent critical organs.

89 This chapter presents a brief introduction of some of the modern radiation therapy

90 delivery techniques used to deliver IMRT. It focuses on methods of delivery that are

91 subsequently used for dose delivery during dosimetric measurements. The two IMRT 2.3. Intensity Modulated Radiation Therapy 6

92 techniques include linac IMRT and Tomotherapy. In addition, the chapter discusses

93 stereotactic radiotherapy (SRS) as this is the method used for small ﬁeld dosimetry.

94 Dosimeters that are currently used for the dosimetric veriﬁcation for these deliveries

95 are also discussed as this provides a focus for comparison. This is followed by an

96 introduction to silicon based dosimeters for high spatial resolution dosimetry and two

97 dimensional array detectors.

98 2.3 Intensity Modulated Radiation Therapy

99 Intensity modulated radiation therapy is a complex radiation delivery. It utilizes an

100 MLC to modulate the radiation beam to achieve a high dose to the tumour while

101 sparing the critical organs close to the irradiated target. IMRT plans tend to produce

102 steep dose fall oﬀ regions at the interface of the target and organ-at-risk. In an

103 IMRT plan, a single ﬁeld is made up of several segments comprised of diﬀerent MLC

104 shapes. The sum of these segments will produce a non-uniformed radiation ﬁeld. Two

105 techniques that are normally used on conventional linear accelerators are the step and

106 shoot delivery and sliding window delivery. Both techniques deliver IMRT treatment

107 with static gantry angles.

108 2.3.1 Step and shoot IMRT delivery

109 In the step and shoot technique, the radiation beam is only switched on when the

110 leaves are stationary. When the leaves are changing from one segment to the next, the

111 radiation beam is switched oﬀ (Boyer et al., 2001). 2.4. Volumetric Modulated Arc Therapy 7

112 2.3.2 Dynamic sliding window IMRT delivery

113 In the dynamic sliding window technique, the radiation beam is kept on all the time

114 as the MLC leaves sweep across the ﬁeld. This technique is sometimes aﬀected by the

115 mechanical limitation of the MLC whereby the leaves may not have reached the right

116 leaf positions, and therefore the radiation beam has to be halted temporarily to wait

117 for the leaf to come into the correct position (Boyer et al., 1992).

118 2.4 Volumetric Modulated Arc Therapy

119 The original concept is based on a paper by Yu (1995) called intensity modulated

120 arc therapy (IMAT). This concept did not include dose rate modulation. Volumetric

121 Modulated Arc Therapy (VMAT) is a special type of IMRT delivery technique where

122 treatment is delivered in a single dynamically modulated arc. In other words, the

123 radiation is delivered as the linac gantry rotates around the patient (Otto, 2008).

124 In VMAT, the dose rate and the gantry rotation speed is also varied to achieve the

125 required beam modulation. Because VMAT essentially uses all available angles in

126 the inverse plan domain of parameters, it has the potential to produce superior dose

127 distributions than IMRT while using the same inverse planning dose objective approach

128 as IMRT. Planning comparisons show some small gains over IMRT (Bzdusek et al.,

129 2009; Matuszak et al., 2010; Yan et al., 2010). The time advantages of not stopping and

130 starting the gantry at ﬁxed gantry locations also suggest VMAT is faster to deliver

131 than IMRT, representing a gain of 30% in patient in room time when all in room

132 procedures are included (Hardcastle et al., n.d.). 2.5. Stereotactic Radiosurgery / Radiotherapy 8

133 2.5 Stereotactic Radiosurgery / Radiotherapy

134 Stereotaxy is a method by which a point is deﬁned within the patient’s body by

135 an external three-dimensional coordinate system (Grosu et al., 2006). Stereotactic

136 radiosurgery (SRS) is the use of radiation ablation in place of conventional surgical

137 excision to remove or modify a benign lesion in the body (Metcalfe et al., 2007).

138 Traditionally, this method requires a delivery of a large dose in a single treatment,

139 resembling a surgical procedure. In malignancy cases, the treatment is carried out

140 using a series of equal dose fractions; this is then termed stereotactic radiotherapy

141 (SRT). The intention of the SRT/SRS treatment is to deliver a concentrated dose to

142 a small volume of tumour tissue, usually located in close proximity to critical organs.

143 Or, in other cases, it is intended as a boost dose to the target volume. The nature of

144 SRT/SRS treatment requires it to have a very high geometric precision, i.e. a tight

145 margin for the planning target volume (PTV) and a sharp dose fall oﬀ (Grosu et al.,

146 2006). Due to the high dose delivery and tight margins required in this technique, the

147 planning and delivery of the treatment requires great precision and accuracy.

148 2.6 Helical Tomotherapy

149 Tomotherapy, literary translated as “slice-therapy” was ﬁrst introduced using the

150 PEACOCK (NOMOS) binary MLC (MIMiC) and retroﬁtted on conventional linacs

151 (Carol, 1995). On this device, the slices were treated with a single gantry rotation.

152 Then the patient was stepped to the next superior-inferior position.

TM 153 Helical Tomotherapy was born in the late 1980s at the University of Wisconsin,

154 Madison. It was designed with the aim to produce non-uniform intensity ﬁelds to

155 achieve a highly conformal dose distribution with a central avoidance structure (Mackie

156 et al., 1993; Mackie, 2006). The concept of helical Tomotherapy is to deliver intensity 2.7. Current quality assurance and dosimetric approaches 9

157 modulated radiation therapy in a continuous rotational manner with the linac mounted

158 on a slip ring gantry, while the patient is translated through the gantry bore. The

159 translation is analogous to a diagnostic computed tomography (CT) scanner, hence the

160 term Tomotherapy. The radiation source used is a fan beam 6 MV linear accelerator.

161 The beam modulation was achieved by a binary multileaf collimator (MLC) seated

162 at the distal end of the linear accelerator. The binary MLC consists of 64 tungsten

163 leaves extending across a ﬁeld of 40 cm in the x-direction at the isocenter. In contrast

164 to the ﬁeld shaping MLC in conventional linacs, the binary MLC has only two states,

165 either closed or opened. Hence, beam modulation is achieved by varying the leaf open

166 time, i.e. temporal modulation. The helical Tomotherapy device also has a linear

167 array of 738 xenon ﬁlled ion chambers that sits opposite the linac source on the gantry

168 ring. This detector is not only used for patient localisation but also allows the helical

169 Tomotherapy to perform MVCT image guided radiation therapy. The original design

170 (Mackie et al., 1993) had a kV x-ray source oﬀset from the MV linac to provide kV

171 CT scans. This as yet has not been implemented and the MV CT scans are used for

172 IGRT.

173 2.7 Current quality assurance and dosimetric ap-

174 proaches

175 Currently, dosimetric veriﬁcation of IMRT plans are patient based. Techniques used

176 are commonly based on a point dose measurement using ion chambers and planar

177 dose veriﬁcation. Films, electronic portal imaging devices (EPIDs), and two dimen-

178 sional diode or ion chamber arrays are often used for planar dose veriﬁcations. Other

179 dosimeters such as diamond detectors, thermoluminescence detectors (TLDs) and gel

180 dosimeter are often used as complimentary dosimeters in IMRT delivery veriﬁcation. 2.7. Current quality assurance and dosimetric approaches 10

181 2.7.1 Ionisation chamber

182 The ionisation chamber, the long established gold standard dosimeter used in radiation

183 therapy, is a highly reliable dosimeter. It measures the number of ion pairs produced in

184 a volume of air due to radiation. In the simplest arrangement, the ionisation chamber

185 exists as two electrode plates spaced apart in air. A large potential (100V - 400V)

186 is applied to the plates. When radiation traverses between the plates, it ionises the

187 air producing free negative electrons and positive ions. The positive and negative

188 charged particles are then swept towards the appropriate electrodes by the electric

189 ﬁeld, producing a steady current ﬂow in the external circuit, which can be measured

190 by an electrometer. The conversion from ionisation to dose is historically understood

191 (Bragg, 1910; Gray, 1929, 1936) and various protocols to correct for detector designs

192 are also well established (Almond et al., 1999; IAEA, 2006)

193 Due to the ﬁnite size of the air volume required for a suﬃcient signal, the ionisation

194 chambers are limited in physical dimension. When used in IMRT ﬁelds where high

195 dose gradients exist, the ionisation chamber tends to overestimate the penumbra width

196 due to its larger volume. Therefore it may not accurately represent areas with high

197 dose gradient fall oﬀ (Bucciolini et al., 2003). Exact output measurements in small

198 SRS ﬁelds are also a challenge (Das et al., 2000, 2008; Fan et al., 2009) due to the

199 volume averaging of the dose.

200 2.7.2 Film dosimetry

201 Radiographic ﬁlm dosimetry has excellent spatial resolution, but it is non-tissue equiva-

202 lent, energy dependent and somewhat eﬀected by processing conditions. Radiochromic

203 (Gafchromic) ﬁlm is almost tissue-equivalent and self-developing, and its ease of han-

204 dling in room light makes this a convenient dosimeter for small ﬁeld dosimetry. All

205 ﬁlm methods provide a 2D pixel intensity map that can be converted to a 2D dose 2.7. Current quality assurance and dosimetric approaches 11

206 map given appropriate calibration. Film methods are non real time and are poten-

207 tially affected by polarization effects and non-uniformities in commercial film scanners

208 Butson et al. (2003b, 2006b); Paelinck et al. (2007).

209 2.7.3 Silicon diode

210 In medical radiation dosimetry, silicon is an attractive semiconductor material as a

211 radiation detector. Silicon diodes are able to achieve high signal response with a small

212 sensitive volume. Hence these detectors can measure to a high spatial resolution. This

213 is because, due to the higher density of a solid, it has 18000 times higher sensitivity

214 than an ionisation chamber with the same volume while the energy required to produce

215 an electron-hole pair is 3.6 eV, 10 times less than those required to ionise air. Another

216 feature that makes the silicon diode a good detector for radiation therapy application

217 is the constancy of the silicon to water electron stopping power ratio for the energy

218 range used in mega voltage radiation therapy (Figure 2.1). Other advantages of the

219 silicon diode are its robustness and mechanical stability, its excellent reproducibility,

220 and its ability to be used in passive mode (Saini & Zhu, 2002; Rosenfeld, 2006).

221 A silicon diode is primarily a p-n junction. A pure bulk silicon is ﬁrst lightly doped

222 with phosphorus or boron to produce n-type or p-type silicon, respectively. Then, a

223 heavily doped impurity of the opposite type is implanted on the surface region to form

224 a p-n junction. The diode type is deﬁned by the bulk material.

225 The principle of operation for a silicon diode as a radiation detector is illustrated

226 in Figure 2.2. At the p-n junction, the majority carriers of each side diﬀuses to the

227 opposite side, resulting in an electric ﬁeld (or build-in potential, Ψo) which prevents

228 further diﬀusion of the majority carriers (Shi et al., 2003). This diﬀusion of the ma-

229 jority carriers also leave behind immobile negatively charged acceptors and positively

230 charge donor ions in the p-andn- side. This layer of the spatially charged region is 2.7. Current quality assurance and dosimetric approaches 12

231 called the ‘depletion layer’, W. The diode is said to be operating in passive mode when

232 no external bias voltage is applied. In this mode, no current will ﬂow in the external

233 circuit. When ionising radiation traverses the diode, it generates electron-hole pairs

234 in the diode. The excess minority carriers (i.e. electrons in the p-side and holes in

235 the n-side) located within the diﬀusion lengths, Ln and Lp, will diﬀuse towards the

236 p-n junction and be swept by the built-in potential towards the electrodes. This ﬂow

237 of charge results in a current ﬂow in the external circuit, which is measurable by the

238 electrometer.

239 Silicon diodes are however not without limitations. The energy, angular, temper-

240 ature, and dose rate dependency require rigorous characterisation (Rice et al., 1987;

241 Beddar et al., 1994). Silicon diodes also over-respond at low photon energies (<150

242 keV) due to the increased photoelectric cross section (Rosenfeld, 2006; Wong et al.,

243 2010). Asymmetric packaging and the inherent anisotropy of the silicon diode will re-

244 sult in the detector’s directional dependence (Westermark et al., 2000; Higgins et al.,

245 2003; Marre & Marinello, 2004; Jursinic, 2009). The use of diodes for in-vivo dosimetry

246 needs to take into account possible temperature dependence (Van Dam et al., 1990;

247 Saini & Zhu, 2002). Dose rate dependence of the detector also needs to be understood,

248 particularly when it is used under pulsed dose rate such as in the linear accelerator

249 Rikner & Grusell (1983); Van Dam et al. (1990); Grusell & Rikner (1993); Shi et al.

250 (2003). Planar arrays of Si diodes are widely used in the quality assurance of IMRT.

4 251 Examples include the Sun Nuclear MapCHECK and the ScandiDos Delta bi-planar

252 system.

4 253 2.7.3.1 Delta system

4 254 The Delta system (Scandidos, Uppsala, Sweden) consists of two planar dosimeter

255 arrays mounted orthogonally. Dose reconstruction software can then generate 3D dose

256 maps. It has 1069 p-type silicon diodes on two orthogonal planes mounted within a 2.7. Current quality assurance and dosimetric approaches 13

Figure 2.1: Silicon/water stopping power ratio.

Figure 2.2: Schematic of a silicon p-n junction diode (Shi et al., 2003). 2.7. Current quality assurance and dosimetric approaches 14

257 polymethylmethacrylate (PMMA) phantom. The diodes are cylindrical in shape with

2 2 258 an active area of 0.78 mm . Diode pitched is 6 mm within the central 6 × 6cm

259 region, and 10 mm for the outer region.

260 Many 2D planar diode or ion chamber arrays were designed for normal incidence

261 beams while veriﬁcation of composite dose distribution is achieved by summing mul-

262 tiple beams. This simpliﬁcation of the QA method lacks the veriﬁcation of gantry,

263 collimator and couch angles (Sadagopan et al., 2009).

4 264 The Delta system can be used in a mode that addresses an additional parameter

265 (i.e. gantry rotation) that exists in VMAT treatment delivery. It has an inclinometer

266 that is attached to the linac gantry providing independent measurements of the gantry

267 rotation for VMAT angle measurement.

268 2.7.4 Diamond detectors

269 Diamond detectors are lauded for their near tissue-equivalence and small sensitive

270 volume, which is particularly useful for small ﬁeld dosimetry and measuring at high

271 dose gradient regions. The detectors were also found to be independent of photon beam

272 quality for the clinical range (Laub et al., 1997; De Angelis et al., 2002). However,

273 one needs to understand the dose-rate dependence and signiﬁcant preirradiation dose

274 eﬀect (Heydarian et al., 1996). Fidanzio et al. (2000) reported a sensitivity variation

275 of 1.8% for the dose per pulse range of 0.068 mGy to 0.472 mGy.

276 2.7.5 Gel dosimetry

277 Gel dosimeters are fabricated from radiosensitive chemicals dissolved in gelatine ma-

278 terial which upon irradiation, polymerised as a function of irradiated dose. Since the

279 ﬁrst introduction of this dosimeter in the 1950s, the gel dosimeter undergoes several

280 modiﬁcations in its formulation to address various limitations of the dosimeter. The 2.7. Current quality assurance and dosimetric approaches 15

281 original ferrous sulphate chemical dosimeter was bog down with post irradiation ion

282 diﬀusion problem (Gore & Kang, 1984; Olsson et al., 1992) while the polymer gel

283 dosimeter is susceptible to atmospheric oxygen inhibiting the polymerisation process

284 (Maryanski et al., 1994). The recent nomoxic gel dosimeter (Fong et al., 2001) are

285 currently used by researchers in the ﬁeld for clinical applications.

286 Various methods can be used to readout gel dosimeters such as the Magnetic Res-

287 onance Imaging (MRI), x-ray computed tomography (CT), optical CT or ultrasound

288 (Gore & Kang, 1984; Maryanski et al., 1994; Hilts et al., 2000; Mather et al., 2002).

289 Gel dosimetry has the advantage of having the same medium for dosimetry and dose

290 scattering. It is near water equivalent and have high spatial resolution in all three

291 dimensions. It is however aﬀected by MRI- and gel-related non-uniformity and arte-

292 facts.(Vergote et al., 2004).

293 2.7.6 Electronic portal imaging device

294 The electronic portal imaging device (EPID) consists of a ﬂat panel amorphous silicon

295 detector array mounted on a retractable arm opposite to the linac beam portal. It

296 measures the exit ﬂuence of the radiation and is usually used for online portal imaging

297 for veriﬁcation of patient localisation. In recent years, its use has been extended to the

298 ﬂuence veriﬁcation of IMRT deliveries. The EPIDs have excellent spatial resolution.

299 However, the use of the EPIDs as an IMRT dosimetric veriﬁcation tool requires com-

300 plicated conversion of image to dose (Warkentin et al., 2003). Van Esch et al. (2004)

301 and Greer et al. (2007) have also achieve success in using this device as an IMRT

302 QA dosimeter device. Challenges in using the EPID as an online patient dosimeter

303 include its oﬀ axis energy response to spectral changes in the beam. Despite these

304 challenges, Partridge et al. (2002) have reported the use of EPID’s for online patient

305 dose veriﬁcation. 2.8. High spatial resolution dosimetry 16

306 2.8 High spatial resolution dosimetry

307 One common feature shared by the various state of the art radiation therapy treatment

308 techniques described in the ﬁrst part of this chapter is that the treatment plans gen-

309 erated with these new methods aim to achieve high target/tumour conformance while

310 avoiding the adjacent organ at risk. This generally produces a dramatic change in

311 dose intensity within a short spatial distance. Therefore, in order to measure the high

312 gradient dose regions generated by IMRT plans, the small ﬁeld dosimetry SRS/SRT

313 plans and the non-conventional dosimetric considerations such that exist in helical To-

314 motherapy treatment plans, one needs to have a detector with high spatial resolution.

315 Hence, the sensitive volume of the detector has to be small in physical size (Kron

316 et al., 1993, 2002).

317 2.8.1 Concept of silicon strip detector

318 Research in high energy physics (HEP) pioneered the development of tracking vertex

319 detectors and data acquisition systems (DAQ). Experience gained from this research

320 was translated into the medical ﬁeld. Strip detectors, which are position sensitive

321 detectors, were ﬁrst used in high energy physics (HEP) experimental research for high

322 precision charge particle tracking in a vertex detector (Damerell, 1995; Rosenfeld et al.,

323 1993).

324 2.8.1.1 Theory of strip detector

325 In its generic form a strip detector is an array of p-n junction diodes. As an example,

326 for a p-type device, a low resistivity p-type wafer is used as the starting bulk material.

327 The p-n junctions are formed by ion implantation of the heavily doped impurities

+ 328 of the opposite side n strips onto the non-segmented wafer substrate. On the back

+ 329 end, a highly doped p implant is used to provide an ohmic contact connected to a 2.8. High spatial resolution dosimetry 17

330 negative voltage, VB (Figure 2.3). When a photon with suﬃcient energy traverses

331 through the silicon, photon interactions and electrons set in motion by the photon

332 creates electron-hole pairs along the tract. The free moving electrons and holes will

+ + 333 then move towards the p and n layers, respectively. A net current ﬂow results from

334 this charge movement.

335 The current ﬂow in the silicon is picked up by the metal contacts that are laid on

+ 336 top of the n strips. The metal contacts are isolated from the implanted strips by a

337 thin passivation layer silicon oxide (SiO2), so called ﬁeld oxide. The strips are wire

338 bonded to a fan-out board connecting to an individual preampliﬁer for each detector.

339 The spatial resolutions of these devices are generally in the order of μm.

+ 340 The thin passivation layer of silicon oxide (SiO2) also exists between the n strips.

341 This passivation layer will inevitably collect positive charges which also increase with

342 accumulated radiation. The presence of this positive charge layer is compensated by a

343 thin layer of electrons in the bulk material. This may eventually create a low resistance

344 interstrip leakage path, eﬀectively shorting the adjacent strips (Damerell 1995). One

+ 345 method to overcome this problem is to create ﬂoating p-type zones between the p

346 strips, also called p-stop layers. However, the addition of these p-stops resulted in

347 additional technical diﬃculties and costs. Another alternative to achieve interstrip

348 insulation is the use of a uniform blanket ion implant performed on the surface of the

349 silicon (p-spray) (Pellegrini et al., 2007).

350 Silicon strip detectors are usually fabricated with a thickness of approximately 300

351 μm. The thinness of a detector is limited by the loss of signal charge, made worse by

352 the poor signal-to-noise ratio due to the increased capacitance from strip to substrate

353 (Damerell, 1995). 2.8. High spatial resolution dosimetry 18

Figure 2.3: Operation principle of a silicon strip detector [modiﬁed from (Damerell, 1995)]. 2.8. High spatial resolution dosimetry 19

354 2.8.1.2 Readout electronics

355 The silicon strip detector typically comprises of 128 channels. The readout of all 128

356 channels simultaneously would be limited by the logistics of the readout electronics.

357 The evolution of the accompanying readout electronics for the silicon strip detectors

358 was driven by various factors:

359 (i) The rapid and diverse application of the detector by the HEP experimental re-

360 searches and the need to equip detectors to work in increasingly varied and

361 extreme hostile environments

362 (ii) The rapid development and commercialisation brought on by the integrated cir-

363 cuit industry

364 Current readout electronics mostly utilises Very Large Scale Integration Applica-

365 tion Speciﬁc Integration Circuit (VLSI ASIC) customised to perform the readout of

366 the silicon strip detector for the intended temporal and noise resolution.

367 2.8.2 Application of high spatial resolution dosimeters in med-

368 ical radiation therapy

369 Application of strip detectors in medical radiation dosimetry were reported by Pappas

2 370 et al. (2008) with their 128 channel strip detector (active area 0.25 × 0.25 mm and

371 pitch 0.3 mm) in stereotactic radiotherapy QA. The use of silicon strip detectors

372 with a pitch of 121 μm for dosimetric characterization and position imaging of Sr-

373 90 for cardiovascular brachytherapy was reported by Caccia et al. (2004). At the

374 Centre for Medical Radiation Physics (CMRP), a 0.2 mm pitch and a sensitive area

2 375 of 0.02 × 2mm high spatial resolution silicon strip detector, referred to as the Dose

376 Magnifying Glass (DMG) was designed, prototyped and used in IMRT, SRS, and

377 helical Tomotherapy QA veriﬁcation (Wong et al., 2010, 2011a,b). 2.9. Concept of a two dimensional array detector 20

378 2.9 Concept of a two dimensional array detector

379 A strip detector provides excellent positional information in a single dimension, but

380 it lacks the capability to produce picture-like two-dimensional information. As IMRT

381 treatment becomes a common treatment delivery method in many radiation therapy

382 delivery centres, quality assurance (QA) of the safe and accurate delivery of IMRT

383 treatments has become ever more important. Routine patient speciﬁc QA is recom-

384 mended for IMRT treatments (Ezzell et al., 2003; Nelms et al., 2011).

385 In its very basic form, the QA of IMRT treatment involves

386 (i) point dose measurements, and

387 (ii) planar dose (or ﬂuence) comparison.

388 For point dose measurement, the common practice is to measure a point dose at

389 a high dose-low gradient region. Planar dose comparison are made using ﬁlms or two

390 dimensional detector arrays.

391 The use of electronic 2D detector arrays are gradually gaining widespread use.

392 They provide eﬃcient means of measuring doses at multiple locations in the ﬁeld,

393 providing real time feed back and performing planar dose comparison simultaneously

394 (Venkataraman et al., 2009). Examples of some of the commercially available 2D de-

395 tector arrays are the ionisation chamber based I’mRT MatriXX, COMPASS 2D trans-

396 mission detector (IBA, Germany), 2D-ARRAY Type 10024 (PTW Freiburg, Germany)

397 (Amerio et al., 2004; Spezi et al., 2005; Stasi et al., 2005; Poppe et al., 2006b) and

398 semiconductor based MapCHECK (Sun Nuclear, Melbourne, Fl) (Jursinic & Nelms,

399 2003; Letourneau et al., 2009). Eﬀorts were also made to utilise the existing electronic

400 portal imaging device (EPID) on the linac as a 2D array detector (Greer et al., 2007).

401 The various existing 2D detector arrays varied in their respective detector/chamber

402 size and detector-to-detector spacing (Table 2.1). Poppe et al. (2007) studied the 2.9. Concept of a two dimensional array detector 21

403 volume averaging eﬀect in large size detectors and detector array gaps. They proposed

404 the use of the Nyquist sampling theorem to estimate the optimal detector pitch for a

405 sample of IMRT plans. In their work, they found that for a moderately complicated

−1 406 head and neck IMRT plan, most of the spatial frequencies were <0.1 mm . Therefore

−1 407 the required sampling frequency should be 0.2 mm , corresponding to 5 mm detector

408 pitch.

409 In general, IMRT QA veriﬁcation using 2D detector arrays for planar dose mea-

410 surement can be divided into three approaches:

411 (a) With the 2D detector array positioned on the linac couch or mounted onto the

412 gantry head with a special jig, in the absence of a patient

413 (b) With the 2D detector array positioned between the MLC collimator and the pa-

414 tient, acting as a transmission detector for on-line and in-vivo measurement, i.e.

415 during patient treatment

416 (c) With the 2D detector array positioned downstream of the patient, such as using

417 an EPID or mounting a detector array on the EPID.

418 The ﬁrst approach, the use of 2D detector arrays on the linac couch for planar

419 dose measurement is one of the most common approaches. This type of QA is patient

420 speciﬁc and usually performed prior to the delivery of the ﬁrst IMRT treatment. It

421 is used to check the ability of the MLC to achieve the intended dose and modulation

422 as planned by the treatment planning system (TPS). The measurement can only be

423 done on a phantom that was CT scanned and recalculated with the patient speciﬁc

424 IMRT plan. The underlying assumption is that IMRT delivery would be reproducible

425 throughout the whole patient treatment period.

426 The third approach, which utilised the EPID can be done with or without the

427 presence of the real patient. It requires accurate reconstruction and recalculation of 2.9. Concept of a two dimensional array detector 22 10 mm variable spacing (7.07 - 10 mm) jected to 10 mm at SDD 100 cm. 6.5 mm Detector pitch 7.5 mm 2 2 0.8 mm 5mm × × ter, 2 mm height tive size/volume mm diameter, 5 mm height) 40) 3.8 mm diame- 27) 5 32) cylindrical (4.5 × × × 371600 (40 4.4 mm pro- -type Si diode 445 0.8 Ion chamber 1027 (32 Ion chambern 729 (27 Multi wire ionisation chamber Planelel paral- chamber ionisation Table 2.1: Speciﬁcation of commercial 2D arrays. Manufacturer Detector typetry,Germany No. of detectorsPTW Detector Freiburg, Germany sensi- Melbourne, Fl Germany IBA Dosimetry, Germany Commercial name I’mRT MatriXX IBA Dosime- 2D-Type 10024 ARRAY MapCHECKDAVID Sun system PTW Nuclear, Freiburg, COMPASS system 2.9. Concept of a two dimensional array detector 23

428 signals which include patient or phantom abrsoption and scattering. If it is being used

429 for in-vivo patient dose monitoring, daily patient variability could easily mask any

430 dose delivery error.

431 The ultimate aim of IMRT QA veriﬁcation, although not fully realized, is to ensure

432 the safe, accurate and reproducible delivery of IMRT treatment onto the patient. This

433 not only means that IMRT QA has to be patient speciﬁc, correct MLC functioning

434 in the pretreatment QA veriﬁcation, but also throughout the whole treatment period.

435 One way to achieve this is via the second approach, where the 2D detector array is

436 positioned upstream of the patient, usually on the linac accessory slot. The detector

437 has to be of the transmission type with minimal beam perturbation. With the on-

438 line and in-vivo measurements, daily and almost real time feedback on the accuracy

439 delivery of IMRT treatment is achievable.

440 The idea of a 2D array transmission detector was ﬁrst conceptualised by Paliwal

441 et al. (1996). They used a commercial oﬀ the shelf dose area product (DAP) meter,

442 commonly used in a ﬂuoroscopy unit in the radiology department to monitor the

443 radiation beam. The concept of a dose area product was further developed by Poppe

444 et al. (2006a) in the design of the DAVID system and by Islam et al. (2009) with the

445 Integral Quality Monitor (IQM). Both systems allow online comparison of the real

446 time, in-vivo measurement of dose area product due to the opened MLC leaves with

447 pre-recorded data. Both groups reported high sensitivity to MLC leaf positioning

448 error down to 1 mm. The DAVID system was eventually used in a clinic for daily

449 IMRT veriﬁcations (Poppe et al., 2010). Although these two systems were highly

450 sensitive to leaf positioning errors, the systems do not have 2D spatial information

451 of the delivered dose making point wise veriﬁcation impossible. Consequently, the

452 detected MLC functioning errors could not be correlated with under- or overdosing of

453 anatomical features in the treatment planning system. In other words, the errors in 2.9. Concept of a two dimensional array detector 24

454 IMRT delivery cannot be correlated to clinical relevant impact or end points.

455 Up until this point, IMRT QA veriﬁcations involved primarily ensuring the correct

456 and reproducible MLC functioning, the stability of the radiation output and the cor-

457 rect setting of device related parameters. The comparison of the measured 2D dose

458 map were usually compared to the TPS generated dose map using performance metrics

459 such as the dose diﬀerence, distance to agreement (DTA) comparison or the Gamma

460 analysis (Low et al., 1998). Nelms et al. (2011) in their recently published paper have

461 quantiﬁed the adequacy of this performance metric (Gamma) and suggested it is not

462 always an ideal tool in the prediction of clinical relevant dose errors. Towards that

463 end, several commercial 2D detector arrays with advanced software reconstruction al-

464 gorithms attempt to close the gap between physics performance metrics with clinically

465 relevant dose errors. These systems measured the 2D dose maps and reconstructed the

466 3D dose distribution using the CT data input from the TPS. Examples of such systems

467 are the COMPASS system (IBA Dosimetry, Germany), DOSIMETRYCHECK (Math

468 Resolutions LLC) and 3DVH (Sun Nuclear Corporation).

469 2.9.1 Pixelated detector

470 On the extreme end of the spectrum of detector size, is the pixelated detector. “Pixel

471 detector” is taken to mean a device equipped with a two dimensional array of detectors.

472 This technology has taken ﬂight in the commercial market particularly in camcorders

473 and digital cameras. The utilisation of a pixelated detector in the medical ﬁeld is not

474 uncommon particularly in medical imaging, such as the positron emission tomography

475 (PET) and gamma cameras. 2.9. Concept of a two dimensional array detector 25

476 2.9.2 MAESTRO project

477 Pixelated silicon detectors were recently proposed and a prototype data acquisition

478 (DAQ) system was developed in the framework of the European project MAESTRO

+ 479 (MAESTRO, 2008). Each pixel is based on an n -p junction surrounded by a guard-

480 ring structure implanted on an epitaxial 50 μm thick p-type silicon layer grown on

481 a Czochralski (Sze, 2001) substrate, a monolithic silicon segmented module. The

482 best compromise between granularity and electronic complexity has been achieved by

2 483 choosing pixels with 2 × 2mm active area and 3 mm pitch. The discrete readout

484 electronics for 441 pixels/channels (21 × 21 pixels) was developed. The system demon-

485 strated good results but is quite bulky due to discrete readout electronics (Menichelli

486 et al., 2007; Talamonti et al., 2007). The next version will utilize nine TERA 06 ASIC

487 chips for readout of segmented detectors. The system in development oﬀered better

488 resolution than 2D discrete diodes which have a spatial resolution of 3 mm. The use

489 of a TERA ASIC chip for the readout of a silicon pixelated detector or strip detector

490 was ﬁrst proposed and utilised by the CMRP. 491 Chapter 3

492 Methodology

493 This chapter describes the two detector systems that were used in this thesis, the

494 Dose Magnifying Glass (DMG) and the Magic Plate (MP). The working principles

495 of the electronic readout system will also be described. In many parts of this thesis,

496 the ﬁlm dosimetry was often used as the comparison dosimeter. The properties and

497 composition of the two types of ﬁlms used in this thesis will also be brieﬂy discussed.

498 3.1 Dose Magnifying Glass

499 3.1.1 Design and fabrication

+ 500 The Dose Magnifying Glass (DMG) is an array of 128 phosphor implanted n strips

+ 501 on a p-type silicon wafer. The sensitive area deﬁned by a single n strip is 20 × 5000

2 2 nd 502 μm for the prototype version and 20 × 2000 μm for the 2 generation DMGs. The

503 thickness of the silicon wafer is 375 μm and the strip pitch is 200 μm.

504 Two types of silicon strip detectors based on p-Si were produced, one of high

0Part of this chapter has been published in Medical Physics: Wong, J. H. D., Carolan, M., Lerch, M. L. F., Petasecca, M., Khanna, S., Perevertaylo, V. L., Metcalfe, P. and Rosenfeld, A. B. (2010). A silicon strip detector dose magnifying glass for IMRT dosimetry. Medical Physics 37(2): 427-439.

26 3.1. Dose Magnifying Glass 27

Figure 3.1: Schematic diagram of the Dose Magnifying Glass (DMG).

505 resistivity (5 k Ω·cm) ﬂoat zone silicon (Sze, 2001), and the other of low resistivity (10

506 Ω·cm) Chochralski silicon (Sze, 2001).

+ 507 The area between strips was implanted with Boron producing a p stop layer to

+ + 508 avoid shorting between adjacent n strips and between n strips with the common

509 electrode. The schematic diagram of the strip detector is presented in Figure 3.1.

+ 510 Aluminium was evaporated on top of the n areas. The detector was used in passive

511 mode and the readout was carried out with the detector conﬁgured as a planar detector

+ + 512 with a common electrode p from the same side as the n strips. 3.2. Magic Plate 28

Figure 3.2: Prototype version of DMG (left) and the 2nd generation DMG (right).

513 3.1.2 Detector packaging

514 The ﬁrst version of DMG is mounted on a ceramic substrate. The ceramic substrate

3 515 was of a higher density (∼ 2.5 - 3.5 g/cm ) material than silicon. The presence of this

516 ceramic substrate in the mounting of the detector caused additional attenuation of

nd 517 the radiation and aﬀected the angular dependence of the detector. The 2 generation

518 DMGs are mounted on a kapton substrate (Figure 3.2). The thin Kapton substrate

3 519 has a density of 1.42 g/cm , and has an eﬀective atomic number, Zeff =6.6(Berger

520 et al., n.d.). This was more tissue-equivalent than ceramic (Al2O3,Zeff = 11.2) and

521 was expected to create less perturbation to the radiation beam. It was therefore

522 also expected to reduce the angular dependency of the detector due to the detector

523 packaging. The detector array was mounted at the end of a 170 mm long Kapton

524 pigtail. This enabled the positioning of the silicon detector in phantoms to provide for

525 suﬃcient scattering.

526 3.2 Magic Plate

527 The Magic Plate (MP) is a 2D array of 11 × 11 silicon diodes covering an area of 10 ×

2 528 10 cm (Figure 3.3). The diodes were mounted on a 0.64 mm thick Kapton substrate 3.2. Magic Plate 29

Figure 3.3: The Magic Plate (MP) 2D array detector.

529 using the ‘drop-in’ technology. This technology was proposed and developed at the

3 530 CMRP. The physical size of a single diode is 1.5 × 1.5 × 0.425 mm .TheMPwas

531 designed to be mounted on the linear accelerator head (accessory mount slot) in line

532 with the radiation beam (Figure 3.4). It was designed to operate as a transmission

533 detector measuring the 2D ﬂuence map of a modulated radiation beam, hence the thin

534 Kapton substrate mounting. However, the MP can also be used as a planar 2D array

535 detector for phantom measurements.

536 3.2.1 Epitaxial diodes

537 The word epitaxy comes from the Greek word which means to arrange upon in an

538 orderly manner (Singh, 2001; Orton, 2004). This process involves growing a high

539 quality thin epitaxial ﬁlm on top of a heavily doped bulk silicon wafer which acts as

540 a crystal seed for the epitaxial growth and later serves as a supporting structure.

541 An epitaxial diode is an attractive alternative to a conventional silicon detector 3.2. Magic Plate 30

Figure 3.4: The Magic Plate mounted on the linac gantry. The mounting plate was a modiﬁed TBI applicator. 3.3. TERA readout system 31

542 due to the thin epitaxial layer which is deemed to be more radiation hard. The ability

543 to reduce the detector sensitive area allows the detector to be made thinner compared

544 to the 300 μm thick silicon detector (Kramberger et al., 2003). The diodes that were

545 used in the MP were grown using the epitaxial-growth technique (Sze, 2001). For the

546 MP diodes, the p-epitaxial layer is a 50 μm thick p-Si layer grown on top of a 375

+ 547 μm thick high resistivity p substrate. The sensitive volume of the individual element

+ 3 548 deﬁned by the n region is 0.5 × 0.5 × 0.05 mm , whilst the detector pitch is 1 cm.

549 The epitaxial diodes are of low resistivity type (100 Ω·cm). In the prototype design of

550 these epitaxial diodes, a 0.7 μm thick layer of SiO2 was grown on top of the epitaxial

551 layer. The silicon oxide (SiO2) layer is generally used as a mask to allow selective

+ 552 implantation of the n region as well as to serve as a protective layer for the silicon

+ 553 diode. The area surrounding the n region and the guard ring was implanted with

+ 11 3 554 Boron producing a p stop layer (0.1 μm and concentration = 10 Ω·cm )toprevent

+ 555 a positive charge building up at the p stop layer.

556 Two types of epitaxial diodes were manufactured, one with a guard ring (Figure

557 3.5) and one without a guard ring (Figure 3.6). In this thesis, the MP mounted with

558 epitaxial diodes that did not have a guard ring was used. The MP was operated in

559 passive mode.

560 3.3 TERA readout system

561 The TERA readout system was employed to readout the current signal from the DMG

562 and the MP. The TERA chip is a Very Large Scale Integration Application Speciﬁc

563 Integration Circuit (VLSI ASIC) that was designed by Istituto Nazionale di Fisica

564 Nucleare (INFN) - Torino Division and University of Torino microelectronics group

565 working on the readout of pixelated ionization or strip chambers for hadron therapy

566 (Bonazzola et al., 1998). It underwent several modiﬁcations during the last decade 3.3. TERA readout system 32

Figure 3.5: Schematic diagram of an epitaxial diode (with guard ring).

Figure 3.6: Schematic diagram of an epitaxial diode (without guard ring). 3.3. TERA readout system 33

Figure 3.7: Schematic diagram of the TERA readout [modiﬁed from (Mazza et al., 2005)].

567 (Mazza et al., 2005). Various versions of the TERA family of ASICSs have been

568 successfully implemented in diﬀerent 2D and 3D “Magic Cube” ionization chambers

569 in hadron therapy (Brusasco et al., 1997; La Rosa et al., 2006, 2008). The latest

570 version (TERA 6.0) of the chip was used by the Scanditronix-Wellh¨ofer (IBA Group)

571 in commercial dosimeters used in Radiotherapy (I’mRT MatriXX and StarTrack) based

572 on gas ionization chambers.

573 3.3.1 How TERA works

574 The structure of the TERA ASIC is based on a current-to-frequency converter followed

575 by a digital counter. It operates based on the charge-balancing or recycling integrator

576 technique (Figure 3.7) (Gottschalk, 1983; Horowitz & Hill, 1989) whereby the circuit

577 counts the number of times a capacitor is charged by an input current and discharged

578 by the circuitry (Bonazzola et al., 1998).

579 When radiation impinges on the silicon detector, it forms electron-hole pairs in

580 the silicon. The free moving charge carriers generate a current which is detected by

581 the readout system. This is the input current, Iin. The input current charged up

582 the 600 fF integrating capacitor (Cint) in the operational transconductance ampliﬁer

583 (OTA). The output waveform from the OTA appears as a voltage ramp, VA.Inthe 3.3. TERA readout system 34

584 comparator, the integrated charge is compared with a threshold voltage, Vth.When

585 the threshold voltage is reached (VA >Vth), the comparator ﬁres a calibrated pulse

586 (VB) to the pulse generator (PG), which triggers the pulse generator to output two

587 pulses; one, to be sent to the digital counter and the second to the subtraction circuit.

588 The pulse issued by the PG to the subtraction circuit will charge up the 200fF

589 subtraction capacitor, Csub. The output responses of the Csub to Vsub is two δ-like

590 current pulses with equal charge but opposite in polarity (Figure 3.8). The amplitude

591 of the Vsub is deﬁned by the diﬀerence of two externally set reference voltages, Vpulse+

592 and Vpulse− . The timing of the two current pulses coincides with the leading (δ+) and

593 trailing (δ-) edge of the voltage pulse. The current pulse with the same polarity to

594 the voltage output of the comparator is shorted to the OTA reference (Vref ), and the

595 other pulse is added algebraically to the input current to subtract a charge quantum

596 from the input current. This results in a sharp decrease of the charge across Cint,

597 proportional to the charge issued by the subtraction circuit (Bonazzola et al., 1998;

598 Mazza et al., 2005).

599 If at the end of the above process, the input voltage to the comparator is still above

600 the threshold voltage, the PG will continue to issue pulses to the counter and the

601 subtraction circuit. Once the integrated charge over Cint is below Vth, the comparator

602 trigger is reset and the PG will stop issuing pulses.

603 Using the charge-balancing technique, the integrated charge over Cint is “sub-

604 tracted” by the algebraic sum of the input current with a negative charge quantum.

605 This allows the Cint to continue to integrate while a ﬁxed amount of charge is being

606 subtracted. Hence, no charge is lost. This is far superior to the conventional resetting

607 of the Cint, whereby dead time is introduced when the capacitor is reset.

608 The charge quantum, Qc is the unit charge needed for 1 count. It is deﬁned by the 3.3. TERA readout system 35

Figure 3.8: Charge subtraction waveform (Mazza et al., 2005).

Table 3.1: Set reference values for the CMRP TERA 03 board. Value Vsupply 5V Vref 2V Vp+ 4.19 V Vp− 1.24 V Vth 2.50 V Cint 600 fF Csub 200 fF

609 following relationship (Eq.3.1),

Qc = Csub · Vsub (3.1)

610 where, Vsub =Vp+ -Vp−. The parameters Vp+ and Vp− are two reference voltages

611 set externally that determine the charge quantum. They can be varied to adjust the

612 Qc depending on the application. However, Qc should be kept within the range of 100

613 fC to 800 fC. The minimum limit of the charge quantum is set by the resolution of

614 the comparator, which is 100 fC. In this thesis, Qc = 590 fC (see Table 3.1 for detail

615 values).

616 The relationship between the input current and pulse frequency is described by 3.3. TERA readout system 36

617 Eq. 3.2:

I Pulse frequency, f = in (3.2) Qc

618 The maximum input current is limited by the pulse generator state machine (Mazza

619 et al., 2005). The current to frequency converter has a frequency limit of 5 MHz.

620 By which, the shortest time between two pulses issued sequentially is 0.2 μs. This

621 puts the limit of the maximum input current at 3 μA(forQc = 600 fC). However,

622 in the event of current overload, the subtraction circuit would not be able to keep

623 up with the integrated charge over Cint. The voltage input to the comparator, VA

624 would continue to remain at a level that is higher than Vth. This prompts the pulse

625 generator to continue to issue pulses at the maximum speed (both to the counter and

626 to the subtraction circuit) until the overload is removed and VA

627 In the case of pulsed linac radiation, a very high current (∼ 7 μA) is delivered

628 within a very short pulse width of 3.5 μs. The pulse period may range from 16 ms

629 to 2.7 ms (equivalent to 100 MU/min to 600 MU/min repetition rate). In this case,

630 within the initial pulse of 3.5 μs, the high input current charges the Cint way beyond

631 the Vth. A single Qc pulse will not be suﬃcient to bring the charge below the threshold

632 value. Hence, the pulse generator will continue to run at a maximum speed, issuing

633 pulses to the subtraction circuit until the overload is removed. Because the linac pulses

634 are brief, with a long pulse period, the subtraction circuit will have suﬃcient time to

635 remove the overload before the next pulse comes on. Hence, no error will result because

636 no charge will be lost (Gottschalk, 1983). However, one should be careful to ensure

637 that the input current producing voltage on the capacitor does not exceed the positive

638 rail voltage. In this case, it would be the power supply, Vsupply to the TERA board.

639 If the input current is driven up the positive rail, the output counts will approach

640 saturation where any increase in the input current cannot produce more counts. 3.3. TERA readout system 37

641 Intuitively, to resolve this problem, one could adjust the Qc to a larger value,

642 therefore subtracting the input current at a faster rate. This will be achieved at the

643 expense of a lower sensitivity of the counts produced.

644 Each TERA ASIC has 64 independent channels. In this thesis, two TERA chips

645 producing 128 channels were used. The channels are coupled to an individual digital

646 counter followed by a 16-bit register with a common load command. This allows the

647 counters to store the counts and readout at a speciﬁc time intervals.

648 3.3.2 Amplitude and timing

649 The accuracy of the comparator circuit deﬁnes the minimum equivalent Qc (100 fC

650 in this case) and hence the lower limit of the dynamic range. The upper limit of

651 the dynamic range is limited by the maximum voltage of the comparator circuit (the

652 supply voltage, 5 V), equating to a maximum charge Qc of 1 pC). The maximum

5 653 deviation from linearity for Qc = 600 fC is 0.5% for a dynamic range of 10 .

654 The data acquisition (DAQ) software was written using LabVIEW 8.6 (National

655 Instruments, USA) and allows on line and oﬀ line data analysis. Two twisted pair

656 ribbon cables, each 15 m long, connect the DAQ computer with the silicon strip

657 detector board. The FE4C front end board [produced by Physalus (Physalus S.r.l.

658 2001)] with two TERA 3.0 chips (PGA 144 package) mounted was controlled by a

659 Nuclear Instrument Card for logging data to the DAQ PC. The DAQ allows real time

660 measurements with a user deﬁned readout frequency of between 250 Hz and 0.5 MHz

661 (eﬀective input current integration time from 4 ms to 2 μs)

662 As mentioned above, a digital counter follows each channel output. The ﬁnal out-

663 put of the chip is a 16-bit number that represents the number of times the integrating

664 capacitor has been discharged. In this case, one would expect to measure between 0

665 and 65536 Qc charge units depending on the level of dark current in the detector (zero 3.3. TERA readout system 38

666 in the case of a passive detector) that would give rise to a minimum >0 counts. After

667 each counter readout clock pulse, the counter is re-zeroed. If more than 65536 pulses

668 are received by the counter with one readout clock cycle, the counter rolls over and

669 starts counting from zero again. The user therefore has to have a systematic method

670 to determine a suitable readout clock cycle (referred to below as the pulse width) to

671 gather good counting statistics while ensuring that the counters do not roll over. The

672 check to determine that the counter has not rolled over can be easily performed by

673 acquiring data at diﬀerent pulse widths and comparing the number of counts acquired.

674 For example, if the number of counts acquired with a 1-second pulse width doubled

675 the counts acquired with a 0.5-second pulse width, the counter has not rolled over at

676 1-second pulse width. In contrary, if the counter had rolled over, the counts acquired

677 with a 1-second pulse width would appear to be lower than those acquired with a

678 0.5-second pulse width. The user therefore should use a shorter pulse width for the

679 measurement.

680 3.3.3 Charge collection in silicon strip detector

681 Estimation of the charge collection in a single strip and the corresponding output

682 frequency of pulses from the TERA chip are important. Assume the linac provides

683 an average dose of 400 cGy/min corresponding to 6.67 cGy/s. This is made up of

684 200 separate 3 - 5 μs pulses every second (Rosenfeld, 2006). This corresponds to

685 0.33 mGy/pulse. For the estimation purpose, assume that the sensitive volume (100%

3 686 charge collection) of the single strip is at least 20 × 5000 × 300 μm . Taking the

3 3 −8 687 density of silicon to be 2.33 × 10 kg/m , the mass of silicon, m is 6.99 × 10

688 kg. Taking into account that the energy required to produce a electron-hole pair in

689 silicon, W = 3.6 eV and the dose in silicon and water are approximately similar for

690 linac photon energies,(Rosenfeld, 2006; Metcalfe et al., 2007) the expected charge is 3.4. Film dosimetry 39

691 (Eq.3.3),

692 Charge collected per linac beam pulse,

D · m (3.3 × 10−4J) × (6.99 × 10−8kg) Q = = =6.32 × 10−12C (3.3) W/e 3.6J/C

693 where, D is the linac dose per pulse, m is the silicon mass, and

3.6eV/ (e-h pair) × 1.602 × 10−19J/eV W/e = =3.6J/C (3.4) 1.602 × 10−19C/e

694 For Qc = 600 fC/count on the TERA, the number of counts on the output of TERA

695 will be about 10 counts/linac pulse. That provides approximately 2000 counts/s for

696 the quoted dose rate. In the case where charge collection is <100% the number of

697 measured counts per second may be less. This justiﬁes the use of the TERA chip for

698 the silicon strip detector in linac radiation ﬁelds.

699 3.4 Film dosimetry

700 3.4.1 Radiographic ﬁlm

701 Radiographic ﬁlm is made up of a clear polyester base coated with radiosensitive emul-

702 sion. This emulsion consists of silver bromide (AgBr) crystals embedded in gelatine

703 material. Absorption of photons or ionising radiation causes the silver bromide to

704 reduce to silver which appears as darkened areas on the ﬁlm. The readout of the ﬁlm

705 is done using an optical densitometer or commercial ﬁlm digitizer such as the Vidar

706 scanner (Vidar Systems Corporation, Hendon, VA). The advantage of ﬁlm is that it

707 has very high spatial resolution. Theoretically, it has resolution the size of a grain (0.2

708 -2μm) (Metcalfe et al., 2007). However, in reality, it is limited by the resolution of

709 the digitiser/readout mechanism. The disadvantages of radiographic ﬁlm are that it is 3.4. Film dosimetry 40

710 a non real time dosimeter and it is energy dependent due to the high atomic number of

711 the silver. Other than radiation photons, ﬁlms are also sensitive to light, which require

712 careful handling of the ﬁlm. The reproducibility or reliability of the radiographic ﬁlm

713 is also heavily dependent on the processor condition.

714 The radiographic ﬁlm that was generally used in this thesis is the enhanced dynamic

715 range (EDR2) ﬁlm (Eastman Kodak Company, Rochester, NY) which comes in a ready

716 packed form. It has a wide dynamic range of up to 600 cGy compared to its predecessor.

717 The EDR2 ﬁlm is widely used in radiation therapy particularly in the QA veriﬁcation

718 of IMRT deliveries (Zhu et al., 2002; Olch, 2002; Childress et al., 2005).

719 3.4.2 Radiochromic ﬁlm

720 Radiochromic ﬁlms are a relatively new dosimeter in medical radiation therapy. These

721 ﬁlms are comprised of a polyester based coated with radiosensitive material which

722 polymerise upon irradiation, producing blue coloration of the ﬁlm. The radiochromic

723 ﬁlms used in this thesis were the Gafchromic EBT and EBT2 ﬁlms (ISP, 2006, 2009).

724 The EBT2 ﬁlm was rolled out in 2009 replacing its predecessor the EBT ﬁlm. The

725 EBT2 ﬁlm was distinguished from its predecessor (EBT) by its yellow color as well as

726 the ﬁlm construction. The physical structure of the EBT and EBT2 ﬁlm is shown in

727 Figure 3.9. The EBT2 ﬁlm has a nominal thickness of 285 μm and is slightly thicker

728 than the EBT ﬁlm. The advantages of using radiochromic ﬁlms are,

729 (i) It is self developing and does not require the use of a conventional processor.

730 (ii) It has high spatial resolution which is only limited by the scanning resolution.

731 (iii) It is not sensitive to ambient light, hence it is easy to handle in room light.

732 (iv) Film composition has a lower Zeff = 6.84 (ISP, 2009) hence making it more

733 tissue equivalent compared to radiographic ﬁlm. 3.4. Film dosimetry 41

734 (v) It is less energy dependent compared to the EBT ﬁlm (Butson et al., 2010).

735 (vi) Readout can be performed using a professional film digitizer or an office flatbed

736 scanner.

737 (vii) It can be cut into any shapes or sizes.

738 (viii) It can be used for measurements in water.

739 The EBT2 ﬁlm however also has its disadvantages:

740 (i) The ﬁlm readout is non real time due to the eﬀect of post-irradiation coloration.

741 Films are generally readout 24 hrs after exposure (Cheung et al., 2005).

742 (ii) It is susceptive to the scanner uniformity and artefacts (Butson et al., 2003b).

743 (iii) Due to the elongated grains, the scanning orientation is critical to ensure correct

744 dose readout (Butson et al., 2006a).

745 (iv) The ﬁlm is susceptive to ﬁnger prints, dust, lint, scratches and moisture. This

746 will result in artefacts during readout. Therefore, careful handling using gloves

747 and lint free cloths are recommended.

748 (v) The EBT2 film suffers from film inhomogeneity up to 3.7% within a piece of

749 ﬁlm, resulting in dose measurement uncertainty of 8.5% at 1 Gy (Hartmann

750 et al., 2010).

751 In short, radiochromic ﬁlm is a useful comparative dosimeter and is used extensively

752 in this thesis. However, careful and proper handling of the ﬁlm is important to ensure

753 correct and consistent readout. 3.4. Film dosimetry 42

Figure 3.9: Physical structure of Gafchromic (a) EBT and (b) EBT2 ﬁlm (ISP, 2006, 2009). 754 Chapter 4

755 Radiation response and basic

756 characterisation of the Dose

757 Magnifying Glass

758 .

759 4.1 Introduction

760 Silicon diodes are widely used for dosimetry in radiation therapy. As with ion chamber

761 dosimetry, silicon diode dosimeters produce current or charge, which is proportional

762 to the dose rate or accumulated dose respectively. One advantage of silicon detectors

763 for radiotherapy is the constancy of the silicon-water electron stopping power ratio

764 over a wide energy range. However, silicon detectors have some disadvantages due

0Part of this chapter has been published in Medical Physics and Radiation Measurements: Wong, J. H. D., Carolan, M., Lerch, M. L. F., Petasecca, M., Khanna, S., Perevertaylo, V. L., Metcalfe, P. and Rosenfeld, A. B. (2010). A silicon strip detector dose magnifying glass for IMRT dosimetry. Medical Physics 37(2): 427-439. Wong, J. H. D., Cutajar, D., Lerch, M. L. F., Petasecca, M., Knittel, T., Carolan, M., Pere- vertaylo, V. L., Metcalfe, P. and Rosenfeld, A. B. (2011). From HEP to Medical Radia- tion Dosimetry - the silicon strip detector Dose Magnifying Glass. Radiation Measurements. DOI:10.1016/j.radmeas.2011.06.031.(Accepted, 14 June 2011)

43 4.2. Materials and methods 44

765 to their dose rate dependence, angular dependence, energy dependence, and radiation

766 damage. The dependence of the silicon diode on these parameters needs to be properly

767 characterised and understood prior to the clinical use of a silicon based detector.

768 In this chapter, the detector response of the Dose Magnifying Glass (DMG) is

769 characterised using similar tests to those carried out when testing the suitability of

770 detectors as dosimeters in the radiation therapy environment. The parameters tested

771 were:

772 (a) percent depth dose

773 (b) dose per pulse dependency

774 (c) stem eﬀect

775 (d) dose linearity

776 (e) energy response dependency, and

777 (f) angular response dependency

778 4.2 Materials and methods

779 4.2.1 Percent depth dose measurement

780 The percent depth dose measures the dose drop oﬀ as a function of distance. This mea-

781 surement represents one of the principle measurements in determining the suitability

782 of a device as a radiation detector. Various literatures have shown that the depth dose

783 measurement made using silicon diodes agrees well with the ion chamber for most MV

784 photon energy beams. However, due to the non tissue equivalence of silicon, a slight

785 over estimation (<2.5%) is often observed at an increased depth in water/solid water

786 (Wilkins et al., 1997; Westermark et al., 2000; Bucciolini et al., 2003). 4.2. Materials and methods 45

Figure 4.1: Sketch of the solid water encapsulation for the prototype DMG.

787 The prototype DMG was sandwiched between two small slabs of solid water (2.5

3 788 × 5.0 × 0.5 cm ) machined to ﬁt the sensitive area (Figure 4.1). The solid water was

789 then sandwiched inside another larger piece of polymethlmetacrylate (PMMA) (7.0 ×

3 790 15.0 × 3.0 cm ), which was machined to accept the strip detector inside its solid water

791 encapsulation. The strip detector with the PMMA housing was placed on top of a 30

3 792 × 30 × 15 cm solid water block. Sections of solid water and PMMA were used to

793 provide scatter around the detector. The percent depth dose proﬁles were obtained

794 using solid water at the depth of 1.0 cm to 1.5 cm (at 0.1 cm intervals), 5 cm, 10 cm,

2 795 15 cm, and 20 cm for a 10 × 10 cm ﬁeld size and at source to surface distance of 100

796 cm. Photon energy of 6 MV was used. The result was compared with a Farmer ion

797 chamber (NE-2571) in a solid water phantom. The active volume of the Farmer ion

798 chamber has an inner diameter of 6.3 mm and an internal length of 24.0 mm.

799 4.2.2 Dose per pulse response measurement

800 Dose per pulse dependency refers to the change of a detector’s sensitivity due to the

801 change of dose rate in a pulsed radiation beam such as those delivered by a medical

802 linear accelerator (Hoban et al., 1994). The dose during each pulse is suﬃciently

803 high to cause a variation in the silicon detector response. It is therefore important to

804 characterize new detectors for dose per pulse sensitivity. This eﬀect was ﬁrst reported 4.2. Materials and methods 46

805 by Rikner & Grusell (1983). They found that n-type silicon diodes are more sensitive to

806 dose per pulse variation compared to p-type diodes, showing an increase in sensitivity

807 with increased dose per pulse. They also found that preirradiation of the diode also

808 reduces the dose per pulse dependence. The dose within a pulse can be changed by

809 changing the source to detector distance (SDD) and introducing attenuators such as

810 wedges or multileaf collimator (MLC) in the way of the radiation beam. However, by

811 the introduction of attenuators in the beam path, the dose per pulse response of the

812 silicon detector would be inﬂuenced by the alteration of the radiation beam energy

813 spectrum. The study of the dose per pulse response of the DMG was divided into two

814 sections;

815 (i) the eﬀect of preirradiation condition and device resistivity at high dose rates,

816 and

817 (ii) the dose per pulse response of a low resistivity, preirradiated device for a larger

818 dose per pulse range.

819 In the later section, the inﬂuence of the alteration of the radiation beam energy

820 spectrum was also studied. The DMGs investigated in this thesis were produced with

821 two types of resistivity, 10 Ω·cm and 5 k Ω·cm. Of these, some of the DMGs were not

822 preirradiated while the others were preirradiated with 1 MeV electrons up to 15 kGy.

823 4.2.2.1 Eﬀect of device resistivity and preirradiation condition on the dose

824 per pulse response at high dose rates

825 Four DMGs were used to study the eﬀect of device resistivity and preirradiation con-

826 dition on the dose per pulse response.

827 (i) high resistivity, preirradiated,

828 (ii) low resistivity, preirradiated, 4.2. Materials and methods 47

829 (iii) high resistivity, unirradiated, and

830 (iv) low resistivity, unirradiated

831 The DMG was positioned at 9 cm depth in an I’mRT phantom and irradiated with

2 832 a10× 10 cm beam. Dose per pulse variation was obtained by varying the SDD from

833 91 cm to 158 cm. At each position, the detector was irradiated with a 6 MV photon

834 beam for a ﬁxed amount of monitor units (MUs). The dose per pulse range measured

−5 −4 835 was 9.45 × 10 to 2.72 × 10 Gy/pulse.

836 The actual dose at each measurement position was obtained with a CC13 ion

837 chamber (Scanditronix Wellh¨ofer) measurement with a similar set up. The ion chamber

838 dose rate dependence was measured to be <0.4% over the range of dose rates used. The

839 dose rate response was taken as the ratio of the DMG readings over the ion chamber

840 readings.

841 A Varian Clinac 21EX (Varian, Palo Alto, USA) linear accelerator was used in

842 these measurements. At the repetition rate of 600 MU/min, the linear accelerator

843 delivered radiation in short beam pulses of 3 μs at a frequency of 360 Hz. The dose

844 rate was deﬁned by the amount of dose delivered by a radiation beam pulse. This was

845 estimated by dividing the dose measured by the ion chamber with the estimated total

846 number of beam pulses. At the source to detector distance of 100 cm the estimated

−4 847 dose per pulse was 2.28 × 10 Gy/pulse.

848 4.2.2.2 Dose per pulse response of low resistivity, preirradiated device for

849 a large dose per pulse range

850 The low resistivity, preirradiated DMG was used to study the dose per pulse response

−7 −4 851 of the device for a larger dose per pulse range of 6.42 × 10 to 2.92 × 10 Gy/pulse.

852 The DMG was positioned at 1.5 cm depth in solid water phantom and irradiated with

2 853 10 × 10 cm beam. The measurements were made in four conditions; 4.2. Materials and methods 48

2 854 (i) in a 10 × 10 cm open ﬁeld,

855 (ii) with a 35 mm thick low melting point alloy (LMA) beam attenuator attached to

856 the linac accessory slot,

857 (iii) with a LMA and 50 mm lead (Pb) block beam attenuator, and

858 (iv) Under a closed MLC.

859 For each of these setups, measurements were made at several SDDs ranging from

860 88 cm to 150 cm. The radiation beam was delivered using a Siemens Oncor linear

861 accelerator with a 300 MU/min repetition rate. The linac beam pulse period is 4.4

862 ms and the estimated dose per pulse measured at SDD 101.5 cm and 1.5 cm depth in

−4 863 solid water phantom is 2.2 × 10 Gy/pulse.

864 4.2.3 Stem eﬀect measurement

865 Stem eﬀect is the radiation induced conductivity in the cable or stem of the radiation

866 detector when exposed to a radiation beam (Beentjes & Garrett, 1966; Ibbott et al.,

867 1975)

868 The stem eﬀect of this device was investigated by irradiating the DMG with a

2 869 rectangular ﬁeld size of 20 × 5cm at a 1.5 cm depth in solid water phantom. The

870 rectangular field size was chosen so that the narrow width of the field size was sufficient

871 to cover the detector area while avoiding the cable connectors when the long axis

872 was orthogonal to the cables. Measurements were made with the long axis of the

873 ﬁeld parallel and then perpendicular to the cables. The diﬀerence between the two

874 measurements was deemed to be due to the stem eﬀect. 4.2. Materials and methods 49

875 4.2.4 Dose linearity measurement

876 Dose linearity measures the linearity of the detector signal as a function of dose. A

877 good detector would require the detector’s response to be linearly proportional to a

878 given dose. A linear dose response also reduces the likelihood of errors and enables

879 faster processing (Metcalfe et al., 2007).

880 The DMG was inserted into the I’mRT phantom (IBA Dosimetry) and positioned

881 at 10 cm depth. PMMA sections were machined speciﬁcally to pack the detector

882 snugly in the I’mRT phantom. The set up was irradiated with 6 MV photons from

883 5 to 400 monitor units (MUs) delivered with a 400 MU/min repetition rate. The

884 corresponding doses delivered to the detector ranged from 3.89 cGy to 311.05 cGy.

885 4.2.5 Energy response measurement

886 Due to the higher atomic number of silicon, silicon diodes tend to display an enhanced

887 response to photoelectric eﬀect at low photon energy <150 keV. This is usually mani-

888 fested in the over-estimation of dose at greater depths in water or solid water phantom

889 and with larger ﬁeld sizes (Westermark et al., 2000; Yin et al., 2004). The energy de-

890 pendence of a silicon diode may also be due to the material around the diode such

891 as the electrode attachment, protective housing and build up material which contain

892 high Z materials (Saini & Zhu, 2007; Jursinic, 2009).

893 To study the energy response of the DMG, the strip detector was irradiated under

894 an orthovoltage machine (Guymay DX 3300) which operated at the range of 50 kV

895 - 250 kV. The orthovoltage machine was calibrated following IPEMB 1996 protocol

896 (IPEMB et al., 1996; IPEM et al., 2005). The absolute calibration accuracy is 1.9%

897 whereas the reproducibility of the calibration is better than 0.5%. The detector was

898 irradiated without any build up. For 50 - 150 kV tube voltage, a 100 mm diameter

899 circular applicator was used and the detector was placed at a focus to surface distance 4.2. Materials and methods 50

900 of 312 mm. For the 200 and 250 kV tube voltage, a 100 mm diameter circular applicator

901 was used with the detector positioned at focus to surface distance of 512 mm. For

902 high energy photons, the detector was irradiated using 6 MV and 10 MV beams from

903 a Varian Clinac 21EX linear accelerator (Varian, Palo Alto, USA) (dmax of 1.5 cm for

904 6 MV and 2.0 cm for 10 MV). The linac was calibrated according to IAEA TRS-398

905 protocol (IAEA, 2006).The relative standard uncertainty of the absolute calibration

906 is approximately 1.7%. 1 Gy water equivalent dose was delivered at all energies. The

907 readings are normalized to 1 at the energy of 6 MV.

908 4.2.6 Angular response measurement

909 Most silicon diodes display angular dependence (Westermark et al., 2000; Higgins

910 et al., 2003) due to the asymmetry in their mounting using high Z elements (Cu),

911 anisothropy in the silicon substrate and the packaging of the detector. Methods have

912 been devised to improve the angular dependence of the silicon diodes (Jursinic, 2009).

913 The angular response of the DMGs was studied to determine the eﬀect of detector

nd 914 packaging for the prototype version and the 2 generation DMG.

915 4.2.6.1 Prototype DMG

916 The design of the prototype detector with its ceramic mounting is expected to cause

917 some angular response to the DMG. The angular response of the prototype DMG was

918 studied in the I’mRT phantom. The whole DMG was inserted into the middle of the

919 I’mRT phantom and positioned at the isocenter. Radiation beams with the ﬁeld size

2 ◦ 920 of 10 × 10 cm were delivered with static gantry angles at the intervals of 15 from

◦ ◦ 921 0 to 90 . The readings were then compared with CC13 ion chamber measurements

922 at the isocenter and at the positions ± 1 cm lateral from the isocenter. The angular

923 response is deﬁned as the ratio of the signal measured by the DMG to the ion chamber 4.2. Materials and methods 51

924 measurements at each point of measurement. For the DMG measurement, the average

925 reading was taken of the detector channels within the distance of ± 1.5 mm from the

926 measurement point. This width is comparable to half of the inner diameter of the

927 CC13 ion chamber. The angular response taken at the 1 cm spatial distance from

928 the center of the detector enabled the investigation of the response of the detector

929 channels as a function of the distance from the center channel.

nd 930 4.2.6.2 2 generation DMG

nd 931 The packaging of the 2 generation DMG was modiﬁed by mounting the silicon wafer

932 on a 0.12 mm thick Kapton substrate. The thin Kapton substrate has a density of

3 933 1.42 g/cm , and has the eﬀective atomic number, Zeff = 6.6 (Hubbell & Seltzer,

934 2004). This is more tissue-equivalent (Zeff = 7.5 for soft tissue) than ceramic (Al2O3,

935 Zeff = 11.2) and is expected to create less perturbation to the radiation beam. It

936 is therefore expected to reduce the angular dependency of the detector due to the

937 detector packaging. The whole detector with the Kapton was encapsulated in a solid

938 water holder which was machined to ﬁt snugly with the device, reducing the presence

939 of air gaps between the solid water holder and the detector (Figure 4.2). The angular

nd 940 response of the 2 generation DMG was investigated in a custom made phantom (SRS

941 phantom) which has a cylindrical body and a hemispherical head. The custom made

942 phantom design will be described in detail in chapter 6. The detector array is mounted

943 on the end of a 170 mm long Kapton pigtail. This enables the positioning of the silicon

944 detector in phantoms to provide for suﬃcient scattering. The angular response of the

945 DMG mounted on Kapton substrate was investigated in two orthogonal directions,

946 i.e. the azimuth angles and polar angles. Refer to Figure 4.3 for pictorial deﬁnition of

947 polar and azimuth angles.

948 For the angular response measurements, the DMG was positioned at SDD 100 cm,

949 9 cm depth of the SRS phantom. For azimuth angle measurements, the couch angle 4.3. Results 52

◦ 2 950 was set to 0 andasmallﬁeldsizeof2.5× 2.5 cm was used. The gantry was rotated

◦ ◦ ◦ 951 from 0 to 350 with stepping angles of 10 intervals and 100 MUs were delivered at

952 each angle.

◦ 953 For measurement of polar angles, the couch was rotated to 270 . Due to the

954 design of the detector mounting on one end of the Kapton pigtail and the logistics

955 of the readout connection/cables at the other end of the Kapton pigtail, polar angle

◦ ◦ 956 measurements were limited to gantry angles of 320 to 180 , clockwise. Measurements

◦ 2 957 were taken in 10 increments, with a 2.5 × 2.5 cm radiation ﬁeld size.

958 The DMG detector response was taken as the average response of the silicon strips

959 located within ± 1 mm of the axis of rotation. The angular dependence of the DMG

960 was taken as a ratio of the DMG response at diﬀerent angles to the DMG response

961 when the radiation beam was perpendicular to the detector plane, corresponding to

◦ 962 the gantry angle 0 (Eq.4.1).

Rθ Angular dependence, Aθ = (4.1) R◦

◦ 963 where, Rθ is the response at gantry angle θ and R◦ is the response at gantry angle

◦ 964 0 .

965 4.3 Results

966 4.3.1 Percent depth dose

967 Figure 4.4 shows the measured percent depth dose proﬁle using the strip detector

968 compared with a Farmer ion chamber (NE-2571) measurement, using solid water for a

2 969 10 × 10 cm ﬁeld size at a source to surface distance of 100 cm. The photon energy was

970 6 MV. The DMG was corrected for the dose per pulse eﬀect using the method described

971 in section 4.3.2.3. The resulting depth dose curve agrees with the Farmer ion chamber 4.3. Results 53

Figure 4.2: The 2nd generation DMG in the solid water holder and the custom made phantom which has a cylindrical body and hemispherical head.

Figure 4.3: Deﬁnition of polar and azimuthal angle. 4.3. Results 54

Figure 4.4: Depth dose curve for a 10 × 10 cm2 ﬁeld size of a 6 MV photon beam comparing DMG with Farmer ion chamber (NE-2571) measurements in solid water.

972 measurements within 0.8% up to 20 cm depth in solid water. The over-response of

973 the DMG measurements at greater depths may be due to the energy response of the

974 silicon detector with lower energy electrons. The D20/10 for the strip detector and the

975 Farmer ion chamber was 0.573 and 0.569, respectively.

976 4.3.2 Dose per pulse response

977 4.3.2.1 Eﬀect of device resistivity and preirradiation condition on the dose

978 per pulse response at high dose rates

979 Figure 4.5 shows the dose rate dependence (expressed as dose per pulse) of the detector

−4 980 normalized to the dose per pulse of 2.28 × 10 Gy/pulse measured at 100 cm SDD.

981 By changing the SDD, the dose rate was varied by a factor of 2.9 (corresponding to

−5 −4 982 dose per pulse of 9.45 × 10 to 2.72 × 10 Gy/pulse). The low resistivity devices

983 appear to have less dose rate dependence compared to the high resistivity devices

984 whilst preirradiating the detector also improved the dose rate dependence. This is in

985 agreement with the published literature (Rikner & Grusell, 1983; Jursinic, 2009). The 4.3. Results 55

Figure 4.5: Dose per pulse response of selected DMGs with diﬀerent resistivity and preirradiation conditions for the dose per pulse range of 9.45 × 10−5 to 2.72 × 10−4 Gy/pulse. The error bars represent the 1 standard deviation of the mean of all the channels in the strip detector.

986 unirradiated, high resistivity device showed the largest dose rate dependence of 8.5%

987 while the preirradiated, high resistivity device showed a slight improvement of 2%.

988 For the low resistivity, unirradiated devices, the maximum dose rate change was found

989 to be 4.1% while its preirradiated counterpart showed a maximum variation of 1.1%.

990 4.3.2.2 Dose per pulse response of a low resistivity, preirradiated device

991 for a large dose per pulse range

992 Figure 4.6 shows the dose per pulse response of a low resistivity, preirradiated device for

−7 −4 993 a dose per pulse range of 6.42 × 10 to 2.92 × 10 Gy/pulse, a dose per pulse change

994 of 454-fold. This large dose per pulse range was achieved by placing beam attenuators

995 such as LMA, Pb block and closing the MLC. For each setup, measurements were

996 made at multiple SDDs creating a dynamic range of dose per pulse that overlapped

997 at either ends. 4.3. Results 56

−4 998 For the dose per pulse >1.01 × 10 Gy/pulse, this device appears to be dose

−5 −5 999 per pulse independent. For the dose per pulse range of 3.92 × 10 to 2.66 × 10

1000 Gy/pulse, the DMG’s sensitivity increases as the dose per pulse increases. At the lower

1001 end of this range (between the measurements setup using the LMA + Pb block and

−5 1002 closed MLC), there is a discontinuity of the data points. At 6.0 × 10 Gy/pulse, the

1003 normalised ratio of DMG/ion chamber gives 0.758 with the LMA + Pb block whilst

1004 under closed MLC condition, the DMG/ion chamber ratio is 0.364. This may indicate

1005 that with the closed MLC setup, there is a signiﬁcant change in the beam spectrum

1006 aﬀecting the silicon response. By changing the SDDs, the dose per pulse was changed

1007 by inversed square distance with the source; the beam spectrum remains the same.

1008 However, when attenuators were introduced, not only was the dose per pulse reduced

1009 but the radiation beam becomes harder as most of the soft x-rays were absorbed by

1010 the MLC. Silicon diode are known to over response with low energy photons, hence

1011 the hardening of the beam resulted in a generally much lower signal to the silicon.

1012 The dose per pulse response under closed MLC was 58%± 6%. This only aﬀects

1013 the measurements when the radiation is highly attenuated, particularly the out of

1014 ﬁeld dose. However, it should be noted that the dose under a closed MLC should

1015 comprised of leaf transmission leakage (which is usually in the order of 2% of an open

1016 ﬁeld (Metcalfe et al., 2007)) and scattered dose from the phantom.

1017 The error bars represent the 1 standard deviation of the mean of all the channels

1018 in the strip detector. The large error bars for the LMA and Pb blocks were partly

1019 attributed to the slight inhomogeneity of the LMA block across the silicon strips,

1020 while the large error bars in the measurements under closed MLCs may be due to

1021 the combined eﬀect of the low signal to noise ratio and the MLC leakage through the

1022 tongue and groves. 4.3. Results 57

Figure 4.6: Dose per pulse response for a low resistivity, preirradiated device for the dose per pulse range of 6.42 × 10−7 to 2.92 × 10−4 Gy/pulse. The dose per pulse responses were normalised to the dose per pulse of 2.2 × 10−4 Gy/pulse.

1023 4.3.2.3 Dose per pulse correction

1024 From the dose per pulse response measured in the previous sections, a correction

1025 function can be derived to correct for the dose per pulse dependence. This is achieved

1026 by fitting a best fit curve to the data points. The dose per pulse effect (α) is a function

1027 of the measured signal (Ni) to the signal at a calibrated dose per pulse, e.g. 2.2 ×

−4 1028 10 Gy/pulse (N◦) ratio (Eq.4.2).

N Dose per pulse eﬀect, α = f i (4.2) N◦

1029 The corrected signal would be (Eq.4.3),

1 Corrected signal, Nc = Ni × (4.3) α Ni/N◦

1030 Figure 4.7 shows the dose per pulse corrected response. The dose per pulse eﬀect

1031 was obtained using an exponential curve ﬁt. However, it was noted that the correction 4.3. Results 58

Figure 4.7: Dose per pulse corrected response.

1032 for the dose under the closed MLC was less satisfactory. The percent depth dose curve

1033 and use of the DMG in IMRT dosimetric veriﬁcation (chapter 5) was corrected for the

1034 dose per pulse eﬀect.

1035 4.3.3 Stem eﬀect

1036 When the DMG was irradiated with the long axis of the ﬁeld parallel to the cables,

1037 the increase in detector measurement was found to be 0.2%± 0.1%. This shows that

1038 the stem eﬀect is quite negligible for a clinically useful ﬁeld size.

1039 4.3.4 Dose linearity

1040 The strip detector response is linear with dose range of 3.89 cGy to 311.05 cGy. The

1041 dose linearity veriﬁcation was carried out at this range because it was deemed to be

1042 within the range of a normal IMRT dose per fraction. This is by no means the dynamic

1043 range limit of the detector as the acquisition rate of the detector readout system can

1044 be varied to accommodate large counts. For this measurement conﬁguration, the slope 4.3. Results 59

Figure 4.8: Dose linearity of the DMG.

1045 of the best linear ﬁt is 1597 counts/cGy with a regression coeﬃcient of 1.00 (Figure

1046 4.8).

1047 4.3.5 Energy response

1048 The energy response of the strip detector was studied using an orthovoltage unit and a

1049 linear accelerator. Figure 4.9 presents the energy response curve for the range of 50 kV

1050 to 10 MV nominal energies (corresponding to 26.8 keV - 2.97 MeV equivalent photon

1051 energies), normalized to 1 at 6 MV nominal photon energy. Error bars represent the

1052 combined uncertainties of the 95% CI of the mean for the measurements and the 0.5%

1053 calibration uncertainty. The ﬁlters for diﬀerent nominal energies are made up of a

1054 combination of copper and aluminum ﬁlters. The photon equivalent energies for the

1055 orthovoltage energies were derived from the half value layers of the ﬁlters (Johns &

1056 Cunningham, 1983). For 6 MV and 10 MV photon beam, the mean energies were

1057 taken from Mohan et al. (1985) Monte Carlo results. The photon energy response

1058 curve showed an enhanced response at low energies up to six times of the response 4.3. Results 60

Figure 4.9: Energy response of the strip detector, normalized to 1 at a 6 MV photon energy. Error bars represent the combined uncertainties of the 95% CI of the mean for three sets of measurements and the 0.5% calibration uncertainty.

1059 at 6 MV. The detector showed an over-response at lower photon energies with the

1060 maximum dose response at 75 kV nominal photon energy. This is due to the increased

1061 photoelectric eﬀect cross-section in silicon at low energies (Meyer et al., 2001). The

1062 energy dependence of the detector in free air geometry was expected as silicon is not

1063 tissue equivalent and therefore sensitive to changes in the energy spectrum. This is

1064 consistent with the literature (Yin et al., 2004; Van Dam et al., 2005; Metcalfe et al.,

1065 2007). The eﬀect of the energy dependence of this silicon detector in megavoltage

1066 beams will not be as signiﬁcant as for measurements under the orthovoltage beams

1067 (Metcalfe et al., 2007). 4.3. Results 61

1068 4.3.6 Angular response

1069 4.3.6.1 Prototype DMG

◦ ◦ 1070 The angular response dependency of the DMG for the incident beam angle of 0 -90

1071 was studied. There was an increased under-response measured by the strip detector

1072 compared to the dose measured by CC13 ion chamber as the incident beam angle

◦ ◦ 1073 changes from 0 to 90 as showed in Figure 4.10. The angular response correction

1074 factor was derived by normalizing the measured dose of DMG to the dose measured

1075 by the CC13 ion chamber. Figure 4.10 shows the mean angular response of three

1076 points of measurement with a spatial distance of 1 cm from each other. The DMG

1077 readings were taken as an average of the channels located within ± 1.5 mm of the

1078 measurement points of 2.8, 12.8 and 22.8 mm along the length of the detector. The

1079 angular response at the point of ± 1 cm away from the center is within 1% of the center.

1080 Therefore, it was deemed justiﬁed to apply the center angular correction factor onto

1081 all 128 detectors (strips) in the IMRT measurements. This is used as a ﬁrst attempt

1082 to correct for the angular response dependence.

1083 Angular response for the centre channels of the strip detector was within 3.1%±

◦ ◦ 1084 0.1% for angle 0 -45 . The largest angular response (28.1%± 0.1%) was found to be at

◦ 1085 the gantry angle of 90 where the beam was parallel to the detector. The uncertainty

1086 reported was the standard error of the mean for the channels located within ± 1.5 mm

1087 of the measurement point and the ion chamber measurements. For this particular set

1088 up the measurement uncertainty for the measurement was <0.5% (standard error).

1089 The observed angular dependence may arise from the combined eﬀects of the linear

1090 air cavity contained within the phantom, the artifact created by ceramic substrate

1091 mounting, and the anisotropies inherent in the silicon chip. The investigation to

1092 minimize the angular dependence in future prototypes by eliminating the small air

1093 volume surrounding the silicon strip detector as well as mounting of the strip detector 4.3. Results 62

Figure 4.10: Angular response correction factor generated as the measured dose by DMG/CC13 ion chamber measurements. As the angle of the incident beam increased from perpendicular beam (θ =0◦)toparallelbeam(θ =90◦), the strip detector under- responded compared to the dose measured by CC13 ion chamber. The maximum angular response 28.1% ± 0.1% was found to be at the parallel incident beam at the gantry angle of 90◦. The diﬀerence between the response of the detector channels at the position of ± 1 cm away from the central axis (mid channels) were <1%.

1094 on Kapton substrate, eliminating the attenuation eﬀect of the ceramic substrate was

1095 carried out and detailed in section 4.3.6.2.

nd 1096 4.3.6.2 2 generation DMG

1097 Figure 4.11 shows the polar and azimuth angular dependency of the DMG. The angular

1098 response was deﬁned as the mean value of all channels within 1 mm from the center

1099 of the detector. The maximum polar angular response was 13.8%± 0.1% at gantry

◦ ◦ ◦ 1100 angle 320 (at couch position 270 ). If the couch position is at 90 , the proﬁle of the

◦ ◦ 1101 polar angular response would be mirrored on the graph’s 0 - 180 axis. The apparent

◦ ◦ 1102 under-response in polar angular response at gantry angles 320 - 350 may be due

1103 to the cylindrical portion of the phantom rather than the intrinsic detector response. 4.4. Conclusion 63

Figure 4.11: Angular response for the 2nd generation DMG.

◦ ◦ 1104 The maximum azimuthal angular response was at gantry angles 90 and 270 ,withan

1105 angular response deviation from unity of 15.3%± 0.1%. Compared to the prototype

1106 version, the maximum azimuthal angular response of the DMG mounted on 0.12 mm of

1107 Kapton substrate was reduced by 12.8%. The polar and azimuthal angular response

◦ 1108 deviation from unity at gantry angle 180 was 6.3%± 0.2%. The resulting angular

1109 response of the DMG is due to the inherent anisotropy of the silicon detector.

1110 4.4 Conclusion

1111 The radiation response and basic characteristics of the Dose Magnifying Glass was

1112 described. The percent depth dose proﬁle for a 6 MV photon energy matches closely

1113 with the measurement with Farmer ion chamber (within 0.8%) up to 20 cm depth in

1114 solid water. The dose per pulse response of four DMGs were studied showing that

1115 low resistivity and preirradiation device is the least dose rate dependent. The use of

1116 attenuators such as MLC will result in much lower detector sensitivity (58%) due to 4.4. Conclusion 64

1117 beam hardening eﬀect. The stem eﬀect of the detector was found to be negligible. The

1118 linearity of the detector with dose is excellent for the dose range of 3 cGy to 300 cGy

1119 typical for a one fraction IMRT delivery. The strip detector showed the typical energy

1120 dependency intrinsic for silicon at lower photon energies in free air geometry. Angular

◦ ◦ 1121 response of the strip detector is within 3.1%± 0.1% for angle 0 -45 while for angle

◦ 1122 90 the response was 28.1%± 0.1%. It is associated with the ceramic packaging of the

nd 1123 detector as well as the inherent anisotropies in the silicon. For the 2 generation DMG,

◦ 1124 the maximum polar angular response was 13.8%± 0.1% at gantry angle 320 when the

◦ 1125 couch position was at 270 . This may be due mainly to geometrical artifacts, rather

1126 than the detector’s response. The maximum azimuth angular response was 15.3%±

◦ ◦ 1127 0.1% at gantry angles 90 and 270 . The improved angular response is partly due to

1128 better detector packaging, which was mounted on a 0.12 mm Kapton substrate. 1129 Chapter 5

1130 Application of the Dose Magnifying

1131 Glass in the dosimetric veriﬁcation

1132 of an intensity modulated radiation

1133 therapy treatment delivery

1134 5.1 Introduction

1135 Intensity modulated radiation therapy (IMRT) enables the delivery of escalated radi-

1136 ation dose to tumor while sparing the critical organs close to the irradiated target. In

1137 doing so, IMRT plans tend to produce steep dose fall oﬀ regions at the interface of

1138 the target and organ-at-risk. Currently, quality assurance (QA) veriﬁcation of IMRT

1139 plans are patient based and the techniques used commonly are based on point dose

1140 measurement using ion chambers and planar fluence verification using films, electronic

0This chapter has been published in Medical Physics: Wong, J. H. D., Carolan, M., Lerch, M. L. F., Petasecca, M., Khanna, S., Perevertaylo, V. L., Metcalfe, P. and Rosenfeld, A. B. (2010). A silicon strip detector dose magnifying glass for IMRT dosimetry. Medical Physics 37(2): 427-439.

65 5.2. Materials and methods 66

1141 portal imaging devices (EPIDs), and two dimensional diode or ion chamber arrays.

1142 The ion chamber, though highly reliable, tends to overestimate the penumbra width

1143 due to its larger volume, therefore may not accurately represent areas with high dose

1144 gradient fall oﬀ (Bucciolini et al., 2003). The use of ﬁlms involved calibration of the

1145 ﬁlm and ﬁlm scanners which delays the QA process. The use of EPIDs as a QA tool

1146 require conversion of image to dose whereas the 2D and 3D diode and ion chamber

1147 arrays currently are limited by their spatial resolution in high dose gradient regions

1148 (Spezi et al., 2005; Poppe et al., 2007). Not withstanding their limitations, these ex-

1149 isting techniques are widely used in many facilities. In order to measure high gradient

1150 dose regions generated by IMRT plans, one needs to have a detector with high spatial

1151 resolution. Hence, the sensitive volume of the detector has to be small in physical

1152 size (Kron et al., 1993) In this chapter, the concept of a silicon strip detector (Dose

1153 Magnifying Glass [DMG]) with high spatial resolution suitable for measuring steep

1154 dose gradient areas such as those formed in an IMRT plan is presented. A uniformity

1155 correction method is proposed and the use of the DMG for penumbra measurements

1156 and application in a clinical IMRT delivery is described.

1157 5.2 Materials and methods

1158 5.2.1 Dose Magnifying Glass

1159 The DMG that was used in this work is the prototype DMG which was mounted on

+ 1160 an alumina substrate. The sensitive area deﬁned by a single n strip is 20 × 5000

2 1161 μm . The resistivity of the silicon is 10 Ω·cm and it was not preirradiated. 5.2. Materials and methods 67

1162 5.2.2 Uniformity measurement

1163 The DMG consisted of a strip silicon detector with 128 pixels. The response from

1164 each detector and its charge ampliﬁer may diﬀer slightly from that or the adjacent

1165 detectors. Hence, taking the raw readings from the detector one would expect slight

1166 non-uniformity between the 128 detector/readout channels. A uniformity correction

2 1167 wasperformedusinga6MVbroadbeamirradiationofa10× 10 cm ﬁeld size at 1.5

1168 cm depth in phantom. There were also some channels which did not operate due to the

1169 PCB circuit board manufacturing imperfection. Data from the non-operational chan-

1170 nels was removed prior to the analysis. For each experiment, the DMG is positioned

1171 at the central axis of the beam and irradiated. The raw data is ﬁrst inspected visually

1172 and the defective channels removed from consideration. The remaining channel read-

1173 ings were then averaged to obtain a mean reading,x ¯. Then, each individual channel’s

1174 reading, xi is normalized tox ¯, to give the response of the channel and thus creat-

1175 ing a correction factor (ui =xi/x¯ ), assuming ﬂatness of radiation ﬁeld. This factor

1176 was applied on all the subsequent measurements to correct for the channel-to-channel

1177 variation in detector and ampliﬁer response.

1178 5.2.3 Penumbra response measurement

2 1179 The penumbra of a 10 × 10 cm ﬁeld size was measured at the dmax (1.5 cm) and

1180 10 cm depth in solid water. A reading was ﬁrst acquired with the detector placed on

1181 the central axis of the beam. A second reading was then acquired with the detector

1182 placed at the ﬁeld edge, with the penumbra positioned at the middle of the strip.

1183 The readings are then normalized to the average readings measured at the central

1184 axis. This measurement technique was necessary because the prototype strip detector

1185 is only 2.56 cm wide and is not able to measure the dose at the central axis as well

1186 as the ﬁeld edge, simultaneously. It is envisaged that clinical strip detector systems 5.2. Materials and methods 68

1187 will incorporate multiple adjacent 128 detector elements allowing the instantaneous

1188 measurement of a full beam proﬁle in a single exposure. Penumbra measurements

1189 were obtained for the secondary collimator jaws and the rounded leaf end of the linac

1190 multileaf collimator (MLC). The measurements are then compared with measurements

1191 in water using Gafchromic EBT (ISP, Wayne, NJ) ﬁlms in a water tank and solid water

1192 phantom.

1193 5.2.4 Clinical application in IMRT ﬁelds

1194 A prostate IMRT plan was selected and copied onto the planning CT images of the

1195 I’mRT phantom and recalculated using a Pinnacle treatment planning system (TPS)

3 1196 with a dose grid of 2 mm . The isocenter of the IMRT plan was purposely shifted

1197 4 cm laterally to obtain a high gradient fall oﬀ at the position of the detector. The

1198 treatment plan was exported to the Aria R & V system. The I’mRT phantom was

1199 set up with the strip detector sandwiched in the middle (9 cm depth). The isocenter

1200 was shifted accordingly and treated. Measurements were also taken with Gafchromic

1201 EBT ﬁlm sandwiched at the depth of 9 cm in the I’mRT phantom to be compared

1202 with the DMG measurements. The EBT ﬁlm was scanned with a ﬂatbed scanner

1203 (Epson Perfection V700) after 24 hours to allow for post-irradiation coloration of the

1204 ﬁlm (Cheung et al., 2005). The ﬁlms were scanned at the scanning resolution of 150

1205 dpi in 48-bit RGB mode and analyzed using image analysis software, Image J 1.39u

1206 (National Institute of Health, USA). Care was taken to scan the ﬁlms at the center of

1207 the scanner and in the same orientation to avoid scanner-induced non-uniformity and

1208 change of pixel values due to ﬁlm scanning orientation (Butson et al., 2006a; Fiandra

1209 et al., 2006). Standard calibration ﬁlms were irradiated in solid water in the same

1210 experimental session. The results from the measurements were compared with the

1211 Pinnacle predicted dose and EBT ﬁlm measurements. 5.3. Results and discussions 69

1212 5.3 Results and discussions

1213 5.3.1 Uniformity

1214 Figure 5.1 shows the detector response by each of the detector channels before and

1215 after uniformity correction normalized to 1 at the average of the channels. The results

1216 presented were the mean of three measurements and error bars represent the 95%

1217 conﬁdence interval (CI) of the mean. The 95% CI was derived by taking the mean

1218 plus and minus 95% conﬁdence coeﬃcient (for 2 degrees of freedom) multiplied by the

1219 standard error of the mean (Dawson & Trapp, 2004). 98% of the working channels

1220 has an uncertainty of less than ± 2% (including set up uncertainties). This represents

1221 the reproducibility uncertainty of the DMG measurements. Before correcting for the

1222 detector response, the coeﬃcient of variation between the detector channels is within

1223 2% whereas after the application of the correction factor, the coeﬃcient of variation

1224 is <0.2%. It is essential to perform a broad beam irradiation to obtain the calibration

1225 factor of the individual channels.

1226 5.3.2 Penumbra measurement

2 1227 The 80-20% penumbra width of a 10 × 10 cm ﬁeld size for the 6 MV beam at

1228 1.5 cm and 10 cm depth was measured using the strip detector and comparison was

1229 made against measurements from Gafchromic EBT ﬁlms (Figure 5.2). The EBT ﬁlm

1230 measurements were performed in a water tank for X-jaw penumbra measurements

1231 while the multileaf collimator (MLC) rounded leaf end penumbra was measured in

1232 solid water phantom. The strip detector and EBT ﬁlm measured the penumbra at

1233 1.5 cm depth for the secondary jaw to be 2.77 ± 0.04 mm and 2.71 ± 0.13 mm,

1234 respectively. Cheung et al. (2006) measured a 80-20% penumbra width of 3.0 ± 0.2

1235 mm at 1.5 cm depth using the EBT ﬁlm. Compared to their measurements, the strip 5.3. Results and discussions 70

Figure 5.1: Detector response before and after uniformity correction. The error bars represent the 95% CI of the mean for three sets of measurements.

1236 detector seems to be able to resolve a narrower penumbra width. For the MLC rounded

1237 leaf end, the strip detector measures a 3.52 ± 0.04 mm penumbra width compared to

1238 the 4.6 ± 0.3 mm MLC penumbra width measured using EBT ﬁlms by Butson et al.

1239 (2003a). Penumbra width measured at 10.0 cm depth are 3.94 ± 0.04 mm for X-jaw

1240 and 5.60 ± 0.04 mm for the rounded left end of the MLC. This is again smaller than

1241 the measurement with the EBT ﬁlms for the X-jaw of 4.50 ± 0.15 mm (Kron et al.,

1242 1993) investigated the 80-20% penumbra width using thermoluminescent dosimetry

1243 extrapolating to inﬁnitesimal small detector found the penumbra width to be 3.2 mm

1244 at 1.5 cm depth and 4.2 mm at 10.0 cm depth. The results are summarized in Table

1245 5.1.

1246 5.3.3 Clinical application in IMRT ﬁelds

1247 A prostate IMRT plan was delivered with a Varian Clinac 21EX linear accelerator 6

◦ ◦ ◦ ◦ ◦ 1248 MV photon beam. The beam angles used were 0 ,45,90, 270 , and 315 ; delivering 5.3. Results and discussions 71

Figure 5.2: Dose profiles showing the penumbra of a 6 MV beam at 1.5 cm depth for the secondary X-jaw and the rounded leaf ends of a multileaf collimator (MLC) using the strip detector and Gafchromic EBT films. The error bars for the EBT profiles represent the 95% CI of the mean of three sets of measurements while the error bars for the DMG measurements represents the 2% reproducibility uncertainty.

Table 5.1: Comparison of the 80-20% penumbra width measurements for a 6 MV beam at 1.5 cm and 10.0 cm depth between the DMG, Gafchromic EBT ﬁlms, and other published literature. The uncertainty reported represent the 95% CI of the mean for three sets of measurements. DMG (mm) EBT ﬁlm (mm) Others (mm) X-Jaw (Symmetric) at 1.5 cm 2.77±0.04 2.71±0.13 3.0±0.2 a depth MLC (Symmetric) at 1.5 cm 3.52±0.04 4.08±0.03 4.6±0.3 b depth X-Jaw (Symmetric) at 10.0 cm 3.94±0.04 4.50±0.15 4.2±0.3 c depth MLC (Symmetric) at 10.0 cm 5.60±0.04 6.61±0.16 ··· depth

a(Cheung et al., 2006) b(Butson et al., 2003a) c(Kron et al., 1993) 5.3. Results and discussions 72

Figure 5.3: IMRT dose proﬁles for each gantry angle and the total summation of all the beams. The error bars represent the reproducibility uncertainty of the measurements at the level of 95% CI of the mean.

1249 a total of 596 MUs. The isocenter of the prostate IMRT plan was purposely shifted 4

1250 cm to one side so that the DMG was positioned at an area of steep gradient. Figure

1251 5.3 shows the DMG measured dose proﬁles obtained from individual beam angle and

1252 the total summation of all the beam angles. The measured dose was corrected for

1253 uniformity, dose per pulse eﬀect, and angular response dependency.

1254 Figure 5.4 shows the comparison of dose proﬁles from the DMG, Gafchromic EBT

1255 ﬁlm and Pinnacle predicted dose. The error bars represent the reproducibility uncer-

1256 tainty of the measuring device, which are in the order of ± 2% for DMG and ± 3% for

1257 EBT ﬁlm. The Pinnacle TPS calculated the doses using collapsed cone convolution

3 1258 dose engine (Mackie et al., 1985; Ahnesjo, 1989) with a dose grid of 2 mm .The

1259 Pinnacle predicted dose points were taken at every 1 mm interval across the length of

1260 2.56 cm. The average diﬀerence between DMG versus Pinnacle and the DMG versus

1261 EBT ﬁlm were 1.1%± 1.8% (1 s.d.) and 1.0%± 1.6% (1 s.d.), respectively. The large

1262 difference between the DMG dose profiles with the Pinnacle and EBT dose profiles 5.3. Results and discussions 73

Figure 5.4: Comparison between Pinnacle predicted dose profiles, measurements with Gafchromic EBT film, and DMG. The error bars represent the reproducibility of the measurements which are in the order of ± 2% for DMG and 3% for EBT film, respectively. On average, the difference between the dose measured using the DMG with the dose points from the Pinnacle dose matrix and the dose measured with the EBT film were 1.1%± 1.8% (1 s.d.) and 1.0%± 1.6% (1 s.d.).

1263 are mainly at the area of steep gradient. The presence of the steep dose gradient was

1264 detected by the EBT ﬁlms. Within the uncertainty of measurement as well as the

2 1265 positioning uncertainty of 1 mm , the three proﬁles agree quite well with each other.

1266 The IMRT plan delivered was a step and shoot IMRT plan. The IMRT segments

1267 in each of the ﬁelds are quite discernable in this measurement due to the time gap

1268 for the MLC to move to position between each segment. However, due to the small

1269 detector size (2.56 cm), the detector may miss out on some segments if they do not

1270 coincide with the location of the detector. The high spatial resolution of the detector

1271 together with the high temporal resolution of the readout and data acquisition system

1272 enabled the measurement of the temporal variation of the dose modulation within a

1273 single IMRT segment. Figure 5.5 (a) - (d) shows the unique capability of this detector

◦ 1274 system. Figure 5.5 (a) shows a single step and shoot IMRT ﬁeld (gantry angle 315 ). 5.3. Results and discussions 74

Figure 5.5: (a) 3D dose profiles of a single step and shoot IMRT field, depicting the modulation of the dose during each segment, within the width of 2.56 cm. (b) Dose profiles of a full IMRT delivery, each gantry angle and modulation in each IMRT segments can be shown temporally. The measurements were performed with the acquisition pulse width of 0.1 s. (c) The temporal cumulative 3D dose profiles of a full IMRT delivery. (d) Temporal dose rate pattern of an IMRT delivery for channel no. 8 and channel no. 127.

1275 IMRT modulation of each segment is clearly depicted. The time for the delivery of each

1276 segment varies but within each segment the beam proﬁle is constant and reproducible.

1277 Figure 5.5 (b) shows the 3D (dose , time , location) dose proﬁles of a full IMRT

1278 delivery, plotted temporally. A total of 1900 samples were acquired at the pulse width

1279 resolution of 0.1 s. Figure 5.5 (c) showed the build up of the cumulative dose, i.e. the

1280 temporal dose growth of the same IMRT delivery.

1281 Various published literature have stressed the importance of the temporal dose

1282 resolution of IMRT as well as it’s eﬃcacy in tumor cell kill (Benedict et al., 1997; Wang 5.3. Results and discussions 75

1283 et al., 2003; Fowler et al., 2004; Altman et al., 2006; Moiseenko et al., 2007). Insofar as

1284 reported in the literature, the postulation of the reduced eﬀect of tumor cell kill due to

1285 protracted IMRT delivery time was drawn from theorectical calculation and in-vitro

1286 cell studies. Fowler et al. (2004) suggested that IMRT delivery that takes longer than

1287 half an hour to deliver per fraction might have a reduced cell killing eﬀect. To the

1288 best of the author’s knowledge, there is as yet no clinical studies that demonstrate

1289 a reduced treatment response from protracted IMRT delivery times. In the light of

1290 the lack of understanding about the temporal eﬀect of radiation dose delivery in the

1291 clinical setting, the DMG may prove to be a valuable tool in investigating the dose rate

1292 variation in real time at diﬀerent points in the radiation ﬁeld simultaneously. Although

1293 the high temporal resolution of the DMG may not address the radiobiology eﬀect of a

1294 long IMRT delivery time (sometimes up to 30 minutes), it could however shed some

1295 light on the dose pattern variation at the level of 0.1 second or lower, which would

1296 be a valuable tool in the dosimetry of volumetric modulation arc therapy (VMAT).

1297 Figure 5.5 (d) shows the temporal dose rate patterns (measured for each 10 seconds)

1298 for channels 8 and 127 of the total IMRT delivery extracted from the measurements.

1299 From this picture, it is clearly seen that there exist time intervals with similar patterns

1300 for both channels 8 and 127 when the dose rate is almost zero. This corresponded to

1301 the time between ﬁeld deliveries. On the other hand, dose rate patterns during the

1302 ﬁeld delivery are quite diﬀerent, for example during the time intervals 40-60 and 120-

1303 130 seconds. Beam oﬀ during gantry movement time has been omitted from the ﬁgure.

1304 All of these variation are within the distance of 2.38 cm between the two points of

1305 measurements. To the author’s knowledge, this is the ﬁrst report in radiation therapy

1306 of a dosimeter allowing direct dose measurements in a particular phantom location in

1307 real time with 0.2 mm spatial resolution and 0.1 seconds temporal resolution. 5.4. Conclusion 76

1308 5.4 Conclusion

1309 The Dose Magnifying Glass comprising a linear array of 128 silicon localized dosimeters

2 1310 (strips) each of them with an active area of 20 × 5000 m and a pitch of 0.2 mm was

1311 tested in a clinical IMRT delivery. The original strip detector had a ± 2% variation

1312 of sensitivity between 128 channels including the DAQ electronics that suggests very

1313 high reproducibility of the strip detectors with the planar microelectronic technology.

1314 A uniformity correction map has been obtained using a ﬂat ﬁeld calibration methods

1315 leading to an excellent uniformity of the response within 0.2%. Penumbra measure-

1316 ments were performed at the depth of dmax (1.5 cm) and 10.0 cm for a 6 MV photon

1317 energy. The measurements were compared with the measurements of Gafchromic EBT

1318 ﬁlm as well as other data published in literature for the same ﬁelds. The DMG with

1319 its physical spatial resolution of 0.2 mm derived a 80-20% penumbra width of 2.77 mm

1320 at 1.5 cm depth and 3.93 mm at 10.0 cm depth, similar to the EBT ﬁlm measurements

1321 and those measured by other dosimeters reported elsewhere. The application of the

1322 DMG for in phantom dosimetry QA in a clinical IMRT delivery was demonstrated.

1323 The dose proﬁles from the measurement were compared with the planned dose from

1324 the treatment planning system as well as measurements made with Gafchromic EBT

1325 ﬁlms at the same depth. The dose proﬁle patterns measured by DMG were able to

1326 reproduce the dose proﬁles predicted by the treatment planning system and ﬁlm mea-

1327 surements. The average diﬀerences were 1.1%± 1.8% (1 s.d.) and 1.0%± 1.6% (1

1328 s.d.) between measurements by the DMG with the Pinnacle predicted dose and the

1329 EBT ﬁlm, respectively. The dose proﬁles agree fairly well with each other within the

2 1330 uncertainty of the measurements as well as the positioning uncertainty of 1 mm .In

1331 addition to a high spatial resolution of 0.2 mm, DMG is capable of providing integral

1332 dose proﬁles for each delivered segment or dose rate measurements at any channel with

1333 temporal resolution of 0.1 second. Such measurements for a typical prostate treatment 5.4. Conclusion 77

1334 delivery demonstrated a temporal pattern in dose rate delivery which is quite diﬀerent

1335 within a short distance between points of measurements in the linear detector array.

1336 The present design of the DMG is suitable for beam characterization and mea-

1337 surement of steep gradient dose proﬁles. Application of this technology with the same

1338 spatial resolution for large radiation ﬁelds is limited by the number of readout channels

1339 and the detector-to-readout system connection logistic. 1340 Chapter 6

1341 Application of the Dose Magnifying

1342 Glass in the dosimetric veriﬁcation

1343 of a stereotactic radiosurgery

1344 treatment delivery

1345 6.1 Introduction

1346 Stereotaxy is a method by which a point is deﬁned within the patient’s body by

1347 an external three-dimensional coordinate system (Grosu et al., 2006). Stereotactic

1348 radiosurgery (SRS), is the use of radiation ablation in place of conventional surgical

1349 excision to remove or modify a benign lesion in the body (Metcalfe et al., 2007).

1350 Traditionally, this method requires a delivery of a large dose in a single treatment,

1351 resembling a surgical procedure. In malignancy cases, the treatment is carried out

0This chapter has been published in Medical Physics: Wong, J. H. D., Knittel, T., Downes, S., Carolan, M., Lerch, M. L. F., Petasecca, M., Perevertaylo, V. L., Metcalfe, P., Jackson, M. and Rosenfeld, A. B. (2011). The use of a silicon strip detector dose magnifying glass in stereotactic radiotherapy QA and dosimetry. Medical Physics 38(3): 1226-1238.

78 6.1. Introduction 79

1352 using a series of equal dose fractions, this is then termed stereotactic radiotherapy

1353 (SRT). The intention of the SRT/SRS treatment is to deliver a concentrated dose to

1354 a small volume of tumor tissue, usually located in close proximity to critical organs.

1355 Or, in other cases, it is intended as a boost dose to the target volume. The nature of

1356 SRT/SRS treatment requires it to have a very high geometric precision, i.e. a tight

1357 margin for the planning target volume (PTV) and a sharp dose fall oﬀ (Grosu et al.,

1358 2006). Due to the high dose delivery and tight margins required in this technique, the

1359 planning and delivery of the treatment requires great precision and accuracy. Existing

1360 problems concerning SRT/SRS treatment are two-fold;

1361 i the use of small ﬁeld size for treatment, and

1362 ii the associated quality assurance (QA) involved in verifying dose delivery.

1363 Due to the exceedingly high dose delivered during a single SRS treatment, this tech-

1364 nique is limited to the treatment of small lesions, usually though not exclusively to

1365 those that are <4.0 cm in diameter (Heydarian et al., 1996). Using small ﬁeld sizes

1366 for radiation delivery raises the question of lateral electronic disequilibrium within the

1367 ﬁeld. This is because the track length of the secondary electron is comparable or larger

1368 than the treatment ﬁeld size. Lateral electronic disequilibrium usually exists at the

1369 ﬁeld edge where more secondary electrons generated by the photons within the ﬁeld

1370 edge are scattered out of ﬁeld, while less electrons are scattered inwards. With large

1371 field sizes, this does not affect the dose at the center of the radiation field. However,

1372 with small field sizes, the effect is significant enough to cause a lower total dose at the

1373 center of the radiation ﬁeld. This condition is worsened by the presence of low density

1374 inhomogeneous media such as air cavities and lung because secondary electrons can

1375 travel further in a low density medium. The presence of ﬁnite size radiation detectors,

1376 in which the Bragg-Gray condition is not satisﬁed, would also cause perturbation to

1377 the radiation ﬁeld (Beddar et al., 1994; Das et al., 2008; Scott et al., 2008, 2009). 6.1. Introduction 80

1378 To accurately predict the dose to be delivered, medical physicists must obtain em-

1379 pirical beam data or make Monte Carlo type calculations. Accurate measurements

1380 of the dose characteristics (percent depth doses, output factors, beam proﬁles) are

1381 needed. This information is then entered into a radiotherapy treatment planning sys-

1382 tem (RTPS) for dose calculation (Pappas et al., 2008). Inaccuracies in the beam data

1383 will produce modeling errors for the dose distribution and dose calculation errors in the

1384 treatment plans. Laub & Wong (2003) reported local discrepancies of >10% between

1385 calculated cross profiles and profiles measured with films, due to insufficient spatial

1386 resolution of the detector that was used to collect the beam data during commission-

1387 ing of the IMRT planning tool. The eﬀect of ﬁnite detector size was manifested in

1388 the artiﬁcial broadening of penumbra width, resulting in a systematic over-irradiation

1389 of the organ-at-risk arising from attempts to ensure suﬃcient PTV coverage during

1390 planning (Garcia-Vicente et al., 1998; Pappas et al., 2008) The absolute dose diﬀerence

1391 was reduced to <2% simply by using a detector with smaller cross section, such as a

3 1392 0.015 cm pin point ion chamber (Laub & Wong, 2003).

1393 According to Pappas et al. (2008), the ideal detector suitable for small ﬁeld size

1394 beam measurements should have the following characteristics:

1395 (i) small sensitive volume to avoid volume averaging eﬀects and the capability of

1396 high spatial resolution,

1397 (ii) features to overcome the problem of positioning accuracy in small ﬁelds,

1398 (iii) tissue equivalent and small perturbation of radiation beam, and

1399 (iv) show energy, dose rate, and directional response dependence that is consistent

1400 and characterisable.

1401 To date, there is no one detector that can satisfy all of the above criteria. Radio-

1402 graphic ﬁlm dosimetry has excellent spatial resolution, but is non-tissue equivalent, 6.1. Introduction 81

1403 is energy dependent and somewhat eﬀected by processing conditions. Radiochromic

1404 (Gafchromic) ﬁlm is almost tissue-equivalent, self-developing and its ease of handling

1405 in room light makes this a convenient dosimeter for small ﬁeld dosimetry. All ﬁlm

1406 methods are non real time and are potentially aﬀected by polarization eﬀects and

1407 non-uniformities in commercial ﬁlm scanners (Butson et al., 2003a, 2006a; Paelinck

1408 et al., 2007). Silicon diodes are able to achieve high spatial resolution with small sensi-

1409 tive volume, but their energy, angular, temperature, and dose rate dependency require

1410 rigorous characterization (Rice et al., 1987; Beddar et al., 1994). Diamond detectors

1411 are lauded for their near tissue-equivalence and small sensitive volume, though one

1412 needs to understand their dose-rate dependence and signiﬁcant preirradiation dose ef-

1413 fect (Heydarian et al., 1996; Pappas et al., 2008). Gel dosimetry has the advantage

1414 of having the same medium for dosimetry and dose scattering, water equivalence, and

1415 having high spatial resolution in all three dimensions. It is however aﬀected by Mag-

1416 netic Resonance Imaging (MRI)- and gel-related non-uniformity and artifacts (Vergote

1417 et al., 2004). In dealing with small ﬁeld dosimetry, accuracy in positioning of detec-

1418 tor, relative to the radiation beam becomes an important factor. It is common for

1419 SRS treatment to have treatment plans that consist of various non-coplanar deliver-

1420 ies. The uncertainties resulting from the collimator, gantry and couch rotations could

1421 certainly inﬂuence the beam targeting accuracy with reference to the lesion. In terms

1422 of dosimetric veriﬁcation, positioning accuracy of a small single detector could pos-

1423 sibly contribute to a large discrepancy in the measured dose. The common practice

1424 is to perform measurements with various detectors as a method to cross check and

1425 complement each other.

1426 In this chapter, the Dose Magnifying Glass (DMG) was used in SRT/SRS dosime-

1427 try. The SRS cones total scatter factor (Scp), proﬁles and penumbra was measured.

1428 The use of this detector in the determination of the center of rotation and the dosi- 6.2. Materials and methods 82

1429 metric veriﬁcation of a simulated clinical SRT treatment was also demonstrated.

1430 6.2 Materials and methods

1431 6.2.1 Dose Magnifying Glass

nd 1432 The DMG that was used in this work is a 2 generation DMG which was mounted on

+ 1433 a 0.12 mm thick Kapton substrate. The sensitive area deﬁned by a single n strip is

2 1434 20 × 2000 m . The resistivity of the silicon is 10 Ω·cm and it was preirradiated with

1435 1 MeV electrons up to 15 kGy.

1436 6.2.2 SRS phantom

1437 A custom made phantom was designed and machined for measurement of SRS deliv-

1438 eries. It was made of water equivalent material (RW3, Scanditronix Wellh¨ofer) which

1439 is primarily made up of white polystyrene with 2.1% titanium oxide (TiO2). It was

1440 machined into a cylinder with a radius of 90 mm and a hemispherical end cap. It

3 1441 has a rectangular cavity (100 × 160 × 20 mm ) in the middle of the phantom that

1442 allows the DMG with its solid water holder (50 mm wide) to ﬁt in accurately together

1443 with some solid water spacers for positional adjustability. The cavity width was made

1444 wider than the DMG holder so as to allow for ± 25 mm lateral shifting of the DMG

1445 within the phantom. The smallest solid water spacers were of 5 mm width, allowing

1446 lateral shifting of the DMG in steps of 5 mm. The center of the sensitive volume of the

1447 silicon wafer was made pass through the center of the hemispherical component. The

1448 SRS phantom has been ﬁtted with an adapter plate which allows it to be bolted onto

1449 the Radionics SRS couch mounting frame (Radionics, Inc., Burlington, MA, USA)

1450 which was in turn attached to the linac couch top (see Figure 6.1). The center of the

1451 phantom was positioned at the isocenter, at a source to detector distance (SDD) of 100 6.2. Materials and methods 83

Figure 6.1: SRS phantom mounted on the Radionics SRS couch mount. A plate with ﬂat top and matching concave bottom was used to achieve a ﬂat calibration phantom. Solid water spacers were used to allow the lateral shifting of the DMG within the phantom.

1452 cm to allow for accurate, reproducible and easy setups for DMG SRS measurements.

1453 6.2.3 Detector relative sensitivity factor measurement

1454 The 128 channels on the DMG are coupled to individual charge ampliﬁers whose

1455 response may be slightly diﬀerent from each other. For all measurements described

1456 in this paper, the DMG was used in conjunction with the SRS phantom which has

1457 a cylindrical body. A ﬂat calibration phantom was achieved by using a plate with a

1458 ﬂat top and matching concave bottom surface over the cylindrical and hemispherical

1459 part of the phantom (Figure 6.1). However, linac beam proﬁles are not perfectly

1460 ﬂat at moderate depths; hence the method described below was used to generate a

1461 sensitivity map of the detector channels. This map was subsequently applied to the

1462 other measurements to correct for the detector response. 6.2. Materials and methods 84

1463 The detector sensitivity map was derived by employing a shifting and multiple

1464 measurements method. The DMG readings, Rni are a product of the detector response,

1465 Gni and dose, Dni. Index ‘n’ represents the measurement sequence (1, 2 or 3) and index

1466 ‘i’ represents the detector elements. However, the detector responses, Gni are aﬀected

1467 by two factors, the inherent detector sensitivity coeﬃcient, si and the variation of the

1468 linac beam proﬁle, gni used for uniformity calibration (Eq.6.1). The two factors needs

1469 to be separated in order to determine the true detector sensitivity factor.

Rni = Gni · Dni = signi · Dni (6.1)

1470 A total of three measurements were made. The ﬁrst measurement was taken with

1471 the DMG at isocenter position, while the second and third measurements were taken

1472 with the DMG shifted laterally by + 5 mm and - 5 mm within the phantom cavity.

1473 For any measurement label ‘x’, the detector to detector relationships were given

1474 initial values. This is achieved by normalizing each detector’s response, Rxi to the

1475 center detector, Rx◦ of the dose proﬁle that is not uniformity corrected (Eq.6.2). The

1476 normalization coeﬃcients, kxi obtained were then applied onto the all the three mea-

1477 surements.

Rxi normalisation coeﬃcient, kxi = (6.2) Rx◦

1478 The next step involves determining the detector response variation due to the linac

1479 beam proﬁle. The normalized DMG readings, R’ni are the product of detector response

1480 purely due to linac dose proﬁle variation, gni and dose, Dni (Eq.6.3).

Rni normalised DMG reading, Rni = (6.3) kxi

1481 By the shifting and multiple measurements method, the detector response and 6.2. Materials and methods 85

1482 measured dose at every 5 mm intervals were derived. The dose points obtained were

1483 fitted with a spline curve to produce an estimated dose profile, Dfiti.

1484 The detector sensitivity coeﬃcient, si for each detector were derived by dividing the

1485 original measurement ‘x’ readings, Rxi by Dﬁti (Eq.6.4). Detector relative sensitivity

1486 factors, fi were obtained by taking the ratio of the sensitivity coeﬃcients of each

1487 individual channel to the sensitivity coeﬃcient of the center channel (Eq.6.5).

Rxi detector sensitivity coeﬃcient, si = (6.4) Dﬁti

si Detector relative sensitivity factor, fi = (6.5) s◦

1488 6.2.4 Angular dependence correction

nd 1489 The characterisation of the angular dependence of the 2 generation DMG was de-

1490 scribed in chapter 4, section 4.3.6.2. Angular dependence was taken as a ratio of the

1491 DMG response at diﬀerent angles to the DMG response when the radiation beam was

◦ 1492 perpendicular to the detector plane, corresponding to a gantry angle 0 (Eq.4.1).

1493 The angular dependence was deﬁned separately for the polar and azimuthal angles.

1494 The angular correction factor (CFp and CFa) is the inverse of the angular dependence

1495 (Eq.6.6).

1 Polar angle correction factor, CFp = and, (6.6a) Aθp 1 Azimuth angle correction factor, CFa = (6.6b) Aθa

1496 However, the ﬁnal angular response dependency was derived from the combined

1497 contribution of the azimuthal and polar angular response. This was then used to 6.2. Materials and methods 86

1498 correct for the data measured in the SRS dosimetric plan veriﬁcation.

1499 It is common for any SRS delivery to incorporate various non-coplanar deliveries.

1500 This is essential to produce a highly focused irradiation on the lesion while reducing

1501 doses to normal tissue by spreading it to larger volumes and to avoid critical structures.

1502 A non-coplanar plan would involve many couch angles. This means that the radiation

1503 beam could reach the detector from many polar or azimuthal angles. Hence, the

1504 angular correction should take into account the contribution of the polar and azimuth

1505 angles based on the position of the couch. The angular correction factor was derived

1506 as Eq.6.7,

⎧ ⎪ · 2 · 2 ⎪ (sin (ϕ) CFp (θ)) +(cos (ϕ) CFa (θ)) ⎪ ⎪ ⎨ for 0◦ ≤ ϕ ≤ 90◦ and CFangle (ϕ, θ)=⎪ (6.7) ⎪ − · 2 − · 2 ⎪ (sin (360 ϕ) CFp (θ)) +(cos (360 ϕ) CFa (θ)) ⎪ ⎩⎪ for 270◦ ≤ ϕ<360◦

1507 where, ϕ is the couch angle and θ is the gantry angle.

1508 6.2.5 Determination of the center of rotation measurement

1509 Positioning uncertainty in small ﬁeld setups using small single element detectors is

1510 one of the challenges involved in the dosimetric veriﬁcation of SRS treatments. The

1511 collimator, gantry and couch rotations involved in an SRS delivery produce an uncer-

1512 tainty in beam targeting and dosimetric veriﬁcation. Understanding of this positioning

1513 variability would allow the evaluation of the uncertainty involved in SRS dosimetry.

1514 The center of rotation (COR) and positioning uncertainty of the linac collimator was

◦ ◦ 1515 determined from the center of the beam proﬁles obtained from the 0 and 180 col-

1516 limator rotation angles (Eq.6.8 and Eq.6.9). The center of the proﬁles was obtained 6.2. Materials and methods 87

1517 from the mid point between the 50% of the intensity proﬁles. The COR of the linac

1518 gantry and couch were determined likewise.

x1 + x2 Center of rotation, x = (6.8) cor 2

x1 − x2 Positioning error, x = (6.9) err 2

1519 where, x1 and x2 are the mid point coordinates between the 50% coordinates of

1520 the intensity proﬁles

1521 6.2.6 SRS cone proﬁles and total scatter factor measurement

1522 The proﬁles and total scatter factor (Scp) for the SRS cones with diameter ranging

1523 from 5 mm to 40 mm were measured. The DMG was positioned at SDD of 100

1524 cm. The measurements were compared with Gafchromic EBT2 (ISP, Wayne, NJ) ﬁlm

1525 measurements and the standard data. The standard data were measured with a CC04

1526 ion chamber (Scanditronix Wellh¨ofer) and veriﬁed with Monte Carlo calculations. The

1527 CC04 ion chamber was used to measure the Scp for the cone diameters of 15 mm to

3 1528 40 mm only. Monte Carlo calculations was made with voxel size of 1 × 1 × 2mm

3 1529 for cone diameter ≤ 10 mm and voxel size of 2 × 2 × 2mm for cones with diameter

1530 of 12.5 - 40 mm. The Scp values based on the measurements was rescaled to match

1531 the Monte Carlo data for the 15 mm diameter cone. For the ﬁlm measurements, a

1532 specially made film insert was made to fit inside the SRS phantom cavity. The film

2 1533 insert had a small slit to ﬁt a piece of 40 × 100 mm EBT2 ﬁlm. It was made from

1534 the same RW3 material as the rest of the SRS phantom.

1535 The ﬁlms were scanned 24 hr after the exposure to allow for the post-irradiation

1536 coloration. An A3 size ﬂatbed scanner (Epson Expression 10000XL) was used. Each 6.3. Results and discussions 88

1537 ﬁlm was scanned six times. Only the last three scans were kept for image analysis.

1538 This was to ensure that the scanner was suﬃciently warmed up and thus ensuring

1539 consistency in the scanned ﬁlms (Fuss et al., 2007). The ﬁlms were scanned in 48-bit

1540 RGB color with scanning resolution of 150 dpi. Care was taken to scan the ﬁlm in the

1541 same orientation at the center region of the scanner to reduce scanner induced non-

1542 uniformity (Butson et al., 2006a). A set of calibration ﬁlms were also irradiated in

1543 the same experimental session. Analysis of the images was done using ImageJ version

1544 1.39U (National Institute of Health, USA) software. The red channel of the last three

1545 scans for each ﬁlm were averaged and used for our analysis (Matney et al., 2010). For

1546 the Scp measurement, dose measured by the EBT2 ﬁlm was averaged over an area of

2 1547 0.5 × 2mm (corresponding to 3 × 12 pixels) at the center of the beam, to match the

1548 sensitive area of the DMG.

1549 6.2.7 Clinical stereotactic arc measurement

1550 A four arc non-coplanar simulated SRS treatment was carried out with a Siemens

◦ 1551 Oncor linear accelerator. A 180-degree gantry arc angle was used, starting at 0 and

◦ ◦ ◦ ◦ 1552 stopping at 180 . The couch angles used were from 270 to 360 at 30 intervals. The

1553 center of the DMG was positioned at isocenter. The whole delivery was carried out

1554 using an 8 mm diameter cone. For each arc, 100 MU was delivered. The measurements

1555 were angular corrected. These measurements were compared with Gafchromic EBT2

1556 ﬁlms using the same protocol. 6.3. Results and discussions 89

Figure 6.2: Detector response of the DMG before (open circle) and after uniformity correction (closed circle).

1557 6.3 Results and discussions

1558 6.3.1 Uniformity

1559 Figure 6.2 shows the detector response before and after the uniformity correction. The

1560 data are normalized to the center detector. Error bars represent 1 s.d. of the mean of

1561 all the detectors. The coeﬃcient of variation (CV) before uniformity correction was

1562 >2% and less than 0.5% after. These sensitivity factors were applied to all subsequent

1563 measurements.

1564 6.3.2 Determination of center of rotation

1565 The center of rotation and positioning error was determined by varying one of the

1566 delivery angles (collimator, gantry or couch) while keeping the other two ﬁxed. The

1567 maximum positioning error due to collimator, gantry, and couch rotation was 0.2 ±

1568 0.1 mm, 0.4 ± 0.1 mm, and 0.4 ± 0.2 mm (1 s.d), respectively. Figure 6.3 shows the 6.3. Results and discussions 90

1569 proﬁles of two acquisitions within the same isocentric setup but with diﬀerent couch

1570 positions. The data were normalized to unity at the peak of the acquisition for couch

◦ 1571 position at 90 . The oﬀset between the two dose proﬁles of opposite gantry angles

1572 shows variability in the isocentric of the couch rotation. This resulted in a lower dose

◦ 1573 measured when the couch was at the position of 90 because the DMG may not be

1574 measuring through the centre of the radial beam but at the slope of the radial beam.

1575 In addition to the measured positioning error due to the rotating component of the

1576 linear accelerator, there is also the uncertainty in the cone positioning in the applicator

1577 of <0.5 mm. With that taken into consideration, the positioning uncertainty added

1578 in quadrature was 0.8 mm for the DMG. This represents the maximum uncertainty

1579 of <2% in the maximum dose measured at the smallest cone diameter of 5 mm. This

1580 test is somewhat similar to the Winston Lutz test (Lutz et al., 1988) that was used to

1581 measure the central axis oﬀset due to the rotating component of the linear accelerator.

1582 The DMG, like ﬁlm is a useful tool for determination of COR and has the advantage

1583 of obtaining immediate proﬁles, and is especially advantageous for the purpose of

1584 mechanical alignment of the SRS cones.

1585 Once the positioning error was determined as described above, the calculated po-

1586 sitioning error was also applied to the EBT2 ﬁlm measurements to estimate the un-

1587 certainty to the measured dose. However, in addition to the 0.8 mm positioning error,

1588 the use of the ﬁlm also involves additional positioning uncertainty of ± 0.5 mm when

1589 it is placed in the phantom. The positioning uncertainty of the EBT2 is therefore 0.9

1590 mm and this translates to 3% dose uncertainty at maximum dose for 5 mm diameter

1591 cone. 6.3. Results and discussions 91

Figure 6.3: Two proﬁles of a 5 mm diameter cone measured with the DMG showing discrepancy in the center of rotation due to couch rotation from 90◦ - 270◦.

1592 6.3.3 SRS cone proﬁles and penumbra

1593 The DMG with its array of 128 channels has a sensitive length of 25.6 mm. This allows

1594 the measurement of the dose proﬁles of the small SRS cones with diameters up to 20

1595 mm without moving the detector. The 80-20% penumbra width and the full width at

1596 half maximum (FWHM) of the cones were also measured. The results were tabulated

1597 in Table 6.1. The average diﬀerences between the EBT2 and DMG measurements are

1598 0.22 ± 0.07 mm and 0.12 ± 0.09 mm for penumbra width and FWHM, respectively.

1599 6.3.4 SRS cone total scatter factor

1600 The Scp of the cones were measured with the DMG (Figure 6.4) for cone diameters

1601 from 5 mm to 40 mm. The measurements were taken as an average of the two central

2 1602 detectors (sensitive area of 0.4 × 2.0 mm ) of DMG. The results were compared with

2 1603 EBT2 measurements (averaged over an area of 0.5 × 2mm) and the standard data

1604 which are based on CC04 ion chamber measurements for SRS cones with diameter 6.3. Results and discussions 92

Table 6.1: The penumbra and FWHM measurements of the dose profiles of 5 mm to 20 mm cone diameter at SDD 100 cm, comparing the DMG and EBT2 measurements. Penumbra (80%-20%) (mm) FWHM (mm) Cone diameter (mm) DMG EBT2 Difference DMG EBT2 Difference 5.0 1.28 1.49 0.21 4.61 4.86 0.25 6.5 1.41 1.60 0.19 6.23 6.36 0.13 8.0 1.55 1.86 0.31 8.24 8.32 0.08 10.0 1.65 1.97 0.32 9.26 9.40 0.14 12.5 1.91 2.11 0.19 12.00 11.82 0.18 15.0 2.05 2.23 0.18 14.43 14.43 0.00 17.5 2.12 2.24 0.12 16.50 16.52 0.02 20.0 2.18 2.44 0.25 19.12 18.93 0.19

1605 >10 mm and Monte Carlo simulations for cones with diameter ≤ 10 mm. The DMG

1606 data are rescaled to the calibrated 15 mm diameter cone of the Monte Carlo data. The

1607 EBT2 data were rescaled to the 15 mm diameter cone value using a ﬁtted curve to

1608 avoid the eﬀect of scatter in the data. The DMG measurements agreed well with the

1609 standard data measurements. The average diﬀerence between the DMG measurements

1610 with the standard data and EBT2 measurements were found to be 1.2 ± 1.1% and 1.9

1611 ± 1.9%, respectively.

1612 The good agreement between the Monte Carlo calculation and the DMG mea-

1613 surements for the Scp is not unexpected since at the center of the beam, the proﬁles

1614 gradient does not change drastically. Volume averaging present at this point of mea-

2 2 1615 surement between a detector with the area of 0.02 × 2mm and 1 × 1mm would not

1616 be signiﬁcantly diﬀerent. However, if the same detector was used to measure the high

1617 gradient penumbra region of a beam proﬁle with radial symmetry, detector with a 0.02

2 1618 × 2mm area may be better due to the 0.02 mm width of the detector than a 1 × 1

2 1619 mm area. In this case averaging eﬀect in the longitudinal direction is not as critical as

1620 in the radial direction and averaging eﬀect of the detector is dominated by the size in

2 1621 the radial direction. This means that a detector with an area of 1 × 1mm will have

2 1622 good agreement with a detector with an area of 0.2 × 2mm in Scp measurements. 6.3. Results and discussions 93

Figure 6.4: SRS cones total scatter factor comparing DMG with EBT2, CC04 measurements and Monte Carlo calculation. The measurements were taken within the spatial distance of ± 0.2 mm from the center axis. The CC04 ion chamber was used to measured the Scp for the cones with 15 mm - 40 mm. The Monte Carlo calculation was made with voxel size of 1 × 1 × 2mm3 for cone diameter ≤ 10 mm while voxel size of 2 × 2 × 2mm3 was used to calculate for the cone diameter >10 cm.

2 1623 However, the detector with an area of 0.2 × 2mm (DMG) will provide a much more

2 1624 realistic beam proﬁle than the detector with an area of 1 × 1mm and should be in a

1625 better agreement with a EBT2 ﬁlm readout with 0.17 mm resolution. This point will

1626 be elaborated further in section 6.3.6.

1627 6.3.5 Clinical SRS application

◦ ◦ ◦ 1628 Four SRS arcs spanning 180 were delivered with couch angles from 270 to 360 at

◦ 1629 every 30 intervals. The relative intensity proﬁle of the sum of the arcs are shown in

1630 Figure 6.5. The relative intensity proﬁles were normalized to 1 at the maximum of

1631 the EBT2 measurement and error bars are the ± 3% and ± 2% dose uncertainties of

1632 the EBT2 and DMG, respectively. DMG measurements were all corrected for angular

1633 dependencies. Within the ± 1 mm positional and 3% dose uncertainty of the measure- 6.3. Results and discussions 94

Figure 6.5: SRS arc (0◦ - 180◦) relative intensity proﬁle.

1634 ment, the EBT2 and DMG agree well with each other. The average diﬀerence above

1635 the 90% intensity was 1.7 ± 0.8%.

1636 6.3.6 On the volume averaging eﬀect of small ﬁeld dosimetry

1637 The averaging eﬀect of the detector should be considered as at least a two dimensional

1638 problem and for the particular ﬁeld and conﬁguration of the detector. For example,

1639 in the case of a SRS radiation ﬁeld with radial symmetry, diﬀerent restriction would

1640 apply to the detector geometry for small beam measurements. The following workings

1641 set out the consideration of the eﬀect of volume averaging of detectors

1642 (a) at the center of the beam (such as the total scatter factor, Scp measurement), and

1643 (b) at the penumbra region.

1644 The volume averaging effect of detectors with different geometrical configuration

1645 was estimated analytically (Section 6.3.6.1). The analytical model used was based on 6.3. Results and discussions 95

Figure 6.6: (a) Pseudo colored image from EBT2 film measurement of a 5 mm diameter beam, normalized to the center of the beam, (b) isodose lines for the 5 mm diameter beam, (c) signal profile alone a horizontal line drawn across the middle of the beam and (d) zoom in to the top beam profile, data points >0.3 were extracted and curve fitted with a 4th degree polynomial function (r2 = 0.998).

1646 the real data measured with an EBT2 ﬁlm (scanned with 150 dpi resolution) (Figure

1647 6.6(a)). Section 6.3.6.2 provides detail of the penumbra consideration.

1648 Consider:

1649 (a) FWHM of beam = 5 mm.

1650 (b) The radiation beam considered is a circular beam with radial symmetry, i.e. the

1651 penumbra drop oﬀ would be almost symmetrical in every direction, as seen in

1652 Figure 6.6 (b).

2 1653 (c) Detector area considered = 2a × 2b mm (where, a = 0.2 mm, b = 1 mm for 6.3. Results and discussions 96

Figure 6.7: Schematic drawing of detector area superimposed on the radial beam contour.

1654 DMG and a = b = 0.5 mm for the Monte Carlo virtual detector) as seen in Figure

1655 6.7.

1656 6.3.6.1 Center of beam consideration (for Scp measurement)

1657 The response of detector is (Eq.6.10)

R = D (r) rdrdθ (6.10) θ r

1658 To simplify the calculation, only the ﬁrst quarter of the detector which is repre-

1659 sented by the areas A1 and A2 is considered [Eq.6.11 and Eq.6.12] (see Figure 6.7).

R = R1 + R2 (6.11) 6.3. Results and discussions 97

a θ1 cosθ R1 = D (r) rdrdθ (6.12a) 0 0 π b 2 sinθ R2 = D (r) rdrdθ (6.12b) θ1 0

1660 where, R1 and R2 represent the detector response corresponding to the detector

1661 areas A1 and A2. For a uniform ﬁeld with a point dose D◦, the response of the detector

1662 is (Eq.6.13)

R = D◦ × ab. (6.13)

1663 From the ﬁlm data (Figure 6.6(a)), the dose proﬁle was plotted across the center

1664 of the beam (Figure 6.6(c)). Since the detector area considered is smaller than the

1665 FWHM of the beam proﬁle, only the central ± 2.5 mm of the proﬁle will be considered.

1666 In Figure 6.6 (d), the data points >0.3 were extracted and plotted as a function

th 1667 of the radial distance, r (mm). The data points were curve ﬁtted with a 4 degree

2 1668 polynomial function (r = 0.998) using Matlab (Eq.6.14). This is the dose function,

1669 D(r) that will be used for the following calculation.

1670 Therefore,

a π b θ1 cosθ 2 sinθ R = kr4 + lr3 + mr2 + nr + p rdrdθ+ kr4 + lr3 + mr2 + nr + p rdrdθ 0 0 θ1 0 (6.14)

−1 b 1671 where θ1 =tan a , coeﬃcients k, l ,m n, p were generated from Matlab curve

1672 ﬁtting toolbox.

R Dose, D = (6.15) ab 6.3. Results and discussions 98

Table 6.2: Dose calculated based on the analytical model showing the effect of dose averaging effect with different detector sizes. Dose,D MC ( 0.5 × 0.5 mm2) 0.9993 DMG (0.2 × 1mm2) 0.9931 Difference (MC-DMG) 0.0062

1673 Results on dose averaging eﬀect is shown in Table 6.2.

1674 This exercise showed that at the center of beam region of a 5 mm diameter beam,

2 1675 the variation in dose measured by diﬀerent detectors with dimension of 1 × 1mm and

2 1676 0.4 × 2mm is in the order of 0.6%. That is why the MC and DMG measurements are

1677 in close agreement. The relative dose averaging eﬀect will be the same for the whole

1678 detector as for a quarter of the detector considered.

1679 6.3.6.2 Penumbra consideration for a beam with radial symmetry

1680 For the volume averaging eﬀect of detectors at the penumbra region, the hypothesis

1681 is that in a radially symmetrical beam, for a detector with dimension of 0.02 × 2

2 1682 mm , the dose gradient in the r-direction is steeper than in the τ-direction (refer

1683 Figure 6.8). Therefore, for a penumbra measurement, a small detector width of 0.02

1684 mm (almost identical to point dose) would be able to represent the real penumbra

1685 proﬁle in the r-direction. In the τ-direction on the other hand, although there would

1686 be some volume averaging, the eﬀect would be less critical due to the shallower dose

1687 gradient. Two methods were used to demonstrate this eﬀect, (i) dose gradient variation

1688 in the orthogonal radial direction corresponding to the length of the detector length of

2 1689 interest and (ii) numerical integration of the dose map using an area of 0.2 × 1mm

2 1690 and 1 × 1mm.

1691 Dose gradient variation in the orthogonal radial directions: Figure 6.8

1692 shows the schematic diagram of the beam proﬁle from the top view and the denotation 6.3. Results and discussions 99

Figure 6.8: Schematic diagram representing the top view of the beam and the denotation of the r- and the τ-directions. The isodose lines are represented as the concentric circles.

1693 of the r- and τ-directions. Two detectors with diﬀerent longitudinal dimensions of 1

1694 mm and 2 mm are considered. However, the dose gradient (Eq.6.16 and 6.17) is only

1695 calculated for one half of the detector length, b.

dD Dose gradient in r-direction = (6.16) dr

ΔD D − D −1 Dose gradient in τ-direction = = n n (6.17) b b

1 1696 Where b = 2 of the longitudinal length of the detector.

1697 Figure 6.9 shows the dose gradient along the r-direction (a) and the τ-direction

1698 (b). The dose gradient is much larger in the r-direction compared to the τ-direction,

1699 particularly at the middle of the penumbra ( r = -2.5 mm), where the dose gradient in

1700 r-direction is 5 times greater than the dose gradient in the τ-direction. It is important

1701 to note that the dose gradient in the τ-direction is almost independent on b-size of

1702 the detector. A detector with a smaller width, 0.2 mm would be able to measure the 6.3. Results and discussions 100

2 1703 penumbra width more accurately, compared to a 1 × 1mm detector (e.g face on

1704 stereotactic diode), even though the length of the detector is 2 mm (e.g. DMG). This

1705 is due to the fact that the dose gradient variation in the τ-direction is not signiﬁcantly

1706 diﬀerent between the two detectors, while the dose gradient in the r-direction varies

1707 more dramatically. It was also conﬁrmed using numerical simulations based on the

1708 dose map obtained with a EBT2 ﬁlm.

1709 Numerical integration of dose: From Figure 6.6 (b), two strips of dose map of

2 2 1710 an area of 2 × 10 mm and 1 × 10 mm were extracted (Figure 6.10). For the 2 × 10

2 2 1711 mm strip, the dose was integrated over an area of 2 × 0.2 mm at 0.2 mm interval

2 2 1712 while for the 1 × 10 mm strip; the dose was integrated over an area of 1 × 1mm at

1713 1 mm interval. The result of the numerical integration is shown in Figure 6.11.

2 1714 Figure 6.11 shows that the eﬀect of volume averaging of 0.2 × 2mm and 1 ×

2 1715 1mm detectors. At the center of the beam proﬁle, the eﬀect of volume averaging

2 2 1716 for the 1 × 1mm detector relative to the 0.2 × 2mm detector is insigniﬁcant in

1717 comparison to the volume averaging eﬀects within the penumbra region.

1718 To summarize, the size of the detector is important in small ﬁeld dosimetry. A

1719 detector with larger dimensions will tend to cause volume averaging of the dose. How-

2 2 1720 ever, when considering detectors with respective sizes of 1 × 1mm and 0.2 × 2mm,

1721 the eﬀect of detector volume averaging should be considered for speciﬁc beam geom-

1722 etry (beam diameter, detector diameter and region of measurement) and the purpose

1723 of the measurement. In the case of a 5 mm diameter circular beam with radial sym-

1724 metry, measurement of the center of the beam such as the total scatter factor, Scp,a

1725 diﬀerence in the longitudinal dimension of 1 mm versus 2 mm would produce only a

1726 small discrepancy, since over most of the region the beam proﬁle does not have a high

1727 dose gradient. However, if the same detectors were to be used for measurement within

1728 the penumbra region where there is a steep dose gradient, a detector with an area of 6.3. Results and discussions 101

1 Figure 6.9: Dose gradient in (a) r- and (b) τ-direction for a detector with 2 longitudinal length of b = 0.5 mm and 1 mm. The x-axis is the radial direction extending from the center of the beam (r = 0 mm) to the edge of the beam (r = - 6 mm). The negative gradient at the region r = -0.5 mm to 0 mm is due to the noise in the EBT2 ﬁlm picked up by the ﬁtted curve. 6.4. Conclusion 102

Figure 6.10: Dose map of (a) 2 × 10 mm2 and (b) 1 × 10 mm2.

2 2 1729 0.2 × 2mm would be more advantageous than detector with an area of 1 × 1mm

1730 because the dose gradient in the transverse direction is signiﬁcantly steeper than that

1731 in the longitudinal direction.

1732 6.4 Conclusion

1733 A silicon strip detector Dose Magnifying Glass (DMG) was tested for a typical SRS

1734 dose deliveries. A custom made SRS phantom was made that allows accurate and

1735 reproducible setups using a commercially available SRS couch mount. The DMG

1736 was used to measure the isocenter displacement due to gantry, collimator and couch

1737 rotations of the linear accelerator. The overall positioning uncertainty of the DMG

1738 due to couch, gantry, collimator rotation and the positioning of the SRS cones was

1739 found to be ± 0.8 mm. This translates to an uncertainty in the measurement of ±

1740 2% for the smallest cone diameter. The SRS cones Scp, proﬁles and penumbra were

1741 measured. DMG with its high spatial resolution of 0.2 mm was able to capture the

1742 dose proﬁle of the smallest SRS cone of 5 mm diameter. The maximum diﬀerence

1743 between DMG and EBT2 ﬁlm measurements for the cone Scp values was 3.8% for cone

1744 diameter of 5 mm. DMG and EBT2 derived penumbra width for SRS cones with 6.4. Conclusion 103

Figure 6.11: Numerical integration of dose for detector of 0.2 × 2mm2 and 1 × 1 mm2 area.

1745 diameter ranged from 5 to 20 mm was found to be 1.77 ± 0.37 mm and 1.99 ± 0.33

1746 mm (1 standard deviation). An analysis of the volume averaging eﬀect of detector was

2 1747 carried out. When considering detectors with respective sizes of 1 × 1mm and 0.2 ×

2 1748 2mm in a small ﬁeld, one has to consider the speciﬁc beam geometry (beam diameter,

1749 detector diameter and region of measurement) and the purpose of the measurement.

1750 For the purpose of measuring Scp at the center of a circular beam, diﬀerence in the

1751 longitudinal dimension of 1 mm versus 2 mm would produce minimal discrepancy due

1752 to the large plateau region at the centre of the beam. However, at the steep dose

2 1753 gradient region, a detector with 0.2 × 2mm would be more advantageous because of

1754 the narrower detector width. The DMG was also tested in a SRS arc delivery where

◦ 1755 four non-coplanar SRS arcs of 0 - 180 were delivered with diﬀerent couch angles. The

1756 maximum difference in profiles between the DMG and the EBT2 film was less than

1757 2.5%. The novelty of the DMG combines the advantage of multiple single element

1758 detectors in measuring a small radiation ﬁeld. In addition, the real time acquisition 6.4. Conclusion 104

1759 of the DMG allows quick evaluation of SRS dose proﬁles. 1760 Chapter 7

1761 Application of the Dose Magnifying

1762 Glass in the quality assurance of

1763 Helical Tomotherapy

1764 7.1 Introduction

1765 Helical tomotherapy is a complex radiation therapy delivery system that integrates

1766 Mega-Voltage CT (MVCT) image guidance with intensity modulated radiation therapy

1767 (IMRT). Tomotherapy is characterized by the delivery of radiotherapy whereby patient

1768 is translated through a rotating fan beam of 6 MV photons analogous to a diagnostic

1769 CT scan. In contrast with conventional linear accelerators, helical tomotherapy does

1770 not have a ﬂattening ﬁlter installed at the distal end of the linear accelerator. As

1771 a result, the beam output from the linac is forward peaked until it is modulated by

1772 the binary collimator situated distal to the beam. Another feature that distinguishes

0This chapter has been published in Medical Physics: Wong, J. H. D., Hardcastle, N., Tome, W. A., Bayliss, A., Tolakanahalli, R., Lerch, M. L. F., Petasecca, M., Carolan, M., Metcalfe, P. and Rosenfeld, A. B. (2011). Independent quality assurance of a helical tomotherapy machine using the dose magnifying glass. Medical Physics 38(4): 2256-2264.

105 7.1. Introduction 106

1773 helical tomotherapy from conventional linacs is its use of a binary multileaf collimator

1774 (MLC) instead of the conventional ﬁeld shaping multileaf collimator. The binary MLC

1775 consists of 64 tungsten leaves that are inter-digitated across the ﬁeld. The maximum

1776 ﬁeld size is 40 cm in the x-direction (with all 64 leaves opened) while the ﬁeld width in

1777 the y-direction is set by the primary collimator to be 1, 2.5 or 5 cm (Balog et al., 2003a).

1778 Each tungsten leaf is 10 cm thick and deﬁnes an average beamlet width of 6.25 mm

1779 (40 cm divided by 64) at the isocenter (Langen et al., 2010). Treatment delivery with

1780 helical tomotherapy consists of multiple gantry rotations around the patient. Each

1781 rotation is planned as 51 projections equally spaced around a full rotation, although

1782 the gantry is in continuous motion. Radiation beam modulation is achieved by varying

1783 the leaf open time as the gantry is rotated, the leaf open time being described as a

1784 fraction of the projection time. This unique mechanism for dose modulation results in

1785 unique dosimetric characteristics that are not present in conventional IMRT delivery.

1786 These characteristics of helical tomotherapy require slightly diﬀerent quality assurance

1787 (QA) considerations dosimetric considerations compared with conventional linac QA.

1788 Previous reports have described QA techniques for helical tomotherapy, some of

1789 which were also summarized in the report of the AAPM task group 148 published

1790 recently (Balog et al., 2003a,b; Fenwick et al., 2004; Balog et al., 2006; Balog & Sois-

1791 son, 2008; Langen et al., 2010). In some of these reports, the on-board megavoltage

1792 (MVCT) detector is employed as a QA tool. The on-board MVCT detector is a linear

1793 array of 738 xenon ﬁlled ion chambers that sits opposite the source on the gantry ring.

1794 It can be used as a direct ﬂuence (in the case of nothing in the beam) or exit ﬂuence (in

1795 the case of patient/phantom in the beam) detector array. Although the CT detector

1796 is a convenient tool for QA, it is not a machine-independent QA tool. The need to

1797 have an independent detector to verify this inherent CT detector was pointed out in

1798 the AAPM Task Group 148 report (Langen et al., 2010). 7.2. Methods and materials 107

1799 In this chapter, three quality assurance tests were performed on a helical tomother-

1800 apy machine using a Dose Magnifying Glass (DMG), namely the MLC alignment, leaf

1801 open time and leaf ﬂuence output factor (LFOF) using the DMG. These were compared

1802 with the standard measurement methods for the tests. The MLC alignment relative

1803 to the gantry was measured and compared with radiographic ﬁlm. The leaf open time

1804 threshold measurements and leaf output factor measurements were compared with the

1805 MVCT detector data.

1806 7.2 Methods and materials

1807 7.2.1 Dose Magnifying Glass

nd 1808 The DMG that was used in this work is a 2 generation DMG which was mounted on

+ 1809 a 0.12 mm thick Kapton substrate. The sensitive area deﬁned by a single n strip is

2 1810 20 × 2000 m . The resistivity of the silicon is 10 Ω·cm and it was preirradiated with

1811 1 MeV electrons up to 15 kGy.

1812 7.2.2 Multileaf collimator (MLC) alignment measurement

1813 The MLC alignment test measures oﬀset of the MLC bank relative to the gantry center

1814 of rotation. This test is usually performed at the time of machine commissioning or

1815 when the MLC bank is replaced. Baseline data is obtained and subsequently compared

1816 with the three-monthly checks (Fenwick et al., 2004). The tolerance limit for MLC

1817 misalignment is ± 1 mm. (Balog et al., 2003a) described a gantry rotate-double

1818 exposure method using ﬁlm to measure the MLC alignment. This method is replicated

1819 in this work for comparison with the measurement using the DMG. A sheet of EDR2

1820 (Eastman Kodak Company, Rochester, NY) ﬁlm was sandwiched between two 6 cm

1821 solid water slabs and placed at machine isocenter (source to detector distance (SDD) 7.2. Methods and materials 108

1822 of 85 cm). A three-step delivery sinogram was written to expose the ﬁlm. The ﬁrst

1823 step is to open and close the two central leaves, 32 and 33, sequentially for a ﬁxed

◦ 1824 duration at gantry angle 0 . The second step is to open and close leaves 28 and 29

◦ 1825 sequentially for a ﬁxed duration at gantry angle 0 . The third and last step is to rotate

◦ 1826 the gantry to 180 and again open and close leaves 28 and 29 for a ﬁxed duration. The

1827 result is a three-peaked dose distribution. The distance between the centres of each

1828 of the outside peak to the centre of the middle peak is measured and the diﬀerence

1829 between these is calculated. This value is then divided by two to obtain the MLC

1830 bank oﬀset relative to the centre of rotation (COR). It is assumed that the distance

1831 between the centre of leaves 28 and 29 and the centre of 32 and 33 does not change

1832 with rotation, therefore any diﬀerence in the peak location is due to an MLC oﬀset

1833 relative to the COR. For a perfectly aligned MLC bank these two distances should be

1834 equal. This measurement was repeated twice. The ﬁlms were processed and scanned

1835 on a Vidar scanner at 142 dpi. The images were analysed in ImageJ software in which

1836 the distances between the peaks was measured.

1837 Due to geometrical limitations of the current DMG prototype, the MLC alignment

1838 test was slightly modiﬁed such that it could be performed with the DMG. The length of

1839 the DMG is 25.6 mm. Each MLC leaf projects to 6.25 mm at 85 cm SSD therefore the

1840 DMG can measure the dose distribution for at most four contiguous leaves without

1841 having to shift the DMG. The ﬁlm test requires coverage of 62.5 mm therefore the

1842 test was simpliﬁed to enable measurement using the DMG without having to shift it

1843 laterally (which would introduce positional uncertainty). The DMG was set up with

1844 its long axis perpendicular the leaf travel direction such that it spanned four leaves.

1845 It was sandwiched between two 6 cm solid water slabs and positioned at SDD 85 cm.

1846 An exposure was made with leaf 32 and 33 opened and closed sequentially at gantry

◦ ◦ 1847 angle 0 . The procedure was repeated with the gantry rotated to 180 . This resulted 7.2. Methods and materials 109

1848 in two peaks (one for each gantry position). The distance between the centres of the

1849 two peaks was measured and divided by two. The measurement was repeated twice.

1850 The positioning of the ﬁlm and the DMG at the isocenter was determined using

1851 the laser guide. The accuracy of this was veriﬁed by measurement of the full width at

1852 half maximum (FWHM) of the radiation proﬁles obtained at opposite gantry angles.

1853 7.2.3 Leaf open time measurement

1854 When the signal is sent for a leaf to move from its closed state to an open state (or

1855 vice versa), a ﬁnite time is needed for the electrical and mechanical/pneumatic system

1856 to physically move the leaf across the ﬁeld. The time required for this activation and

1857 transition is called leaf latency (Kapatoes et al., 2001b; Balog et al., 2003b). The

1858 leaf latencies of the leaves are determined at the commissioning of the tomotherapy

1859 machine. It is also recommended as a three-monthly check QA (Fenwick et al., 2004).

1860 There are two leaf latency corrections used in the treatment planning and delivery

1861 software. The ﬁrst is the ‘leaf activation latency’ which is composed of the delay

1862 between the signal being sent for a leaf to open or close and the actual time it takes to

1863 open a leaf. If the leaf activation latency is not taken into account, the leaves would

1864 open and close at a slightly oﬀset angle to the intended angle of delivery (Kapatoes

1865 et al., 2001b). The second is the ‘leaf open time threshold’, which is the time below

1866 which the actual leaf open time is not equal to the programmed leaf open time. This

1867 correction is applied during the end of planning calculations. If the leaf open time

1868 threshold is not applied then the dose calculation would be including all beamlets

1869 whose leaf open time is less than what is physically achievable. Application of the

1870 leaf open time threshold results in all beamlets whose leaf open time is less than the

1871 threshold being discarded. During commissioning, the leaf open time threshold is

1872 measured by taking the signal measured for some fraction of the programmed open 7.2. Methods and materials 110

1873 time divided by the signal measured for the 100% open time. This is usually done

1874 with the on-board CT detector array. The CT detector array reads the exit radiation

1875 as it passes through the phantom and couch. The detector is synchronized with the

1876 linac beam pulses, therefore data are recorded at every 3.3 ms (Van de Vondel et al.,

1877 2009).

1878 For measurements of the leaf open time threshold, the DMG was positioned at the

◦ 1879 depth of 1.5 cm and SDD of 85 cm. The gantry was set to 0 and the DMG was

1880 positioned within the trajectory of the center leaf (leaf 32). To enable measurement of

1881 the real time transition of the leaf motion, the DMG was rotated to align the long axis

1882 of the DMG parallel to the direction of the leaf travel. The Y-jaw of the collimator was

1883 set to 2.5 cm. A procedure was created to deliver radiation with projection times of 50,

1884 100, 200, 303 and 400 ms with a static gantry. For a helical tomotherapy treatment,

1885 the projection time is the time taken for the gantry to rotate through one projection

1886 (1/51) of the full rotation. The modulation factor in tomotherapy is deﬁned as the

1887 maximum leaf open time divided by average leaf open time. In this thesis, the average

1888 leaf open time is equal to the actual leaf open projection time. For each projection

1889 time, the modulation factor was varied from 1 to 20 (that is, from 100% to 5% of the

1890 projection time). The eﬀect of this leaf open time is best described with the following

1891 example. For a 200 ms programmed leaf open time, a modulation factor of 2 would

1892 produce a 100 ms programmed open time, whereas a modulation of 10 would equal to

1893 a 20 ms programmed open time. For each leaf open time, the leaf was programmed

1894 to open and close ﬁve times to obtain repeated measurements, with a delay of three

1895 times the projection time between each leaf opening giving a ‘duty cycle’ of 25%. This

1896 gave a distinct separation of the individual leaf projection. The DMG readout was set

1897 to readout data every 3 ms. The same procedure was repeated for leaf 4 and 61.

1898 During the leaf open time threshold measurement, the CT detector data that mea- 7.2. Methods and materials 111

1899 sured the exit radiation was extracted from the tomotherapy computer database for

1900 comparison with the DMG measurements. The detector data was extracted after each

1901 beam delivery. A compression factor of one (that is, no compression) was applied to all

1902 measurements to maximize the number of data points. The binary data was converted

1903 into numerical format using the TDAT software (Tomotherapy Inc., Middleton, WI,

1904 USA) and analyzed using Matlab (The MathsWorks Inc.).

1905 7.2.4 Leaf ﬂuence output factor measurement

1906 The third QA test performed measured the leaf ﬂuence output factor (LFOF) of an

1907 individual opened leaf. The energy ﬂuence under a speciﬁc leaf of interest (LOI) is

1908 dependent on the state of its adjacent leaves (Balog et al., 1999). The output of a

1909 particular leaf is greater when the adjacent leaf or leaves are opened. This eﬀect needs

1910 to be incorporated into the treatment planning system to ensure a correct delivery

1911 model. The LFOF measurements are performed during commissioning and are gen-

1912 erally required on the installation of a new MLC bank. The DMG was positioned at

1913 SDD 64 cm at 1.5 cm depth with the DMGs long axis perpendicular to the direction

1914 of the leaf travel. The leaf width projected at this SDD is 4.7 mm therefore the DMG

1915 spanned ﬁve leaves. According to Balog et al. (2003b), the opened leaves beyond the

1916 two adjacent leaves do not increase the LFOF signiﬁcantly. This was attributed to

1917 the lack of ﬂattening ﬁlter which would otherwise produce more extra-focal radiation

1918 passing through the opening of the multiple opened leaves, hence contributing to the

1919 increase in the LFOF. In this work, the LFOF of three leaves (33, 47 and 62) was

◦ 1920 measured. The gantry was set to 0 . The center of the DMG was positioned at the

1921 center of leaf 33’s trajectory. A delivery sinogram was programmed to deliver a se-

1922 quence of diﬀerent combinations of leaf openings as shown in Table 7.1. The sequence

1923 was delivered at the rate of 1 s/projection. The DMG was set to acquire data at 100 7.2. Methods and materials 112

Table 7.1: Programmed leaf opening conﬁguration for LFOF measurement for LOI = 33. The same delivery sinogram was also used with leaf 47 and 62 as the LOI. The symbol “|” denotes an open leaf. Leaf Projection no. 31 32 33 34 35 1 | 2 || 3 || 4 ||| 5 | 6 || 7 || 8 ||| 9 | 10 || 11 || 12 |||

1924 ms intervals. Hence, for each conﬁguration, 10 acquisitions were obtained. To avoid

1925 coinciding with the leaf transition period, only the central eight acquisitions were av-

1926 eraged. Three measurements were made for each LOI and the LFOF was computed

1927 as the mean ± 1 standard deviation of the mean of three measurements.

1928 The DMG measurements were compared with the MVCT detector measurements

1929 acquired with the tomotherapy couch retracted exposing the MVCT detector to the

1930 direct radiation beam. The LFOF is deﬁned by Eq.7.1.

Fluenceab LFOFab = for one adjacent leaf opening (7.1a) (Fluencea + Fluenceb) Fluenceabc LFOFabc = for both adjacent leaves opening (Fluencea + Fluenceb + Fluencec) (7.1b)

1931 where ‘a’, represent the opened LOI and ‘b’, ‘c’ represent the opened adjacent

1932 leaves. 7.3. Results and discussions 113

1933 7.3 Results and discussions

1934 7.3.1 Multileaf collimator alignment

1935 The ﬁlm measurement produced three darkened areas on the EDR2 ﬁlm, shown in

1936 Figure 7.1(a). Each of these peaks has a dip in the middle as a result of the overlapping

1937 penumbra from delivering the radiation by sequential opening and closing of the two

1938 adjacent leaves. The same dip was observed in the DMG measurement, shown in

1939 Figure 7.1(b). For the ﬁlm analysis, the distance between the two outside peaks and

1940 the center peak was measured as X1 and X2. The MLC alignment error is deﬁned as

1941 half of the diﬀerence between X1 and X2. For the DMG measurement, the MLC error

1942 is deﬁned as the half of the distance between the centers of the proﬁles obtained at

◦ ◦ 1943 gantry angles 0 and 180 . The EDR2 ﬁlm and the DMG agree with each other within

1944 the measurement uncertainty with MLC alignment error of 0.71 ± 0.09 mm and 0.55

1945 ± 0.10 mm, respectively. The uncertainties quoted represent the 1 standard deviation

1946 of the three measurements. This error is within the tolerance limit of ± 1.5 mm as

1947 recommended by Balog & Soisson (2008). The diﬀerences between the FWHM of the

◦ ◦ 1948 proﬁles taken with the gantry positioned at 0 and 180 was found to be 0.05 mm and

1949 0.09 mm for the ﬁlm and DMG measurements, respectively. These corresponded to

1950 0.34 mm and 0.61 mm shift from the isocenter, which is within the accuracy of the

1951 laser guide and the physical setup uncertainty.

1952 7.3.2 Leaf open time threshold

1953 The average of the ﬁve central silicon strips was used for the leaf open time computa-

1954 tion. The leaf open time was calculated as the full width at half maximum (FWHM) of

1955 each individual projection. There were ﬁve projections within a modulation bracket.

1956 The mean and 1 standard deviation of the ﬁve projections was calculated. The results 7.3. Results and discussions 114

Figure 7.1: (a) EDR2 film measured leaf alignment test. The distances from the two outside peaks to the central peak were X1 = 24.94 ± 0.09 mm and X2 = 23.53 ± 0.09 mm. The MLC alignment error was taken as half the difference between these two distances (dX = 0.71 ± 0.09 mm) and (b) DMG measured leaf alignment test. Half the distance between the centre of each profile gives the MLC alignment error (dX = 0.55 ± 0.10 mm).

1957 are depicted in Figure 7.2. The graphs were plotted with the actual percent open time

1958 versus the programmed percent open time for 50, 100, 200 and 303 projection times

1959 for leaves 4, 32 and 61. From the ﬁgure, it is clear that there exist non linearity for

1960 the projection times <200 ms. Balog et al. (2003b) reported that the leaf latency is

1961 linear within the range of 25% - 80% of the programmed percent open time for 200

1962 ms projection time. This is extended to 10% - 90% for the 400 ms projection time.

1963 Various reports have quoted a leaf open time of 20 ms (Balog et al., 2003b; Fenwick

1964 et al., 2004; Balog & Soisson, 2008). Using the DMG, the leaf open time was taken

1965 to be the time taken for the actual leaf open time to equal the programmed leaf open

1966 time. From Figure 7.2 this was found to be approximately 40%, 20% and 10% for 50

1967 ms, 100 ms and 200 ms projection times, respectively. This equates to an open time

1968 threshold of approximately 20 ms. Leaf open time threshold aﬀects shorter projection

1969 time, as the time for the leaf to move from one state to the other constitutes a larger

1970 fraction of the programmed projection time. This was clearly demonstrated in the 50 7.3. Results and discussions 115

Figure 7.2: Actual measured leaf open times plotted against programmed leaf open times for (a) 50 ms, (b) 100 ms, (c) 200 ms and (d) 303 ms projection time. A linear curve has been ﬁtted for guidance. The error bars represent 1 s.d. of the mean of the ﬁve projections within the same modulation bracket. 7.3. Results and discussions 116

Figure 7.3: Comparison of the DMG and MVCT measured leaf open time for leaf 32 (a) 50 ms, (b) 100 ms, (c) 200 ms and (d) 303 ms projection time. A linear curve has been ﬁtted for guidance. 7.3. Results and discussions 117

1971 ms and 100 ms leaf open time. Leaf open time curves for each leaf was slightly diﬀerent

1972 and appears to be geographically related as shown in Figure 7.2 (a) and (b) where leaf

1973 4 and 61 latency proﬁles were generally slightly lower than the latency proﬁle of leaf

1974 32. However, the point at which the actual leaf open time is equal to the programmed

1975 leaf open time is approximately the same for all three leaves. Therefore a leaf open

1976 time threshold equal to the highest time at which actual equals programmed leaf open

1977 time would account for the leaf open time of all leaves.

1978 Figure 7.3 shows the comparison between the DMG and tomotherapy CT detector

1979 measurements showing good agreements between the two detectors. The maximum

1980 diﬀerence between the two detectors was <5% for the leaf open times greater than 20

1981 ms. The close agreement between the two detectors demonstrates the suitability of

1982 the DMG as an independent QA for the tomotherapy.

1983 According to Kapatoes et al. (2001a), the eﬀect of leaf activation latency on the to-

◦ 1984 tal delivery is a rotational shift of <2 , which is insigniﬁcant in terms of total deposited

1985 dose and is taken into account in the treatment delivery software. Two methods were

1986 also suggested to reduce the eﬀect of leaf open time on the total dose delivery: a)

1987 inclusion of leaf open time in the EOP consideration: projection times <20 ms are

1988 discarded from the planning sinogram, and (b) increase the programmed leaf open

1989 time at the expense of increased treatment time and decreased throughput. However,

1990 as a result, the treatment plan would generally be less complex, reducing the dynamic

1991 range of beam modulation available for treatment planning. The former method which

1992 discards projection times less than the leaf open time threshold, is used in the current

1993 treatment planning software version.

1994 The DMG and its readout system have a spatial resolution of 0.2 mm and temporal

1995 resolution of 3 ms therefore the dose proﬁles in the direction of leaf travel as the leaf

1996 traversed the ﬁeld was recorded at a series of time points (Figure 7.4). The binary 7.3. Results and discussions 118

Figure 7.4: DMG measurement of the leaf opening and closing for a 200 ms projection time. Each row of the ribbon represents a 3 ms acquisition. Red arrows point to the direction of leaf travel.

1997 MLC leaf motion was captured, depicting a wave-like proﬁle across the 128 silicon

1998 strips for the ﬁrst 6 - 9 ms. The DMG data was acquired at 3 ms interval, while the

1999 linac pulse repetition frequency was set to 300 Hz (eﬀective pulse period of 3.3 ms),

2000 the DMG recorded a slightly lower signal at every 10 acquisitions due to the missed

2001 radiation pulse.

2002 7.3.3 Leaf ﬂuence output factor

2003 The LFOF for three leaves (leaf 31, 47 and 62) representing diﬀerent position of

2004 tomotherapy lateral proﬁle were measured. The LFOF due to the opening of the

2005 upper leaf, lower leaf and both adjacent leaves to the leaf of interest (LOI) were shown

2006 in Table 7.2. As expected, the highest LFOF was measured when both adjacent leaves

2007 were opened simultaneously. LFOF for LOI and its upper leaf opening were also higher

2008 than the lower leaf opening. This is due to the sloping tomotherapy beam proﬁle that

2009 decreases with lateral distance from the central axis. The DMG measurements showed

2010 a larger error compared to the tomo detector measurements. This is because the signal 7.4. Conclusion 119

Table 7.2: LFOF measurements by DMG and tomo detectors. The uncertainties represent 1 s.d. of the mean. upper leaf lower leaf both leaves LOI DMG CT detector DMG CT detector DMG CT detector L33 1.042± 1.049± 1.053± 1.051± 1.082± 1.081± 0.007 0.001 0.008 0.002 0.006 0.002 L47 1.052± 1.056± 1.052± 1.039± 1.084± 1.077± 0.014 0.000 0.015 0.001 0.013 0.001 L62 1.055± 1.062± 1.037± 1.036± 1.081± 1.079± 0.006 0.001 0.007 0.001 0.006 0.001

2011 measured by the DMG is three orders of magnitude lower than the signals recorded by

2012 the tomo detectors. Despite the lower signal, the maximum error of 1.5% corresponds

2013 to <9 digital counts or three beam pulses, showing the high accuracy of the DMG

2014 measurements. Good agreement within error can be seen between the DMG and the

2015 CT detectors.

2016 7.4 Conclusion

2017 Helical tomotherapy with its unique dose modulation mechanism resulted in unique

2018 dosimetric considerations and QA procedures that are not present in convention linac

2019 QA procedures. Current QA procedures require the use of the on-board MVCT detec-

2020 tor ring therefore could lack the rigor of machine-independence. This study presents

2021 results of routine QA procedures obtained using a novel solid state dosimeter.

2022 The 128 channel silicon strip detector, dose magnifying glass (DMG) was used to

2023 perform three tomotherapy QA tests: MLC leaf alignment, leaf open time threshold

2024 and leaf ﬂuence output factor. The MLC alignment error was found to be 0.71 ± 0.09

2025 mm compared to 0.55 ± 0.10 mm with EDR2 ﬁlm measurements. This alignment

2026 error was within the tolerance limit of 1.5 mm. The leaf open times of 50 ms to 400

2027 ms were measured and were shown to be in good agreement with the CT detector 7.4. Conclusion 120

2028 measurements. The real-time leaf motion was captured at high temporal resolution

2029 of 3 ms. MLC leaf open time threshold was found to be approximately 20 ms. The

2030 LFOF measured with the DMG ranged from 1.036 ± 0.007 to 1.084 ± 0.013 and agreed

2031 within error limits with the LFOF measured with the CT detectors. The DMG was

2032 found to be a suitable QA tool for the independent veriﬁcation of selected tomotherapy

2033 QA procedures, allowing real time acquisitions of high spatial and temporal resolution

2034 measurements and validation of current QA measurement techniques that utilize the

2035 on-board CT detector ring. 2036 Chapter 8

2037 Radiation response and basic

2038 characterisation of the Magic Plate

2039 8.1 Introduction

2040 The use of epitaxial technology for radiation detectors in medical radiation therapy

2041 is still rare. In the framework of the European Integrated Project MAESTRO, a

2042 dosimeter utilising this technology was designed and prototyped (Menichelli et al.,

2043 2007). The preliminary characterisation of this device was reported by Talamonti

2044 et al. (2007) and Bruzzi et al. (2007). The Magic Plate (MP) is a novel 2D radiation

2045 detector array based on the proprietary epitaxial diode design and the ‘drop-in’ diode

2046 mounting technique of the CMRP. It would require a full characterisation prior to

2047 its used in a clinical setting. This chapter describes the following characterisation

2048 measurements that were carried out:

2049 (a) Radiation damage eﬀect

2050 (b) Percent depth dose

2051 (c) Dose linearity

121 8.2. Materials and methods 122

2052 (d) Energy dependence

2053 (e) Temperature dependence

2054 (f) Field size dependence

2055 (g) Dose per pulse response

2056 (h) Angular response, and

2057 (i) Beam perturbation studies

2058 8.2 Materials and methods

2059 8.2.1 Packaging of the Magic Plate

2060 The design of the epitaxial diode attempts to minimise energy dependence of the diodes

2061 by avoiding the use of metal contacts above the sensitive area. This was achieved by

2 + 2062 using thin aluminium contacts on the periphery of the 0.5 × 0.5 mm n ion implanted

2063 regions. The trade oﬀ in this design was that the p-n junctions of the diodes were

2064 exposed to light. Due to the sensitivity of the silicon detectors to visible light, the MP

2065 needs to be packaged in a way that it is properly sealed from light photons. Figure

2066 8.1 shows the MP packaging when used as (a) a transmission detector mounted on

2067 the linac accessory slot and for (b) measurement using the I’mRT phantom (IBA

2068 Dosimetry). Figure 8.2 shows the schematic diagram of the packaging.

2069 For the transmission mode, the MP was sandwiched between two pieces of clear

2070 plastic sheets (100 μm), solid water plates (1 mm) and black plastic sheets (80 μm).

2071 The clear plastic sheets protect the MP diodes from dust and accidental contacts. The

2072 two solid water plates serve three purposes, (i) to shield the detector from ambient

2073 lights, (ii) for mechanical strength and protection of the MP and (iii) as scattering 8.2. Materials and methods 123

Figure 8.1: MP packaging: (a) MP mounted on the Y-shaped Perspex frame. The readout electronics box is connected to the MP on the left. (b) MP sandwiched between two 5 mm solid water slabs for measurements in the I’mRT phantom.

2074 material to increase the signal generated in the diodes. Finally, a black plastic sheet

2075 was used to wrap the MP and the solid water plates together to reduce light leaking

2076 to the detector. The whole assembly was then clamped between two 6 mm Perspex

2077 slabs shaped like a “Y”. For detail design of the perspex frame and solid water slabs,

2078 see appendix A.

2079 The MP can also be used as a 2D planar diode array for dose measurement in

2080 phantom. Two pieces of 5 mm solid water plates were machined to ﬁt the Kapton

2081 substrate. The width of the solid water plates was 160 mm, the same as the width of

2082 the kapton substrate the I’mRT phantom’s cavity. This packaging allowed the MP to

2083 be positioned at any depths in the phantom.

2084 8.2.2 Radiation damage studies

2085 The sensitivity of silicon radiation diodes may change over time. This is because

2086 ionising radiation creates defects in the silicon lattices. Depending on the type of 8.2. Materials and methods 124

Figure 8.2: Schematic diagram of the MP packaging for the use in (a) transmission mode and (b) in an I’mRT phantom.

2087 radiation source, either point defects or cluster defects (Konozenko et al., 1971; Kraner,

2088 1984; Damerell, 1995; Moll, 1999) are created. These defects become recombination

2089 centres for the minority charge carriers, reducing the minority carriers’ lifetime and the

2090 sensitivity of the detector (Grusell & Rikner, 1984). The eﬀect of radiation damage on

2091 silicon diodes were also manifested in the dose per pulse dependence (Rikner & Grusell,

2092 1983; Grusell & Rikner, 1984; Wilkins et al., 1997; Saini & Zhu, 2004), increased of

2093 dark current and temperature dependence (Grusell & Rikner, 1986; Van Dam et al.,

2094 1990; Saini & Zhu, 2002). In most published literature, the eﬀect of radiation damage

2095 on silicon diodes was shown as a decrease in the diode sensitivity (Rikner & Grusell,

2096 1983; Bruzzi et al., 2007).

2097 During the initial stage of the MP characterisation, the MP diodes showed an ab-

2098 normal behaviour. The detector sensitivities appeared to increase with accumulated

2099 radiation. This is in contrast with the radiation damage eﬀect that was widely pub-

2100 lished in existing literatures where the detector sensitivity decreased with accumulated

2101 radiation. In the general literature, the discussions were generally based on low resis- 8.2. Materials and methods 125

2102 tivity diodes (in the order of 0.1 - 0.5 Ω·cm). The resistivity of the epitaxial diode

3 2103 used in the Magic Plate detectors is in the order of 10 higher than most commercial

2104 diodes. This behaviour observed in the MP diodes bears similarity to the radiation

2105 damage eﬀect in a metal oxide semiconductor (MOS) structure (Hughes & Benedetto,

2106 2003; Oldham & McLean, 2003).

2107 The design of the epitaxial detector was brieﬂy described in chapter 3. The epitaxial

2108 diode have a thick layer of a ﬁeld oxide (SiO2) which functions as a passivation layer on

2109 the surface of the p-Si diode. At the Si-SiO2 interface, positive charges (holes trapped

2110 in the border of the ﬁeld oxide closed to the interface) are always present and will

2111 increase with radiation (Pellegrini et al., 2007). The presence of these positive charges

2112 are compensated by a thin layer of mobile electrons in the epitaxial layer due to the

2113 repulsion of the holes (Damerell, 1995; Rosenfeld et al., 1993). During the initial

2114 investigation, the sensitivities of the epitaxial diodes were found to be continually

2115 increasing at the rate of 0.5% per Gy. This indicated that the positive charge build

2116 up at the Si-SiO2 interface had resulted in an increase in the depletion region under

+ + 2117 the ﬁeld oxide between the n core region and the p spray region. This eﬀectively

2118 increased the sensitive volume of the diode.

2119 To resolve this issue, the MP was given a 41.5 kGy irradiation to saturate the

2120 interface traps and stabilize the detector sensitivity. The dose was delivered in two

2121 parts. The ﬁrst 1.3 kGy was delivered using 6 MV photon energy of medical linear

2122 accelerators and the rest was delivered by a high dose rate (3 kGy/hr) cobalt-60 source

2123 at the Gamma Technology Research Irradiator (GATRI) facility, Australian Nuclear

2124 Science and Technology Organisation (ANSTO). 8.2. Materials and methods 126

Table 8.1: Measurement setup for the dose per pulse measurement. Measurement setup Setup description SDD range Dose per (cm) pulse range (Gy/pulse) Free air geometry MP sandwiched between two 60.5 - 100.0 1.07 × 10−4 - pieces of 1 mm solid water phan- 3.39 × 10−4 tom and suspended in air Open ﬁeld 1.5 cm depth in solid water phan- 87.0 - 154.4 1.20 × 10−4 - tomwitha10× 10 cm2 open ﬁeld 3.79 × 10−4 2.0cmPbattenua- 1.5 cm depth in solid water phan- 87.0 - 154.4 3.53 × 10−5 - tion tom, radiation beam blocked by 1.17 × 10−4 2.0 cm Pb block 4.0cmPbattenua- 1.5 cm depth in solid water phan- 87.0 - 154.4 1.33 × 10−5 - tion tom, radiation beam blocked by 4.45 × 10−5 4.0 cm Pb block 6.0cmPbattenua- 1.5 cm depth in solid water phan- 87.0 - 154.4 4.80 × 10−6 - tion tom, radiation beam blocked by 1.70 × 10−5 6.0 cm Pb block Under closed mul- 1.5 cm depth in solid water phan- 87.0 - 154.4 1.71 × 10−6 - tileaf collimator tom, MLC closed 5.81 × 10−6 (MLC)

2125 8.2.3 Dose per pulse response measurement

2126 For the dose per pulse measurement, a 222-fold dose per pulse change was achieved

−6 −4 2127 (1.71 × 10 to 3.79 × 10 Gy/pulse). This was done with six measurement setups

2128 involving the use of attenuators such as lead (Pb) blocks and under a closed multileaf

2 2129 collimator (MLC) (Table 8.1). The MP was irradiated with a 10 × 10 cm ﬁeld size

2130 beam for a ﬁxed number of monitor units (MUs).

2131 Two pieces of Pb blocks with the thicknesses of 2.0 and 4.0 and a minimum area

2 2132 of 8 × 8cm were used. The Pb blocks were placed downstream from the linac beam

2133 output using a modiﬁed total body irradiation (TBI) tray (Figure 8.3). At this source

2134 to attenuator distance of 62 cm, the linac beam was completely blocked by the Pb

2135 block. However, the use of Pb attenuators and closing the MLC hardened the linac

2136 beam spectrum. This may result in a diﬀerent energy response in the silicon detector, 8.2. Materials and methods 127

Figure 8.3: Modiﬁed accessory tray to position the Pb attenuator closed to linac gantry.

2137 particularly to the lower photon energies. To address this issue, ﬁve dose rates (or

2138 dose per pulse data points) were obtained by varying the source to detector distances

2139 (SDDs) in each setup (Table 8.1). A dynamic range of dose per pulse was achieved

2140 with a slight overlapping of the adjacent dose per pulse ranges. In this way, the eﬀect

2141 of the change of the beam spectrum on the detector’s dose per pulse response was able

2142 to be isolated from the actual dose per pulse response.

2143 Four dosimeters were used: a Farmer ion chamber (NE-2571, Nuclear Enterprise),

2144 a CC13 ion chamber (Scanditronix Welh¨ofer), a highly doped p-type commercial diode

2145 (EFD-3G, Scanditronix Welh¨ofer) and the MP. See Table 8.2 for the speciﬁcations of

2146 the respective dosimeters. The Farmer ion chamber was used as the reference chamber

2147 for all the measurements in phantoms while the CC13 ion chamber was used as the

2148 reference chamber for the free air geometry measurements. The CC13 ion chamber

2149 was also cross calibrated with the Farmer ion chamber. 8.2. Materials and methods 128

Table 8.2: Speciﬁcations of the dosimeters used. Dosimeter NE-2571 CC13 EFD-3G MP Farmer Compact Commercial Detector type type ion MP chambers diode chamber High resistivity Low resistivitya (100 Ω·cm) Device details ··· ··· p-type silicon p-type epitaxial diode diode Sensitive volume 0.6 cm3 0.13 cm3 2.9 × 10−4 cm3 1.25 × 10−5 cm3 6.3 mm 6.0 mm Sensitive area 2.5 mm dia. 0.5 × 0.5 mm2 dia.b dia. Sensitive thick- 24.0 mm 5.8 mm 0.06 mm 0.05 mm ness/length 1.5 × 1.5 × 0.05 Physical size ··· ··· 7.0 mm dia. mm3

aActual resistivity was not available. bdiameter

2150 The MP dose per pulse response was measured and compared with the EFD-3G

2151 diode. The MP and the Farmer ion chamber measurements were made at the depth

2152 of 1.5 cm in the solid water phantom while the EFD-3G diode was measured at the

3 2153 same depth in a small water tank of 29 × 20 × 17 cm .

2154 The number of dose pulses delivered was calculated from the information in the

2155 linac manual. The Varian Clinac 21EX (Varian, Palo Alto, USA) linear accelerator

2156 has a pulse repetition frequency of 360 Hz for a 6 MV photon energy. From the

−4 2157 known dose rate, the dose per pulse was estimated to be 2.78 × 10 Gy/pulse at

2158 1.5 cm depth and the SSD of 100 cm. The average linac pulse rate was assumed

2159 constant between measurements, i.e. although the linac slightly modiﬁes its pulse rate

2160 to maintain constant dose rate.

2161 The diode sensitivity is the ratio of the charge measured by the silicon diode to 8.2. Materials and methods 129

2162 the charge measured by the ion chamber at the same position and setup (Eq.8.1).

Rdetector count Sdetector,f,a = unit = (8.1) Rion chamber f,a nC

2163 Where,

2164 Detector = EFD-3G diode or MP

2165 f = source to detector distance (cm)

2166 a = conﬁguration (free air geometry, open ﬁeld, Pb or closed MLC)

2167 The dose per pulse response was taken as the detector sensitivity of the MP or the

2168 EFD-3G diode at any dose per pulse normalised to 1 at the dose per pulse of 2.78 ×

−4 2169 10 Gy/pulse.

2170 8.2.4 Percent depth dose measurement

2171 The percent depth dose proﬁle of a 6 MV photon energy was measured in a solid

2 2172 water phantom. A 10 × 10 cm ﬁeld size was used and measurements were made at

2173 selected depths from 1.2 to 300 mm. Due to the thin epitaxial layer (50 μm), the MP

2174 can be used to measure the build up region of the depth dose curve. The result was

2175 compared with the Attix chamber measurements for the build up region (0 - 15 mm

2176 depth) [courtesy of Dr. Bradley Oborn, Illawarra Cancer Care Centre, Wollongong]

2177 and the standard data measured using a CC13 ion chamber (3 - 300 mm depth) in

2178 water.

2179 8.2.5 Dose linearity measurement

2180 The MP was set up in a solid water phantom at the SSD of 10 cm and depth of 1.5 cm

2 2181 in a 10 × 10 cm radiation ﬁeld. The dose linearity was measured for the dose range

2182 of 5 to 1000 cGy using a 6 MV photon energy of a linac. 8.2. Materials and methods 130

Figure 8.4: Schematic drawing of the ‘face-up’ and ‘face-down’ conﬁgurations for the MP diode.

2183 8.2.6 Energy dependence measurement

2184 The energy response of the MP was studied using an orthovoltage machine (Guymay

2185 DX 3300) for the nominal energies of 75 kV - 250 kV and a Varian Clinac 21EX linear

2186 accelerator (Varian, Palo Alto, USA) for the photon energies of 6 MV and 10 MV. For

2187 the 50 - 150 kV tube voltages, a 100 mm diameter circular applicator was used and

2188 the detector was placed at a focus to surface distance of 312 mm. For the 200 and

2189 250 kV tube voltages, 100 mm diameter circular applicator was used with detector

2190 positioned at the focus to surface distance of 512 mm. For the high energy photons,

2191 the detector was irradiated at the depth of maximum dose (dmax of 1.5 cm for 6 MV

2192 and 2.0 cm for 10 MV).

2193 Energy response of the MP was evaluated for four diﬀerent geometries (Table 8.3)

2194 and the readings were normalized to 1 at the energy of 6 MV. The ‘face-up’ and

2195 ‘face-down’ conﬁgurations are shown in Figure 8.4.

2196 8.2.7 Temperature dependence measurement

2197 The variation of silicon diode sensitivity with temperature is well known (Van Dam

2198 et al., 1990; Saini & Zhu, 2002). Temperature dependence of silicon diode is a major

2199 concern when the diode is use for in-vivo measurements due to the large temperature 8.2. Materials and methods 131

Table 8.3: Energy dependence measurement conﬁgurations. Geometry Description A ‘face-up’, in phantom∗ MP sandwiched between two pieces of 1 mm solid water plates and position in the ‘face- up’ orientation on top of 11.6 cm of solid water backscatter material B ‘face-up’, free air geometry MP sandwiched between two pieces of 1 mm solid water plates in the ‘face-up’ orientation, without the backscatter material C ‘face-down’, in phantom∗ MP sandwiched between two pieces of 1 mm solid water plates and position in the ‘face- down’ orientation on top of 11.6 cm of solid water backscatter material D ‘face-down’, free air geometry MP sandwiched between two pieces of 1 mm solid water plates in the ‘face-down’ orientation, without the backscatter material

∗Build up thickness is photon energy speciﬁc. For the kV energies, a minimum build up of 1 mm solid water was always present. For the 6 MV and 10 MV photon energies, the measurement depths were at 1.5 cm and 2.0 cm.

2200 variation from ambient room temperature to the patient body temperature (Jursinic,

2201 2001; Welsh & Reinstein, 2001; Marre & Marinello, 2004). The MP was designed to be

2202 used as a 2D transmission detector and planar dose measurement in phantoms. The

2203 temperature variation in these measurement conditions is expected to be less dramatic.

2204 The temperature dependence of the MP was measured between the temperatures

◦ 2205 rangeof20to27C. The experiment was setup as in Figure 8.5. The two pieces

2206 of 1 mm solid water plates were removed. Temperature control was achieved using

2207 a thermoelectric temperature controller (Model TC-36-25) and a Peltier cell. Two

2208 temperature probes were used. Probe T1 was the temperature controller probe. It

2209 was attached to a metal plate which functioned as a heat transfer tool. Probe T2 was

2210 the temperature monitoring probe attached over the top layer of the clear plastic sheet

2211 above the kapton substrate. The whole set up was covered with a white photographic

2212 cloth to protect the detector from the ambient light. The radiation source was provided 8.2. Materials and methods 132

Figure 8.5: Experiment setup for the temperature dependence measurement.

2213 by the linac beam. The dose rates were varied by changing the SDD [approximately

2214 94.5, 128.5 and 161.5 cm]. Five temperature points were measured in the temperature

◦ 2215 rangeof20to27C. Readings were taken 10 minutes after each temperature change

2216 to allow for the temperature in the system to reach equilibrium.

2217 8.2.8 Field size dependence measurement

2218 The ﬁeld size dependence of the MP was evaluated for selected square ﬁeld sizes from

2 2 2219 5 × 5cm to 40 × 40 cm . Measurements were made in two conditions;

2220 (a) In a solid water phantom at the SDD of 101.5 cm, 1.5 cm depth, and

2221 (b) In the free air geometry at the SDD of 58 cm

2222 The MP measurements were compared to the standard data measured using a

2223 Farmer ion chamber at the SDD of 101.5 cm and 1.5 cm depth in a solid water

2224 phantom. 8.2. Materials and methods 133

2225 8.2.9 Angular response measurement

2226 To date, most of the large area 2D detector arrays were designed for orthogonal beam

2227 measurements due to the angular dependence of the arrays. IMRT treatment veriﬁca-

2228 tions are often performed with all beams delivered from a ﬁxed gantry angle. However,

2229 recently, various groups have explored the use of these arrays for rotational beam mea-

2230 surements (Bhardwaj et al., 2009; Jursinic et al., 2010). The use of 2D array detectors

2231 for rotational beam measurements requires the characterisation of the angular depen-

2232 dence of the devices. Bhardwaj et al. (2009) showed that the MATRIXX system (IBA

2233 Dosimetry, Germany) has a maximum angular response of 7.7% when beams were

◦ 2234 delivered at gantry angle 180 . Jursinic et al. (2010) found that the metal contacts of

2235 the diode resulted in ± 20% of angular dependence in the MapCHECK (Sun Nuclear,

2236 Melbourne, Fl) detector. Modiﬁcation of the detector design by inserting copper pieces

2237 to oﬀset the diode asymmetry and Lucite sheets to ﬁll in air gaps reduces the angular

2238 dependence to ± 2% (Jursinic et al., 2010).

2239 The angular response of the MP was carried out using the I’mRT phantom (IBA

2240 Dosimetry, Germany). The MP was sandwiched between two pieces of 5 mm thick

2241 solid water phantoms. The MP was positioned at the depth of 9 cm in the phantom

2 2242 at the SDD of 100 cm. A ﬁxed number of MUs were delivered with a 16 × 16 cm

2243 radiation ﬁeld size. The larger ﬁeld size was chosen in order to cover all the 11 × 11

2 2244 MP diodes extending over an area of 10 × 10 cm . A Varian Clinac iX (Varian, Palo

2245 Alto, USA) linear accelerator was used for this measurement. The angular dependence

2246 of the MP was studied with the MP in the faced up conﬁguration for the gantry angles

◦ ◦ ◦ 2247 of 0 to 180 at 15 intervals. The asymmetric design of the epitaxial diode (Figure

2248 8.4) may result in directional dependence when the beam is entering from the ‘face-

◦ ◦ 2249 up’ or ‘face-down’ conﬁguration. In addition, for the gantry angles 120 to 180 ,the

2250 radiation beam passing through the linac couch may be attenuated. This may result 8.2. Materials and methods 134

2251 in a diﬀerent silicon response due to the energy dependence of the silicon. To isolate

2252 the angular response of the epitaxial diode due to the silicon anisotropy, the MP

◦ ◦ 2253 was irradiate with the ‘face-down’ conﬁguration for the gantry angles 0 to 90 and

2254 compared with the corresponding angular response made with the MP in the ‘face-up’

2255 conﬁguration.

2256 Gafchromic EBT2 ﬁlm (ISP,Wayne, NJ) was used as the benchmarking dosimeter

2 2257 in this study. The EBT2 ﬁlms were cut into the size of 101.5× 120.0 mm and po-

2258 sitioned at the centre of the I’mRT phantom at the depth of 9 cm. The EBT2 ﬁlms

2259 were scanned with an A3 ﬂatbed scanner (Epson Expression 10000XL). Each ﬁlm was

2260 scanned six times. Only the last three scans were kept for image analysis. This was

2261 to ensure that the scanner was suﬃciently warmed up and thus ensuring consistency

2262 in the scanned ﬁlms (Fuss et al., 2007). The ﬁlms were scanned in 48-bit RGB colour

2263 with the scanning resolution of 96 dpi (equivalent to pixel size of 0.265 mm). Care

2264 was taken to scan the ﬁlm in the same orientation at the centre region of the scanner

2265 to reduce scanner induced non-uniformity (Butson et al., 2006a). A set of calibration

2266 ﬁlms was also irradiated in the same experimental session. Analysis of the images was

2267 done using ImageJ version 1.43U software (National Institute of Health, USA) and

2268 Matlab (The MathWorks Inc., Natick, MA). The red channel of the last three scans

2269 were averaged for each ﬁlm (Matney et al., 2010) and the pixel intensities were con-

2270 verted to dose values based on the calibration curve. A 3 by 3 pixels two dimensional

2271 median ﬁlter was applied onto the image to reduce image noise (Andres et al., 2010).

2272 The data points located at the x- and y-coordinates that corresponded to the spatial

2273 positions of the MP diodes were sampled.

2274 The angular response of the MP diodes was evaluated solely in the azimuth di-

2275 rection. For simplicity, the angular coeﬃcients for the 11 diodes positioned along the

2276 same column were averaged. 8.2. Materials and methods 135

2277 The pixel intensities at the film edges were affected by image artifacts. The films

2278 were cut to the width of 101.5 mm to economise on the use of ﬁlms. The drawback

2279 of this was that the ﬁlm edges were very close to the region of interest. This resulted

st th 2280 in incorrect pixel values for the regions that coincide with the 1 and 11 columns

2281 on the MP detector (at ± 5 cm oﬀ axis distances). Hence, only the mean values of

nd th 2282 thedetectorswithinthe2 to 10 columns were used to evaluate for the angular

nd 2283 dependence. Both data sets were ﬁtted with 2 degree polynomial ﬁts and the mean

2284 signal for the detectors located at the ± 5 cm oﬀ axis distance were extrapolated from

2285 the curve ﬁtted equations.

2286 The angular coeﬃcient of the MP diode was taken as the ratio of the signal mea-

2287 sured by the MP over the EBT2 dose values (Eq.8.2).

MPi counts angular coeﬃcient,kai = unit = (8.2) EBT2i cGy

2288 Where, the index i represent the ith detector column. The angular response of

2289 the MP diode was taken as the ratio of the angular coeﬃcient of an arbitrary gantry

◦ 2290 angle, θ to the angular coeﬃcient of the gantry angle 0 (Eq.8.3). The angular response

2291 correction for the MP diodes is the inverse of the angular response.

kaθi angular response,Caθi = (8.3) ka0i

2292 8.2.10 Beam perturbation measurements

2293 The use of the MP as a transmission detector may attenuate the beam or inﬂuence

2294 the secondary electron contamination of the photon beam (Poppe et al., 2010). This

2295 necessitates the investigation of the radiation beam perturbation eﬀect. Currently,

2296 there are three other transmission detectors that were reported in the literature,

2297 (i) the DAVID system (Poppe et al., 2006b, 2010), 8.2. Materials and methods 136

2298 (ii) the Integral Quality Monitoring system (IQM) (Islam et al., 2009), and

2299 (iii) the IBA COMPASS detector (Venkataraman et al., 2009).

2300 All of these devices showed increased surface dose and dose within the build up

2301 region. For the dose beyond the dmax, the presence of transmission devices caused

2302 beam attenuation up to 7% (Islam et al., 2009; Venkataraman et al., 2009; Poppe

2303 et al., 2010).

2304 The beam perturbation study for the MP was carried out to study two possible

2305 inﬂuences of the MP on the radiation beam:

2306 (i) surface dose and dose within the build up region of a depth dose curve, and

2307 (ii) transmission factor at the depth of 10 cm in a solid water phantom

2308 8.2.10.1 Surface dose and build up region measurements

2309 The surface dose in the build up region (0 to 15 mm depth) of a 6 MV photon energy

2310 were measured. Measurements were made in a solid water phantom using a Markus

2311 type (PTW, Freiburg, Germany) parallel plate ion chamber. Markus type chambers

2312 are known for their over response due to the small guard ring (Nilsson & Montelius,

2313 1986; Rubach et al., 1986). Various methods were developed to correct for this over

2314 response (Velkley et al., 1975; Gerbi & Khan, 1990; Mellenberg, 1990; Rawlinson et al.,

2315 1992). In this thesis, the Rawlinson correction (Rawlinson et al., 1992) was used to

2316 correct for the Markus chamber measurements. The Markus chamber used was an

2317 original Markus chamber, type 23343. The chamber dimension used for the Rawlinson

2318 correction calculation was obtained from Chen et al. (2010).

2 2 2319 Two ﬁeld sizes (20 × 20 cm and 30 × 30 cm ) were measured at two source to

2320 surface distances (SSDs) of 80 cm and 90 cm. Measurements were made with the MP

2321 mounted in the accessory tray (MP ﬁeld) and with an open ﬁeld. The measurements

2322 were normalised to the depth of maximum dose. 8.3. Results 137

2323 8.2.10.2 Transmission factor measurements

2324 When a block tray or physical wedge is placed on the linac accessory tray, a tray or

2325 wedge transmission factor is applied to correct for the attenuation of the photon beam

2326 quality due to the tray/wedge. The tray transmission factor is deﬁned as a ratio of

2327 the dose in a phantom with and without a tray in the beam for the same number of

2328 monitor units delivered at a ﬁxed source to detector distance and depth in phantom

2329 (Sharma & Johnson, 1994; van Kleﬀens et al., 2000).

2330 The transmission factor of the MP was measured at the SDD of 100 cm and the

2 2 2331 depth of 10 cm in solid water phantom for three ﬁeld sizes (10 × 10 cm ,20× 20 cm

2 2332 and 30 × 30 cm ). A CC13 ion chamber was used for this measurement.

2333 8.3 Results

2334 8.3.1 Radiation damage

2335 Figure 8.6 shows the detector sensitivity as a function of accumulated radiation dose.

2336 The positive charge build up at the Si-SiO2 layer due to the accumulated radiation

2337 caused the apparent increase in the sensitivity of the detector by a factor of two after

2338 a dose of 41.5 kGy. Figure 8.7 shows the medium term reproducibility of the MP up to

2339 64 days, post irradiation of the 41.5 kGy. The post irradiation coeﬃcient of variation

2340 was 2.1%. This shows that the MP sensitivity was stable after it was given a dose of

2341 41.5 kGy. However, it is prudent to perform periodic dose recalibration if the detector

2342 is being used for absolute dose measurements. No observation in radiation damage of

2343 the kapton was recorded. 8.3. Results 138

Figure 8.6: Sensitivity of the detector as a function of the accumulated radiation dose. The error bars represent the ± 1 standard deviation of the channels located within the central 20-80% region of the 11 × 11 MP detector. The left axis shows the detector sensitivity expressed as counts/1Gy while the right axis shows the detector sensitivity normalised to the ﬁrst irradiation of 1 Gy.

Figure 8.7: Reproducibility of the MP. The error bars represent ± 1 standard deviation of the mean of three measurements. 8.3. Results 139

2344 8.3.2 Dose per pulse response

2345 Figure 8.8 showed the dose per pulse response measurements for a highly doped com-

2346 mercial p-type Si diode (EFD-3G) and the epitaxial diode of the MP. The EFD-3G is

2347 an unshielded diode with a thin sensitive volume of 0.06 mm. The diode showed dose

2348 per pulse independence for almost all dose rates except for the measurements made in

2349 the free air geometry where the diode showed a dropped in diode sensitivity relative

2350 to the ion chamber measurements.

2351 The epitaxial diode has a thin sensitive volume of 0.05 mm and is lightly doped

2352 (high resistivity = 100 Ω·cm). It exhibited a larger dose per pulse dependence com-

2353 pared to the EFD-3G diode. The onset of the drop in the epitaxial diodes’ sensitivity

−4 2354 was observed at the dose per pulse of 1 × 10 Gy/pulse.

2355 The MP’s dose per pulse dependence was shown as a drop in the diode sensitivity

2356 relative to the ion chamber measurements at the same dose rate. The same trend was

2357 also observed in the EFD-3G diode at a higher dose rate (with the free air geometry

2358 setup). However, this trend was in contrast with the dose per pulse dependence of the

2359 Dose Magnifying Glass (DMG) which showed an increase of diode sensitivity as the

2360 dose rate increases (see Chapter 4). The reasons/explanations for both of these trends

2361 (either increased or decreased of diode sensitivity with increased dose rates) have been

2362 detailed in various literature (Wirth & Rogers, 1964; Long et al., 1983; Fjeldly et al.,

2363 2001; Alexander, 2003; Shi et al., 2003). It will be qualitatively discussed in the

2364 following sections. The discussion will be divided into four diﬀerent sections based on

2365 the diﬀerent dose rate regimes illustrated in Figure 8.9 and Table 8.4.

2366 The dose rate dependence of commercial silicon diodes to pulsed ionising radiation

2367 has been widely reported (Van Dam et al., 1990; Grusell & Rikner, 1993; Wilkins

2368 et al., 1997; Shi et al., 2003; Saini & Zhu, 2004). Most literature reported an increase

2369 in diode sensitivity with increased dose rates. The explanation for this eﬀect was based 8.3. Results 140

Figure 8.8: Dose per pulse response for the MP and the commercial diode, normalised to the dose per pulse of 2.78 × 10−4 Gy/pulse.

Figure 8.9: Division of the dose per pulse measurements into four (A to D) sections. 8.3. Results 141

Table 8.4: Dose per pulse dependence sections. Section Measurement setup A 2-, 4- and 6 cm Pb attenuated (in phantoms) B Open ﬁeld, in phantom (in phantoms) C Free air geometry D Under closed MLC (in phantoms)

2370 on the Shockley-Read-Hall (SRH) recombination model, whereby the recombination

2371 of minority charge carriers occurred through mid gap traps (Shockley & Read, 1952).

2372 The SRH model postulated that the lifetime of excess minority carriers increased

2373 proportionally with the increased of the injection level. Low resistivity (highly doped)

2374 silicon diode is less prone to the variation of dose rate. This was due to the larger

2375 numbers of recombination and regeneration (RG) centres and the smaller change of

2376 injection level with respect to the same dose rate. Preirradiation of the diode also

2377 reduces the dose rate dependence. Radiation increases bulk defects and creates more

2378 RG centres, thereby reducing the initial excess minority carrier lifetime (Rikner &

2379 Grusell, 1983). Therefore, most commercial diodes are of the low resistivity type and

2380 radiation hardened. Section A: Both the MP and the EFD-3G diode showed dose rate independence. At low dose rates (or low injection levels, where the number of excess charge carriers <

Ln = Dnτn (8.4a) Lp = Dpτp (8.4b)

2381 At low dose rates, RG centres are greater than excess minority carriers, the minority

2382 carrier lifetime was constant due to the SRH recombination statistics with mid band 8.3. Results 142

Table 8.5: Deﬁnition of symbols.

Ln Diffusion length for electron in p-type Si; Lp Diffusion length for hole in n-type Si; Dn Electron diffusion coefficient in p-type Si; (kT/q) μn k Boltzmann constant 1.38 × 10−23 (joules/◦K) 2 μn Electron mobility Unit = cm /(v-s) T Temperature Unit = degree K Dp Hole diffusion (kT/q) μp 2 μp Hole mobility Unit = cm /(v-s) τn Minority carrier lifetime of electron in p-type τp Minority carrier lifetime of hole in n-type

2383 traps. The diode is dose rate independent. As the dose rates increases, the excess

2384 minority carrier increases. There were not suﬃcient RG centres to sink the excess

2385 minority charge carriers generated, resulted in an increased minority carrier lifetime

2386 and diﬀusion length. Diode sensitivity was independent of the dose rate only if the

2387 diﬀusion length becomes longer than the thickness of the diodes sensitive volume.

2388 The caveat in this process was that the model that was used to describe it was a

2389 “long” diode model. This model assumed that the size of the base of the diode was

2390 much larger than the diﬀusion length (Alexander, 2003). The increase of the minority

2391 carrier lifetime was manifested in the increase of the diﬀusion length and increased

2392 diode sensitivity. The design of the DMG mimics a“long” diode where the width of

2393 the sensitive volume, determined by the diﬀusion length, Ln in passive mode was much

2394 shorter than the thickness of the diode. In contrast, the design of the epitaxial diode

2395 is at the other extreme where the diﬀusion length is larger than the physical thickness

2396 of epitaxial layer (50 μm). As a result of that, the increase of minority carrier lifetime

2397 with higher injection levels would not change the diode sensitivity since the diﬀusion

2398 length was larger the physical thickness of the epitaxial diode.

2399 Section B: The EFD-3G appeared to be dose rate independent while the epitaxial

2400 diodes showed a decrease in sensitivity at higher dose rates. The dose rate response 8.3. Results 143

Figure 8.10: Minority carrier lifetime as a function of ionising radiation dose rate in a 16 −3 highly doped n-type silicon (NA =10 cm ) [reproduced from (Alexander, 2003)].

2401 of the epitaxial diode seen here may be rare in the existing publication on the silicon

2402 radiation diode used for medical radiation dosimetry. Nonetheless, this behaviour was

2403 not unique to the epitaxial diode as it was reproduced by the commercial EFD-3G

2404 diode in the measurements made in free air geometry. This phenomenon was explained

2405 by Alexander (2003) and (Fjeldly et al., 2001).

2406 At extreme high dose rates (excess minority carriers >>the density of the majority

2407 charge carriers) and particularly for diodes with low doping concentration, the SRS

2408 recombination model would start to fail at lower dose rates. This is brought on as

2409 the Auger recombination becomes more dominant (Figure 8.10) (Fjeldly et al., 2001).

2410 Auger recombination is a band-to-band recombination which requires the participation

2411 of three carriers (either two electrons and one hole or two holes and one electron) (Sze,

2412 1981). The electron and hole recombines and the excess energy is transferred to a

rd 2413 3 carrier which signiﬁcantly reduces the minority carrier lifetime. The onset of the

2414 Auger recombination is dependent on the doping concentration with higher resistivity

2415 material beginning at a lower dose rate than a lower resistivity material, such as in

2416 the case of the epitaxial diode with a resistivity of 100 Ω·cm (Alexander, 2003). 8.3. Results 144

2417 Section C: These points represent measurements in free air geometry where the

2418 true dose is hard to deﬁne. Both MP epitaxial diode and the EFD-3G diode showed

2419 a decreasing sensitivity with increased dose rates, albeit at a much higher sensitivity

2420 (up to a factor of 1.4) relative to the normalisation point. In reality, the injection level

2421 in the free air geometry setup is much higher than the in phantom open ﬁeld setup

2422 due to the shorter SDD. However, since dose is deﬁned in phantom at a certain depth

2423 and SSD, the actual dose in the free air geometry measurement cannot be determined

2424 due to three factors,

2425 (i) Charge particle equilibrium (CPE) condition does not exist in the free air ge-

2426 ometry. The measured dose in air by a large diameter ion chamber without a

2427 build up cap and dose in silicon would also be largely aﬀected by the Si/air mass

2428 energy absorption coeﬃcient ratio.

2429 (ii) Without the CPE in a full scattering condition, the actual dose measured by the

2430 silicon diode would comprise of only a fraction of the dose at dmax. However,

2431 since the ion chambers and the diodes have diﬀerent sensitive volumes, the eﬀect

2432 of volume averaging in the larger detectors would give a lower reading.

2433 (iii) Measurement of the ﬂuence at short SDD also exposed the detectors to a larger

2434 portion of the low energy photons scattered by the components in the linac head

2435 (Baumgartner et al., 2009). This would induced an over-response in the silicon

2436 detectors at photons energy <150 keV.

2437 Section D: The attention is drawn to the large error bars of the measurements

2438 made under a closed MLC. Under the closed MLC (in phantom), both the MP and the

2439 commercial diode showed dose per pulse independence for the range of SDD measured.

2440 The dose per pulse response between the MP and the commercial diode showed an

2441 average diﬀerence of 15.9%. The MP appears to be under-responding by 8.9% while 8.3. Results 145

2442 the commercial diode over-responded by 7%. The large error bar of 15.5% showed

2443 the measurement error due to the MLC leakage. This value was obtained by taking

2444 multiple measurements with the diode position translated perpendicular to the long

2445 axis of the MLC leaves by up to 0.3 mm, thereby measuring variation between the dose

2446 directly under a MLC leaf and the MLC leakage dose. The same error was applied

2447 onto MP measurements.

2448 Uncertainty determination: The error bars for all (except for MLC related

2449 measurements) MP and commercial diode measurements were 0.002 and 0.004, re-

2450 spectively. This represents the maximum uncertainty of the measurements which

2451 comprised of the 1 standard deviation of the mean of three measurements for either

2452 MP (or diode) and the ion chamber added in quadrature. For measurements with

2453 closed MLC, the uncertainty due to the positioning error leading to MLC leakage was

2454 included, giving a total uncertainty of 0.155.

2455 Dose per pulse correction: The need to apply a dose per pulse response correc-

2456 tion to the measured data is determined by the measurement setup and the magnitude

2457 of the dose per pulse eﬀect.

2458 When the MP is used as a transmission device at a ﬁxed position at SDD of 58

2459 cm, no dose per pulse correction is required for the in ﬁeld measured ﬂuence. When

2460 the MP is used as a 2D planar dose array in the I’mRT phantom at a depth of 9 cm

2461 with an isocentric technique, the SDD to the central diode remains the same SDD of

2462 100 cm. In the worst case scenario, where the radiation beam is delivered from gantry

◦ ◦ 2463 angle 90 or 270 , the distance between the row of detector nearest to the linac source

2464 and the row furthest away from the linac source is ± 5 cm at a depth of 9 cm. From

2465 a tissue-maximum-ratio (TMR) chart, this represents a change of ± 14% of the dose,

2466 which correlates to ± 1.5% of diﬀerence in dose per pulse correction. This is within

2467 the ± 2% of the ﬁeld ﬂatness and symmetry. In this case, the dose per pulse correction 8.3. Results 146

2468 may not be necessary.

2469 In situations where the dose per pulse correction was required, the dose per pulse

2470 correction factor can be derived from these measurements. The method for dose per

2471 pulse correction was described in chapter 4.

2472 8.3.3 Percent depth dose

2473 Figure 8.11 shows the depth dose curve measured with the MP and the CC13 ion

2474 chamber. The MP appears to be slightly over-responding compared to the ion chamber

2475 measurements for all depths beyond the dmax with the maximum over-response of

2476 1.74% at 10 cm depth. This may be due to the dose per pulse response of the MP.

2477 The under response of the diode at high dose rates resulted in an under-estimation of

2478 the dose at dmax and over-estimation of the dose beyond dmax. The depth dose curve

2479 was subsequently corrected for the dose rate eﬀects and the resulting curve agrees with

2480 the CC13 data within 0.7%. The dose per pulse response of silicon diode aﬀecting the

2481 percent depth dose measurements were also reported by Wilkins et al. (1997).

2482 Figure 8.12 shows the build up region, comparing the MP measurement with the

2483 Attix chamber and the CC13 ion chamber. The MP data was dose per pulse corrected.

2484 Based on this ﬁgure, the water equivalent depth of the MP at 1 mm depth was 1.9

2485 mm.

2486 8.3.4 Dose linearity

2487 The dose linearity was excellent for the measured dose range of 5 to 1000 cGy as shown

2488 in Figure 8.13. 8.3. Results 147

Figure 8.11: Depth dose curve for a 10 × 10 cm2 ﬁeld size of a 6 MV photon energy measured with a CC13 ion chamber and the MP.

Figure 8.12: Build up region of the depth dose curve, comparing the MP, CC13 and Attix chamber measurements (courtesy of Dr. Bradley Oborn, ICCC and CMRP). 8.3. Results 148

Figure 8.13: The dose linearity of the MP.

2489 8.3.5 Energy dependence

2490 The energy response of the MP was studied using an orthovoltage unit and a linear

2491 accelerator. Figure 8.14 present the energy response curve for the range of 75kV to

2492 10 MV nominal energies (corresponding to 26.8 keV - 2.97 MeV equivalent photon

2493 energies), normalized to 1 at 6 MV nominal photon energy (‘face-up’, in phantom

2494 geometry). The photon energy response curve showed an enhanced response at low

2495 energies up to 7.5 times of the response at 6 MV for the ‘face-down’, in phantom

2496 geometry. The detector showed and over-response at lower photon energies with the

2497 maximum dose response at 75 kV nominal photon energy. This is due to the increased

2498 photoelectric eﬀect cross-section in silicon at low energies (Meyer et al., 2001).

2499 Energy dependence of the detector in the free air geometry was expected as silicon

2500 is not tissue equivalent and therefore sensitive to changes in the energy spectrum.

2501 The MP used in the free air geometry had a lower signal response due to the lack of

2502 backscatter electron contribution.

2503 The detector’s response in the ‘face-up’ conﬁguration was lower than in the ‘face- 8.3. Results 149

2504 down’ conﬁguration. This may be due to the presence of the 0.375 mm silicon sub-

2505 strate. However, this silicon substrate aﬀects the detector response in diﬀerent ways

2506 depending whether the detector is in the free air geometry or used with the solid water

2507 phantom.

2508 Detector irradiated faced up with the solid water phantom: For the low

2509 electron energies of <150 keV, the range of electrons in silicon is less than 0.35 mm.

2510 All back scattered electrons generated in the phantom were stopped in the silicon

2511 substrate. Hence, the detector response in the ‘face-up’ conﬁguration was lower.

2512 Detector irradiated faced down in free air geometry: The increased of

2513 the detector response at with the ‘face-down’ conﬁguration may be due to the dose

2514 enhancement eﬀect of the higher Z of silicon, i.e. additional number of electrons were

2515 forward scattered from the silicon substrate to the sensitive volume of the epitaxial

2516 layer downstream. With decreasing photon energies, this eﬀect is gradually reduced

2517 as the competing eﬀect of the x-ray attenuation due to the 0.375 mm silicon substrate.

2518 Figure 8.15 showed the energy response in the megavoltage (MV) region. Compared

2519 to the energy response at low photon energies (Figure 8.14), the energy response of the

2520 epitaxial diodes measured in the solid water phantom was reversed. In a solid water

2521 phantom, the epitaxial diode in ‘face-down’ conﬁguration showed a lower response

2522 compared to those used in ‘face-up’ conﬁguration. For the in phantom measurement

2523 at SSD 100 cm, the epitaxial diode in the ‘face-down’ conﬁguration under-responded by

2524 9.7%± 2.1% and 7.0%± 2.1% for the 6 and 10 MV, respectively. This may be because

2525 the detector response is dominated by the x-ray and secondary electrons attenuation

2526 of the 0.375 mm silicon substrate. In free air geometry, dose enhancement is the

2527 dominant eﬀect.

2528 In free air geometry, the MP was sandwiched between two pieces of 1 mm solid

2529 water phantom. This setup does not have electronic equilibrium; hence the signal 8.3. Results 150

Figure 8.14: The energy dependence of the MP, measured with four diﬀerent setups; comparing ‘face-up’/‘face-down’ conﬁgurations. The measurements were made in free air geometry and with a solid water phantom.

2530 measured by the MP comprised only a fraction of the dose at a full scattering, CPE

2531 condition. In this case, the measured signal were 53.0%± 1.5% and 37.7%± 2.5%

2532 (mean error) for the 6 and 10 MV photon energies, respectively.

2533 8.3.6 Temperature dependence

2534 The detector response of the MP as a function of temperature was measured for the

◦ 2535 temperature ranged between 20 to 27 C (Figure 8.16). The sensitivity variation with

2536 temperature (svwt) was taken as the ratio of the detector response at temperature, t

◦ 2537 to the detector response at a reference temperature, tref (23.32 C). The temperature

2538 coeﬃcients was obtained from the linear regression of the data and were found to

◦ ◦ ◦ 2539 be 0.55%/ C, 0.50%/ C and 0.54%/ C for the high-, medium- and low dose rate,

2540 respectively. 8.3. Results 151

Figure 8.15: The energy response of the MP in ‘face-up’ and ‘face-down’ conﬁgurations for 6 and 10 MV photon energies. The detector response was normalised to 1 at the reference setup of SSD 100 cm and dmax of 1.5 cm and 2.0 cm for the 6 and 10 MV photon energies, respectively.

Figure 8.16: Sensitivity variation with temperature (svwt) of the MP. The detector response was normalized to 1 at 23.32◦C. Error bars represent 1 s.d. of the mean of three measurements. 8.3. Results 152

2541 8.3.7 Field size dependence

2542 Figure 8.17 shows the ﬁeld size dependence of the MP at two diﬀerent SDDs. The

2543 output factor measured using the Farmer chamber was also included for comparison.

2544 The ﬁeld size dependence of the MP measured at SDD 101.5 cm agreed well with

2545 the Farmer chamber measurement within 1.1%. The ﬁeld size dependence of the MP

2546 measured at the SDD of 58 cm showed a much steeper slope. This may be because

2547 at shorter SDD and in the free air geometry, the amount of collimator scatter and

2548 phantom scatter contribution to the ﬁeld size factor was vastly diﬀerent to those

2549 measured in phantom with full scattering condition.

2550 In a phantom, the response of the silicon diode was determined by the secondary

2551 electrons whereas in the free air geometry, the diode response was determined by

2552 photons, i.e. the mass energy absorption coeﬃcients of the silicon. The resulting

2553 energy dependence is due to scattered radiation. At smaller ﬁeld sizes, the lack of

2554 phantom scatter resulted in a much lower ﬁeld size factor, whereas for ﬁeld size >10

2 2555 × 10 cm , the higher contribution of the collimator scatter due to the shorter SDD

2556 resulted in a higher ﬁeld size factor. Asuni et al. (2011) performed Monte Carlo

2557 studies on the contaminant electrons due to transmission detector. They found that

2558 the transmission detector increased the contaminant electrons at shorter SSD of 70

2559 cm and increasing with larger ﬁeld sizes. However, most contaminant electrons are

2560 of low energies with large angular spread. Hence, it does not contribute much to the

2561 dose at large SSD (e.g. 100 cm). They also looked at the contribution of contaminant

2562 electrons by various components in the linac head. For an open ﬁeld, the phase space

2563 ﬁle showed that the major diﬀerence in the contributors were the secondary jaws and

2564 the air column between the linac and the surface of the phantom. At SSD of 70 cm,

2565 the jaws contributed up to 15% more contaminant electrons whereas the air column

2 2566 contribution was 16.8% higher at SSD of 100 cm for a 10 × 10 cm ﬁeld size. In 8.3. Results 153

Figure 8.17: The ﬁeld size dependence of MP at SDD of 101.5 and 58 cm compared with the standard data measured using the Farmer chamber measured at SDD of 101.5 cm. The data was normalised to 1 at the ﬁeld size of 10 × 10 cm2.

2567 addition, the mean energy scored for the short SSD was also found to be slightly

2568 lower.

2569 8.3.8 Angular dependence

2570 Figure 8.18 shows the mean angular response of the MP diodes located on three

2571 detector columns representing the centre and two lateral edges of the MP (-5, 0 and

2572 +5 cm oﬀ axis distance). The angular response of the MP diodes was <2.7% for gantry

◦ ◦ ◦ ◦ 2573 angles 0 to 60 (corresponding mirror angles of 300 to 360 ) and 4.5% at gantry angle

◦ ◦ 2574 75 (mirror angle of 285 ). Due to the asymmetric geometry of the MP diode and the

◦ ◦ 2575 inherent silicon anisotropy, the mean angular dependence for gantry angles 90 to 270

◦ 2576 was 10.5%± 0.7% (1 s.d.). At the gantry angle of 180 , the mean angular response

2577 was 10.8%± 0.7%.

2578 Silicon anisotropy is described as the non-symmetrical geometry of the silicon diode.

2 2579 The MP epitaxial silicon diode has a physical size of 1.5 × 1.5 mm and a total 8.3. Results 154

2580 thickness of 425 μm, of which the epitaxial layer (the thickness of the sensitive volume)

+ 2581 made up the top 50 μm. The n implant on the epitaxial surface covers an area of 0.5

2 2582 × 0.5 mm in the centre of the epitaxial layer. This means that the sensitive volume in

2583 the epitaxial diode is not symmetrical in the vertical direction and horizontal direction.

2584 In the vertical direction, the bulk silicon wafer which the epitaxial silicon was grown

2585 upon does not contribute to the generation of photocurrent. In the lateral direction,

2586 the quasi-neutral region adjacent to the sensitive volume is not fully depleted in a

2587 passive mode and is also larger than the diﬀusion length.

2588 The uncertainties of the measurements were taken as the ± 1 standard deviation

2589 of the mean of 11 diodes and the equivalent pixel positions on the ﬁlm. The average

2590 uncertainty was ± 2.3% and ± 0.7% for the ﬁlm and MP measurements, respectively.

2591 8.3.9 Beam perturbation study

2592 8.3.9.1 Surface dose and build up region of the depth dose curve

2593 The presence of the MP as a transmission detector increases the surface dose and

2594 the dose within the build up region of the depth dose curve (Figure 8.19). This was

2595 attributed to the increased electron contamination due to the presence of the MP. Table

2596 8.6 shows the surface dose diﬀerence due to the presence of the MP in percentage. For

2 2597 the 20 × 20 cm ﬁeld size, the surface dose increased by 3.4% and 7.3% for the SSD of

2598 90 cm and 80 cm, respectively. This was 4.7% and 10.9% lower than the COMPASS

2599 system (Venkataraman et al., 2009) for the same ﬁeld size and SSDs. The electron

2600 contamination increases with larger ﬁeld size and with shorter SSDs. For the largest

2601 ﬁeld size investigated at the SSD of 80 cm, the presence of the MP would increase the

2602 dose to the superﬁcial tissues by 12.1%. 8.3. Results 155

Figure 8.18: Mean angular response of the MP for the detectors located at the central column (0 cm) and at ± 5 cm oﬀ axis distance.

Table 8.6: Difference between the surface doses measured with open fields and a MP fields in percent. The values in the parentheses are the measured surface dose normalised to the maximum dose. MP field - open field (surface dose) (%) Field size (cm2) SSD 90 cm SSD 80 cm 20 × 20 3.4 (28.5) 7.3 (29.2) 30 × 30 6.8 (38.4) 12.1 (40.5) 8.4. Conclusion 156

Figure 8.19: Dose at the build up region for a 20 × 20 cm2 and 30 × 302 cm at (a) SSD of 90 cm and (b) SSD of 80 cm. The measurement uncertainty was ± 0.5%, representing 1 standard deviation of the mean of three measurements.

2603 8.3.9.2 Transmission factor

2 2 2604 The transmission factors for three ﬁeld sizes of 10 × 10 cm ,20× 20 cm and 30 ×

2 2605 30 cm were measured. Average transmission factor was 0.990 ± 0.002 (1 standard

2606 deviation) for 6 MV beam. Compared to the other transmission detectors, the DAVID

2607 and IQM systems attenuate a 6 MV beam by 7% while the COMPASS system reported

2608 a 3.3% beam attenuation.

2609 8.4 Conclusion

2610 The radiation response and basic characterisation of the MP was carried out. The

2611 MP was irradiated up to 41.5 kGy to stabilise the epitaxial diodes. Medium term

2612 reproducibility of the preirradiated MP was 2.1%. The MP showed decreased dose per

−4 2613 pulse response at higher dose rates (dose per pulse >1 × 10 Gy/pulse) while at lower

2614 dose rates typical for IMRT deliveries the MP appears to be dose rate independent.

2615 MP measured depth dose measurement agrees with the ion chamber depth dose mea- 8.4. Conclusion 157

2616 surement within 0.7%. The dose linearity of the MP was excellent for the measured

2617 dose range of 0 to 10 Gy. MP epitaxial diode showed an enhanced response up to a

2618 factor of 7.5 at low photon energies that were consistent with other silicon devices.

◦ 2619 The average temperature coeﬃcient was 0.53 ± 0.03%/ C. The ﬁeld size dependence

2620 of the MP at the shorter SDD of 58 cm was much higher than those measured at SDD

2621 100 cm. This was due to the extra contribution of contaminant electrons by the linac

2622 jaws at shorter SDD. MP showed angular response dependency due to the anisotropy

2623 of the silicon diode with the maximum angular response of 10.8% at gantry angle

◦ ◦ 2624 180 . Angular dependence was within 3.5% for the gantry angles ± 75 .Inthebeam

2 2625 perturbation study, the surface dose increased by 12.1% for a 30 × 30 cm ﬁeld size

2626 at the SSD of 80 cm whilst the transmission factor for the MP was 1%. 2627 Chapter 9

2628 Investigation of the Magic Plate in

2629 clinical application

2630 9.1 Introduction

2631 Intensity modulated radiation therapy (IMRT) is a complex treatment delivery tech-

2632 nique demanding rigorous quality assurance and dosimetric veriﬁcation of the treat-

2633 ment delivery (Ezzell et al., 2003; Nelms et al., 2011). The dosimetric veriﬁcation

2634 of an IMRT or volumetric modulated arc therapy (VMAT) plan often requires com-

2635 parison of a 3D dose map produced by a treatment planning system (TPS). This is

2636 conventionally achieved by means of a point dose measurements at isocentre with an

2637 ion chamber and a spatial fluence or dose profile verification usings two dimensional

2638 dosimeters such as ﬁlms (Zhu et al., 2002; Childress et al., 2005), electronic portal

2639 imaging device (EPID) (Van Esch et al., 2004; van Zijtveld et al., 2006; Lee et al.,

2640 2009) or 2D electronic arrays (Letourneau et al., 2004; Jursinic et al., 2010). To date,

2641 the only true 3D dosimeter is the gel dosimetry (Duthoy et al., 2003). However, the

2642 use of this dosimeter in most clinics is still not widespread due to the complexity of

2643 the readout techniques (Vergote et al., 2004; Baldock & et al., 2010).

158 9.1. Introduction 159

2644 In the absence of a true 3D dosimeter, dosimetric veriﬁcation of IMRT treatments

2645 is conﬁned to two dimensions. The use of ﬁlm dosimetry is still common due to

2646 its simplicity of use and high spatial resolution, limited only by the scanning resolu-

2647 tion in practice. However, this dosimeter is not real time and is susceptible to ﬁlm

2648 inhomogeities, energy dependence (for radiographic ﬁlms) and scanner induced non

2649 uniformities for self processing polymer ﬁlms (e.g. Gafchromic ﬁlms).

2650 The use of 2D electronic arrays is becoming widespread due to the real time feed-

2651 back. Comprehensive and fast dose map comparison software packages are becoming

2652 a feature of many of these devices. Most of the 2D electronic arrays are angular de-

2653 pendent (Bhardwaj et al., 2009; Jursinic et al., 2010), hence the veriﬁcation of IMRT

2654 deliveries are often performed with all the beams delivered orthogonal to the detector

2655 plane.

2656 The measured dose distribution is then used to compare with the TPS generated

2657 field-by-field dose distributions or composite field dose distribution. Comparison of the

2658 measured with the TPS predicted distributions is usually performed using the Gamma

2659 index method (Low et al., 1998).

2660 The advent of the rotational modulated radiation therapy such as volumetric mod-

2661 ulated arc therapy (VMAT) further complicates the dosimetric veriﬁcation process

2662 by the introduction of simultaneous gantry rotation and dose rate modulation during

2663 treatment delivery (Otto, 2008). Dosimetric veriﬁcation with all the beams delivered

2664 orthogonal to the detector plane may no longer be adequate (Nelms & Feygelman,

2665 2010). However, the use of 2D arrays in rotational treatment delivery requires char-

2666 acterisation of the angular dependence of the device (Jursinic et al., 2010). Currently,

2667 there are two commercial ’pseudo 3D’ dosimetry systems that attempt to address this

4 2668 issue, the Delta system (Scandidos, Uppsala, Sweden) and the ArcCHECK system

4 2669 (Sun Nuclear, Melbourne, Fl). The Delta system measures IMRT or VMAT dose 9.1. Introduction 160

2670 in two orthogonal planar arrays and the 3D dose distribution is reconstructed. An

2671 inclinometer is attached to the linac gantry to record the gantry angles and correct for

2672 the angular dependence of the silicon diodes (Bedford et al., 2009; Feygelman et al.,

2673 2009). The ArcCHECK system is designed with the diodes mounted in a “helical

2674 array” (Feygelman et al., 2010).

2675 Up until now, dosimetric veriﬁcation of rotational IMRT deliveries has commonly

2676 focused on dose measurement in solid water phantoms. The phantom and detectors

2677 were positioned on the linac couch. Quality assurance of IMRT treatment had been

2678 approached in two other ways,

2679 (i) with the 2D detector array positioned between the MLC collimator and patient.

2680 The array functions as a transmission detector for on-line and in-vivo measure-

2681 ment (i.e. during patient treatment) (Poppe et al., 2006b; Islam et al., 2009;

2682 Venkataraman et al., 2009), and

2683 (ii) with the 2D detector array positioned downstream of the patient, such as using

2684 an EPID or mounting a detector array on the EPID (Greer et al., 2007; Lee et al.,

2685 2009).

2686 These approaches eliminate the need for the angular correction for the devise used

2687 and introduce the possibility of a real time, in-vivo IMRT dose veriﬁcation. Clini-

2688 cally, the ability to perform real time and in-vivo measurements means that the daily

2689 treatment delivery can be measured to ensure accurate and reproducible delivery. Any

2690 error in the treatment delivery can be detected and addressed at the earliest time. The

2691 measured ﬂuence can be input into a 3D dose reconstruction software to determine

2692 the eﬀect of the dose errors relative to the anatomical features based on the CT data.

2693 Corrective action can be carried out with adaptive radiotherapy.

2694 The Magic Plate (MP) is a 2D detector array with 121 epitaxial silicon diodes

2695 mounted on a 0.6 mm Kapton substrate using the ‘drop-in’ technique. It was designed 9.2. Material and methods 161

2696 as a dosimetric tool for IMRT/VMAT quality assurance. It can function as a trans-

2697 mission device and also for dose measurement in solid water phantoms. This chapter

2698 describes the implementation of the MP in the dosimetric veriﬁcation of a clinical

2699 IMRT treatment plan both as a transmission device and for dose measurement in the

2700 phantom.

2701 9.2 Material and methods

2702 9.2.1 Uniformity and absolute dose calibration

2703 The MP comprises of 11 × 11 individual epitaxial diodes covering an area of 10 × 10

2 2704 cm . The detector response of each diode is slightly diﬀerent from each other. A uni-

2705 formity calibration needs to be performed to calibrate the detector response. Achieving

2706 this is complicated by the non uniformed linac radiation ﬁeld at most intermediate

2707 depths and in free air geometry.

2708 The photon radiation output from a linear accelerator target is forward peaked

2709 until it reaches a conical shape ﬂattening ﬁlter whereby the beam is attenuated to

2710 produce a flat uniform field. However, the cross beam profiles of a linac radiation

2711 ﬁeld are not entirely ﬂat but have a slightly higher dose near the edges compared to

2712 the central axis. These dose regions are referred to as the “dose horns” (Metcalfe

2713 et al., 2007). The dose horns exist because the ﬂattening ﬁlter is designed to achieve a

2714 relatively uniform radiation ﬁeld for all ﬁeld sizes at 10 cm depth. The horns contain

2715 low energy components due to the thinner edge of the ﬂattening ﬁlter. These low

2716 energy components are attenuated more rapidly through the medium, resulting in a

2717 relatively ﬂatter cross beam proﬁles at approximately 10 cm depth. Therefore, a simple

2718 assumption of a flat radiation beam profile is not suitable for a large radiation field.

2719 A more comprehensive method to determine the detector response and subsequently 9.2. Material and methods 162

Figure 9.1: Sketch of the MP coordinate system. The diode at coordinate [0,0] is the centre diode.

2720 to calibrate the detector in terms of ﬁeld uniformity is needed.

2721 Uniformity calibration of the MP employed a multiple shift and irradiation method.

2722 The MP was positioned at a source to detector distance (SDD) of 100 cm and 1.5 cm

2723 depth in a solid water phantom. The central diode was aligned with the central axis

2724 cross hair. The x- and y- axis of the MP was also aligned parallel to the x- and y- axis

2725 of the laser (Figure 9.1). The MP was irradiated with ﬁxed monitor units (MUs) using

2 2726 a40× 40 cm ﬁeld size. Then, it was shifted and irradiated at four other positions

2727 sequentially (Table 9.1). The ﬁnal aim of this exercise was to obtain a factor that

2728 correlates the sensitivity of each diode relative to the sensitivity of the central diode.

2729 The method assumed that,

2730 (i) the ﬁve calibration beams were consistent and reproducible. Thus, the same

2731 amount of dose would be given to all the ﬁve diodes that were positioned at the

2732 beam’s centre axis. 9.2. Material and methods 163

Table 9.1: MP positioning for the uniformity calibration of the diodes. Position Description A MP centre diode (0,0) positioned at centre axis of beam B MP shifted laterally by -1 cm relative to position A (diode [1,0] at the radiation beam central axis) C MP shifted laterally by +1 cm relative to position A (diode [-1,0] at the radiation beam central axis) D MP shifted longitudinally by +1 cm relative to position A (diode [0,-1] at the radiation beam central axis) E MP shifted longitudinally by -1 cm relative to position A (diode [0,+1] at the radiation beam central axis)

2733 (ii) The detectors’ sensitivities were consistent throughout the measurements.

2734 The MP readings measured by the TERA readout, is a product of two components,

2735 the detector response and the radiation dose proﬁle. The method to determine the

2736 detectors’ sensitivity is similar to the method described in chapter 6 (section 6.2.3)

2737 and is brieﬂy described here.

2738 A ﬁxed dose was delivered to the diode at the beams’ central axis. This allowed

2739 the determination of the response of this diode. By shifting the MP laterally by 1

2740 cm, the beam central axis was positioned on the adjacent diode. This enables the

2741 determination of the dose value and detector response at 1 cm away from the beam

2742 centre axis. The process of calculating the detector response and the dose value was

2743 done iteratively in the x- and y- direction until all the diode’s responses were obtained.

2744 The relative sensitivity factor was taken as the ratio of the individual detector response

2745 to the central detector’s response. This is also the uniformity correction factor.

2746 The MP can be used for absolute dose measurements by calibrating it with a

2747 known dose. Radiation was delivered using a 6 MV photon energy from a Clinac iX

2748 linear accelerator (Varian, Palo Alto, USA). The dose calibration coeﬃcient was taken

2749 as a ratio of the given dose divided by the total number of counts measured by the

2750 central detector (unit = cGy/count). This dose calibration coeﬃcient was then used 9.2. Material and methods 164

2751 to convert the measured signal to dose in all subsequent measurements.

2752 When the MP was used in conjunction with the I’mRT phantom (IBA Dosime-

2753 try, Germany), both the uniformity and the absolute dose calibration of the MP was

2754 performed in this phantom to avoid the need to switch between phantoms.

2755 9.2.2 Dose measurement in solid water phantom

2756 The MP was sandwiched between two pieces of 5 mm solid water machined to ﬁt into

2757 the cavity of the I’mRT phantom. The MP and the I’mRT phantom were aligned

2758 to the isocentre (SDD 100 cm). A 9 ﬁeld IMRT head and neck treatment plan (the

2759 original site of the plan was a left tonsil) was delivered in two modes:

◦ 2760 (i) with all the beams delivered at the gantry angle 0 ,and

2761 (ii) with the beams delivered at the actual treatment gantry angles.

2762 Field by ﬁeld measurements were made. The composite IMRT ﬁeld was a sum-

2763 mation of all the individual IMRT ﬁelds. For the treatment delivered with the actual

2764 gantry angles, the MP measurements were corrected for angular dependence.

2765 The MP measured dose distributions were compared with the Pinnacle TPS pre-

2766 dicted and the Gafchromic EBT2 (ISP, Wayne, NJ) ﬁlm measured dose distribution.

2767 The Philips Pinnacle (Philips Radiation Oncology Systems, Milpitas, CA) TPS cal-

2768 culated the doses using collapsed cone convolution dose engine (Mackie et al., 1985;

3 2769 Ahnesjo, 1989) with a dose grid of 2 mm . It was exported as a DICOM dose ﬁle. Due

2770 to the large amount of data, a graphic user interface (GUI) was written with Matlab

2771 (The MathWorks Inc., Natick, MA) to handle this work (see appendix B for the GUI

2772 and Matlab scripts).

2773 Gafchromic EBT2 film measurements were also made for all IMRT fields. The films

2 2774 were cut to the size of 101.5 × 120.0 mm and positioned at the centre of the I’mRT 9.2. Material and methods 165

2775 phantom at the depth of 9 cm. The ﬁlms were scanned 24 hr after the exposure to allow

2776 for the post-irradiation coloration. The EBT2 ﬁlms were scanned with an A3 ﬂatbed

2777 scanner (Epson Expression 10000XL). Each ﬁlm was scanned six times. Only the last

2778 three scans were kept for image analysis. This was to ensure that the scanner was

2779 suﬃciently warmed up and thus ensuring consistency in the scanned ﬁlms (Fuss et al.,

2780 2007). The ﬁlms were scanned in 48-bit RGB colour with scanning resolution of 96 dpi

2781 (equivalent to pixel size of 0.265 mm). The ﬁlms were scanned in the same orientation

2782 at the centre region of the scanner to reduce scanner induced non-uniformity (Butson

2783 et al., 2006a). A set of calibration ﬁlms were also irradiated in the same experimental

2784 session. Analysis of the images was done using the ImageJ version 1.43U software

2785 (National Institute of Health, USA) and Matlab. The red channel of the last three

2786 scans for each ﬁlm were averaged and used for the analysis (Matney et al., 2010). The

2787 EBT2 pixel intensities were converted to dose values based on a calibration curve.

2788 During the image analysis, a 3 by 3 pixel two dimensional median ﬁlter was applied

2789 to the image to reduce image noise (Andres et al., 2010).

2790 9.2.3 Fluence measurements in transmission mode

2791 For the ﬂuence mode measurements, the MP was sandwiched between two pieces of

2792 1 mm solid water material. This was then mounted on the linac accessory slot using

2793 a modiﬁed total body irradiation (TBI) tray and the Y-shaped Perspex frame. The

2794 source to detector distance (SDD) at this point is 58 cm. The EBT2 ﬁlm measurement

2795 were also made at the SDD of 58 cm. For the ﬁlm measurement, the linac gantry was

◦ 2796 rotated to 180 angle and the EBT2 ﬁlms were placed between the two 1 mm solid

2797 water slabs on the TBI and the Y-shaped Perspex frame.

2798 These measurements were also compared with the TPS predicted at the SDD of

2799 58 cm. The TPS dose distribution was calculated in a thin phantom volume with a 9.3. Results and discussion 166

2800 thickness of 14 mm located at the source to surface distance (SSD) of 57.9 mm.

2801 9.2.4 Performance index used for the dose distribution com-

2802 parison

2803 The Gamma index method (Low et al., 1998) was employed as the dose distribution

2804 comparison index. The Gamma index compares two dose distributions in two terms,

2805 (i) dose deviation (DD) in percentage of the maximum dose, and

2806 (ii) Distance-to-Agreement (DTA) in the unit of mm.

2807 The Gamma index criterion used for the comparison in this thesis was 3% DD

2808 and 3 mm DTA. These criterions were chosen because they were the most prevalent

2809 standard for acceptance testing of IMRT QA (Nelms & Simon, 2007). The dose values

2810 in the MP and the EBT2 ﬁlm measurements were calibrated absolute dose values.

2811 9.3 Results and discussion

2812 9.3.1 Uniformity correction

2813 Figure 9.2 shows the cross plane and in plane proﬁles of the MP diodes before and

2814 after uniformity correction. The detector response were normalised to 1 at the central

2815 axis. The coeﬃcient of variation of the diodes was ± 2.8% before correction and

2816 ± 0.6% after uniformity correction. The reproducibility of the uniformity correction

2817 factors over a period of four measurement sessions was found to be ± 1.1% (1 standard

2818 deviation). However, it was decided that it would be prudent to perform a uniformity

2819 calibration within each measurement session. 9.3. Results and discussion 167

Figure 9.2: Mean cross plane and in plane proﬁles of the MP diodes before and after the uniformity correction. The error bars represent the ± 1 standard deviation of the mean of 11 detectors located within the same row or column.

2820 9.3.2 Dose distribution comparison

2821 The following sections detail the results of the dose distribution comparisons measured

2822 in three conﬁgurations:

2823 (i) In the I’mRT phantom with IMRT ﬁelds delivered through a ﬁxed gantry angle

◦ 2824 of 0 ,

2825 (ii) In the I’mRT phantom with IMRT ﬁelds delivered with the actual treatment

2826 gantry angles, and

2827 (iii) With the dosimeters (MP or EBT2 ﬁlms) positioned at the SDD of 58 cm, as a

2828 transmission detector.

2829 Three data sets were compared; the MP, the EBT2 ﬁlm measurements and the TPS

2830 predicted dose distributions. The MP and EBT2 data were calibrated and comparison

2831 were made using absolute dose values. A Gamma criterion of 3% DD and 3 mm

2832 DTA were used as the performance index. The horizontal and vertical dose proﬁles

2833 corresponding to the MP diode positions were also plotted for comparison. 9.3. Results and discussion 168

2 2834 The grid sizes used for the Gamma analysis were 0.265, 1 and 10 mm for the

2835 EBT2, TPS and MP, respectively. The current prototype design of the MP has a

2836 limited spatial resolution (10 mm detector-to-detector distance). No attempt was

2837 made to interpolate the MP measurements to smaller grid sizes using 2D interpolation

2838 techniques. The 10 mm detector pitch was too sparse for an accurate interpolation

2839 of IMRT dose modulation. Hence, the Gamma analysis was only performed on each

2840 of the 121 MP dose points when compared to the EBT2 and TPS dose distributions.

2841 Gamma analysis for the EBT2 and TPS dose distributions were compared with a

2 2842 dose grid of 1 mm . Dose distributions were compared in terms of ﬁeld-by-ﬁeld dose

2843 comparison, and the composite IMRT ﬁeld.

2844 Figure 9.3 shows the measurement of a single IMRT ﬁeld delivered at gantry angle

◦ 2845 0 . The MP and the EBT2 ﬁlms were positioned at SDD of 100 cm in the I’mRT

2846 phantom. TPS versus MP and EBT2 shows a fairly good gamma agreement with pass

2847 rates ≥ 90% for a gamma criteria of 3% DD and 3 mm DTA. Figure 9.3 (II) and (III)

2848 shows the MP, EBT2 and TPS predicted dose proﬁles corresponding to the MP diodes

2849 positions. The large discrepancies between the film’s profiles to the other two profiles

2850 in (III.column = -50 mm and 50 mm) is due to ﬁlm artifacts. The width of ﬁlms were

2851 101.5 mm, the artifacts were caused by the edges of the ﬁlms. This problem can be

2852 avoided by increasing the width of the ﬁlms.

2853 The horizontal and vertical proﬁles plots showed a good agreement between the

2854 three data sets. The EBT2 ﬁlm was the dosimeter with the highest spatial resolution

2 2855 (0.265 mm grid size). It was able to model the steep dose gradients and sharp dose

2856 modulations of IMRT ﬁeld. It was followed by the TPS dose distribution which was

3 2857 calculated with a 2 mm dose grid. The MP was limited in the spatial resolution

2858 (10 mm detector-to-detector distance). It was not able to reproduce the high dose

2859 modulations of IMRT delivery. However, the active area of the epitaxial diodes is very 9.3. Results and discussion 169

Figure 9.3: Dose measurement for a single IMRT field delivered with gantry angle set to 0◦. (I.a) shows the normalised cross plane dose profile at an off axis distance of 0 mm. (I.b to d) shows the three planar dose distributions of the MP, EBT2 and the Pinnacle TPS. Gamma analysis of the (I.e) TPS versus MP (pass rate = 97.5%), and (I.f) TPS versus EBT2 (pass rate = 92.4%) dose distributions. (II) shows the horizontal and (III) vertical dose profiles for the three data sets. 9.3. Results and discussion 170

2 2860 small (0.5 × 0.5 mm ). The small sensitive volume of the epitaxial diodes will have

2861 minimal detector volume averaging eﬀect. This was shown as the good agreement

2862 between the MP and the EBT2 ﬁlm measurements. Compared to the conventional

2863 single point dose veriﬁcation, the convenience of having 121 accurate dose points at

2864 various region of IMRT plan simulataneously would deﬁnitely be useful in IMRT QA

2865 veriﬁcations.

2866 Figure 9.4 shows the results of a composite IMRT delivery. The 9-ﬁeld IMRT plan

◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ 2867 was delivered through the gantry angles of 20 ,60, 100 , 150 , 180 , 210 , 260 , 300 ,

◦ 2868 and 340 . The TPS predicted and MP measured dose distributions were the sum of

2869 all the gantry angles. The MP measurements were corrected for angular dependence,

2870 described in chapter 8 (section 8.2.9).

2871 The summary of the gamma analysis results is tabulated in Table 9.2. Comparing

2872 the three measurement conﬁgurations, IMRT ﬁelds delivered with the gantry angle set

◦ 2873 to gantry 0 showed the highest gamma agreement with the mean pass rates of 95.7%±

2874 3.6% (1 standard deviation). The MP measured dose distributions were calibrated

2875 for ﬁeld uniformity and absolute dose. No other correction except for the angular

2876 dependence correction was applied. The TPS predicted dose distribution was corrected

2877 for the linac couch attenuation. The good agreement for the treatment delivered with

◦ 2878 ﬁxed gantry angle of 0 shows that the eﬀect of the dose per pulse or energy dependence

2879 may be negligible in this type of measurement setup.

2880 The delivery of IMRT ﬁelds with the actual treatment gantry angles introduces

2881 additional positional uncertainty. A slight vertical positional oﬀset of the dosimeter

2882 relative to the isocenter may produce a larger deviation in the dose comparison. This

2883 is reﬂected as the slightly lower pass rates in the gamma analysis with the mean pass

2884 rates of 86.8%± 4.6%. 9.3. Results and discussion 171

Figure 9.4: Dose measurement for a composite IMRT plan delivered with the actual treatment gantry angles. (I.a) shows the normalised cross plane dose profile at an off axis distance of 0 mm. (I.b to d) shows the three planar dose distributions of the MP, EBT2 and the Pinnacle TPS. The MP measurements were corrected for angular dependence. Gamma analysis of the (I.e) TPS versus MP (pass rate = 80.2%), and (I.f) TPS versus EBT2 (pass rate = 82%) dose distributions. (II) shows the horizontal and (III) vertical dose profiles for the three data sets. 9.3. Results and discussion 172 TPS vs. EBT2 MP vs. EBT2 MP vs. TPS )SDD58cm ◦ * 52.9 90.9 54.7 ··· TPS vs. EBT2 * MP vs. EBT2 ··· Measurement configuration MP vs. TPS TPS vs. EBT2 MP vs. EBT2 TPS 80.210.8 89.7 4.2 92.0 2.6 97.2 2.1 94.0 2.5 93.4 2.3 65.0 94.0 8.7 62.5 3.3 7.9 Actual treatment gantry angles Fixed gantry angle (gantry angle 0 *No values were available because EBT2 measusurements were not made for this field. The author ran out of EBT2 film while doing this IMRT field MP vs. G180G210G300 90.9G340 86.0G260 89.3G20 93.4 88.4G60 88.4 70.2G100 85.1 91.4G150 90.9 66.9 87.6 92.6 71.9 90.1 62.8Average 97.5 89.1 90.1 87.6 95.9 94.8 96.7 88.4 99.2 95.9 92.6 89.3 95.9 91.7 93.7 93.4 95.8 96.7 92.1 93.6 90.1 97.5 92.0 92.6 99.2 94.3 99.2 89.2 97.5 95.0 96.3 93.4 92.6 54.5 94.8 63.6 96.4 66.1 93.8 93.4 62.8 92.6 75.2 99.2 66.8 95.0 57.0 65.0 97.5 66.1 59.7 73.6 63.3 90.1 65.0 90.1 97.5 57.1 54.6 58.1 1s.d. Composite 87.6 81.8 90.9 99.2 95.9 92.0 78.5 93.4 81.2 measurement. The average and 1 s.d. of values represents the statistics of 8 measured IMRT fields. Table 9.2: Gamma analysismeasurements. (3% DD,3 mm DTA) of IMRT dose distributions comparing TPS calculations, MP and EBT2 9.3. Results and discussion 173

2885 The results for the measurements acquired at SDD of 58 cm shows a much larger

2886 disagreement in the gamma analysis. In general, the gamma analysis of the MP versus

2887 EBT2 was still very high ( >90%). The agreements were poorer when comparing the

2888 TPS predicted dose distributions with the measured dose distributions of MP and

2889 EBT2. It is unclear at this point as to the true reasons for this discrepancy. However,

2890 some possible reasons are:

2891 (i) The TPS dose calculation engine was designed and modelled to calculate dose

2892 in water-like phantoms at moderate source to surface distances (SSDs). The

2893 attempt to generate a dose distribution with the measurement conditions may

2894 render inaccurate dose calculations due to the lack of scattering material and

2895 shorter SDD.

2896 (ii) MP was shown to have ﬁeld size dependence at shorter SDD that were quite

2897 diﬀerent to those at SDD 100 cm.

2898 (iii) The MP being a silicon based detector was shown to be energy dependent. At

2899 shorter SDDs, the contribution of the low energy photons may be much higher

2900 (Baumgartner et al., 2009).

2901 A possible solution to address the inadequacy of the TPS dose calculation engine

2902 in calculating the dose distribution measured with the MP as a transmission detector

2903 is to use Monte Carlo calculation methods. The impact of the ﬁeld size and energy

2904 dependence of the MP is not directly measurable in these measurements. Separate

2905 studies may be required to investigate the eﬀect of these parameters on the MP mea-

2906 surements when used as a transmission device. These studies however are not within

2907 the scope of this thesis.

2908 While the poor agreement of the MP measurements with the TPS predicted dose

2909 distributions remains, the value of the MP used as a transmission detector for real 9.4. Conclusion 174

2910 time patient monitoring is diminished. If the discrepancies are explained by Monte

2911 Carlo calculation, the role of MP as a transmission detector is enhanced. The ability

2912 to monitor the beam-by-beam patient treatment in real time and comparing it with

2913 Monte Carlo simulated transmission ﬁeld could be useful in identifying treatment er-

2914 rors. The measured IMRT delivery can be fed back into a three dimensional computed

2915 tomography (CT) patient data to recalculate the actual treated 3D dose distribution

2916 with respect to the patient anatomy. This would allow one to correlate physics perfor-

2917 mance index such as the gamma index with clinically relevant biological consequences

2918 (Nelms et al., 2011).

2919 9.4 Conclusion

2920 The Magic Plate is a 2D array silicon detector based on epitaxial technology mounted

2921 on a 0.6 mm Kapton substrate. It comprises of 11 × 11 detectors covering an area

2 3 2922 of 10 × 10 cm . The sensitive volume of each diode is 0.5 × 0.5 × 0.05 mm .The

2923 diodes are spaced at 10 mm apart. A correction method to correct for ﬁeld uniformity

2924 was described and the coeﬃcient of variation of the diodes was ± 0.6% after the

2925 uniformity correction. The application of the MP in phantom dosimetry QA in a

2926 clinical IMRT delivery was demonstrated. A 9-ﬁeld head and neck IMRT plan was

◦ 2927 delivered initially with all the beams delivered with gantry angle set to 0 and then

2928 at the actual treatment gantry angles. MP measurements were compared with the

2929 Philips Pinnacle (version 9.0) predicted dose distributions and EBT2 measurements.

2930 Gamma criteria of 3%DD and 3 mm DTA was used as the performance index. The dose

◦ 2931 distributions delivered with the gantry angles set to 0 showed the highest agreement

2932 with the mean pass rate of 95.7%. IMRT plans delivered with the actual treatment

2933 gantry angles gave a mean pass rate of 86.8%. This may be due to the combined

2934 positional uncertainty due to the detector position and gantry rotation. The MP was 9.4. Conclusion 175

2935 also used as a transmission detector to measure the exit ﬂuence at the SDD of 58

2936 cm. The dose distribution comparison showed good agreement with the EBT2 ﬁlm

2937 measurements ( >90%). Agreement with the TPS predicted dose distribution was

2938 less satisfactory. This is because the TPS dose calculation engine was not design

2939 to calculate dose at such conditions and the possible eﬀects of ﬁeld size and energy

2940 dependence of the MP. The Pinnacle system is optimsed to predict predict dose in

2941 patients and the extra focal modelling of this device is that of a simple gaussian

2942 (Jursinic & Mackie, 1996). The secondary scatter is more complex than this and

2943 hence a Monte Carlo simulation may provide a better comparison alternative. 2944 Chapter 10

2945 Conclusion

2946 Two novel dosimetry systems based on silicon substrates were studied in this thesis.

2947 These two systems were successfully characterised in terms of the radiation response in

2948 high dose rate and spatially and temporally modulated radiation ﬁelds. They were then

2949 successfully used for dosimetric veriﬁcation of intensity modulated radiation therapy

2950 (IMRT), stereotactic radiosurgery (SRS) and helical Tomotherapy Megavoltage x-ray

2951 beams, respectively. This chapter summarises the main ﬁndings of this thesis in terms

2952 of the Dose Magnifying Glass (DMG) and the Magic Plate (MP). Advantages and

2953 limitations of both systems are discussed. Potential future developments that may

2954 enhance the applications of both systems are explored.

2955 10.1 Dose Magnifying Glass

2956 The DMG is a high spatial resolution detector based on strip detector technology.

2957 It has 128 silicon strips extending to a length of 25.6 mm. The sensitive area of a

2958 single strip is deﬁned by the length (2 or 5 mm) and width (0.02 mm) of a single strip

2959 element. The pitch of the silicon strips is 0.2 mm. The high spatial resolution of the

2960 DMG is ideal for the measurement of high dose gradients. In this thesis, the radiation

176 10.1. Dose Magnifying Glass 177

2961 response of the DMG was characterized (chapter 4). The dose per pulse response of

2962 four DMGs was studied showing that low resistivity and preirradiation of the device

2963 produces the least dose rate dependence. The use of attenuators such as a multileaf

2964 collimator results in much lower detector sensitivity (58%) due to beam hardening

2965 eﬀect. The percent depth dose proﬁle of a 6 MV photon beam matches closely with

2966 measurements from a Farmer thimble ion chamber (within 0.8%) up to 20 cm depth

nd 2967 in solid water. The 2 generation DMG mounted on a 0.12 mm Kapton substrate has

◦ ◦ 2968 a maximum angular response of 15.3%± 0.1% at gantry angles 90 and 270 .

2969 In chapter 5, The DMG was used to measure penumbra widths. It measured a

2970 80-20% penumbra width of 2.77 mm at 1.5 cm depth and 3.93 mm at 10.0 cm depth.

2971 It was subsequently used for the dosimetric veriﬁcation of a 5 ﬁeld prostate IMRT

2972 plan. DMG was able to reproduce the IMRT dose proﬁles that were predicted by a

2973 treatment planning system (Philips Pinnacle). The dose proﬁles also matched EBT

2974 ﬁlm measurements. The average diﬀerences were 1.1%± 1.8% (1 s.d.) and 1.0%±

2975 1.6% (1 s.d.) between measurements by the DMG with the Pinnacle predicted dose

2976 and the EBT ﬁlm, respectively. Using the DMG, it was found that a typical prostate

2977 treatment delivery demonstrated a temporal pattern in dose rate delivery which is

2978 quite diﬀerent within a short distance (25.6 mm) between points of measurements.

2979 DMG is most useful in small ﬁeld dosimetry where accurate dose proﬁles and output

2980 factor measurements are often limited by the detector size (Das et al., 2000; Laub &

2981 Wong, 2003). This was demonstrated with the use of a DMG in the measurement of

2982 the total scatter factor, Scp for stereotactic radiosurgery (SRS) cones with diameter

2983 down to 5 mm (chapter 6). The DMG measurements agreed with the Monte Carlo

3 2984 calculation of the Scp using a voxel size of 1 × 1 × 1mm with an average diﬀerence

2985 of 1.2 ± 1.1%. An analysis of the volume averaging eﬀect of detector was carried

2 2 2986 out considering detectors with respective sizes of 1 × 1mm and 0.2 × 2mm.It 10.1. Dose Magnifying Glass 178

2987 was found that the DMG was most useful when used for measuring the steep dose

2988 gradient due to its narrow detector width, particularly in a 5 mm diameter beam.

2989 When used to measure the Scp at the beam’s central axis, the diﬀerences due to the

2990 volume averaging eﬀect was minimal. This is because of the the large plateau region

2991 at the centre of the beam. DMG derived 80-20% penumbra width of 1.77 ± 0.37 mm

2992 (1 standard deviation) for SRS cones with diameter ranges from 5 to 20 mm.

2993 The DMG coupled with an electronic readout system based on the TERA chip al-

2994 lowed fast acquisitions, hence the detector was determined as having a high temporally

2995 resolution. The high spatial and temporal resolution characteristics were successfully

2996 deployed for as an independent quality assurance tool for the veriﬁcation of helical To-

2997 motherapy binary multileaf collimator (MLC) machine parameters (chapter 7). The

2998 MLC alignment error was found to be 0.71 ± 0.09 mm. The real-time leaf motion was

2999 captured showing a leaf opening and closing across the ﬁeld within a time of 6 - 9 ms.

3000 MLC leaf open time threshold was found to be approximately 20 ms. The maximum

3001 leaf ﬂuence output factor measured was 1.097 for both adjacent leaves opened.

3002 The design of the DMG with its maximum measurement length of 25.6 mm limited

3003 its use in the dosimetric veriﬁcation of large ﬁeld IMRT or TomoTherapy treatment

3004 deliveries. Future work on this detector system would involve positioning multiple

3005 DMGs adjacent to each other, extending the usable length of the detector. This

3006 would require the development of a corresponding electronic readout system with more

3007 channels that would need to manage the larger amount of data output.

3008 Strip detectors being a linear array of detectors provide excellent spatial resolution

3009 in one dimension. However, these types of devices do not provide two dimensional

3010 spatial information, such that required for the dosimetry of an IMRT treatment. To

3011 achieve this, one would need a two dimensional detector array. Chapter 8 and 9 of

3012 this thesis described one of such device, called the Magic Plate (MP). 10.2. Magic Plate 179

3013 10.2 Magic Plate

3014 The Magic Plate is a 2D array detector based on epitaxial technology. The prototype

3015 used in this thesis has 121 detectors spaced at a pitch of 1 cm. The diodes were

3016 mounted on a thin 0.6 mm Kapton substrate using the ‘drop-in’ technique. This

3017 design allows the device to function as a transmission detector when mounted on the

3018 accessory slot of a linear accelerator. The ‘drop-in’ mounting technique avoided the use

3019 of high atomic number wire bonding on the diode and reduced the energy dependence

3020 of the diode. The thin Kapton substrate caused minimal beam perturbation when

3021 it was placed in the linac accessory slot. The MP also lends itself to be used as a

3022 conventional 2D planar detector array for dose measurements in solid water phantoms.

3023 The radiation response and basic characteristics of the MP were described in chap-

3024 ter 8. The MP showed decreased dose per pulse response at higher dose rates (dose

−4 3025 per pulse >1 × 10 Gy/pulse) due to the Auger recombination eﬀect. At lower dose

3026 rates typical for IMRT deliveries the MP appears to be dose rate independent. The

◦ 3027 temperature coeﬃcient of the MP was 0.5%/ C. Hence, temperature dependence of

3028 the MP would be minimal in most phantom based applications where the variation

3029 of the ambient temperature would not be signiﬁcant. Angular dependence of the MP

◦ 3030 was within 3.5% for the gantry angles ± 75 whilst the maximum angular response

◦ 3031 was 10.8% at of gantry angle 180 . When the MP was used as a transmission device,

2 3032 the surface dose increased by 12.1% for a 30 × 30 cm ﬁeld size at a SSD of 80 cm.

3033 However, beyond the dmax, the transmission factor for the MP was negligible.

3034 The MP was also used to verify a clinical IMRT treatment delivery in chapter 9.

3035 Planar dose measurements were made using the I’mRT phantom at the SDD of 100

3036 cm and with the MP used as a transmission detector. The Gamma index of 3% dose

3037 deviation and 3 mm distance to agreement was employed as the performance metric.

3038 Comparison was made between the MP, the Philips Pinnacle treatment planning sys- 10.2. Magic Plate 180

3039 tem (TPS) predicted and the EBT2 ﬁlm measured dose distributions. MP measured

3040 IMRT dose distribution showed reasonably good agreement with EBT2 ﬁlm measure-

3041 ments. The TPS predicted dose distributions also matched MP at most dose points

3042 using the Gamma analysis for the dose measurements in the solid water phantom.

3 3043 The MP individual detectors have a small sensitive volume (0.5 × 0.5 × 0.05 mm )

3044 of the epitaxial diodes. Due to this high spatial resolution, the diodes have minimal

3045 volumetric averaging eﬀect on the dose measured. The prototype design of the MP is

3046 currently limited primarily by the sparse pitch between each diode (i.e. 1 cm). Poppe

3047 et al. (2007) recommended a 2D array detector with resolution of 5 mm to be adequate

3048 for IMRT dose modulation measurement. This may be adequate at the normal SDD

3049 of 100 cm. However, at shorter SDD such as the MP’s use as a transmission device,

3050 the spatial resolution of the 2D transmission detector array may need to be stepped

3051 down to 2.5 mm or less.

3052 Future development of the MP system should be directed towards the implemen-

3053 tation of a larger number of diodes at a lower detector pitch. A suggested detector

3054 pitch of 2.5 mm should ensure appropriate spatial resolution of the detector array for

3055 most IMRT applications.

3056 To enable the real time dosimetric veriﬁcation of a volumetric modulated arc ther-

3057 apy (VMAT), an inclinometer should be incorporated for real time gantry angle in-

3058 formation. The rotational treatment technique also lends itself to development of an

3059 MP readout system in a wireless format to avoid excess cable wind up.

3060 In chapter 9, the dose distributions measured with MP and EBT2 ﬁlms were com-

3061 pared with the TPS predicted dose distribution. The Gamma index was employed as

3062 the comparison metric. This metric only compares the agreement of two dose distri-

3063 bution, regardless of the impact on the biological relevant parameters (Nelms et al.,

3064 2011). The use of MP as a transmission detector could be further coupled with the de- 10.2. Magic Plate 181

3065 velopment of software packages. This would allow correlation of physics performance

3066 metrics such as the Gamma index or an alternate metric to quantitatively predict the

3067 clinically relevant dosimetric implications of any mismatch between MP (i.e. the dose

3068 delivered) and the TPS (i.e. the dose planned). It may be possible at some future date

3069 to integrate the MP results into an adaptive radiotherapy process, where deviations

3070 are accounted for in the subsequent treatment fractions. Appendices

182 Appendix A

SolidWorks drawing

Detector holders and phantoms with special geometries were designed and machined speciﬁcally for particular parts of this thesis. The design of these devices often require high precision and details. A 3D computed aided drawing package, SolidWorks (Das- sault Systemes SolidWorks Corporation, Massachusetts, USA) was used to made the drawings.

A.1 The 2nd generation Dose Magnifying Glass de-

tector holder

The 2nd generation Dose Magnifying Glass (DMG) holder was used for the work described in chapter 6 and 7 (Figure A.1).

A.2 SRS phantom

The SRS phantom was used in the work described in chapter 6 (Figure A.2).

183 A.2. SRS phantom 184

C A B 150.400 A4 6 VIEW A

REVISION

SCALE 1 : 2

5 45 45 SHEET 1 OF 22 8 5 DO NOT SCALE DRAWING DMG4-assembled DWG NO. SCALE:1:5 TITLE: 130 DEBUR AND BREAK SHARP EDGES 4 300 320.700 MATERIAL: WEIGHT: DATE FINISH: SIGNATURE 170 3 NAME Q.A UNLESS OTHERWISE SPECIFIED: DIMENSIONS ARE IN MILLIMETERS SURFACE FINISH: TOLERANCES: LINEAR: ANGULAR: MFG CHK'D APPV'D DRAWN 5.700 4.300

A 36

2 2 50 1 1 A B C D

Figure A.1: The 2nd generation Dose Magnifying Glass detector holder A.2. SRS phantom 185

180

140

20 REVISION 20

SHEET 1 OF 1 SHEET 1

20 20 100

10 10 DO NOT SCALE DRAWING SCALE DO NOT 119 DWG NO. SCALE:1:3 TITLE: 160 Assem_complete_v2 20 EDGES DEBUR AND AND DEBUR BREAK SHARP 21 456 solid water MATERIAL: WEIGHT: WEIGHT:

DATE 180 FINISH:

SIGNATURE

9 R 3

NAME 20 Q.A TOLERANCES: LINEAR: ANGULAR: UNLESS OTHERWISE SPECIFIED: MILLIMETERS IN ARE DIMENSIONS FINISH: SURFACE MFG CHK'D APPV'D 70 DRAWN

R 0

7 R 2

100 180 12 1 SolidWorks Student License Academic Use Only

Figure A.2: SRS phantom A.3. Magic Plate Y-shape perspex frame 186

A.3 Magic Plate Y-shape perspex frame

The perspex frame was designed to hold the Magic Plate onto the modiﬁed TBI tray. It was used in the work described in chapter 8 and 9 (Figure A.3).

A.4 Magic Plate holder to be used with the I’mRT

phantom

The solid water plates (Figure A.4) were used in conjunction with the I’mRT phantom in the work described in chapter 9. A.4. Magic Plate holder to be used with the I’mRT phantom 187 C A B A4 6

REVISION

210 3 SHEET 1 OF 5 DO NOT SCALE DRAWING YShapePerspex 225

DWG NO. SCALE:1:5 TITLE:

1.500

48 48 8 DEBUR AND BREAK SHARP EDGES 4 MATERIAL: WEIGHT:

DATE 120 FINISH: SIGNATURE 3 270 NAME 215 Q.A UNLESS OTHERWISE SPECIFIED: DIMENSIONS ARE IN MILLIMETERS SURFACE FINISH: TOLERANCES: LINEAR: ANGULAR: MFG CHK'D APPV'D DRAWN

211

25 5

2 2

60 6 60 6 1 1 A B C D

Figure A.3: Magic Plate perspex frame A.4. Magic Plate holder to be used with the I’mRT phantom 188 C A B A4 6

REVISION

4.700

SHEET 1 OF

48 48 2.950 5

DO NOT SCALE DRAWING 4.690 ImrtInsert_assembly DWG NO. SCALE:1:5 TITLE: DEBUR AND BREAK SHARP EDGES 4 249.500 MATERIAL: WEIGHT: 160 DATE

3 FINISH: SIGNATURE A 3 NAME

5 160 Q.A UNLESS OTHERWISE SPECIFIED: DIMENSIONS ARE IN MILLIMETERS SURFACE FINISH: TOLERANCES: LINEAR: ANGULAR: MFG CHK'D APPV'D DRAWN

2 2 0.610 DETAIL A SCALE 2 : 1 1 1 A B C D

Figure A.4: Solid water plates used with an I’mRT phantom. Appendix B

Matlab Scripts

Matlab scripts and graphic user interfaces (GUIs) (Figure B) was designed using Mat- lab (The MathsWorks Inc.) to manage the large amount of data involved in chapter 9.

B.1 Uniformity correction script

This script was used to correct for MP uniformity described in chapter 9.

1 function [ CorrectionMap1D]= MP Tera6IterativeUniformityCorrection ... ( CalFile , FaceUp ) 2 %Definition 3 %CalFile contains a 5 x 121 matrix of the five calibration fields 4 %FaceUp = 1 i f the MP was irradiated in `Face−up' condition and 0 if ... it was irradiated in `Face−down ' condition 5 6 subFileArray = CalFile ; 7 [r subFileArray , c subFileArray]=size (subFileArray ) ; 8 %reshape the matrix into 2D array based on `face−up ' or `face−down ' ... condition 9 for i =1: r subFileArray 10 if FaceUp ==1 11 [Temp]=MP reshape2DFaceUp ( subFileArray (i ,:) ) ; 12 else 13 [Temp]=MP reshape2DFaceDown ( subFileArray (i ,:) ) ; 14 end 15

189 B.1. Uniformity correction script 190

Figure B.1: Matlab GUI for processing and comparing the three data sets, (i) Magic Plate, (ii) EBT2 ﬁlm and (iii) Pinnacle predicted and (iv) Monte Carlo simulated dose distributions. B.1. Uniformity correction script 191

16 subFileArray2D{ i } ( : , : ) =Temp ; 17 end 18 19 A = subFileArray2D{1}; 20 B = subFileArray2D{2}; 21 C = subFileArray2D{3}; 22 D = subFileArray2D{4}; 23 E = subFileArray2D{5}; 24 25 % % Sensitivity = S 26 27 Dose = zeros (11 ,11) ; 28 G=zeros (11 ,11) ; 29 CenterPixel = 6; 30 31 % With Step 1 & step 2 moves , the MP was shifted −1 channel on the ... x−axis , 32 %the interation would be 33 34 i=CenterPixel ; 35 j=CenterPixel ; 36 %define the dose at the central axis to 100. 37 Dose ( i , j ) = 100; 38 G( i , j )=A( i , j ) / Dose ( i , j ) ; 39 40 DoseperCount = Dose ( i , j ) /A( CenterPixel , CenterPixel ) ; 41 42 for n=1:5 43 Dose ( i , j−n)= B(i ,j−n+1) /G( i , j−n+1); 44 G( i , j−n)=A(i ,j−n) / Dose ( i , j−n); 45 n=n+1; 46 end 47 48 % With Step 1 & step 3 moves , the MP was shifted +1 channel on the x−axis 49 % relative to the center pixel , the interation would be 50 51 for n=1:5 52 Dose ( i , j+n)= C( i , j+n−1) /G( i , j +n−1) ; 53 G( i , j+n)=A( i , j+n) / Dose ( i , j+n) ; 54 n=n+1; 55 end 56 57 % With Step 1 & step 4,5 moves , the MP was shifted −1 channel on the ... y−axis 58 % relative to the center pixel , the interation would be 59 60 for j = 1:11 61 for n=1:5 62 Dose ( i−n,j)= D(i−n+1,j ) /G(i−n+1,j ) ; 63 G( i−n,j)=A(i−n , j ) / Dose ( i−n,j); 64 65 Dose ( i+n , j )= E( i+n−1,j ) /G( i+n−1,j ) ; B.2. Gamma analysis script 192

66 G( i+n , j )=A( i+n , j ) / Dose ( i+n , j ) ; 67 n=n+1; 68 end 69 j=j+1; 70 end 71 72 %Sensitivity factor 73 74 for i = 1:11 75 for j = 1:11 76 CorrectionMap (i , j )=G(i , j ) /G( CenterPixel , CenterPixel ) ; 77 end 78 end 79 %Rearrange the correction map into 1D array. 80 if FaceUp ==1 81 CorrectionMap1D=reshape ( flipud (CorrectionMap ) ,1 ,121) ; 82 else 83 CorrectionMap1D=reshape ( flipud ( fliplr (CorrectionMap ) ) ,1 ,121) ; 84 end

B.2 Gamma analysis script

The Gamma index was used as the performance matric to compare the dose distributions measured by the Magic Plate (MP), EBT2 ﬁlm and the treatment planning system (TPS). The Matlab script was based on Gamma index equation of Low et al. (1998), adapted from Scherman (2009).

1 %Definition 2 %nMP i s the MP dose values normalised to the maximum dose v a l u e . 3 %nFilm is the Film dose values normalised to the MP' s maximum dose v a l u e . 4 %LDCutoff − low dose cutoff criterion (%), is user defined and can be ... implemented , if desired. 5 %gridsize − comparison grid size (in mm) used. Both data sets will be ... interpolated to this grid size. 6 %CDose − dose deviation criterion (%) is user defined on the GUI 7 %CDTA − DTA criterion (mm) is user defined GUI 8 %AboveCutoffPercent − percentage of the pixels > the LDCutoff. 9 10 G=zeros( size (nMP)); 11 Ga=zeros( size (nFilm)); 12 13 %for interpolating to smaller grid size. 14 for i =1:size (nMP,1) 15 for j=1:size (nMP,2) B.2. Gamma analysis script 193

16 if nMP ( i , j )>LDCutoff 17 for k=1:size (nFilm ,1) 18 for l=1:size (nFilm ,2) 19 if nFilm (k , l )>LDCutoff 20 r2 =((i−k)* gridsize ) ˆ2+((j−l)* gridsize )ˆ2; 21 d2 = (nMP ( i , j )−nFilm (k , l ) ) ˆ2; 22 Ga(k , l )=r2 /CDTAˆ2+d2 / CDose ˆ2; 23 else 24 Ga ( k , l ) =NaN ; 25 end

26 end 27 end 28 else 29 for k=1:size (nFilm ,1) 30 for l=1:size (nFilm ,2) 31 Ga ( k , l ) =NaN ; 32 end 33 end 34 end 35 G( i , j )=min ( min (Ga) ) ; 36 end 37 end 38 G=sqrt (G); 39 40 % for Gamma analysis on the 121 detector position only. 41 %MPaxis − axis definition for MP 42 %Filmaxis interpV − Vertical axis definition for Film 43 %Filmaxis interpH − Horizontal axis definition for Film 44 for i =1:size (nMP,1) 45 for j=1:size (nMP,2) 46 if nMP ( i , j )>LDCutoff 47 for k=1:size (nFilm ,1) 48 for l=1:size (nFilm ,2) 49 if nFilm (k , l )>LDCutoff 50 r2 =(MPaxis ( i )−Filmaxis interpV (k))ˆ2+(MPaxis( j )− ...... 51 Filmaxis interpH ( l )) ˆ2; 52 d2 = (nMP ( i , j )−nFilm (k , l ) ) ˆ2; 53 Ga(k , l )=r2 /CDTAˆ2+d2 / CDose ˆ2; 54 else 55 Ga ( k , l ) =NaN ; 56 end 57 end 58 end 59 else 60 for k=1:size (nFilm ,1) 61 for l=1:size (nFilm ,2) 62 Ga ( k , l ) =NaN ; 63 end 64 end 65 end 66 G( i , j )=min ( min (Ga) ) ; B.2. Gamma analysis script 194

67 end 68 end 69 G=sqrt (G); 70 71 % calculate gamma 72 for i=1:size (G,1) 73 for j=1:size (G,2) 74 if G( i , j )≤1 75 count (i , j )=1; 76 else 77 count (i , j )=0; 78 end 79 end 80 end 81 %Gamma passing rate 82 PassRate = ... (sum(count (:) ) /(( size (G,1)* size (G,2) )* AboveCutoffPercent /100) )*100; References

Ahnesjo, Anders. 1989. Collapsed cone convolution of radiant energy for photon dose calculation in heterogeneous media. Medical Physics, 16(4), 577–592.

AIHW, & AACR. 2010. Cancer in Australia: an overview 2010. Tech. rept.

Alexander, D. R. 2003. Transient ionizing radiation eﬀects in devices and circuits. Nuclear Science, IEEE Transactions on, 50(3), 565–582.

Almond, Peter R., Biggs, Peter J., Coursey, B. M., Hanson, W. F., Huq, M. Saiful, Nath, Ravinder, & Rogers, D. W. O. 1999. AAPM’s TG-51 protocol for clinical reference dosimetry of high-energy photon and electron beams. Medical Physics, 26(9), 1847–1870.

Altman, Michael B., Chmura, Steven J., Deasy, Joseph O., & Roeske, John C. 2006. Optimization of the temporal pattern of radiation: An IMRT based study. Interna- tional Journal of Radiation Oncology*Biology*Physics, 66(3), 898–905.

Amerio, S., Boriano, A., Bourhaleb, F., Cirio, R., Donetti, M., Fidanzio, A., Garelli, E., Giordanengo, S., Madon, E., Marchetto, F., Nastasi, U., Peroni, C., Piermattei, A., Freire, C. J. Sanz, Sardo, A., & Trevisiol, E. 2004. Dosimetric characterization of a large area pixel-segmented ionization chamber. Medical Physics, 31(2), 414–420.

Andres, C., del Castillo, A., Tortosa, R., Alonso, D., & Barquero, R. 2010. A compre-

195 References 196

hensive study of the Gafchromic EBT2 radiochromic ﬁlm. A comparison with EBT. Medical Physics, 37(12), 6271–6278.

Asuni, G., Jensen, Martin, & McCurdy, B M C. 2011. A Monte Carlo investigation of contaminant electrons due to a novel in vivo transmission detector. Physics in Medicine and Biology, 56(4), 1207.

Baldock, C., & et al. 2010. Polymer gel dosimetry. Physics in Medicine and Biology, 55(5), R1.

Balog, J. P., Mackie, T. R., Reckwerdt, P., Glass, M., & Angelos, L. 1999. Character- ization of the output for helical delivery of intensity modulated slit beams. Medical Physics, 26(1), 55–64.

Balog, John, & Soisson, Emilie. 2008. Helical Tomotherapy Quality Assurance. In- ternational Journal of Radiation Oncology*Biology*Physics, 71(1, Supplement 1), S113–S117.

Balog, John, Mackie, T. R., Pearson, D., Hui, Susanta, Paliwal, Bhudatt, & Jeraj, Robert. 2003a. Benchmarking beam alignment for a clinical helical tomotherapy device. Medical Physics, 30(6), 1118–1127.

Balog, John, Olivera, Gustavo, & Kapatoes, Jeﬀ. 2003b. Clinical helical tomotherapy commissioning dosimetry. Medical Physics, 30(12), 3097–3106.

Balog, John, Holmes, Tim, & Vaden, Richard. 2006. Helical tomotherapy dynamic quality assurance. Medical Physics, 33(10), 3939–3950.

Battista, Jerry J., Rider, William D., & Van Dyk, Jake. 1980. Computed tomography for radiotherapy planning. International Journal of Radiation Oncol- ogy*Biology*Physics, 6(1), 99–107. References 197

Baumgartner, Andreas, Steurer, Andreas, & Josef Maringer, Franz. 2009. Simulation of photon energy spectra from Varian 2100C and 2300C/D Linacs: Simpliﬁed estimates with PENELOPE Monte Carlo models. Applied Radiation and Isotopes, 67(11), 2007–2012.

Beddar, A. S., Mason, D. J., & O’Brien, P. F. 1994. Absorbed dose perturbation caused by diodes for small ﬁeld photon dosimetry. Medical Physics, 21(7), 1075–1079.

Bedford, James L., Lee, Young K., Wai, Philip, South, Christopher P., & Warrington, Alan P. 2009. Evaluation of the Delta4 phantom for IMRT and VMAT veriﬁcation. Physics in Medicine and Biology, 54(9), N167–N176.

Beentjes, L. B., & Garrett, F. A. 1966. RADIATION INDUCED CONDUCTIVITY AND THE STEM EFFECT. American Journal of Roentgenology, 96(3), 791–793.

Benedict, Stanley H., Lin, Peck-Sun, Zwicker, Robert D., Huang, David T., & Schmidt- Ullrich, Rupert K. A. 1997. The biological eﬀectiveness of intermittent irradiation as a function of overall treatment time: Development of correction factors for linac-based stereotactic radiotherapy. International Journal of Radiation Oncol- ogy*Biology*Physics, 37(4), 765–769.

Berger, M.J., Coursey, J.S., Zucker, M.A., & Chang, J. Stopping-Power and Range Tables: Electrons, Protons, Helium Ions.

Bhardwaj, Anup Kumar, Sharma, Suresh Chander, Rana, Bhupendra, & Shukla, Arvind. 2009. Study of 2D ion chamber array for angular response and QA of dynamic MLC and pretreatment IMRT plans. Reports of Practical Oncology & Radiotherapy, 14(3), 89–94.

Bonazzola, G. C., Cirio, R., Donetti, M., Marchetto, F., Mazza, G., Peroni, C., & Zampieri, A. 1998. Performances of a VLSI wide dynamic range current-to-frequency References 198

converter for strip ionization chambers. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equip- ment, 405(1), 111–120.

Boyer, Arthur L., Antonuk, Larry, Fenster, Aaron, Van Herk, Marcel, Meertens, Harm, Munro, Peter, Reinstein, Larry E., & Wong, John. 1992. A review of electronic portal imaging devices (EPIDs). Medical Physics, 19(1), 1–16.

Boyer, Arthur L., Biggs, Peter J., Galvin, James M., Klein, Eric E., LoSasso, Thomas, Low, Daniel, Mah, Katherine, & Yu, Cedric. 2001. BASIC APPLICATIONS OF MULTILEAF COLLIMATORS. Tech. rept.

Bragg, W. H. 1910. XXXIX. The consequence of the corpuscular hypothesis of the γ and X rays, and the range of β rays. The London, Edinburgh and Dublin philosophical magazine and journal of science, 20(117), 385–416.

Brahme, A., Roos, J.E., & Lax, I. 1982. Solution of an integral equation encountered in rotation therapy. Physics in Medicine and Biology, 27(10), 1221.

Brusasco, C., Cattai, A., Cirio, R., Dellacasa, G., Donetti, M., Isoardi, P., Marchetto, F., Peroni, C., Rolando, V., Ruspa, M., Solano, A., & Zambernardi, C. 1997. Strip ionization chambers as 3-D detector for hadron therapy. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 389(3), 499–512.

Bruzzi, M., Bucciolini, M., Casati, M., Menichelli, D., Talamonti, C., Piemonte, C., & Svensson, B. G. 2007. Epitaxial silicon devices for dosimetry applications. Applied Physics Letters, 90(17), 172109–3.

Bucciolini, M., Buonamici, F. Banci, Mazzocchi, S., De Angelis, C., Onori, S., & References 199

Cirrone, G. A. P. 2003. Diamond detector versus silicon diode and ion chamber in photon beams of diﬀerent energy and ﬁeld size. Medical Physics, 30(8), 2149–2154.

Butson, M. J., Yu, P. K. N., & Cheung, T. 2003a. Rounded end multi-leaf penumbral measurements with radiochromic ﬁlm. Physics in Medicine and Biology, 48(17), N247–N252.

Butson, M. J., Cheung, T., & Yu, P. K. 2006a. Scanning orientation eﬀects on Gafchromic EBT ﬁlm dosimetry. Australasian Physical and Engineering Sciences in Medicine, 29(3), 281–4.

Butson, Martin J., Yu, Peter K. N., Cheung, Tsang, & Metcalfe, Peter. 2003b. Ra- diochromic ﬁlm for medical radiation dosimetry. Materials Science and Engineering: R: Reports, 41(3-5), 61–120.

Butson, Martin J., Cheung, Tsang, & Yu, Peter K. N. 2006b. Weak energy dependence of EBT gafchromic ﬁlm dose response in the 50 kVp-10MVp X-ray range. Applied Radiation and Isotopes, 64(1), 60–62.

Butson, Martin J., Yu, Peter K. N., Cheung, Tsang, & Alnawaf, Hani. 2010. En- ergy response of the new EBT2 radiochromic ﬁlm to x-ray radiation. Radiation Measurements, 45(7), 836–839.

Bzdusek, Karl, Friberger, Henrik, Eriksson, Kjell, Hardemark, Bjorn, Robinson, David, & Kaus, Michael. 2009. Development and evaluation of an eﬃcient approach to volumetric arc therapy planning. Medical Physics, 36(6), 2328–2339.

Caccia, M., Alemi, M., Bianchi, C., Bulgheroni, A., Cappellini, C., Conte, L., Kucewicz, W., Prest, M., Vallazza, E., & Sampietro, C. 2004. Imaging of [beta] particle sources used in medical applications with position sensitive Silicon sensors. References 200

Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spec- trometers, Detectors and Associated Equipment, 525(1-2), 294–297.

Carol, Mark P. 1995. Peacock[TM]: A system for planning and rotational delivery of intensity-modulated ﬁelds. International Journal of Imaging Systems and Technol- ogy, 6(1), 56–61.

Chen, F. Q., Gupta, R., & Metcalfe, P. 2010. Intensity modulated radiation therapy (IMRT) surface dose measurements using a PTW advanced Markus chamber. Australasian physical & engineering sciences in medicine / supported by the Aus- tralasian College of Physical Scientists in Medicine and the Australasian Association of Physical Sciences in Medicine, 33(1), 23–34.

Cheung, Tsang, Butson, Martin J., & Yu, Peter K. N. 2005. Post-irradiation coloura- tion of Gafchromic EBT radiochromic ﬁlm. Physics in Medicine and Biology, 50(20), N281–N285.

Cheung, Tsang, Butson, Martin J., & Yu, Peter K. N. 2006. Measurement of high energy x-ray beam penumbra with Gafchromic[TM] EBT radiochromic ﬁlm. Medical Physics, 33(8), 2912–2914.

Childress, Nathan L., Salehpour, Mohammad, Dong, Lei, Bloch, Charles, White, R. Allen, & Rosen, Isaac I. 2005. Dosimetric accuracy of Kodak EDR2 ﬁlm for IMRT veriﬁcations. Medical Physics, 32(2), 539–548.

Damerell, C J S. 1995. Vertex detectors: The State of the art and future prospects.

Das, Indra J., Downes, M. Beverly, Kassaee, Alireza, & Tochner, Z. 2000. Choice of Radiation Detector in Dosimetry of Stereotactic Radiosurgery-Radiotherapy. Jour- nal of Radiosurgery, 3(4), 177–186. References 201

Das, Indra J., Ding, George X., & Ahnesjo, Anders. 2008. Small ﬁelds: Nonequilibrium radiation dosimetry. Medical Physics, 35(1), 206–215.

Dawson, Beth, & Trapp, Robert G. 2004. Basic & Clinical Biostatistics. McGraw-Hill Professional.

De Angelis, C., Onori, S., Pacilio, M., Cirrone, G. A. P., Cuttone, G., Raﬀaele, L., Bucciolini, M., & Mazzocchi, S. 2002. An investigation of the operating characteristics of two PTW diamond detectors in photon and electron beams. Medical Physics, 29(2), 248–254.

Duthoy, Wim, De Gersem, Werner, Vergote, Koen, Coghe, Marc, Boterberg, Tom, De Deene, Yves, De Wagter, Carlos, Van Belle, Simon, & De Neve, Wilfried. 2003. Whole abdominopelvic radiotherapy (WAPRT) using intensity-modulated arc therapy (IMAT): ﬁrst clinical experience. International Journal of Radiation Oncol- ogy*Biology*Physics, 57(4), 1019–1032.

Ezzell, Gary A., Galvin, James M., Low, Daniel, Palta, Jatinder R., Rosen, Isaac, Sharpe, Michael B., Xia, Ping, Xiao, Ying, Xing, Lei, & Yu, Cedric X. 2003. Guidance document on delivery, treatment planning, and clinical implementation of IMRT: Report of the IMRT subcommittee of the AAPM radiation therapy com- mittee. Medical Physics, 30(8), 2089–2115.

Fan, J., Paskalev, K., Wang, L., Jin, L., Li, J., Eldeeb, A., & Ma, C. 2009. Determina- tion of output factors for stereotactic radiosurgery beams. Medical Physics, 36(11), 5292–5300.

Fenwick, J. D., Tom´e, W. A., Jaradat, H. A., Hui, S. K., James, J. A., Balog, J. P., DeSouza, C. N., Lucas, D. B., Olivera, G. H., Mackie, T. R., & Paliwal, B. R. References 202

2004. Quality assurance of a helical tomotherapy machine. Physics in Medicine and Biology, 49(13), 2933–2953.

Feygelman, V., Zhang, G., Stevens, C., & Nelms, B. 2010. SU-DD-A1-03: On Rota- tional Measurements with the ArcCHECK 3D Diode Array. Pages 3088–3088 of: Medical Physics, vol. 37. AAPM.

Feygelman, Vladimir, Forster, Kenneth, Opp, Daniel, & Nilsson, G¨orgen. 2009. Eval- uation of a biplanar diode array dosimeter for quality assurance of step-and-shoot IMRT. Journal of Applied Clinical Medical Physics, 10(4), 64–78.

Fiandra, Christian, Ricardi, Umberto, Ragona, Riccardo, Anglesio, Silvia, Giglioli, Francesca Romana, Calamia, Elisa, & Lucio, Francesco. 2006. Clinical use of EBT model Gafchromic[trademark sign] ﬁlm in radiotherapy. Medical Physics, 33(11), 4314–4319.

Fidanzio, Andrea, Azario, Luigi, Miceli, Roberto, Russo, Aniello, & Piermattei, An- gelo. 2000. PTW-diamond detector: Dose rate and particle type dependence. Med- ical Physics, 27(11), 2589–2593.

Fjeldly, T. A., Deng, Y., Shur, M. S., Hjalmarson, H. P., Muyshondt, A., & Ytterdal, T. 2001. Modeling of high-dose-rate transient ionizing radiation eﬀects in bipolar devices. Nuclear Science, IEEE Transactions on, 48(5), 1721–1730.

Fong, Peter, M., Keil, Derek K., Does, Mark D., & Gore, J. C. 2001. Polymer gels for magnetic resonance imaging of radiation dose distributions at normal room atmo- sphere. Physics in Medicine and Biology, 46(12), 3105.

Fowler, Jack F., Welsh, James S., & Howard, Steven P. 2004. Loss of biological eﬀect in prolonged fraction delivery. International Journal of Radiation Oncol- ogy*Biology*Physics, 59(1), 242–249. References 203

Fuss, Martina, Sturtewagen, Eva, Wagter, Carlos De, & Georg, Dietmar. 2007. Dosi- metric characterization of GafChromic EBT ﬁlm and its implication on ﬁlm dosimetry quality assurance. Physics in Medicine and Biology, 52(14), 4211–4225.

Garcia-Vicente, F., Delgado, J. M., & Peraza, C. 1998. Experimental determination of the convolution kernel for the study of the spatial response of a detector. Medical Physics, 25(2), 202–207.

Gerbi, Bruce J., & Khan, Faiz M. 1990. Measurement of dose in the buildup region using ﬁxed-separation plane-parallel ionization chambers. Medical Physics, 17(1), 17–26.

Gore, J. C., & Kang, Y. S. 1984. Measurement of radiation dose distributions by nuclear magnetic resonance (NMR) imaging. Physics in Medicine and Biology, 29(10), 1189.

Gottschalk, Bernard. 1983. Charge-balancing current integrator with large dynamic range. Nuclear Instruments and Methods in Physics Research, 207(3), 417–421.

Gray, L. H. 1929. The Absorption of Penetrating Radiation. Proceedings of the Royal Society of London. Series A, 122(790), 647–668.

Gray, L. H. 1936. An Ionization Method for the Absolute Measurement of γ-Ray Energy. Proceedings of the Royal Society of London. Series A - Mathematical and Physical Sciences, 156(889), 578–596.

Greer, Peter B., Vial, Philip, Oliver, Lyn, & Baldock, Clive. 2007. Experimental investigation of the response of an amorphous silicon EPID to intensity modulated radiotherapy beams. Medical Physics, 34(11), 4389–4398.

Grosu, Anca-Ligia, Kneschaurek, Peter, & Schlegel, Wolfgang. 2006. Stereotactic Radiotherapy/Radiosurgery. Pages 267–276 of: Schlegel, Wolfgang, Bortfeld, T., References 204

Grosu, Anca-Ligia, Brady, L.W, Heilmann, H.-P., Molls, M., & Sartor, K. (eds), MEDICAL RADIOLOGY, Radiation Oncology. Berlin Heidelberg: Springer-Verlag.

Grusell, E., & Rikner, G. 1984. Radiation Damage Induced Dose Rate Non-Linearity in an N-Type Silicon Detector. Acta Oncologica, 23(6), 465 – 469.

Grusell, E., & Rikner, G. 1986. Evaluation of temperature eﬀects in p-type silicon detectors. Physics in Medicine and Biology, 31(5), 527–534.

Grusell, E., & Rikner, G. 1993. Linearity with dose rate of low resistivity p-type silicon semiconductor detectors. Physics in Medicine and Biology, 38(6), 785–792.

Hardcastle, Nicholas, Tome, Wolfgang A., Foo, Kerwyn, Miller, Andrew, Carolan, Martin, & Metcalfe, Peter. Comparison of prostate imrt and vmat biologically optimised treatment plans. Medical Dosimetry, In Press, Corrected Proof.

Hartmann, Bernadette, Martisikova, Maria, & Jakel, Oliver. 2010. Technical Note: Homogeneity of Gafchromic[TM] EBT2 ﬁlm. Medical Physics, 37(4), 1753–1756.

Heydarian, M., Hoban, P. W., & Beddoe, A. H. 1996. A comparison of dosimetry techniques in stereotactic radiosurgery. Physics in Medicine and Biology, 41(1), 93–110.

Higgins, P. D., Alaei, P., Gerbi, B. J., & Dusenbery, K. E. 2003. In vivo diode dosimetry for routine quality assurance in IMRT. Medical Physics, 30(12), 3118–3123.

Hilts, M., Audet, C., Duzenli, C., & Jirasek, A. 2000. Polymer gel dosimetry using x-ray computed tomography: a feasibility study. Physics in Medicine and Biology, 45(9), 2559.

Hoban, P. W., Heydarian, M., Beckham, W. A., & Beddoe, A. H. 1994. Dose rate References 205

dependence of a PTW diamond detector in the dosimetry of a 6 MV photon beam. Physics in Medicine and Biology, 39(8), 1219–1229.

Horowitz, Paul, & Hill, Winﬁeld. 1989. The Art of Electronics. Cambridge University Press.

Hubbell, J.H., & Seltzer, S.M. 2004. Tables of X-Ray Mass Attenuation Coeﬃcients and Mass Energy-Absorption Coeﬃcients.

Hughes, H. L., & Benedetto, J. M. 2003. Radiation eﬀects and hardening of MOS technology: devices and circuits. Nuclear Science, IEEE Transactions on, 50(3), 500–521.

IAEA. 2006. Absorbed Dose Determination in External Beam Radiotherapy: An In- ternational Code of Practice for Dosimetry based on Standards of Absorbed Dose to Water. Tech. rept. Vienna.

Ibbott, Geoﬀrey S., Barnes, J. Edward, Hall, G. Richard, & Hendee, William R. 1975. Stem corrections for ionization chambers. Medical Physics, 2(6), 328–330.

IPEM, Aukett, R. J., Burns, J. E., Greener, A. G., Harrison, R. M., Moretti, C., Nahum, A. E., & Rosser, K. E. 2005. Addendum to the IPEMB code of practice for the determination of absorbed dose for x-rays below 300 kV generating potential (0.035 mm Al??4 mm Cu HVL). Physics in Medicine and Biology, 50(12), 2739– 2748.

IPEMB, Klevenhagen, S. C., Aukett, R. J., Harrison, R. M., Moretti, C., Nahum, A. E., & Rosser, K. E. 1996. The IPEMB code of practice for the determination of absorbed dose for x-rays below 300 kV generating potential (0.035 mm Al - 4 mm Cu HVL; 10 - 300 kV generating potential). Physics in Medicine and Biology, 41(12), 2605–2625. References 206

Islam, Mohammad K., Norrlinger, Bernhard D., Smale, Jason R., Heaton, Robert K., Galbraith, Duncan, Fan, Cary, & Jaﬀray, David A. 2009. An integral quality monitoring system for real-time veriﬁcation of intensity modulated radiation therapy. Medical Physics, 36(12), 5420–5428.

ISP. 2006. GAFCHROMIC EBT - SELF-DEVELOPING FILM FOR RADIOTHER- APY DOSIMETRY. Tech. rept. NJ.

ISP. 2009. GAFCHROMIC EBT2 - SELF-DEVELOPING FILM FOR RADIOTHER- APY DOSIMETRY. Tech. rept. NJ.

Johns, Harold Elford, & Cunningham, John Robert. 1983. The Physics of Radiology. Springﬁeld: Charles C Thomas.

Jursinic, P. A., & Mackie, T. R. 1996. Characteristics of secondary electrons produced by 6, 10 and 24 MV x-ray beams. Physics in Medicine and Biology, 41(8), 1499– 1509.

Jursinic, Paul A. 2001. Implementation of an in vivo diode dosimetry program and changes in diode characteristics over a 4-year clinical history. Medical Physics, 28(8), 1718–1726.

Jursinic, Paul A. 2009. Angular dependence of dose sensitivity of surface diodes. Medical Physics, 36(6), 2165–2171.

Jursinic, Paul A., & Nelms, Ben E. 2003. A 2-D diode array and analysis software for veriﬁcation of intensity modulated radiation therapy delivery. Medical Physics, 30(5), 870–879.

Jursinic, Paul A., Sharma, Renu, & Reuter, Jim. 2010. MapCHECK used for rotational IMRT measurements: Step-and-shoot, Tomotherapy, RapidArc. Medical Physics, 37(6), 2837–2846. References 207

Kapatoes, J. M., Olivera, Gustavo H., Balog, J. P., Keller, H., Reckwerdt, P., & Mackie, T. R. 2001a. On the accuracy and eﬀectiveness of dose reconstruction for tomotherapy. Physics in Medicine and Biology, 46(4), 943.

Kapatoes, J. M., Olivera, Gustavo H., Ruchala, Kenneth J., & Mackie, T. R. 2001b. On the veriﬁcation of the incident energy ﬂuence in tomotherapy IMRT. Physics in Medicine and Biology, 46(11), 2953–2965.

Konozenko, I.D., Semenyuk, A. K., & Khivrich, V.I. 1971. Radiation Defects in Si of High Purity. Radiation Eﬀects, 8, 121–127.

Kramberger, G., Contarato, D., Fretwurst, E., H??nniger, F., Lindstr??m, G., Pin- tilie, I., R??der, R., Schramm, A., & Stahl, J. 2003. Superior radiation tolerance of thin epitaxial silicon detectors. Nuclear Instruments and Methods in Physics Re- search Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 515(3), 665–670.

Kraner, H. W. 1984. Radiation damage in silicon detectors. Nuclear Instruments and Methods in Physics Research, 225(3), 615–618.

Kron, T., Rosenfeld, A., Lerch, M., & Bazley, S. 2002. Measurements in Radiotherapy Beams using On-line MOSFET Detectors. Radiation Protection Dosimetry, 101(1- 4), 445–448.

Kron, Tomas, Elliott, Andrew, & Metcalfe, Peter. 1993. The penumbra of a 6-MV x-ray beam as measured by thermoluminescent dosimetry and evaluated using an inverse square root function. Medical Physics, 20(5), 1429–1438.

La Rosa, A., Garella, M. A., Bourhaleb, F., Cirio, R., Donetti, M., Giordanengo, S., Givehchi, N., Marchetto, F., Martin, F., Meyroneinc, S., Peroni, C., & Pitt?, G. 2006. A pixel ionization chamber used as beam monitor at the Institut CurieCentre References 208

de Protontherapie de Orsay (CPO). Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equip- ment, 565(2), 833–840.

La Rosa, A., Donetti, M., Borri, M., Rivero, F., Attili, A., Bourhaleb, F., Cirio, R., Garella, M. A., Giordanengo, S., Givehchi, N., Mazza, G., Marchetto, F., Pardo, J., Pecka, A., & Peroni, C. 2008. Characterization of a front-end electronics for the monitoring and control of hadrontherapy beams. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associ- ated Equipment, 586(2), 270–275.

Langen, Katja M., Papanikolaou, Niko, Balog, John, Crilly, Richard, Followill, David, Goddu, S. Murty, Grant, Walter, Iii, Olivera, Gustavo, Ramsey, Chester R., & Shi, Chengyu. 2010. QA for helical tomotherapy: Report of the AAPM Task Group 148. Medical Physics, 37(9), 4817–4853.

Laub, Wolfram U., & Wong, Tony. 2003. The volume eﬀect of detectors in the dosimetry of small ﬁelds used in IMRT. Medical Physics, 30(3), 341–347.

Laub, Wolfram U., Kaulich, Theodor W., & Nusslin, Fridtjof. 1997. Energy and dose rate dependence of a diamond detector in the dosimetry of 425 MV photon beams. Medical Physics, 24(4), 535–536.

Lee, Christopher, Menk, Fred, Cadman, Patrick, & Greer, Peter B. 2009. A simple approach to using an amorphous silicon EPID to verify IMRT planar dose maps. Medical Physics, 36(3), 984–992.

Letourneau, Daniel, Gulam, Misbah, Yan, Di, Oldham, Mark, & Wong, John W. 2004. Evaluation of a 2D diode array for IMRT quality assurance. Radiotherapy & Oncology, 70(2), 199–206. References 209

Letourneau, Daniel, Publicover, Julia, Kozelka, Jakub, Moseley, Douglas J., & Jaf- fray, David A. 2009. Novel dosimetric phantom for quality assurance of volumetric modulated arc therapy. Medical Physics, 36(5), 1813–1821.

Long, David M., Florian, Joseph R., & Casey, Richard H. 1983. Transient Response Model for Epitaxial Transistors. Nuclear Science, IEEE Transactions on, 30(6), 4131–4134.

Low, Daniel A., Harms, William B., Mutic, Sasa, & Purdy, James A. 1998. A technique for the quantitative evaluation of dose distributionsnelmnel. Medical Physics, 25(5), 656–661.

Lutz, Wendell, Winston, Ken R., & Maleki, Nasser. 1988. A system for stereotactic radiosurgery with a linear accelerator. International Journal of Radiation Oncol- ogy*Biology*Physics, 14(2), 373–381.

Mackie, T. R. 2006. History of tomotherapy. Physics in Medicine and Biology, 51(13), R427–R453.

Mackie, T. R., Scrimger, J. W., & Battista, J. J. 1985. A convolution method of calculating dose for 15-MV x rays. Medical Physics, 12(2), 188–196.

Mackie, T. Rock, Holmes, Timothy, Swerdloﬀ, Stuart, Reckwerdt, Paul, Deasy, Joseph O., Yang, James, Paliwal, Bhudatt, & Kinsella, Timothy. 1993. Tomother- apy: A new concept for the delivery of dynamic conformal radiotherapy. Medical Physics, 20(6), 1709–1719.

MAESTRO. 2008. European Integrated Project MAESTRO.

Marre, D., & Marinello, G. 2004. Comparison of p-type commercial electron diodes for in vivo dosimetry. Medical Physics, 31(1), 50–56. References 210

Maryanski, M. J., Schulz, R.J., Ibbott, G.S., Gatensby, J.C., Xie, J., Horton, D., & Gore, J.C. 1994. Magnetic resonance imaging of radiation dose distributions using a polymer-gel dosimeter. Physics in Medicine and Biology, 39(9), 1437.

Mather, Melissa, L., Whittaker, Andrew K., & Baldock, C. 2002. Ultrasound evaluation of polymer gel dosimeters. Physics in Medicine and Biology, 47(9), 1449.

Matney, Jason, E., Parker, Brent, C., Neck, Daniel, W., Henkelmann, Greg, & Rosen, Isaac, I. 2010. Evaluation of a commercial flatbed document scanner and radiographic film scanner for radiochromic EBT film dosimetry. Journal of Applied Clin- ical Medical Physics, 11(2), 198–208.

Matuszak, Martha M., Yan, Di, Grills, Inga, & Martinez, Alvaro. 2010. Clinical Appli- cations of Volumetric Modulated Arc Therapy. International Journal of Radiation Oncology*Biology*Physics, 77(2), 608–616.

Mazza, G., Cirio, R., Donetti, M., LaRosa, A., Luparia, A., Marchetto, F., & Peroni, C. 2005. A 64-Channel Wide Dynamic Range Charge Measurement ASIC for Strip and Pixel Ionization Detectors. Nuclear Science, IEEE Transactions on, 52(4), 847–853.

Mellenberg, Jr David E. 1990. Determination of build-up region over-response corrections for a Markus-type chamber. Medical Physics, 17(6), 1041–1044.

Menichelli, David, Bruzzi, Mara, Bucciolini, Marta, Talamonti, Cinzia, Casati, Marta, Marrazzo, Livia, Tesi, Mauro, Piemonte, Claudio, Pozza, Alberto, Zorzi, Nicola, Brianzi, Mirko, & De Sio, Antonio. 2007. Design and development of a silicon- segmented detector for 2D dose measurements in radiotherapy. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 583(1), 109–113. References 211

Metcalfe, P., Kron, T., & Hoban, P. 2007. The Physics of Radiotherapy X-Rays and Electrons. Madison, Wisconsin: Medical Physics Publishing.

Meyer, Philippe, Regal, Raymond, Jung, Monique, Siﬀert, Paul, Mertz, Luc, & Con- stantinesco, Andre. 2001. Feasibility of a semiconductor dosimeter to monitor skin dose in interventional radiology. Medical Physics, 28(10), 2002–2006.

Mohan, Radhe, Chui, Chen, & Lidofsky, Leon. 1985. Energy and angular distributions of photons from medical linear accelerators. Medical Physics, 12(5), 592–597.

Moiseenko, Vitali, Duzenli, Cheryl, & Durand, Ralph E. 2007. In vitro study of cell survival following dynamic MLC intensity-modulated radiation therapy dose delivery. Medical Physics, 34(4), 1514–1520.

Moll, Michael. 1999. Radiation Damage in Silicon Particle Detectors - microscopic defects and macroscopic properties. Ph.D. thesis, University of Hamburg, Hamburg.

Nelms, B., & Feygelman, V. 2010. SU-DD-A1-02: The X’s and O’s of 3D Dosimetry Phantoms. vol. 37. AAPM.

Nelms, Benjamin E., Zhen, Heming, & Tom´e, Wolfgang A. 2011. Per-beam, planar IMRT QA passing rates do not predict clinically relevant patient dose errors. Medical Physics, 38(2), 1037–1044.

Nelms, Benjamin Edward, & Simon, Jeﬀ A. 2007. A Survey on IMRT QA Analysis. Journal of Applied Clinical Medical Physics; Vol 11, No 2 (2010).

Nilsson, Bo, & Montelius, Anders. 1986. Fluence perturbation in photon beams under nonequilibrium conditions. Medical Physics, 13(2), 191–195.

Olch, Arthur J. 2002. Dosimetric performance of an enhanced dose range radiographic References 212

ﬁlm for intensity-modulated radiation therapy quality assurance. Med. Phys., 29(9), 2159–2168.

Oldham, T. R., & McLean, F. B. 2003. Total ionizing dose eﬀects in MOS oxides and devices. Nuclear Science, IEEE Transactions on, 50(3), 483–499.

Olsson, L. E., Westrin, B A, Fransson, A., & Nordell, B. 1992. Diﬀusion of ferric ions in agarose dosimeter gels. Physics in Medicine and Biology, 37(12), 2243.

Orton, John. 2004. The story of semiconductors. New York: Oxford University Press.

Otto, Karl. 2008. Volumetric modulated arc therapy: IMRT in a single gantry arc. Medical Physics, 35(1), 310–317.

Paelinck, L., Neve, W. De, & Wagter, C. De. 2007. Precautions and strategies in using a commercial ﬂatbed scanner for radiochromic ﬁlm dosimetry. Physics in Medicine and Biology, 52(1), 231–242.

Paliwal, Bhudatt R., Zaini, Mehran, McNutt, Todd, Fairbanks, E. Jeﬀrey, & Kitchen, Rebecca. 1996. A consistency monitor for radiation therapy treatments. Medical Physics, 23(10), 1805–1807.

Pappas, E., Maris, T. G., Zacharopoulou, F., Papadakis, A., Manolopoulos, S., Green, S., & Wojnecki, C. 2008. Small SRS photon ﬁeld proﬁle dosimetry performed using a PinPoint air ion chamber, a diamond detector, a novel silicon-diode array (DOSI), and polymer gel dosimetry. Analysis and intercomparison. Medical Physics, 35(10), 4640–4648.

Partridge, M., Ebert, M., & Hesse, B. M. 2002. IMRT veriﬁcation by three-dimensional dose reconstruction from portal beam measurements. Medical Physics, 29(8), 1847– 1858. References 213

Pellegrini, G., Fleta, C., Campabadal, F., Miñano, M., Lozano, M., Rafí,J.M.,& Ullán, M. 2007. Technology of p-type microstrip detectors with radiation hard p- spray, p-stop and moderated p-spray insulations. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associ- ated Equipment, 579(2), 599–603.

Pollack, Alan, Zagars, Gunar K., Smith, Lewis G., Lee, J. Jack, von Eschenbach, An- drew C., Antolak, John A., Starkschall, George, & Rosen, Isaac. 2000. Preliminary Results of a Randomized Radiotherapy Dose-Escalation Study Comparing 70 Gy With 78 Gy for Prostate Cancer. Journal of Clinical Oncology, 18(23), 3904–3911.

Poppe, B., Thieke, C., Beyer, D., Kollhoff, R., Djouguela, A., Rühmann, A., Willborn, K. C., & Harder, D. 2006a. DAVID - a translucent multi-wire transmission ionization chamber for in vivo verification of IMRT and conformal irradiation techniques. Physics in Medicine and Biology, 51(5), 1237–1248.

Poppe, Björn, Blechschmidt, Arne, Djouguela, Armand, Kollhoff, Ralf, Rubach, Antje, Willborn, Kay C., & Harder, Dietrich. 2006b. Two-dimensional ionization chamber arrays for IMRT plan verification. Medical Physics, 33(4), 1005–1015.

Poppe, Björn, Djouguela, Armand, Blechschmidt, Arne, Willborn, Kay, Rühmann, Antje, & Harder, Dietrich. 2007. Spatial resolution of 2D ionization chamber arrays for IMRT dose verification: single-detector size and sampling step width. Physics in Medicine and Biology, 52(10), 2921–2935.

Poppe, Bj¨orn, Looe, Hui Khee, Chofor, Ndimofor, R¨uhmann, Antje, Harder, Dietrich, & Willborn, Kay C. 2010. Clinical performance of a transmission detector array for the permanent supervision of IMRT deliveries. Radiotherapy and Oncology, 95(2), 158–165. References 214

Rawlinson, J. Alan, Arlen, Dianne, & Newcombe, Dalton. 1992. Design of parallel plate ion chambers for buildup measurements in megavoltage photon beams. Medical Physics, 19(3), 641–648.

Rice, R. K., Hansen, J. J., Svensson, G. K., & Siddon, R. L. 1987. Measurements of dose distributions in small beams of 6 MV X-rays. Physics in Medicine and Biology, 32(9), 1087–1099.

Rikner, G., & Grusell, E. 1983. Eﬀects of radiation damage on p-type silicon detectors. Physics in Medicine and Biology, 28(11), 1261–1267.

Rosenfeld, A. B., Pugatch, V. M., Zinets, O. S., Litovchenko, P. G., Barabash, L. I., Pavlenko, Yu N., & Vasilyev, Yu O. 1993. Strip detector for short-range particles. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spec- trometers, Detectors and Associated Equipment, 326(1-2), 234–238.

Rosenfeld, Anatoly B. 2006. Electronic dosimetry in radiation therapy. Radiation Measurements, 41(Supplement 1), S134–S153.

Rubach, Antje, Conrad, F, & Bichsel, Hans. 1986. Dose build-up curves for cobalt- 60 irradiation: a systematic error occurring with pancake chamber measurements. Physics in Medicine and Biology, 31(4), 441.

Sadagopan, Ramaswamy, Bencomo, Jose A., Martin, Rafael Landrove, Nilsson, Gor- gen, Matzen, Thomas, & Balter, Peter A. 2009. Characterization and clinical evaluation of a novel IMRT quality assurance system. Journal of Applied Clinical Medical Physics, 10(2), 104–119.

Saini, Amarjit S., & Zhu, Timothy C. 2002. Temperature dependence of commercially available diode detectors. Medical Physics, 29(4), 622–630. References 215

Saini, Amarjit S., & Zhu, Timothy C. 2004. Dose rate and SDD dependence of commercially available diode detectors. Medical Physics, 31(4), 914–924.

Saini, Amarjit S., & Zhu, Timothy C. 2007. Energy dependence of commercially available diode detectors for in-vivo dosimetry. Medical Physics, 34(5), 1704–1711.

Scherman, Jonas Bengtsson. 2009. Developement and evaluation of methods for comparison of dose distributions in radiotherapy using calculated, synthetic and simulated measured dose distributions. Ph.D. thesis, Lund University, Lund.

Scott, Alison J. D., Nahum, Alan E., & Fenwick, John D. 2008. Using a Monte Carlo model to predict dosimetric properties of small radiotherapy photon ﬁelds. Medical Physics, 35(10), 4671–4684.

Scott, Alison J. D., Nahum, Alan E., & Fenwick, John D. 2009. Monte Carlo modeling of small photon ﬁelds: Quantifying the impact of focal spot size on source occlusion and output factors, and exploring miniphantom design for small-ﬁeld measurements. Medical Physics, 36(7), 3132–3144.

Sharma, S. C., & Johnson, M. W. 1994. Recommendations for measurement of tray and wedge factors for high energy photons. Medical Physics, 21(4), 573–575.

Shi, Jie, Simon, William E., & Zhu, Timothy C. 2003. Modeling the instantaneous dose rate dependence of radiation diode detectors. Medical Physics, 30(9), 2509–2519.

Shockley, W., & Read, W. T. 1952. Statistics of the Recombinations of Holes and Electrons. Physical Review, 87(5), 835.

Singh, Jasprit. 2001. Semiconduction Devices: Basic Principles. New York: John Wiley & Sons, Inc. References 216

Spezi, E., Angelini, A. L., Romani, F., & Ferri, A. 2005. Characterization of a 2D ion chamber array for the veriﬁcation of radiotherapy treatments. Physics in Medicine and Biology, 50(14), 3361–3373.

Stasi, M, Giordanengo, S, Cirio, R, Boriano, A, Bourhaleb, F, Cornelius, I, Donetti, M, Garelli, E, Gomola, I, Marchetto, F, Porzio, M, Sanz Freire, C J, Sardo, A, & Peroni, C. 2005. D-IMRT veriﬁcation with a 2D pixel ionization chamber: dosimetric and clinical results in head and neck cancer. Phys. Med. Biol., 50, 4681??4694.

Sze, Simon M. 1981. Physics of semiconductor devices. New York: John Wiley & Sons, Inc.

Sze, Simon M. 2001. Semiconductor Devices: Physics and Technology, 2nd Edition. Wiley.

Talamonti, C., Bruzzi, M., Bucciolini, M., Marrazzo, L., & Menichelli, D. 2007. Prelim- inary dosimetric characterization of a silicon segmented detector for 2D dose veriﬁ- cations in radiotherapy. Nuclear Instruments and Methods in Physics Research Sec- tion A: Accelerators, Spectrometers, Detectors and Associated Equipment, 583(1), 114–118.

Van Dam, J., Leunens, G., & Dutreix, A. 1990. Correlation between temperature and dose rate dependence of semiconductor response; inﬂuence of accumulated dose. Radiotherapy and Oncology, 19(4), 345–351.

Van Dam, J., Bosmans, H., Marchal, G., & Wambersie, A. 2005. Characteristics of dosemeter types for skin dose measurements in practice. Radiat Prot Dosimetry, 117(1-3), 185–189.

Van de Vondel, Iwein, Tournel, Koen, Verellen, Dirk, Duchateau, Michael, Lelie, References 217

Steven, & Storme, Guy. 2009. A diagnostic tool for basic daily quality assurance of a Tomotherapy Hi Art machine. J. Appl. Clin. Med. Phys., 10(4), 151–164.

Van Esch, Ann, Depuydt, Tom, & Huyskens, Dominique Pierre. 2004. The use of an aSi-based EPID for routine absolute dosimetric pre-treatment verification of dynamic IMRT fields. Radiotherapy and Oncology, 71(2), 223–234. van Kleffens, Herman, Venselaar, Jack, Heukelom, Stan, Jager, Niek, Mijnheer, Ben, van der Laarse, Rob, van Gasteren, Hans, & Westermann, Carel. 2000. Dependence of the tray transmission factor on collimator setting and sourcesurface distance. Medical Physics, 27(9), 2117–2123. van Zijtveld, Mathilda, Dirkx, Maarten L. P., de Boer, Hans C. J., & Heijmen, Ben J. M. 2006. Dosimetric pre-treatment verification of IMRT using an EPID; clinical experience. Radiotherapy and Oncology, 81(2), 168–175ortal dosimetry.

Velkley, D. E., Manson, D. J., Purdy, J. A., & Oliver, Jr G. D. 1975. Build-up region of megavoltage photon radiation sources. Medical Physics, 2(1), 14–19.

Venkataraman, Sankar, Malkoske, Kyle E., Jensen, Martin, Nakonechny, Keith D, Asuni, Ganiyu, & McCurdy, B M C. 2009. The inﬂuence of a novel transmission detector on 6 MV x-ray beam characteristics. Physics in Medicine and Biology, 54(10), 3173.

Vergote, K., Deene, Y. De, Duthoy, W., Gersem, W. De, Neve, W. De, Achten, E., & Wagter, C. De. 2004. Validation and application of polymer gel dosimetry for the dose veriﬁcation of an intensity-modulated arc therapy (IMAT) treatment. Physics in Medicine and Biology, 49(2), 287–305.

Wang, Jian Z., Li, X. Allen, D’Souza, Warren D., & Stewart, Robert D. 2003. Im- pact of prolonged fraction delivery times on tumor control: A note of caution for References 218

intensity-modulated radiation therapy (IMRT). International Journal of Radiation Oncology*Biology*Physics, 57(2), 543–552.

Warkentin, B., Steciw, S., Rathee, S., & Fallone, B. G. 2003. Dosimetric IMRT veriﬁcation with a ﬂat-panel EPID. Medical Physics, 30(12), 3143–3155.

Welsh, Keith T., & Reinstein, L. E. 2001. The thermal characteristics of diﬀerent diodes on in vivo patient dosimetry. Medical Physics, 28(5), 844–849.

Westermark, M., Arndt, J., Nilsson, B., & Brahme, A. 2000. Comparative dosimetry in narrow high-energy photon beams. Physics in Medicine and Biology, 45(3), 685– 702.

Wilkins, David, Li, X. Allen, Cygler, Joanna, & Gerig, Lee. 1997. The eﬀect of dose rate dependence of p-type silicon detectors on linac relative dosimetry. Medical Physics, 24(6), 879–881.

Wirth, J. L., & Rogers, S. C. 1964. The Transient Response of Transistors and Diodes to Ionizing Radiation. Nuclear Science, IRE Transactions on, 11(5), 24–38.

Wong, J. H. D., Carolan, M., Lerch, M. L. F., Petasecca, M., Khanna, S., Perevertaylo, V. L., Metcalfe, P., & Rosenfeld, A. B. 2010. A silicon strip detector dose magnifying glass for IMRT dosimetry. Medical Physics, 37(2), 427–439.

Wong, J. H. D., Hardcastle, N., Tome, W. A., Bayliss, A., Tolakanahalli, R., Lerch, M. L. F., Petasecca, M., Carolan, M., Metcalfe, P., & Rosenfeld, A. B. 2011a. Independent quality assurance of a helical tomotherapy machine using the dose magnifying glass. Medical Physics, 38(4), 2256–2264.

Wong, J. H. D., Knittel, T., Downes, S., Carolan, M., Lerch, M. L. F., Petasecca, M., Perevertaylo, V. L., Metcalfe, P., Jackson, M., & Rosenfeld, A. B. 2011b. The use References 219

of a silicon strip detector dose magnifying glass in stereotactic radiotherapy QA and dosimetry. Medical Physics, 38(3), 1226–1238.

Yan, Guanghua, Lu, Bo, Kozelka, Jakub, Liu, Chihray, & Li, Jonathan G. 2010. Calibration of a novel four-dimensional diode array. Medical Physics, 37(1), 108– 115.

Yin, Z., Hugtenburg, R. P., & Beddoe, A. H. 2004. Response corrections for solid- state detectors in megavoltage photon dosimetry. Physics in Medicine and Biology, 49(16), 3691–3702.

Yu, C. X. 1995. Intensity-modulated arc therapy with dynamic multileaf collimation: an alternative to tomotherapy. Physics in Medicine and Biology, 40(9), 1435–1449.

Zhu, X. R., Jursinic, P. A., Grimm, D. F., Lopez, F., Rownd, J. J., & Gillin, M. T. 2002. Evaluation of Kodak EDR2 ﬁlm for dose veriﬁcation of intensity modulated radiation therapy delivered by a static multileaf collimator. Medical Physics, 29(8), 1687–1692.