UNIVERSITY OF SOUTHAMPTON EEG connectivity measures and their application to assess the depth of anaesthesia and sleep by Giulia Lioi A thesis submitted in partial fulfillment for the degree of Doctor of Philosophy in the Engineering and the Environment Institute of Sound and Vibration Research January 2018 Declaration of Authorship I, Giulia Lioi, declare that this thesis titled, `EEG connectivity measures and their application to assess the depth of anaesthesia and sleep' and the work presented in it are my own. I confirm that: This work was done wholly or mainly while in candidature for a research degree at this University. Where any part of this thesis has previously been submitted for a degree or any other qualification at this University or any other institution, this has been clearly stated. Where I have consulted the published work of others, this is always clearly at- tributed. Where I have quoted from the work of others, the source is always given. With the exception of such quotations, this thesis is entirely my own work. I have acknowledged all main sources of help. Where the thesis is based on work done by myself jointly with others, I have made clear exactly what was done by others and what I have contributed myself. iii iv Parts of this work have been published as: Conference Papers Lioi G, Bell SL, Smith DC, Simpson DM. Characterization of functional brain connectivity networks in slow wave sleep. PGBiomed 2015, Liverpool, Germany. Lioi G, Bell S L and Simpson D M 2016. Changes in Functional Brain Connectivity in the Transition from Wakefulness to Sleep in different EEG bands. MEDICON 2016 Paphos, Cyprus, 2016. Ed E Kyriacou et al (Berlin: Springer) IFMBE Pro- ceedings, 57,pp 38. Lioi G, Bell SL, Smith DC, Simpson DM. Changes in EEG directional connectivity during a slow induction of propofol anesthesia. EMBEC and NBC 2017 (IFMBE Conference),Tampere, Finland.(Abstract) Journal Papers Lioi G, Bell SL, Smith DC, Simpson DM. Directional Connectivity in the EEG is able to discriminate wakefulness from NREM sleep. Physiol. Meas. 38 (2017) 18021820 Lioi G, Bell SL, Smith DC, Simpson DM. Changes in EEG directional connectivity during slow induction of anaesthesia. British Journal of Anaesthesia (In Prepa- ration) Signed: Date: \And the LORD God caused a deep sleep to fall upon Adam, and he slept; and He took one of his ribs, and closed up the flesh in its place" Genesis (2:21) \I like the scientific spirit-the holding off, the being sure but not too sure, the willingness to surrender ideas when the evidence is against them: this is ultimately fine-it always keeps the way beyond open-always gives life, thought, affection, the whole man, a chance to try over again after a mistake-after a wrong guess" Walt Whitman Abstract General anaesthesia has been used for more than two centuries to guarantee uncon- sciousness, analgesia and immobility during surgery, yet our ability to evaluate the level of anaesthesia of the patient remains insufficient. This contributes on one hand to oc- casional episodes of intraoperative awareness and recall and on the other to `controlled' drug over-dosage that increases hospital costs and patients recovery times. At present parameters used in clinical practice to monitor anaesthesia are indirect measures of the state of the brain, which is the target organ of anaesthetics. The lack of a reliable monitor of anaesthetic depth has led to considerable effort to develop new monitor- ing methods based on electrophysiological measurements. This progress has produced a series of depth of anaesthesia monitors based on various features of the electroencephalo- gram (EEG) signal. Even though these indexes are practically useful, their theoretical and physiological validity is poorly evidenced and they suffer from some practical limi- tations. As a result, their clinical uptake has been quite low. In recent years increasing attention has been given to brain connectivity as a powerful tool to investigate the com- plex behaviour of the brain. Theoretical and experimental findings have identified the disruption of brain connectivity as a crucial mechanism of anaesthetic-induced loss of consciousness. In this work a novel index of anaesthetic depth based on brain connec- tivity estimated from non-invasive scalp recordings (EEG) is proposed. Firstly, robust estimators of directed connectivity were identified in the framework of multivariate au- toregressive (MVAR) models. With a series of simulation studies the performances of these methods in estimating causal connections were assessed in particular with re- spect to the deleterious effects of instantaneous connectivity due to volume conduction. Recently published solutions were also tested (and rejected). From a comparison of connectivity measurements in simulations, MVAR based estimators were most robust to the effects of volume conduction than conventional coherence measurements. Next the performances of directed connectivity estimators were tested in two experimental studies on NREM sleep and on anaesthesia. Features that exhibited the most robust changes with the individual level of consciousness were identified and their performances in discriminating wakefulness from anaesthesia tested on ten patients undergoing a slow induction of propofol anaesthesia. The performance of the proposed method were also compared with established depth of anaesthesia indexes such as Bispectral Index (BIS) or Auditory Evoked Potentials (AEP). Results suggest that EEG connectivity features are sensitive to the anaesthetic induced changes and that they have the potential to be integrated in future monitors of intra-operative awareness and anaesthetic adequacy. Acknowledgements Firstly I would like to express my sincere gratitude to my supervisors Steven Bell, David Simpson and David Smith. I thank them in the first place for giving me the opportunity to work in their group and for the constant support and positive presence during these years. They have always encouraged my intellectual freedom, supported me in attending meetings and conferences and stimulated the exchange of ideas. I deeply thank them for their wisdom, generosity and humanity: they have made my PhD a cheerful journey and taught me to be a better scientist and person. I also would like to thank the members of my internal review panel Professor Paul White and Dr. Thomas Blumensath for their useful comments to my work. My days in the Signal Processing and Control Group have been joyful also thanks to my fellow labmates: Luigi, Dario, Michele Iodice, Michele Zilletti thanks for all the laughs! I am very grateful to Alessandro Beda, Nadjia Cristinne Carvalho, Hyorrana and Cris- tiano to have warmly welcomed me in UFMG and helped me with the experiments: their support and kindness have been very precious for me and never to be forgotten. My deepest gratitude full of good memories and a little of nostalgia goes to the Shaftes- bury 65 crew: Damien, Domenico, Tual, Alessandra, Laura, Jana, Armand, Tobias, Xander and all the persons gravitating towards this unusual collection of people that have been my family here in Southampton and made these years unforgettable. Finally, I would like to thank my parents and my brother that have always believed in me and supported me with their unlimited love, regardless of the distances that stand between us. vii Contents Declaration of Authorship iii Abstract vi Acknowledgements vii List of Figures xiii List of Tables xxv Abbreviations xxvii Symbols xxix 1 Introduction1 1.1 Aims and Hypothesis of the Research....................3 1.2 Thesis Overview................................4 1.3 Original Contributions.............................5 2 Mechanisms and Electrophysiological signatures of Anaesthesia9 2.1 Component and Mechanisms of Anaesthesia.................9 2.1.1 Clinical Signs and EEG patterns of General Anaesthesia and their relation to Sleep............................ 10 2.1.2 Models of anaesthetic action..................... 12 2.2 Electrophysiological Measures for Anaesthesia Monitoring......... 13 2.2.1 Bispectral Index............................ 14 2.2.2 Auditory Evoked Potentials...................... 16 2.2.3 Comparison of DoA monitors..................... 18 2.3 Anaesthesia and Brain Connectivity..................... 19 3 Measuring coupling and causality with Multivariate Autoregressive modeling 23 3.1 Non-directed measures of coupling: Coherence and Partial Coherence... 24 3.2 Directed measures of causality derived from MVAR models........ 28 3.2.1 Interpretation in the sense of Granger Causality.......... 32 3.2.2 Differences between P DC and DC .................. 32 ix Contents x 3.2.3 An Illustrative Example........................ 33 3.3 MVAR Model Identification and Validation................. 41 3.3.1 MVAR model estimation....................... 41 3.3.1.1 MVAR estimation algorithm................ 42 3.3.1.2 Model Order Selection.................... 45 3.3.2 Model Validation............................ 46 3.4 Summary.................................... 51 4 The Impact of Instantaneous Effects on the Estimation of Scalp Con- nectivity 53 4.1 The impact of instantaneous causality on the estimation of DC and PDC between scalp channels............................. 53 4.1.1 The extended MVAR model...................... 55 4.1.2 Objectives................................ 59 4.1.3 Methods................................. 60 Simulation study I....................... 60 Simulation study II....................... 62 Application to recorded EEG data.............. 65 4.1.4