UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Modulation of frequency-dependent EEG connectivity in Parkinson's disease Tropini, Giorgia 2010

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2010_fall_tropini_giorgia.pdf [ 4.28MB ]
Metadata
JSON: 24-1.0071053.json
JSON-LD: 24-1.0071053-ld.json
RDF/XML (Pretty): 24-1.0071053-rdf.xml
RDF/JSON: 24-1.0071053-rdf.json
Turtle: 24-1.0071053-turtle.txt
N-Triples: 24-1.0071053-rdf-ntriples.txt
Original Record: 24-1.0071053-source.json
Full Text
24-1.0071053-fulltext.txt
Citation
24-1.0071053.ris

Full Text

MODULATION OF FREQUENCY-DEPENDENT EEG CONNECTIVITY IN PARKINSON'S DISEASE by  GIORGIA TROPINI B.Sc. (Hons), University of British Columbia, 2006  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE  in The Faculty of Graduate Studies (Neuroscience)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  July 2010 © Giorgia Tropini, 2010  ABSTRACT Traditional models of Parkinson's disease (PD) have emphasized the progressive degeneration of dopaminergic projections to the basal ganglia (BG). Since the advent of deep brain stimulation (DBS) surgery for PD that allows direct BG recordings, it has become apparent that PD is also characterized by abnormal oscillatory activity within BG-thalamocortical loops. Altered β band activity in particular has been shown to correlate with bradykinesia and rigidity, and appears to be suppressed by both levodopa medication and high-frequency subthalamic nucleus (STN) DBS. However, recent animal studies have suggested that the primary site of DBS action is the cortex, thus implying that cortical areas might play a greater role than previously recognized in the modulation of abnormal PD rhythms. This thesis aimed to investigate cortical connectivity modulation in frequency bands (α, β) that have been described in oscillatory models of PD, and to understand the effects of levodopa on connectivity. We utilized a sparse Multivariate Autoregressive (mAR)-based Partial Directed Coherence (PDC) method to assess frequency-dependent EEG connectivity in PD subjects and controls performing a visually guided task previously shown to modulate abnormal oscillatory activity in the STN of PD patients. In addition, we utilized traditional spectral analysis to evaluate task-dependent power modulation in five electrode regions of interest. While spectral analysis revealed power modulation differences between PD and control subjects, it showed relatively modest differences between regions. In contrast, PDC-based analysis revealed complex, regiondependent alterations of directional connectivity in PD subjects as compared to normal subjects. Connectivity was particularly altered posteriorly, suggesting abnormalities in visual and visuomotor processing. Moreover, connectivity measures correlated with motor Unified Parkinson‟s Disease Rating Scores (UPDRS) in PD subjects withdrawn from medication. Levodopa administration only partially restored connectivity, and in some cases resulted in further exacerbation of abnormalities.  ii  Overall, we suggest that the use of a PDC-based method might be ideally suited to investigate temporally-sensitive, directional connectivity changes in both the healthy and the diseased state using non-invasive EEG. Our findings have implications for the investigation of abnormal rhythms not only in PD, but also in other conditions characterized by altered oscillatory activity, such as epilepsy or schizophrenia.  iii  TABLE OF CONTENTS Abstract ............................................................................................................................... ii Table of Contents ............................................................................................................... iv List of Tables ..................................................................................................................... vi List of Figures ................................................................................................................... vii List of Abbreviations ....................................................................................................... viii Acknowledgments.............................................................................................................. ix Dedication ........................................................................................................................... x Co-Authorship Statement................................................................................................... xi Chapter 1: Introduction and Purpose ............................................................................ 1 1.1 Parkinson‟s Disease - Symptoms, Neuropathology and Therapeutics ............... 1 1.2 The Classical Model of the Basal Ganglia .......................................................... 3 1.3 Basal Ganglia Oscillations .................................................................................. 4 1.4 Network-wide Oscillations - The Role of the Cortex ......................................... 7 1.5 Study Motivation and Aims ................................................................................ 9 1.6 Therapeutic Implications .................................................................................. 13 1.7 References ......................................................................................................... 15 Chapter 2: Partial Directed Coherence-Based Information Flow in Parkinson's Disease Patients Performing a Visually-guided Motor Task ............................................ 20 2.1 Preamble ........................................................................................................... 20 2.2 Summary ........................................................................................................... 20 2.3 Introduction ....................................................................................................... 21 2.4 Methods............................................................................................................. 25 2.5 Results ............................................................................................................... 32 2.6 Discussion ......................................................................................................... 35 2.7 Acknowledgment .............................................................................................. 38 2.8 References ......................................................................................................... 39 Chapter 3: Altered Cortical Connectivity in Parkinson‟s Disease Patients Performing a Visually Guided Task ..................................................................................................... 41 3.1 Preamble ........................................................................................................... 41 3.2 Introduction ....................................................................................................... 42 3.3 Methods............................................................................................................. 45 Subjects ..................................................................................................................... 45 Behavioral Paradigm ................................................................................................. 47 Data Collection and Pre-processing .......................................................................... 48 Data Analysis ............................................................................................................ 50 3.4 Results ............................................................................................................... 53 Behavioural Measurements ....................................................................................... 53 Temporal Dynamics- Spectrograms ......................................................................... 54 Temporal Dynamics- PDCograms ............................................................................ 55 Quantitative Analysis - Power Spectra ..................................................................... 56 Quantitative Analysis - PDC Spectra and Correlation with UPDRS ........................ 56 3.5 Discussion ......................................................................................................... 65 3.6 Conclusions ....................................................................................................... 71 iv  3.7 References ......................................................................................................... 72 Chapter 4: General Discussion and Conclusions ........................................................ 76 4.1 Introduction ....................................................................................................... 76 4.2 Summary of Findings ........................................................................................ 77 4.3 Study Significance ............................................................................................ 79 4.4 Study Limitations .............................................................................................. 83 4.5 Future Directions .............................................................................................. 86 4.6 References ......................................................................................................... 89 Appendices ........................................................................................................................ 92 Appendix I: STN LFP Data Collection ......................................................................... 92 I.1 Overview ............................................................................................................. 92 I.2 Methods ............................................................................................................... 92 I.3 Outcomes and Discussion .................................................................................... 95 I.4 References ........................................................................................................... 96 Appendix II: Ethics Certificates.................................................................................... 97 II.1 Ethics Certificates for EEG Data Collection ...................................................... 97 II.2 Ethics Certificates for DBS LFP Data Collection ............................................ 101 Appendix III: Additional Materials for Chapter 3 ...................................................... 103 III.1 Details for all subjects ..................................................................................... 103 III.2 Response times for all included subjects ........................................................ 105 III.3 Temporal Dynamics - Additional PDCograms ............................................... 106 III.4 Quantitative Analysis-Correlations of PDC connectivity with OFF UPDRS 109 III.5 Quantitative Analysis- Correlation of PDC connectivity with PD subject gender ................................................................................................................................. 116  v  LIST OF TABLES Table 3-1. Summary of patient details. ............................................................................. 46 Table A-1. Details for all healthy controls...................................................................... 103 Table A-2. Details for all PD subjects. ........................................................................... 104  vi  LIST OF FIGURES Figure 1-1. The Classical Basal Ganglia Model. ............................................................... 3 Figure 1-2. Anatomical connections and oscillatory connections in the BG. .................... 4 Figure 1-3. STN LFP in a PD patient................................................................................. 6 Figure 2-1.Visually guided choice reaction paradigm (Tropini et al. 2009, © [2009] IEEE)................................................................................................................................. 29 Figure 2-2. Flow Analysis Schematics. ........................................................................... 32 Figure 2-3. Average spectrograms and PDCograms for each electrode region. .............. 34 Figure 2-4.Average information flow into the electrode regions of interest for all subject groups. ............................................................................................................................... 35 Figure 3-1: Schematics of joystick task. .......................................................................... 47 Figure 3-2. Headplot of electrode regions. ...................................................................... 49 Figure 3-3. Spectrograms for all groups. ......................................................................... 59 Figure 3-4. PDCograms for all groups, a). ....................................................................... 60 Figure 3-5. PDCograms for all groups, b). ...................................................................... 61 Figure 3-6. Power Spectra for all groups. ........................................................................ 62 Figure 3-7. PDC Spectra for the PRE phase of the task. ................................................. 63 Figure 3-8. PDC Spectra for the MOV phase of the task. ............................................... 64 Figure A-1. Additional PDCograms for all groups, a). .................................................. 106 Figure A-2. Additional PDCograms for all groups, b). ................................................. 107 Figure A-3. Additional PDCograms for all groups, c). .................................................. 108 Figure A-4. PRE Central to FCentral Correlation Graph, 8-30 Hz................................ 109 Figure A-5. PRE Central to FCentral Correlation Graph, 13-30 Hz.............................. 109 Figure A-6. MOV Central to FCentral Correlation Graph, 8-12 Hz.............................. 112 Figure A-7. MOV Central to FCentral Correlation Graph, 8-30 Hz.............................. 113 Figure A-8. MOV Central to FCentral Correlation Graph, 13-30 Hz............................ 113  vii  LIST OF ABBREVIATIONS 6-OHDA- 6-Hydroxydopamine  mAR- Multivariate AutoRegressive  AAR- Automatic Artifact Removal  MEG- Magnetoencephalography  AR- AutoRegressive  PD- Parkinson's Disease  a.u. Arbitrary Units  PDC- Partial Directed Coherence  BG- Basal Ganglia  SL- Synchronization Likelihood  BIC- Bayesian Information Criterion  SMA- Supplementary Motor Area  DA- Dopamine  smAR- Sparse Multivariate  DBS- Deep Brain Stimulation  Autoregressive  ECT- ElectroConvulsive Therapy  STN- Subthalamic Nucleus  EEG- Electroencephalography  TMS- Transcranial Magnetic  ERD- Event Related Desynchronization  Stimulation  ERS- Event Related Synchronization  TTL- Transistor-Transistor Logic  fMRI- Functional Magnetic Resonance  UPDRS- Unified Parkinson's Disease  Imaging  Rating Score  GC- Granger Causality GPi- Globus Pallidus (Internal) GPe- Globus Pallidus (External) LFP - Local Field Potential LID- Levodopa-Induced Dyskinesia LQA- Local Quadratic Approximation M1- Primary Motor Cortex  viii  ACKNOWLEDGMENTS I would first like to thank my supervisor, Dr. Martin McKeown, for his generous support and encouragement throughout my graduate studies. Thanks to your guidance, I have learnt a great deal about Parkinson's disease and biomedical engineering techniques, and developed skills that will be extremely helpful in my future research endeavors in medicine. I am also thankful for your medical guidance and emotional support while I was recovering from the concussion earlier this year. You really helped me through difficult times and made it easier to get back on track afterwards. Thank you also to Dr. Jane Wang and Joyce Chiang for their help with analysis methods, and for their insight and helpful comments. Thanks also go to my committee members Dr. Romeo Chua, Dr. Raul de la FuenteFernandez and Dr. Debbie Giaschi for their time, guidance and useful suggestions. I would also like to thank our collaborator Dr. Chris Honey for allowing me to perform recordings during deep brain stimulation surgery for Parkinson's disease patients, and Dr. Neil Simms for showing me around in the OR and assisting with setup. While working at the Pacific Parkinson's Research Centre, I have been extremely fortunate to be surrounded by very supportive staff, clinicians and fellow students. Thank you especially to Dr. Edna Ty for her help recruiting subjects and assistance running experiments, to Shamin Babul for being the "guardian angel" looking after all of us students in various financial and bureaucratic matters, and to Edwin Mak for his help building and repairing electrical equipment. Thanks also to Sam, Magda, Vignan, James, Vesta, Shawna, Agnes and Ali for your camaraderie, help and encouragement that have made the last three years truly enjoyable and memorable. I am particularly indebted to all the patients and their family members that gave their time to participate in my experiments, my work would have not been possible without all of you. Last but not least, I would like to thank my parents, Ian, my sister Carolina, Titta and Nonno Beppe for their constant support, encouragement and love, I could not have done it without you. A special thanks goes to my Nonnetta, you have been my inspiration while working on Parkinson's disease, and this work is dedicated to you.  ix  DEDICATION  For Nonnetta  x  CO-AUTHORSHIP STATEMENT Chapters 2 and 3 of this thesis involve work that was performed under the supervision of Dr. Martin J. McKeown, and in collaboration with Dr. Jane Z. Wang and PhD student Joyce Chiang. I was responsible for designing and implementing the experimental paradigm and related Graphical User Interface in Matlab, as well as setting up the synchronized EEGVisual Stimuli system. I also personally recruited subjects N001 and P001-P005, while the remaining subjects were recruited by Dr. Edna Ty. I collected all patient and control subject data myself, and was responsible for training subjects and setting up the EEG cap with the assistance of Dr. Ty. I obtained written and verbal informed consent from all subjects, and assisted with assessment of Parkinson's disease (PD) symptoms from PD subjects off their medication. While the analysis method (smAR-PDC) utilized in chapters 2 and 3 was developed by our collaborators Dr. Wang and Ms Chiang, I implemented and optimized the analysis in Matlab after pre-processing all of the EEG data, under the supervision of Dr. McKeown. I created all of the figures for both chapters, and was responsible for the interpretation and writing of both manuscripts, save for the technical methods section in Chapter 2, and for comments and suggestions from my research supervisor and collaborators. In addition, I was responsible for setting up the portion of the study conducted in collaboration with Dr. Chris Honey, and involving recordings during DBS surgery for PD (discussed in Appendix I). Specifically, I prepared and obtained Ethics Board Approval  xi  for the study from both UBC and the Operating Room (OR) at VGH, and was responsible for having our recording apparatus tested and vetted both during the Ethics Approval process and separately the day before the surgery. I also set up and tested the portable system that involved a recording unit synchronized to a visual stimuli computer. In addition, I recruited subjects with the help of Dr. Honey and Dr. Ty, obtained informed consent and collected data during DBS surgery.  xii  CHAPTER 1:  1.1  INTRODUCTION AND PURPOSE  Parkinson’s Disease - Symptoms, Neuropathology and Therapeutics  Parkinson‟s disease (PD) is a progressive neurodegenerative movement disorder, characterized by the motor symptoms of rest tremor, rigidity and bradykinesia. Other symptoms include postural instability, dyskinesias, autonomic dysfunction, as well as cognitive and psychiatric changes (Samii et al. , 2004). PD occurrence in industrialized countries is 0.3% of the overall population, a prevalence that rises to 1% in the population aged 60 years and older. Men tend to be more affected by the disorder than women, while occurrence is similar among different ethnicities (Samii et al. , 2004). Progression of the disease is characterized by asymmetry, with one side of the body being initially more affected, although over time clinical manifestations become progressively more bilateral. Post-mortem analysis of tissue from affected patients also shows the presence of intraneuronal inclusions known as Lewy bodies (Braak et al. , 2003). The primary pathology in PD is the progressive depletion of dopamine-secreting neurons in the substantia nigra pars compacta (SNc). In the healthy state, these fibers project to the striatum, a part of the basal ganglia (BG), which are a collection of deep structures that play an important role in the generation and control of movement. Clinical signs of the disease are generally evident when about 60-80% of dopaminergic fibers are lost (Fearnley and Lees, 1991). Since the 1960's, the 'gold standard' of pharmacological treatments for PD has been the administration of levodopa, a synthetic analogue of the precursor of dopamine 1  (DA), L-dopa. Unlike DA, levodopa can reach the brain by crossing the blood-brain barrier (BBB) upon entering the blood circulation after oral administration. Once in the brain, levodopa is then decarboxylated by the enzyme aromatic amino acid decarboxylase (AADC) to form DA (Aminoff, 1994). Surgical intervention was also introduced as a form of treatment for PD in the 1940‟s and 1950‟s. (Samii et al. , 2004). The most commonly used forms of surgery were thalamotomy (the ablation of the areas of the thalamus responsible for motor relay to the cortex), and pallidotomy (the surgical destruction of small areas of the globus pallidus), with the goal of removing the tonic inhibition of motor output caused by the disease. By the 1980's and 1990's, ablative surgery was largely substituted by functional neurosurgery, with the advent of high frequency deep brain stimulation (DBS) of the subthalamic nucleus (STN) or the internal globus pallidus (GPi)(Perlmutter and Mink, 2006). DBS has been particularly effective in treating patients that develop side effects while being pharmacologically treated, as it reduces the time a patient spends in the "off" state when medication wears off. In addition, DBS allows medication doses to be reduced thus lessening the extent of side effects (Jaggi et al. , 2004, Perlmutter and Mink, 2006). While PD can be managed with pharmacological or surgical treatments, both of these approaches can result in serious side effects, such as psychosis, compulsive behaviours (i.e. gambling), depression, mania, and even suicide (Lader, 2008, Temel et al. , 2009).  2  Figure 1-1. The Classical Basal Ganglia Model. Shown are excitatory (GLU) and inhibitory (GABA) connections. (Adapted from PHYL 301 Class Notes, UBC 2004).  1.2  The Classical Model of the Basal Ganglia  The classical model of the BG (Figure 1-1) involves the striatum projecting to the motor cortex via a relay to the globus pallidus internal (GPi) first, followed by the VL/VA nuclei of the thalamus (direct pathway), as well as through a relay to the external globus pallidus (GPe), the subthalamic nucleus (STN), the GPi and finally the VL/VA thalamus (indirect pathway)(Albin et al. , 1989). The cortex also influences BG activity through striatal inputs. Dopaminergic inputs from the SNc normally promote excitation of the direct pathway (through D1 receptors in the striatum) and inhibition of the indirect pathway (through D2 receptors), thus inducing an overall increase in thalamic activity and motor output. When dopamine (DA) is depleted, as is the case in PD, the thalamus is tonically inhibited and motor output decreases (Albin et al. , 1989, Obeso et al. , 2002). Thus, the classical model of the BG explains the poverty of movement observed in PD 3  patients via firing rate changes caused by DA depletion (Albin et al. , 1989, Arbuthnott and Garcia-Munoz, 2009).  1.3  Basal Ganglia Oscillations  Since the advent of functional neurosurgery, and through electrophysiological studies in rodents and primates, electrical recordings have revealed the presence of oscillatory activity in BG-thalamocortical circuits (Brown and Williams, 2005, Gatev et al., 2006). Several circuits within the BG have been shown to possess intrinsic pacemaker oscillatory activity in both cell cultures and animal recordings: the GPe-STN circuit (Figure 1-2, connection "1"), cortical- STN circuits ("2"), STN-GPi ("3"), BG-thalamic connections ("4" and "5") as well as striatal-cortical connections ("6"). (Gatev et al. , 2006).  Figure 1-2. Anatomical connections and oscillatory connections in the BG. Red arrows indicate excitatory connections, while black arrows denote inhibitory ones. Green arrows and numbers refer to connections and internuclear mechanisms that involve circuit oscillations (reprinted from Gatev et al. , 2006)  4  While the loss of DA in PD has been shown to result in changes in firing rates and increased bursting in BG neurons (Wichmann and DeLong, 2003), recent studies in PD patients undergoing DBS surgery have revealed the presence of abnormal oscillatory activity in the BG (Brown and Williams, 2005). Altered oscillatory activity has been recorded both via microelectrode recording of pairs of neuronal units (Levy et al. , 2000), as well as from macroelectrode recording of local field potentials (LFPs) which provide information on local synchronization of neuronal activity, as the timing of neuronal discharges is closely related to fluctuations in the LFP (Brown, 2003). Oscillatory activity in the BG can be subdivided into roughly four bands, <8 Hz, α (8-12 Hz), β (13-30 Hz), and >60 Hz, although inconsistencies remain in the definition of these frequency bands (Brown, 2003, Brown and Williams, 2005). LFP BG activity in the β band has been best characterized in the literature. It has mainly been described in studies utilizing recordings in the STN (Brown et al. , 2001, Kühn et al. , 2004, Priori et al. , 2004, Trottenberg et al. , 2007, Williams et al. , 2002, Williams et al. , 2003, Williams et al. , 2005), and in the GPi (Brown et al. , 2001, Priori et al. , 2002, Silberstein et al. , 2003) of PD patients undergoing DBS surgery. Synchronization of neuronal activity in the β band is suppressed by levodopa medication (Brown et al. , 2001, Levy et al. , 2000, Levy et al. , 2002a, Levy et al. , 2002b, Williams et al. , 2002), and dopaminergic treatment has been associated, albeit inconsistently, with synchronization at higher (60-90 Hz) frequencies (Brown et al. , 2001, Brown and Williams, 2005, Cassidy et al. , 2002, Foffani et al. , 2005, Fogelson et al. , 2006, Williams et al. , 2002) (Figure 1-3).  5  Figure 1-3. STN LFP in a PD patient. (A) LFP following overnight withdrawal from medication. (B) LFP after levodopa administration.(C) and (D) are the corresponding power spectra. All data is from the same contact (#12) in left STN (reprinted from Brown and Williams, 2005, Copyright (2005), with permission from Elsevier).  In addition, β-band LFP oscillations in PD patients are suppressed by behaviourally relevant stimuli, and both prior to, and during, externally- and self-paced movements (Amirnovin et al. , 2004, Doyle et al. , 2005, Kühn et al. , 2004, Williams et al. , 2003). Suppression of activity in the 8-30 Hz band during movement has also been reported in the putamen of an epileptic patient via stereoencephalographic (SEEG) recordings (Sochurkova and Rektor, 2003) and in the striatum of healthy primates (Courtemanche et al. , 2003). Thus, it has been suggested that suppression of activity in this 'idling' frequency range might aid motor processing (Brown and Williams, 2005). For example, it has been shown that the latency of LFP β-power suppression correlates with mean reaction time in PD patients performing a visually guided choice reaction task (Kühn et al. , 2004). Similarly, in a choice reaction paradigm in which both 100% and 50% predictive warning cues were presented prior to the go cue, the duration of β-power suppression following the 50% predictive warning cue was less than in the case of the 6  100% predictive cue, again pointing to a role for β-power suppression in movement preparation (Williams et al. , 2003). Thus, if β-power suppression has a role in motor processing, abnormal and persistent synchronization in this range might disturb the proper rate coding that is necessary for the generation of movement. However, it is not clear whether excessive synchronization in this range is a purely pathological aspect of PD or merely an exacerbation of normal physiology (Brown and Williams, 2005). Nevertheless, excessive synchronization in this range has been dubbed 'antikinetic' (Brown and Williams, 2005), as direct stimulation of the STN in PD patients at 10 Hz (Timmermann et al. , 2004) as well as at 20 Hz (Chen et al. , 2007, Fogelson et al. , 2005) has been shown to slow motor performance and worsen bradykinesia. In contrast, high frequency stimulation (HFS, 130 Hz) of the STN via DBS surgery has been shown to result in suppression of βband LFP activity that correlates with increased movement amplitude in a simple wrist pronation-supination motor task (Kühn et al. , 2008). In addition, the degree of levodopainduced LFP β-band power suppression in the STN has been shown to correlate with clinical improvements in both bradykinesia and rigidity, but not tremor (Kühn et al. , 2009). 1.4  Network-wide Oscillations - The Role of the Cortex  Abnormal oscillatory activity is not limited to the STN or GPi, but instead appears to be present throughout the entire BG-thalamocortical circuitry (Brown et al. , 2001, Brown and Williams, 2005, Cassidy et al. , 2002, Gatev et al. , 2006, Williams et al. , 2002). In particular, coupling of activity between the STN and the GPi, and STN and cortex in the β band has been reported (Klostermann et al. , 2007, Marsden et al. , 2001, 7  Williams et al. , 2002). Interestingly, a recent LFP study (Eusebio et al. , 2009) showed that the BG-cortical network has a tendency to resonate at ~20 Hz in PD patients off medication, suggesting that a resonance phenomenon might be responsible for the amplification and propagation of oscillatory activity that is synchronized in the β band. In addition, the study showed that DA might act to increase damping in the circuit and thus reduce resonance in the network (Eusebio et al. , 2009). In agreement with the network-wide nature of these abnormal oscillations, recent work has also shown that activity in the STN can be modulated via both direct and indirect motor cortical stimulation. A recent study showed a decrease in STN β-power following transcranial magnetic stimulation (TMS) of the primary motor cortex (M1) and the supplementary motor area (SMA) (Doyle Gaynor et al. , 2008). Similarly, β-band activity showed event related desynchronization (ERD) during motor imagery of a warned time reaction motor task (Kühn et al. , 2006). In fact, the role of the cortex in the parkinsonian state has recently been under investigation, suggesting a review of traditional BG models that focused mainly on the direct and indirect pathways. A recent study has shown that high-frequency DBS induces antidromic stimulation of layer V (the input layer to the BG)  cortical cells in 6-  hydroxydopamine (6-OHDA) lesioned, as well as healthy, anesthetized rodents (Li et al. , 2007). In addition, the size of the antidromic evoked potential correlates with therapeutic improvements of akinesia in awake freely moving rats (Dejean et al. , 2009). A study involving optical deconstruction of the parkinsonian neural circuit in awake 6-OHDA lesioned rats further confirmed that the primary site of DBS action is not the STN, but rather the motor cortex (Gradinaru et al. , 2009). Using optical activation of channel  8  proteins that govern membrane excitability in neurons, researchers in the Deisseroth Group were able to directly inhibit or activate target fibers ('optical HFS') in the BGthalamocortical circuits. They showed that activation or silencing of the STN had no effect on PD behavioural symptoms shown by the rodents, while stimulation of layer V neurons in M1 resulted in significant behavioural improvements (Gradinaru et al. , 2009). Thus, it has been proposed that the next model of the BG should place more emphasis on the action of the cortex on striatal cells and its overall effect on the BG loops, especially in the context of the generation and propagation of abnormal firing patterns that appear to be incompatible with movement (Arbuthnott and Garcia-Munoz, 2009). Indeed, additional cortical inputs to the BG that were not originally included in the Albin, Young and Penney model (Albin et al. , 1989) have been recently described. One such connection is the 'hyperdirect' pathway that connects the cortex to the STN and that might have an important function in the selection of appropriate motor programs in the control of voluntary movements (Aron et al. , 2007, Nambu et al. , 2002), as well as in the inhibition of initiated responses (Aron and Poldrack, 2006).  1.5  Study Motivation and Aims  As the cortex appears to play a greater role than previously recognized in the modulation of abnormal rhythms in PD, further exploration into the state of cortical areas in the parkinsonian state is warranted. In particular, this thesis aimed to investigate cortical connectivity modulation in the frequency bands that have previously been described in oscillatory models of BG in PD (β-band in particular, but also α-band and  9  frequencies < 10 Hz), and to understand the effects of levodopa medication on connectivity. We chose to utilize electroencephalography (EEG) to investigate the rhythms of interest primarily because of its high temporal resolution (~ms) that would be able to capture rapid oscillatory changes in neuronal activity. In addition, the EEG might be biased to detect widespread activity that has synchronized over large scales (Volkmann, 1998), and would thus be appropriate to study network-wide oscillations in the BGthalamocortical circuitry. Thus, not only would our investigation assess frequencydependent connectivity in cortical areas, but it would also allow us to indirectly investigate activity in the BG-thalamocortical loops, especially given that a strong correlation between cortical and BG activity has been found in previous studies (Cassidy et al. , 2002, Klostermann et al. , 2007, Marsden et al. , 2001, Williams et al. , 2002). In addition, EEG measurements are non-invasive, and thus allowed us to collect data not only from PD patients but also from control subjects. This is particularly important given that a vast majority of the studies conducted on abnormal oscillatory activity in BG-cortical loops have utilized invasive LFP recordings during DBS surgery, and thus could not directly compare PD patients to healthy controls. The non-invasive nature of the EEG also allowed us to assess PD patients at less advanced stages of the disease than those typically seen in candidates for functional neurosurgery. A number of previous studies have investigated the relationship between activity in the STN and cortical activity via concurrent LFP and scalp EEG recordings (Cassidy et al. , 2002, Fogelson et al. , 2006, Klostermann et al. , 2007, Lalo et al. , 2008, Marsden et al. , 2001, Williams et al. , 2002). However, these studies utilized a limited number of  10  EEG electrodes (at most 6 electrodes) due to difficulties in placing electrodes in the presence of surgical dressings. In addition, these investigations mainly utilized coherence analysis as a measure of correlation between activity in the STN and limited areas in the cortex, and thus only provided a general estimation of connectivity between the BG and cortical areas. A few EEG studies have investigated cortico-cortical functional connectivity in the context of abnormal oscillatory activity in PD subjects only (no controls were included), utilizing coherence as a measure of coupling between cortical regions (Cassidy and Brown, 2001, Silberstein et al. , 2005). However, coherence analysis provides a measure of correlation of activity between areas of interest, but does not provide directional (causal) information that is particularly important when assessing biological systems. In addition, coherence is sensitive to volume conduction effects (Nunez et al. , 1997, Pereda et al. , 2005, Srinivasan et al. , 2007, Winter et al. , 2007), thus introducing confounds in the interpretation of results. A few other studies used magnetoencephalography (MEG) to measure functional connectivity in PD and controls using a synchronization likelihood (SL) technique ( a general measure of linear and non-linear temporal correlations between time series) (Stoffers et al. , 2008a, Stoffers et al. , 2008b). However, SL does not allow one to obtain directional connectivity information, and in addition these studies only investigated the resting state. This thesis aimed to expand on these previous studies by using a method of analysis that would not only reveal correlations between EEG nodes or cortical areas, but could also infer causal dependencies between them. In collaboration with the Wang  11  Group (UBC Electrical and Computer Engineering), we implemented a partial directed coherence (PDC) method based on a sparse multivariate autoregressive model (smAR). The use of PDC-based methods has recently increased in popularity in the analysis of EEG data for neuroscience studies (Astolfi et al. , 2007, Sun et al. , 2009, Witte et al. , 2009), as PDC allows to measure the strength of directional connectivity between two nodes or regions of interest (and thus provide a measure of causality) while at the same time discounting the effects of all other neighboring regions (Baccalá and Sameshima, 2001). In addition, since PDC-based methods take into account the influence of other nodes when assessing the connectivity between a given pair of signals, they are more robust to volume conduction effects that afflict traditional coherence-based techniques. The development of the method and its preliminary application to EEG data from PD subjects and controls is described in Chapter 2. Chapter 3 of this thesis subsequently describes the full application of the method to assess the modulation of abnormal brain rhythms in PD patients ON and OFF medication and in healthy controls. In particular, we chose to investigate modulation of activity during a simple choice reaction task based on a paradigm previously shown to modulate STN LFP β-band activity (Amirnovin et al. , 2004) We hypothesized that connectivity would be altered in PD in the frequency bands of interest and that levodopa medication would restore connectivity. We chose to investigate not only the β (13-30 Hz) frequency but also the α (8-12 Hz) and θ (4-7 Hz) frequencies, as recent work has suggested that abnormal STN oscillations might not only involve the β-band but also lower frequencies (Priori et al. , 2004).  12  In addition, we initiated a collaboration with VGH Neurosurgery in order to test a small subset of PD patients both before and during STN DBS neurosurgery, so that we could better compare information derived from LFP recordings to that inferred via EEG techniques. A description of methods and results from this portion of the study is contained in Appendix I. 1.6  Therapeutic Implications  The capability to non-invasively assess abnormal oscillations and modulation of connectivity in PD via EEG recordings can benefit existing therapies for PD, and can potentially open new therapeutic avenues. For example, our findings can have several potential implications for DBS surgery: (1) After DBS implementation, stimulator settings could be set to minimize abnormal connectivity measures derived from the EEG, as opposed to optimizing stimulation parameters based on crude clinical tests; (2) Since DBS can have beneficial effects on symptoms normally unresponsive to medication, such as sleep and postural symptoms, benefits of surgery could be predicted a priori by checking for persistent excessive abnormal rhythms in the EEG even after medication; (3) As adaptive changes in the brain are seen after DBS, changes in efficacy of DBS over time could be assessed by non-invasive EEG; (4) The extent of altered rhythms and connectivity could be used as a functional marker for disease severity even in PD subjects too mildly affected to be normally considered for surgery. In addition, an increased understanding of cortical oscillations and their modulation can potentially aid and direct the use of TMS as a non-invasive therapy for PD, given that recent work has shown a suppression of β-band oscillations in the STN 13  following TMS of SMA and M1(Doyle Gaynor et al. , 2008). Similarly, dorsal column stimulation (DCS) has recently been shown to restore locomotion in two rodent models of PD, potentially by indirectly activating large cortical areas and subsequently increasing cortical input to the striatum, in turn activating striatal projection neurons (Fuentes et al. , 2009). Although DCS would need further, long-term testing in primate models of PD before it can be evaluated for clinical use, an increased understanding of oscillatory activity in the cortex could aid future therapeutic efforts. Another alternative therapy that has been shown to have beneficial effects in PD patients that are refractory to other types of therapeutic interventions is electroconvulsive therapy (ECT). It has been hypothesized that ECT might act in PD by enhancing DA neurotransmission and increasing sensitivity of DA receptors (Popeo and Kellner, 2009). However, it is also conceivable that ECT might -at least temporarily- disrupt abnormal oscillations in the BG-thalamocortical loops. Thus, methods capable of detecting changes in modulation of frequency-dependent connectivity could be beneficial in assessing potential post-ECT functional changes in cortical activity.  Ultimately, this project aimed to increase our understanding of how connectivity is modulated in PD in the context of altered BG-cortical oscillations, with the long-term goal to facilitate existing treatments for this disease, and potentially lead to novel therapeutic approaches. In addition, the development of methods to non-invasively assess altered connectivity and abnormal synchronization patterns can be beneficial to the study of other conditions characterized by altered oscillatory neuronal activity, such as epilepsy or schizophrenia.  14  1.7  References  Albin RL, Young AB, Penney JB. The functional anatomy of basal ganglia disorders. Trends Neurosci. 1989 Oct;12(10):366-75. Aminoff MJ. Treatment of Parkinson's disease. West J Med. 1994 Sep;161(3):303-8. Amirnovin R, Williams ZM, Cosgrove GR, Eskandar EN. Visually guided movements suppress subthalamic oscillations in Parkinson's disease patients. J Neurosci. 2004 Dec 15;24(50):11302-6. Arbuthnott G, Garcia-Munoz M. Dealing with the devil in the detail - some thoughts about the next model of the basal ganglia. Parkinsonism Relat Disord. 2009 Dec 1;15 Suppl 3:S139-42. Aron AR, Poldrack RA. Cortical and subcortical contributions to Stop signal response inhibition: role of the subthalamic nucleus. J Neurosci. 2006 Mar 1;26(9):2424-33. Aron AR, Behrens TE, Smith S, Frank MJ, Poldrack RA. Triangulating a cognitive control network using diffusion-weighted magnetic resonance imaging (MRI) and functional MRI. J Neurosci. 2007 Apr 4;27(14):3743-52. Astolfi L, De Vico Fallani F, Cincotti F, Mattia D, Marciani MG, Bufalari S, et al. Imaging functional brain connectivity patterns from high-resolution EEG and fMRI via graph theory. Psychophysiology. 2007 Nov 1;44(6):880-93. Baccalá LA, Sameshima K. Partial directed coherence: a new concept in neural structure determination. Biol Cybern. 2001 Jun 1;84(6):463-74. Braak H, Del Tredici K, Rub U, de Vos RA, Jansen Steur EN, Braak E. Staging of brain pathology related to sporadic Parkinson's disease. Neurobiol Aging. 2003 Mar-Apr;24(2):197-211. Brown P, Oliviero A, Mazzone P, Insola A, Tonali P, Di Lazzaro V. Dopamine dependency of oscillations between subthalamic nucleus and pallidum in Parkinson's disease. J Neurosci. 2001 Feb 1;21(3):1033-8. Brown P. Oscillatory nature of human basal ganglia activity: relationship to the pathophysiology of Parkinson's disease. Mov Disord. 2003 Apr;18(4):357-63. Brown P, Williams D. Basal ganglia local field potential activity: character and functional significance in the human. Clinical Neurophysiology. 2005 Nov 1;116(11):2510-9. Cassidy M, Brown P. Task-related EEG-EEG coherence depends on dopaminergic activity in Parkinson's disease. Neuroreport. 2001 Mar 26;12(4):703-7. Cassidy M, Mazzone P, Oliviero A, Insola A, Tonali P, Di Lazzaro V, et al. Movement-related changes in synchronization in the human basal ganglia. Brain. 2002 Jun 1;125(Pt 6):1235-46. Chen CC, Litvak V, Gilbertson T, Kühn A, Lu CS, Lee ST, et al. Excessive synchronization of basal ganglia neurons at 20 Hz slows movement in Parkinson's disease. Experimental Neurology. 2007 May 1;205(1):214-21. Courtemanche R, Fujii N, Graybiel AM. Synchronous, focally modulated beta-band oscillations characterize local field potential activity in the striatum of awake behaving monkeys. J Neurosci. 2003 Dec 17;23(37):11741-52.  15  Dejean C, Hyland B, Arbuthnott G. Cortical effects of subthalamic stimulation correlate with behavioral recovery from dopamine antagonist induced akinesia. Cereb Cortex. 2009 May 1;19(5):1055-63. Doyle Gaynor LMF, Kühn AA, Dileone M, Litvak V, Eusebio A, Pogosyan A, et al. Suppression of beta oscillations in the subthalamic nucleus following cortical stimulation in humans. European Journal of Neuroscience. 2008 Oct 1;28(8):1686-95. Doyle LM, Kuhn AA, Hariz M, Kupsch A, Schneider GH, Brown P. Levodopa-induced modulation of subthalamic beta oscillations during self-paced movements in patients with Parkinson's disease. Eur J Neurosci. 2005 Mar;21(5):1403-12. Eusebio A, Pogosyan A, Wang S, Averbeck B, Gaynor LD, Cantiniaux S, et al. Resonance in subthalamocortical circuits in Parkinson's disease. Brain. 2009 Aug 1;132(Pt 8):2139-50. Fearnley JM, Lees AJ. Ageing and Parkinson's disease: substantia nigra regional selectivity. Brain. 1991 Oct;114 ( Pt 5):2283-301. Foffani G, Ardolino G, Rampini P, Tamma F, Caputo E, Egidi M, et al. Physiological recordings from electrodes implanted in the basal ganglia for deep brain stimulation in Parkinson's disease. the relevance of fast subthalamic rhythms. Acta Neurochir Suppl. 2005;93:97-9. Fogelson N, Kuhn AA, Silberstein P, Limousin PD, Hariz M, Trottenberg T, et al. Frequency dependent effects of subthalamic nucleus stimulation in Parkinson's disease. Neurosci Lett. 2005 Jul 1-8;382(1-2):5-9. Fogelson N, Williams D, Tijssen M, van Bruggen G, Speelman H, Brown P. Different functional loops between cerebral cortex and the subthalmic area in Parkinson's disease. Cereb Cortex. 2006 Jan;16(1):6475. Fuentes R, Petersson P, Siesser WB, Caron MG, Nicolelis MA. Spinal cord stimulation restores locomotion in animal models of Parkinson's disease. Science. 2009 Mar 20;323(5921):1578-82. Gatev P, Darbin O, Wichmann T. Oscillations in the basal ganglia under normal conditions and in movement disorders. Mov Disord. 2006 Oct 1;21(10):1566-77. http://www3.interscience.wiley.com/journal/76507419/home Gradinaru V, Mogri M, Thompson KR, Henderson JM, Deisseroth K. Optical deconstruction of parkinsonian neural circuitry. Science. 2009 Apr 17;324(5925):354-9. Jaggi JL, Umemura A, Hurtig HI, Siderowf AD, Colcher A, Stern MB, et al. Bilateral stimulation of the subthalamic nucleus in Parkinson's disease: surgical efficacy and prediction of outcome. Stereotact Funct Neurosurg. 2004;82(2-3):104-14. Klostermann F, Nikulin VV, Kühn AA, Marzinzik F, Wahl M, Pogosyan A, et al. Task-related differential dynamics of EEG alpha- and beta-band synchronization in cortico-basal motor structures. Eur J Neurosci. 2007 Mar 1;25(5):1604-15. Kühn AA, Williams D, Kupsch A, Limousin P, Hariz M, Schneider G-H, et al. Event-related beta desynchronization in human subthalamic nucleus correlates with motor performance. Brain. 2004 Apr 1;127(Pt 4):735-46. Kühn AA, Doyle L, Pogosyan A, Yarrow K, Kupsch A, Schneider G-H, et al. Modulation of beta oscillations in the subthalamic area during motor imagery in Parkinson's disease. Brain. 2006 Mar 1;129(Pt 3):695-706.  16  Kühn AA, Kempf F, Brücke C, Gaynor Doyle L, Martinez-Torres I, Pogosyan A, et al. High-frequency stimulation of the subthalamic nucleus suppresses oscillatory beta activity in patients with Parkinson's disease in parallel with improvement in motor performance. J Neurosci. 2008 Jun 11;28(24):6165-73. Kühn AA, Tsui A, Aziz T, Ray N, Brücke C, Kupsch A, et al. Pathological synchronisation in the subthalamic nucleus of patients with Parkinson's disease relates to both bradykinesia and rigidity. Experimental Neurology. 2009 Jan 14;215(2):380-7. Lader M. Antiparkinsonian medication and pathological gambling. CNS Drugs. 2008;22(5):407-16. Lalo E, Thobois S, Sharott A, Polo G, Mertens P, Pogosyan A, et al. Patterns of bidirectional communication between cortex and basal ganglia during movement in patients with Parkinson disease. J Neurosci. 2008 Mar 19;28(12):3008-16. Levy R, Hutchison WD, Lozano AM, Dostrovsky JO. High-frequency synchronization of neuronal activity in the subthalamic nucleus of parkinsonian patients with limb tremor. J Neurosci. 2000 Oct 15;20(20):7766-75. Levy R, Ashby P, Hutchison WD, Lang AE, Lozano AM, Dostrovsky JO. Dependence of subthalamic nucleus oscillations on movement and dopamine in Parkinson's disease. Brain. 2002a Jun 1;125(Pt 6):1196209. Levy R, Hutchison WD, Lozano AM, Dostrovsky JO. Synchronized neuronal discharge in the basal ganglia of parkinsonian patients is limited to oscillatory activity. J Neurosci. 2002b Apr 1;22(7):2855-61. Li S, Arbuthnott GW, Jutras MJ, Goldberg JA, Jaeger D. Resonant antidromic cortical circuit activation as a consequence of high-frequency subthalamic deep-brain stimulation. J Neurophysiol. 2007 Dec 1;98(6):3525-37. Marsden JF, Limousin-Dowsey P, Ashby P, Pollak P, Brown P. Subthalamic nucleus, sensorimotor cortex and muscle interrelationships in Parkinson's disease. Brain. 2001 Feb 1;124(Pt 2):378-88. Nambu A, Tokuno H, Takada M. Functional significance of the cortico-subthalamo-pallidal 'hyperdirect' pathway. Neurosci Res. 2002 Jun;43(2):111-7. Nunez PL, Srinivasan R, Westdorp AF, Wijesinghe RS, Tucker DM, Silberstein RB, et al. EEG coherency. I: Statistics, reference electrode, volume conduction, Laplacians, cortical imaging, and interpretation at multiple scales. Electroencephalogr Clin Neurophysiol. 1997 Nov 1;103(5):499-515. Obeso JA, Rodriguez-Oroz MC, Rodriguez M, Arbizu J, Gimenez-Amaya JM. The basal ganglia and disorders of movement: pathophysiological mechanisms. News Physiol Sci. 2002 Apr;17:51-5. Pereda E, Quiroga RQ, Bhattacharya J. Nonlinear multivariate analysis of neurophysiological signals. Prog Neurobiol. 2005 Jan 1;77(1-2):1-37. Perlmutter JS, Mink JW. Deep brain stimulation. Annu Rev Neurosci. 2006;29:229-57. Popeo D, Kellner CH. ECT for Parkinson's disease. Med Hypotheses. 2009 Oct;73(4):468-9. Priori A, Foffani G, Pesenti A, Bianchi A, Chiesa V, Baselli G, et al. Movement-related modulation of neural activity in human basal ganglia and its L-DOPA dependency: recordings from deep brain stimulation electrodes in patients with Parkinson's disease. Neurol Sci. 2002 Sep;23 Suppl 2:S101-2.  17  Priori A, Foffani G, Pesenti A, Tamma F, Bianchi AM, Pellegrini M, et al. Rhythm-specific pharmacological modulation of subthalamic activity in Parkinson's disease. Exp Neurol. 2004 Oct;189(2):369-79. Samii A, Nutt JG, Ransom BR. Parkinson's disease. Lancet. 2004 May 29;363(9423):1783-93. Silberstein P, Kuhn AA, Kupsch A, Trottenberg T, Krauss JK, Wohrle JC, et al. Patterning of globus pallidus local field potentials differs between Parkinson's disease and dystonia. Brain. 2003 Dec;126(Pt 12):2597-608. Silberstein P, Pogosyan A, Kühn AA, Hotton G, Tisch S, Kupsch A, et al. Cortico-cortical coupling in Parkinson's disease and its modulation by therapy. Brain. 2005 Jun 1;128(Pt 6):1277-91. Sochurkova D, Rektor I. Event-related desynchronization/synchronization in the putamen. An SEEG case study. Exp Brain Res. 2003 Apr;149(3):401-4. Srinivasan R, Winter WR, Ding J, Nunez PL. EEG and MEG coherence: measures of functional connectivity at distinct spatial scales of neocortical dynamics. J Neurosci Methods. 2007 Oct 15;166(1):4152. Stoffers D, Bosboom JLW, Deijen JB, Wolters EC, Stam CJ, Berendse HW. Increased cortico-cortical functional connectivity in early-stage Parkinson's disease: an MEG study. Neuroimage. 2008a Jun 1;41(2):212-22. Stoffers D, Bosboom JLW, Wolters EC, Stam CJ, Berendse HW. Dopaminergic modulation of corticocortical functional connectivity in Parkinson's disease: an MEG study. Experimental Neurology. 2008b Sep 1;213(1):191-5. Sun Y, Zhang H, Feng T, Qiu Y, Zhu Y, Tong S. Early cortical connective network relating to audiovisual stimulation by partial directed coherence analysis. IEEE Trans Biomed Eng. 2009 Nov 1;56(11 Pt 2):27214. Temel Y, Tan S, Visser-Vandewalle V, Sharp T. Parkinson's disease, DBS and suicide: a role for serotonin? Brain. 2009 Oct;132(Pt 10):e126; author reply e7. Timmermann L, Wojtecki L, Gross J, Lehrke R, Voges J, Maarouf M, et al. Ten-Hertz stimulation of subthalamic nucleus deteriorates motor symptoms in Parkinson's disease. Mov Disord. 2004 Nov;19(11):1328-33. Trottenberg T, Kupsch A, Schneider G-H, Brown P, Kühn AA. Frequency-dependent distribution of local field potential activity within the subthalamic nucleus in Parkinson's disease. Experimental Neurology. 2007 May 1;205(1):287-91. Volkmann J. Oscillations of the human sensorimotor system as revealed by magnetoencephalography. Mov Disord. 1998;13 Suppl 3:73-6. Wichmann T, DeLong MR. Functional neuroanatomy of the basal ganglia in Parkinson's disease. Adv Neurol. 2003;91:9-18. Williams D, Tijssen M, Van Bruggen G, Bosch A, Insola A, Di Lazzaro V, et al. Dopamine-dependent changes in the functional connectivity between basal ganglia and cerebral cortex in humans. Brain. 2002 Jul 1;125(Pt 7):1558-69.  18  Williams D, Kühn A, Kupsch A, Tijssen M, van Bruggen G, Speelman H, et al. Behavioural cues are associated with modulations of synchronous oscillations in the human subthalamic nucleus. Brain. 2003 Sep 1;126(Pt 9):1975-85. Williams D, Kühn A, Kupsch A, Tijssen M, van Bruggen G, Speelman H, et al. The relationship between oscillatory activity and motor reaction time in the parkinsonian subthalamic nucleus. Eur J Neurosci. 2005 Jan 1;21(1):249-58. Winter WR, Nunez PL, Ding J, Srinivasan R. Comparison of the effect of volume conduction on EEG coherence with the effect of field spread on MEG coherence. Stat Med. 2007 Sep 20;26(21):3946-57. Witte H, Ungureanu M, Ligges C, Hemmelmann D, Wüstenberg T, Reichenbach J, et al. Signal informatics as an advanced integrative concept in the framework of medical informatics. New trends demonstrated by examples derived from neuroscience. Methods Inf Med. 2009 Jan 1;48(1):18-28.  19  CHAPTER 2: PARTIAL DIRECTED COHERENCE-BASED INFORMATION FLOW IN PARKINSON'S DISEASE PATIENTS PERFORMING A VISUALLY-GUIDED MOTOR TASK1 2.1  Preamble  Chapter 2 describes the preliminary implementation of a sparse, multivariate autoregressive (mAR) partial directed coherence (PDC) method that was initially developed by our collaborators from the Wang Group (UBC Electrical and Computer Engineering). In this chapter, we present the theoretical basis for the method, and test its validity and applicability to the study of oscillatory activity in PD. In particular, we analyzed a preliminary EEG data set from PD subjects OFF medication and healthy controls performing a visually guided task that had previously been shown to modulate βband activity in LFP recordings of PD patients (Amirnovin et al., 2004).  2.2  Summary  We propose a partial directed coherence (PCD) method based on a sparse multivariate autoregressive (mAR) model to investigate patterns of information flow in electroencephalography (EEG) recordings in Parkinson‟s disease (PD) patients performing a visually-guided motor task. The use of a sparsity constraint on the mAR matrix addresses issues such as sample size, model order selection and number of parameters to be estimated, particularly when the number of EEG channels used is large and the window size is small in order to capture dynamic changes. The proposed PDC-  1  A version of this chapter has been published: Tropini G, Chiang J, Wang Z, McKeown MJ. Partial directed coherence-based information flow in Parkinson's disease patients performing a visually-guided motor task. Conf Proc IEEE Eng Med Biol Soc. 2009;2009:1873-8, © [2009] IEEE.  20  based information flow analysis demonstrated distinctly altered patterns of connectivity between PD patients off medication and healthy subjects, particularly with respect to net information outflow from the left sensorimotor (L Sm) region, which might indicate excessive spreading of activity in the diseased state. In addition, PDC-based analysis proved to be more sensitive to temporally dynamic and region-specific changes than traditional spectral analysis. We suggest that sparse mAR-PDC is a suitable technique to investigate altered connectivity in Parkinson's disease.  2.3  Introduction  Recent trends in neurobiology and signal processing research show an increased interest in understanding the functional connectivity of brain regions that can be inferred from electroencephalography (EEG) data. Since the EEG remains the most widespread technology capable of recording brain activity at msec resolution, several mathematical methods have been proposed to infer connectivity changes that occur at rapid time scales, such as correlation, coherence and Granger causality (Baccalá and Sameshima, 2001, Pereda et al. , 2005). Disrupted brain connectivity is being increasingly recognized in pathological conditions such as Parkinson‟s disease, epilepsy, schizophrenia and Alzheimer‟s Disorder (AD). In fact, synchronous oscillations in the β (13-30 Hz) and γ (30-100 Hz) bands of the EEG may underlie cognitive functions such as object perception, selective attention, working memory, as well as consciousness (Varela et al. , 2001). Moreover, synchronization between different brain regions may facilitate functional coupling that  21  could underlie integrative processes (Brown and Williams, 2005, Engel and Singer, 2001). Parkinson‟s disease (PD) is an excellent model to assess abnormal synchronization and connectivity in humans. Patients undergoing surgery for deep brain stimulation (DBS) have electrodes placed in deep brain structures such as the subthalamic nucleus (STN), providing a unique opportunity to assess local field potential (LFP) oscillatory behavior in brain circuits that are not normally accessible by non-invasive (scalp) EEG recordings. These neuronal recordings demonstrate that a lack of the neurotransmitter dopamine, characteristic of the disease, results in an abnormal synchronization of neuronal activity within basal ganglia (BG) structures such as the STN and the globus pallidus external (GPe) (Brown and Williams, 2005). In particular, excessive synchronization in the β-band in these brain structures appears to suppress movement and could thus be involved in PD symptoms such as bradykinesia (slowness of movement) (Brown and Williams, 2005, Brown, 2006, Gatev et al. , 2006). Moreover, the PD state results in large-scale oscillatory activity that can readily propagate through the BG circuitry and spread to numerous cortical areas through BG-thalamocortical connections (Gatev et al. , 2006). Thus, an increased understanding of patterns of cortical synchronization and connectivity in PD could potentially allow us to infer modulation of activity in subcortical structures within the BG. In fact, animal models suggest that there may be a stronger correlation between cortical and BG activity than previously thought (Magill et al. , 2006, Tseng et al. , 2001), and therefore the EEG might be an accurate, albeit indirect, marker of abnormal BG oscillations, despite the fact that it still largely reflects cortical activity. Most significantly, work in PD patients undergoing DBS surgery  22  of the STN and performing a choice reaction task has shown that patterns of event related synchronization (ERS) and desynchronization (ERD) in the  and β-bands were robustly detected simultaneously in both scalp and depth recordings (Klostermann et al. , 2007). In order to investigate functional connectivity changes in PD, we employed a partial directed coherence (PDC) based methodology. Traditional methods of EEG analysis such as coherence and Granger causality (GC) are limited by the fact that they can only investigate pair-wise connectivity while neglecting the possible influence(s) from other nodes. While coherence is unable to identify the direction of information flow between cortical regions (Pereda et al. , 2005), GC is capable of this distinction. GC is based on comparing the variance of residuals from a scalar autoregressive (AR) application to one signal x(t), with that from a bivariate AR application to x(t) and a potentially driving signal y(t) (Ding et al. 2006). However, traditionally GC has only been used in multiple pair-wise analyses when considering the multichannel case (Baccalá and Sameshima, 2001). To overcome these limitations, partial directed coherence (PDC) examines multichannel time series and allows the simultaneous modeling of all channels with a multivariate autoregressive (mAR) model (Baccalá and Sameshima, 2001). This has the advantages of allowing differentiation of direct and indirect causal influences among interacting entities and giving more reliable estimations of causality than bivariate techniques. Moreover, since neural signals often exhibit frequency-specific oscillatory activity, the ability to provide spectral information of causal relations makes PDC an attractive tool for neuroscience studies. Finally, since PDC takes into account influences of all other channels when assessing the connectivity between any given pair of signals, it  23  will be more robust to volume conduction effects (i.e. the simultaneous detection of multiple loci of scalp activity that, although generated from the same subcortical source, might erroneously be attributed to separate sources) that may affect standard coherence techniques. However, the computation of PDC poses several technical challenges in real applications, especially when the number of EEG channels of interest is relatively large in practice (e.g. ~20). As noted in (Pereda et al. , 2005), the successful estimation of PDC critically depends upon the proper fitting of the mAR model to the data, which in turn is dependent upon the number of channels, optimal mAR model order selection and sample sizes of the training data. In general, a higher model order allows more intricate data dynamics to be captured and gives a higher frequency resolution, but at the expense of a greater number of parameters to be estimated. Furthermore, the full-connectivity assumption of PDC analysis may be questionable in EEG results, as the EEG typically demonstrates “small world” network properties, i.e. most nodes are not directly connected to one another, but long range connections between local clusters of nodes can be present (e.g. via the corpus callosum) (Ferri et al. , 2008, Micheloyannis et al. , 2006). Additionally, a sparse model requires the estimation of fewer parameters, of critical importance when data samples are at a premium. The above observations motivated us to incorporate a sparse mAR model into the current PDC technique as has recently been suggested for fMRI analysis (Valdes-Sosa et al. , 2005). In this paper, we utilized a sparse mAR-based PDC technique in which PDC estimates are derived from sparse mAR coefficient matrices given by penalized  24  regression. Penalized regression effectively reduces the number of free parameters to be estimated, which is particularly important when the data are of limited sample size. The sparse-mAR based PDC model was used to determine patterns of information flow in PD patients performing a visually guided motor task using a joystick. The task was based upon the paradigm employed by Amirnovin et al. (Amirnovin et al. , 2004), which had shown modulation of beta-band synchronization during movement in local field potential (LFP) recordings of PD patients undergoing DBS surgery.  2.4  Methods  We first review traditional multivariate autoregressive models and least squarebased parameter estimation techniques. We then introduce the concept of sparse mAR and present penalized regression methods for solving such sparse problems. The definition of partial directed coherence is then given in Section 2.4-A. Finally, Sections 2.4, C-F describe the experimental paradigm and data collection as well as analysis techniques. A. Sparse mAR Model In a regular mAR model, the multivariate time series at each time point is represented as a linear, weighted sum of its previous time points, and it can be formulated as P  y (t )     A ry  t   r    e (t )  (1)  r 1  where the observation y(t) is a d-dimensional vector at time t, p denotes the order of the mAR model, and the vector e(t) represents white Gaussian noise. The mAR  25  coefficient Ar is a d x d matrix, where the element Ar (i, j) measures the influence that variable j exerts on variable i after r time points. In a regression framework, Eqn 1 can be rewritten as Ζ  X  E  (2)  where: Z  Y p  1: N  [ y ( p  1) , y ( p  2 ) , ..., y ( N ) ] , T  X  [ Y p : N  1 , Y p  1: N  2 , ..., Y 1: N  p ] ,    [ A 1 , A 2 , ..., A p ] , T  E  [ e ( p  1) , e ( p  2 ) , ...e ( N ) ] . T  Eqn. 2 can be solved using the maximum likelihood (ML) approach. Under the iid (independent and identically distributed) white noise assumption of E, this is equivalent to minimizing the mean square error: ˆ  a rg m in ( Z  X   2  (3)    It is worth emphasizing that the performance of the ML estimator is highly dependent on the sample size N and the number of parameters to be estimated. In real applications, the available data points are often limited, which in turn leads to poor estimation accuracy. Furthermore, the estimated coefficients yielded by the least square (LS) approach in (3) are typically non-zero, which makes neurobiological interpretation of results (e.g. identifying brain connectivity patterns in EEG studies) difficult. Such nonzero observations are also against the sparsity assumption in brain connectivity networks. To address these issues, a possible solution is to impose a sparsity constraint on the mAR coefficients (i.e. Ar matrix) and perform variable selection using penalized regression methods (Valdes-Sosa et al. , 2005). The basic idea of penalized regression is 26  to maximize the likelihood while at the same time, penalize complex models. In mathematical terms, penalized regression can be expressed as the minimization of the penalized least square function: d  ˆ  a rg m in ( Z  X  ) β  2     2    p(   j  ),  (4)  i 1  where λ is the regularization parameter which controls the amount of penalization imposed on the solution. In this paper, the value of λ is determined using the Bayesian information criterion (BIC). „ p ( β j ) ‟ is the penalty function applied to each regression coefficient. Several different penalty functions have been introduced in the literature, including ridge, LASSO and SCAD. An overview of these penalty functions can be found in (Fan and Li, 2001). In this paper, the popular LASSO penalization p(  j  )  j  ,  (5)  is chosen because of its ability to automatically set small coefficients to zero, which effectively yields sparse solutions. This special property, also known as the sparsity property, is particularly useful in variable selection problems (Fan and Li, 2001). To solve the optimization problem in (4), we use a Local Quadratic Approximation (LQA) algorithm, proposed by Fan and Li (Fan and Li, 2001). LQA first casts the problem of penalized least square minimization presented in Eqn. 4 into a penalized likelihood maximization problem. It further addresses the issue of singularity at the origin that exists in penalty functions such as LASSO and SCAD by locally approximating „ p ( β j ) ‟ with a quadratic function. The resulting penalized likelihood function becomes both differentiable and concave, and it can be easily solved using  27  gradient-based optimization methods such as the Newton-Raphson algorithm. A detailed description of the LQA algorithm can be found in (Fan and Li, 2001).  B. Partial Directed Coherence PDC can be considered as the frequency-domain representation of Granger causality. It involves the transformation of the mAR coefficients in (2) into the frequency domain via the Fourier transform p  A( f )  I     A re   i 2  fr  (6)  r 1  where I is a d x d identity matrix. The estimate of PDC from the node node  yj  yi  to the  is defined as    y j  yi  A j ,i ( f )  (f )     d m 1  .  A m ,i ( f )  (7)  2  PDC takes on a value between 0 and 1. It essentially measures the relative interaction strengths with respect to a given source signal. The PDC the pairwise relatedness between  yi  and  yj  yj  yi  (f )  describes  as a function of frequency after discounting  the effect of other simultaneously observed series.  28  250 ms 2500 ms 2000 ms (max) 1000 ms 1500 ms  Figure 2-1.Visually guided choice reaction paradigm (Tropini et al. 2009, © [2009] IEEE).  C. Subjects Nine volunteers with mild to moderately severe, clinically diagnosed PD participated in our study. All patients stopped their levodopa medication overnight for a minimum of 12hrs before the study. We also recruited nine healthy aged-matched volunteers without active neurological disorders. In each group, eight subjects were righthanded and one was left-handed.  D. Experimental Design and Data Collection Subjects performed a visually guided joystick task based on the paradigm described in Amirnovin et al. (2004) (Amirnovin et al. , 2004). Briefly, subjects used a joystick with their dominant hand to select one of four randomly activated (yellow) targets as quickly as possible (Figure 2-1). Targets turned green once they were correctly selected. A total of 68 trials were presented, with an equal number of targets in each direction. Target presentation was pseudo-randomized so that the same sequence of trials was presented to each subject.  29  Subjects were fitted with an EEG cap with 20 active channels using the international 10-20 placement system, referenced to the mastoids. Artifacts due to eye movements were recorded by surface electrodes placed above and below the eyes. Data were recorded at 1000 Hz and aligned to task-related events via TTL pulse timestamps sent from the stimulus computer (via Matlab commands) to the EEG system through the parallel port.  E. Data Analysis Data were down-sampled to 250Hz, and eye and EMG-derived artifacts were removed via the Automatic Artifact Removal (AAR) toolbox v1.3 (Release 09.12, Gomez-Herrero 2007) of the EEGLab open source Matlab Toolbox (Delorme and Makeig, 2004). The denoised data were then bandpassed at 1-100 Hz. Next, data were normalized to unit variance and subsequently averaged over four electrode regions (Fronto-central (F-Central): Fp1, Fp2, F3, F4, F7, F8, Fz, Left sensorimotor (L Sm): C3, P3, T7, P7, Right sensorimotor (R Sm): C4, P4, T8, P8, and Central: Cz, Pz).  F.1. Spectral analysis For each subject group (Normals, PD) and each subject, trials were analyzed separately by computing spectrograms over the 1-50 Hz frequency range for each individual trial. For analysis purposes, a trial was defined as the time from the initial presentation of the fixation cross to the end of the immediately following inter-trial interval (i.e. from  30  cross to cross as depicted in Figure 2-1). The window size used was 32 samples and the window was 16. Individual trial spectrograms for each electrode region were then averaged across all subjects within one group. In order to ensure that trials were all the same length, they were truncated to the shortest trial length (5.816 s).  F.2. Sparse mAR-based PDC analysis Information flow between pairs of electrode regions was determined by computing the PDC spectrum, based on a 5th order sparse mAR model as described in Sections 2.4 A-B, for each individual trial. In order to ensure that trials were all the same length, they were again truncated to the shortest trial length as described above. The window size for the PDC spectrum analysis was 120 samples (= 0.48s) and the window shift was 16 samples (= 0.064s) and the PDC was computed over the 1-50 Hz frequency range. Individual trial PDC spectrograms (PDCograms) were then averaged across all subjects within each group, for each direction of flow between pairs of electrode regions (e.g. L SmF-Central, and F-CentralL Sm). We then sought to determine the overall information flow into and out of each electrode region. For each region of interest, the sum of the average information flow towards each of the remaining regions was subtracted from the sum of the average information flow into the region of interest over the 1-50 Hz frequency range and over the length of a trial. The final values were then displayed as Delta PDCograms, where positive values in the color map indicate net flow of information into the region of  31  interest, while negative values indicate information outflow from the same region (Figure 2-2). 2.5  Results  A. Spectral Analysis The spectral analysis showed a decrease in beta-band power over the 15-30 Hz range in correspondence to the onset of the “Go” cue in the F-Central, L Sm and R Sm regions (Figure 2-3). This beta band power decrease was especially evident in the normal subjects, and rebounded to rest levels in correspondence to the end of movement. In contrast, blunted modulation was present at the same latency in PD subjects. Spectral analysis of the Central region in control subjects showed increased power in the 1-10 Hz range in correspondence to movement. The overall power in the Central region was also decreased as compared to the power in the other regions. In PD, there was an overall increase in 1-10 Hz power across the entire trial, although a modest power increase, mimicking that seen in control subjects, was also observed in correspondence to movement.  Figure 2-2. Flow Analysis Schematics. Representation of the information flow into and out of the regions of interest (F-Central, L Sm, R Sm, Central). As depicted in the Delta PDCograms (Fig. 4), colours in the red range (positive values) represent net inflow, while colours in the blue range (negative values) represent net outflow (Tropini et al. 2009, © [2009] IEEE).  32  B. PDC-based Information Flow Analysis Information flow analysis revealed striking differences between control subjects and PD subjects, particularly with respect to net inflow and outflow to the L Sm and Central regions (Figure 2-4). In control subjects, a net outflow in the 10-20 Hz range was observed in the L Sm region, while a net inflow was present in higher (>35 Hz) and lower (<10Hz) frequency ranges. In contrast, PD subjects showed an increased outflow in the 10-20 Hz range, and a corresponding decreased inflow in the same higher and lower frequency ranges. Interestingly, the patterns of information flow in the Central region reversemirrored those observed in the L Sm region. In particular, the net inflow pattern in the 10-20 Hz range observed in the control group was greatly increased in the PD group, except during the peri-movement interval. PD subjects also showed an increased inflow to the R Sm region in the <10 Hz frequency range as compared to normal subjects. Moreover, PD subjects showed an increased net inflow to the F-Central region over a wide frequency range (> 30 Hz) as compared to control subjects. The pair-wise flow analysis shown in the individual PDCograms (Figure 2-3) also revealed peculiar patterns of information flow, particularly with respect to the L and R Sm areas and the F-Central region. In PD subjects, the directional flow between the L and R Sm areas was in fact greatly decreased as compared to control subjects at frequencies <20 Hz. In addition, PDC analysis revealed asymmetry in pairwise connectivity between the L Sm and F-Central region in normal subjects. In fact, the decrease in beta power observed during movement rebounded much later in the L Sm F-Central direction of flow as compared to the F-Central L Sm connection.  33  NORMALS  PD  Figure 2-3. Average spectrograms and PDCograms for each electrode region. Diagonal terms are spectrograms; non-diagonal items are PDCograms. Red vertical bars are mean response times. “Cue” indicates presentation of the targets, while “Go” indicates target activation. Nondiagonal colorbars indicate absolute connectivity (arbitrary units), while diagonal colorbars indicate absolute power in dB. The blue arrows point to reduced connectivity between L and R Sm below 20 Hz. (Tropini et al. 2009, © [2009] IEEE).  34  Figure 2-4.Average information flow into the electrode regions of interest for all subject groups. Positive values indicate net inflow, while negative values indicate net outflow. Red vertical bars indicate the mean response times. “Cue” indicates presentation of the target circles, while “Go” indicates the target activation (Tropini et al. 2009, © [2009] IEEE).  This pattern of connectivity may underlie the greater drive from motor areas recruited by the task to the frontal and prefrontal areas not only during the movement phase but also in the recovery phase immediately following completion of the task.  2.6  Discussion  Our results indicate that healthy and PD subjects show considerable differences in the modulation of beta band activity in correspondence to movement, as shown by spectral analysis. While control subjects down-regulate beta power during movement, PD subjects are unable to similarly modulate activity in this frequency range. This result is consistent with LFP studies that have suggested that excessive synchronization in the beta band in PD subjects might have an anti-kinetic effect, thus being responsible for PD symptoms such as bradykinesia (slowness of movement) (Brown and Williams, 2005, Brown, 2006, Gatev et al. , 2006). From a behavioral point of view, the PD-affected 35  individuals showed increased response times (albeit not significantly) as compared to the control group (PD: 0.91 ± 0.12 s, Normals: 0.76 ± 0.16 s, p=0.1134, Student's t-test). The observed increase in net outflow from the L Sm region in the diseased state might indicate an excessive spread of activation from the primary motor areas involved in the contralateral movement, to neighboring regions. For example, current fMRI work in our group has demonstrated a wider spread of activation in PD subjects (as measured by amplitude and spatial variance in the BOLD response) in bilateral cerebellar hemispheres, primary motor cortex (M1) and supplementary motor area (SMA) during a visuo-motor tracking task using a pressure-responsive bulb (Ng et al. , 2010). Previous cognitive studies have also shown a wider spread of activation in the prefrontal cortex of PD subjects off medication (Monchi et al. , 2004). Thus, while it might be difficult - in part due to the nature of EEG data itself - to assess with certainty the absolute significance of neuronal activity modulation during a motor task, our results appear to be consistent with a body of knowledge that involved the use of different experimental techniques (fMRI, LFP recordings). Overall, our results suggest that the use of a sparse mAR, PDC-based technique might be better-suited than spectral analysis to detect task-dependent activity changes in different brain regions. The PDC-based information flow analysis was in fact able to reveal region- and frequency-dependent modulation of activity otherwise undetectable through simple spectral analysis, given that spectrogram results were not considerably different across most regions (in particular F-Central, L Sm and R Sm) for a given subject group. Thus, spectral analysis might be primarily sensitive to large-scale power changes, and therefore might not reveal specific patterns of activity modulation that differ in the  36  normal and diseased states as well as across regions. Moreover, PDC-based analysis allows us to infer directional (and thus causal) information in the detected patterns of connectivity, thus providing important biological information. The use of a sparse mAR model is invaluable in that the sparsity constraint reduces the risk of over-fitting the data, thus leading to higher sensitivity in the detection of connectivity changes that can occur during a motor task. This is especially true given that activity modulation across the motor task interval is quite rapid (even as detected by spectral analysis), and thus short window sizes are needed to allow sufficient temporal resolution, which can lead to a significant increase in the number of parameters to be estimated. In order to reduce computational demands, we employed an averaging method to reduce the number of nodes used in the mAR model and obtain a limited number of regions of interest but we are currently exploring other ways to better define these brain areas, such as for example the use of ICA-based techniques. To further reduce the number of parameters to be estimated, only four areas of interest were included in our study, since they were most relevant to the analysis of the motor task presented in our study. However, future averaging or clustering methods might allow us to better identify other potential regions of interest.  Overall, the use of a sparse mAR-based PDC technique was able to reveal patterns of activity modulation that are quite different between healthy controls and PD subjects, and that would not otherwise be apparent when utilizing traditional spectral analysis methods. Although future work is needed to further optimize the choice of electrode  37  regions or nodes used in the analysis, sparse mAR-based PDC appears to be well-suited to detect temporally-sensitive connectivity changes in both the healthy and the diseased state. Thus, this technique might be applicable to the investigation of brain activity patterns in other neurological conditions, such as epilepsy or schizophrenia, which are characterized by rapid and/or task-dependent connectivity changes.  2.7  Acknowledgment  The authors would like to thank Dr. Edna Ty and Vignan Yogendrakumar for assistance in patient recruitment and data collection.  38  2.8  References  Amirnovin R, Williams ZM, Cosgrove GR, Eskandar EN. Visually guided movements suppress subthalamic oscillations in Parkinson's disease patients. J Neurosci. 2004 Dec 15;24(50):11302-6. Baccalá LA, Sameshima K. Partial directed coherence: a new concept in neural structure determination. Biol Cybern. 2001 Jun 1;84(6):463-74. Brown P, Williams D. Basal ganglia local field potential activity: character and functional significance in the human. Clinical Neurophysiology. 2005 Nov 1;116(11):2510-9. Brown P. Bad oscillations in Parkinson's disease. J Neural Transm Suppl. 2006(70):27-30. Delorme A, Makeig S. EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. J Neurosci Methods. 2004 Mar 15;134(1):9-21. Ding M, Chen Y, Bressler SL. Granger causality: basic theory and application to neuroscience. Handbook of Time Series Analysis. Wiley-VCH Verlage. 2006; 437–460. Engel AK, Singer W. Temporal binding and the neural correlates of sensory awareness. Trends Cogn Sci. 2001 Jan 1;5(1):16-25. Fan J, Li R. Variable selection via non concave penalized likelihood and its oracle properties. Journal of the American Statistical Association. 2001; (456):1348–1360. Ferri R, Rundo F, Bruni O, Terzano MG, Stam CJ. The functional connectivity of different EEG bands moves towards small-world network organization during sleep. Clin Neurophysiol. 2008 Sep;119(9):202636. Gatev P, Darbin O, Wichmann T. Oscillations in the basal ganglia under normal conditions and in movement disorders. Mov Disord. 2006 Oct 1;21(10):1566-77. Klostermann F, Nikulin VV, Kühn AA, Marzinzik F, Wahl M, Pogosyan A, et al. Task-related differential dynamics of EEG alpha- and beta-band synchronization in cortico-basal motor structures. Eur J Neurosci. 2007 Mar 1;25(5):1604-15. Magill PJ, Sharott A, Bolam JP, Brown P. Delayed synchronization of activity in cortex and subthalamic nucleus following cortical stimulation in the rat. The Journal of Physiology. 2006 Aug 1;574(Pt 3):929-46. Micheloyannis S, Pachou E, Stam CJ, Vourkas M, Erimaki S, Tsirka V. Using graph theoretical analysis of multi channel EEG to evaluate the neural efficiency hypothesis. Neurosci Lett. 2006 Jul 24;402(3):273-7. Monchi O, Petrides M, Doyon J, Postuma RB, Worsley K, Dagher A. Neural bases of set-shifting deficits in Parkinson's disease. Journal of Neuroscience. 2004 Jan 21;24(3):702-10. Ng B, Palmer S, Abugharbieh R, McKeown MJ. Focusing Effects of L-Dopa in Parkinson's Disease. Human Brain Mapping. 2010 Jan;31(1):88-97. Pereda E, Quiroga RQ, Bhattacharya J. Nonlinear multivariate analysis of neurophysiological signals. Prog Neurobiol. 2005 Jan 1;77(1-2):1-37.  39  Tseng KY, Kasanetz F, Kargieman L, Riquelme LA, Murer MG. Cortical slow oscillatory activity is reflected in the membrane potential and spike trains of striatal neurons in rats with chronic nigrostriatal lesions. J Neurosci. 2001 Aug 15;21(16):6430-9. Valdes-Sosa PA, Sanchez-Bornot JM, Lage-Castellanos A, Vega-Hernandez M, Bosch-Bayard J, MelieGarcia L, et al. Estimating brain functional connectivity with sparse multivariate autoregression. Philos T Roy Soc B. 2005 May 29;360(1457):969-81. Varela F, Lachaux JP, Rodriguez E, Martinerie J. The brainweb: phase synchronization and large-scale integration. Nat Rev Neurosci. 2001 Apr 1;2(4):229-39.  40  CHAPTER 3: ALTERED CORTICAL CONNECTIVITY IN PARKINSON’S DISEASE PATIENTS PERFORMING A VISUALLY GUIDED TASK2  3.1  Preamble  Chapter 2 described the development of a novel sparse mAR-based PDC method that could assess frequency-dependent connectivity changes in the EEG, and its application to preliminary data from PD subjects OFF medication and control subjects performing a visually guided task. This initial testing revealed distinct differences in the EEG-based connectivity between PD and healthy subjects. PDC-based analysis was also more sensitive to region-specific changes than spectral analysis, and because of its ability to provide directional connectivity information we proposed that it might be a suitable technique to investigate altered cortical connectivity in PD. In Chapter 3, we optimized the PDC-based analysis of EEG data and applied it to a full EEG data set from PD patients both ON and OFF medication, as well as healthy controls. We optimized our analysis by calculating the connectivity with respect to baseline levels, and by performing a quantitative analysis of both the movement preparation and execution phases of the task. In addition, we correlated connectivity values from PD patients with motor Unified Parkinson's Disease Rating Score (UPDRS) clinical measures, in order to better relate changes in connectivity in PD to the underlying clinical symptoms.  2  A version of this chapter will be submitted for publication: Tropini G, Chiang J, Wang ZJ and McKeown MJ (2010). Altered Cortical Connectivity in Parkinson's Disease Patients Performing a Visually Guided Task.  41  3.2  Introduction  There is growing evidence that the Parkinsonian state results in abnormal oscillatory activity in the basal ganglia (BG) and related BG-cortical loops. Local field potential (LFP) recordings in the subthalamic nucleus (STN) and the internal segment of the globus pallidus (GPi) during deep brain stimulation surgery (DBS) for Parkinson‟s disease (PD) have revealed the presence of exaggerated synchrony in the 13-30 Hz frequency range (Brown et al. , 2001, Brown and Williams, 2005). These beta-band oscillations are suppressed by behaviourally relevant stimuli (Amirnovin et al. , 2004, Kühn et al. , 2004, Williams et al. , 2003) and levodopa medication (Brown et al. , 2001, Levy et al. , 2002, Williams et al. , 2002). In addition, the degree of levodopa-induced beta-band power suppression in STN LFPs correlates with clinical improvements in both rigidity and bradykinesia (Kühn et al. , 2009). Moreover, direct stimulation of the STN at 20 Hz has been shown to slow motor performance, thus providing further evidence that abnormal BG oscillatory activity in the beta-band may contribute to the slowing of movement typical of PD (Chen et al. , 2007). In contrast, high frequency deep brain stimulation (100-180 Hz) of the STN results in suppression of beta-band LFP activity that also correlates with improvements in motor performance (Kühn et al. , 2008). Although prior work on oscillatory activity in PD has emphasized findings from STN recordings, there is increasing recognition that the cortex might be involved in the generation and propagation of abnormal rhythms to a greater degree than has been previously recognized (Arbuthnott and Garcia-Munoz, 2009). In fact, abnormal oscillatory activity in PD is not limited to the STN and GPi and on the contrary is thought to be a network phenomenon, present throughout the BG-thalamocortical circuitry  42  (Brown et al. , 2001, Brown and Williams, 2005, Cassidy et al. , 2002, Gatev et al. , 2006, Williams et al. , 2002). In addition, high-frequency DBS induces antidromic stimulation of layer V cortical cells in anesthetized healthy and 6-OHDA-lesioned rodents (Li et al. , 2007) and the size of the antidromic evoked potential correlates with therapeutic improvements from akinesia in awake freely moving rats (Dejean et al. , 2009). Further insight into DBS mechanisms was obtained by using optical deconstruction of the parkinsonian neural circuitry in rats, in recent work that confirmed that the primary site of DBS action is not the STN, but the motor cortex (Gradinaru et al. , 2009). Furthermore, in humans, single-pulse transcranial magnetic stimulation (TMS) of the primary motor area and supplementary motor area in PD patients at rest and off levodopa medication, resulted in a decrease of beta-band activity in the STN (Doyle Gaynor et al. , 2008). These results warrant further exploration into the role of cortical oscillatory activity in PD. Several previous studies have mainly focused on the investigation of cortical-STN coherence via concurrent scalp EEG (with a limited number of electrodes due to experimental limitations) and depth STN recordings in PD patients (Cassidy et al. , 2002, Fogelson et al. , 2006, Klostermann et al. , 2007, Lalo et al. , 2008, Marsden et al. , 2001, Williams et al. , 2002). However, only a few EEG studies have investigated functional cortico-cortical connectivity in the context of abnormal oscillatory activity in PD, again utilizing coherence as a measure of coupling between cortical regions (Cassidy and Brown, 2001, Silberstein et al. , 2005). Other studies used magnetoencephalography (MEG) to measure functional connectivity in PD using a synchronization likelihood (SL) technique (a general measure of linear and non-linear temporal correlations between time  43  series) but only investigated the resting state (Stoffers et al. , 2008a, Stoffers et al. , 2008b). Our study aimed to expand on these previous studies by using a method of analysis that would allow us not only to establish interrelations between EEG nodes or cortical areas, but also to infer causal dependencies between them. Partial directed coherence (PDC) methods appear particularly well-suited to investigation of cortical oscillatory activity from the EEG (Astolfi et al. , 2007, Sun et al. , 2009, Witte et al. , 2009) as they allow the measurement of the strength of directional connectivity between two nodes or regions of interest, (and hence infer a measure of causality) while at the same time discounting the effect of all other neighboring nodes (Baccalá and Sameshima, 2001). Since PDC-based methods take into account the influence of other nodes when assessing the connectivity between any given pair of signals, they are also more robust to volume conduction effects that affect standard coherence techniques (Nunez et al. , 1997, Srinivasan et al. , 2007, Winter et al. , 2007). We suggest that the use of a sparse mAR model to calculate the PDC (Tropini et al. , 2009) is more appropriate for EEG analysis as a full-connectivity assumption (i.e. the assumption that each EEG electrode contains activity that is functionally connected to all other nodes) is at odds with studies demonstrating that the EEG has “small world” network properties. That is, most nodes are not directly connected to one another, but cluster of nodes may communicate via long-range connections (Ferri et al. , 2008, Micheloyannis et al. , 2006). In order to assess cortical connectivity in the context of modulation of abnormal brain rhythms in PD, we investigated movement preparation and execution during a choice reaction task while recording EEG data. This type of paradigm has been  44  commonly used in studies investigating LFP activity in the BG, and has been shown to modulate beta-band activity in STN recordings (Amirnovin et al. , 2004, Cassidy et al. , 2002, Klostermann et al. , 2007, Kühn et al. , 2004, Williams et al. , 2003). In contrast to prior EEG studies on abnormal oscillatory activity in PD (Cassidy and Brown 2000, Silberstein et al 2005) we also assessed modulation of brain rhythms in normal controls, and thus provided a direct comparison of cortical connectivity in PD and in the healthy state. 3.3  Methods Subjects  Ten volunteers with clinically diagnosed PD participated in our study (two women, eight men, mean age 60.50 ± 9.71 years, one left-handed). All patients were recruited from the Movement Disorders clinic at the University of British Columbia, and had mild to moderately severe PD (Hoehn and Yahr Stage 1-3); their clinical details are summarized in Table 3-1. Exclusion criteria included atypical Parkinsonism, presence of other neurological conditions, and visual conditions requiring correction beyond the use of eye glasses or contact lenses. All patients were on levodopa medication. We also recruited ten healthy aged-matched volunteers (five women, five men, mean age 59.60 ± 8.96 years, one left-handed) without active neurological disorders. All patients stopped their anti-parkinson medication overnight for a minimum of 12hrs before the study. The mean Unified Parkinson‟s Disease Rating Scale (UPDRS) motor score during the “off medication” (PD-OFF) state was 26.00 ± 8.15. After completing the task in the PD-OFF state, patients were given their usual morning dose of levodopa (180 ± 58.69 mg). After an interval of approximately 1hr, introduced to ensure 45  that the medication reached peak dose, subjects repeated the task while on medication (PD-ON). Control subjects performed the task only once. However, we deemed practice effects to have reached saturation in both control and PD subjects based on a sufficient number of practice trials, as determined in previous work from our group (Palmer et al. , 2009a, Palmer et al. , 2009b, Palmer et al. , 2010). The study was approved by the University of British Columbia Clinical Research Ethics Board (CREB), and all subjects gave written and oral consent prior to participation.  Case  Disease Duration (Years)  Age  Gend er  Handedness  Motor UPDRS Part III (OFF)  Levodopa daily dose (mg)  P004 P005 P006 P007 P008 P010 P011 P012 P014 P016  4 4 5 8 12 9 16 7 4 16  64 44 75 66 62 64 58 55 47 70  M M M M M M F M F M  R R R R R R R R L R  30 16 35 25 17 19 22 31 24 41  600 300 400 450 800 800 900 500 400 950  AVG STDEV  8.50 4.72  60.5 9.71  26.00 610.00 8.15 234.28 Table 3-1. Summary of patient details. Additional details for recruited patients that were not ultimately included in the study can be found in Table A-2 in the Appendix.  46  Behavioral Paradigm  Figure 3-1: Schematics of joystick task.  Subjects performed a visually guided choice reaction task using a joystick (“Attack 3”, Logitech) while sitting 1.5 meters away from a large screen. The task was modified from a paradigm described in a recent study (Amirnovin et al. , 2004). Briefly, each trial began with the presentation of a small fixation cross at the centre of the screen. After a 250 ms delay four circular targets appeared (Cue) around the central cross (Figure 3-1). After a delay of 2500 ms, a randomly selected target turned yellow (Go cue). Simultaneously, a white cursor appeared on top of the cross, and subjects were instructed to use the joystick to guide the cursor over the target while moving as fast as possible, and to respond within 2000 ms of the go cue. Once the target was reached, it turned green for 1000 ms, and during this time subjects were instructed to return the joystick to its resting position. The stimuli were then erased from the screen, and the fixation cross was again presented for 1500 ms (the inter-trial interval). A total of 68 trials were presented, with an equal number of targets in each direction (North, South, East and West). Target presentation was pseudo-randomized so that the same sequence of trials was presented to each subject. Before the experiment, subjects also underwent basic joystick training, as  47  well as specific training on the task (12 trials, as determined in pilot trials). If the subjects moved the joystick prematurely, or failed to reach the target on time, the trial was automatically marked for removal from subsequent analysis. Response times were recorded. Rest data were also collected for 1 minute while subjects were fixating on the central cross, at the beginning of the experimental task. The visual stimuli presented during the task were designed in Matlab (The Mathworks Inc., Massachusetts), using the Psychophysics Toolbox Version 3.  Data Collection and Pre-processing Subjects were fitted with an EEG cap (Synamps2 Quik-Cap, Neuroscan, Compumedics, Texas USA) with 19 active channels using the international 10-20 placement system, referenced to the mastoids. Artifacts due to eye movements were recorded by surface electrodes placed above and below the eyes (Xltek, Ontario, Canada). EEG data were recorded at 1000 Hz and stored via the Synamps2 amplifier system and Scan4.3 software (Neuroscan, Compumedics, Texas USA). Alignment of EEG data to task-related events was performed by sending Matlab-coded TTL pulses via parallel port from the stimulus computer to the Neuroscan Synamps2 system. TTL pulses were then recorded as event-specific timestamps by the Scan 4.3 software.  48  Figure 3-2. Headplot of electrode regions.  Data were down-sampled to 250 Hz, and eye and EMG-derived artifacts were removed via the Automatic Artifact Removal (AAR) toolbox v1.3 (Release 09.12, Gomez-Herrero 2007) of the EEGLab open source Matlab Toolbox (Delorme and Makeig, 2004). The denoised data were then bandpassed at 1-100 Hz. Next, data were normalized to unit variance, de-trended („detrend‟ function in Matlab) and averaged over five electrode regions. (Figure 3-2. Fronto-Central (FCentral): Fp1, Fp2, F3, F4, F7, F8, Fz; Left Sensorimotor (L Sm): C3, P3, T7, P7; Right Sensorimotor (R Sm): C4, P4, T8, P8; Central: Cz, Pz and Occipital: O1, O2). Fp1/Fp2, as well as F3/F4 and F7/F8 have all been shown to correspond to frontal areas (superior frontal gyrus, middle frontal gyrus and inferior frontal gyrus respectively, (Homan et al. , 1987), while C3/C4 correspond to Brodmann area 4 (Homan et al. , 1987) and P3/P4 to Brodmann area 7 (involved in visuo-motor coordination, Homan et al. , 1987) and were thus grouped together with parietal (P7/P8) and temporal (T7/T8) electrodes as 'sensorimotor' areas for simplicity and ease of computation. O1 and O2 49  correspond to Brodmann area 17, the primary visual cortex (Homan et al. , 1987). Last, Cz and Pz correspond to central and parietal midline areas (Steinmetz et al. , 1989).  Data Analysis Overview Data analysis involved two main components: analysis of the temporal dynamics of connectivity modulation during each experimental trial interval, and a frequencydependent, quantitative analysis of connectivity measures during both the preparatory and movement phases of the motor task. In addition to using smAR-PDC based analysis, we also utilized spectral analysis in order to both assess power changes in the regions of interest, and to compare our techniques with more traditional methods of analysis. All analyses were conducted using the Matlab Signal Processing Toolbox, or via custom Matlab scripts (Matlab, The Mathworks, Inc., Massachusetts).  Temporal dynamics SmAR-PDC based Connectivity Analysis Directional connectivity between pairs of electrode regions (5x5 connections, minus 5 non-meaningful auto-connections) was determined by computing the partial directed coherence (PDC) connectivity spectrum, based on a 5th order sparse mAR model for each individual trial (Tropini et al. , 2009). Briefly, smAR-PDC involves first fitting of the desired nth order mAR model to the data. The sparsity constraint on the mAR model is obtained via penalized regression, which limits the number of parameters to be estimated and thus reduces computational demands. Once the mAR coefficients are  50  determined, they are transformed into the frequency domain via Fourier transform, and the estimate of the PDC between two nodes (or regions of nodes) is calculated (Cf. Tropini et al. , 2009 for a detailed discussion). The PDC takes a value between 0 and 1 and describes the pair-wise strength of directional interaction between a node i and a node j as a function of frequency, while at the same time discounting the effect of other simultaneously active nodes. For analysis purposes, a trial was defined as the time from the initial presentation of the fixation cross to the end of the immediately following inter-trial interval (i.e. from cross to cross as depicted in Figure 3-1). In order to ensure that trials were all the same length, they were truncated to the shortest trial length (5816 ms). Baseline data were obtained from the rest recordings performed at the beginning of the experimental task, while subjects where fixating on the cross presented at the centre of the screen. For each trial and each electrode region, the baseline PDC was subtracted from task PDC in order to produce baseline-normalized connectivity values. The window size for the PDC connectivity spectrum was 120 samples (48 ms), the window shift was 16 samples (64 ms) and the PDC was computed over the 1-50 Hz frequency range. Individual trial PDC spectrograms (PDCograms) were then averaged across all subjects within each group, for each direction of connectivity between pairs of electrode regions. Each PDCogram data point thus provided information on the strength of directional connectivity with respect to rest, at a specific frequency and time, for a pair of nodes or electrode regions.  51  Spectral analysis For each subject group (Controls, PD-OFF, PD-ON) and each subject, trials were analyzed separately by computing short-time Fourier transform spectrograms („spectrogram‟ function in Matlab) over the 1-50 Hz frequency range for each individual trial and for the baseline period. As for the smAR-PDC connectivity analysis, trials were truncated to the shortest trial length to ensure consistency. The window size used was 32 samples and the window shift was 16 samples. For each trial and each electrode region, the mean baseline log power spectrum was subtracted from the task spectral estimate to obtain normalized values. Individual trial spectrograms for each electrode region were then averaged across all subjects within one group to obtain group values.  Quantitative analysis of the movement preparation (PRE) and movement (MOV) phases - Directional Connectivity and Power Spectra We then sought to quantitatively assess the results of the temporal dynamics analysis, by calculating the average power spectrum and average PDC connectivity spectrum with respect to rest during both the preparatory and movement phases of the motor task. The preparatory phase of movement („PRE‟) was defined as the portion of the interval preceding the “Go” cue (0 to 2750 ms), and the movement (MOV‟) phase was defined as the part of interval immediately following the “Go” cue and lasting until the average response time for all groups (2750 ms to 3590 ms). Response time was defined as the time the subject had reached the target by using the joystick to guide the white cursor. The average power spectra and the average PDC connectivity spectra over the 150 Hz frequency range (with 1 Hz resolution) were extracted from the temporal dynamics  52  data for PRE and MOV for each individual subject, and mean group values were then computed. Error bars were obtained via leave-one-out cross-validation. Individual subject PDC connectivity spectra values for PD-OFF were then averaged over different frequency bands (: 4-7 Hz, : 8-12 Hz, : 13-30 Hz and -: 830 Hz) and correlated with UPDRS III OFF motor scores, using a robust regression („robustfit, Matlab). Analysis was performed for each connection pair separately. Significance was established with a false discovery rate (FDR) corrected alpha level of 0.025 (alpha level of 0.05 corrected for 20 connections, 4 frequency bands, 2 task intervals ). 3.4  Results  Behavioural Measurements PD-OFF subjects showed increased response times as compared to both PD-ON and age matched controls, although differences in response times were not significant (Controls: 0.76  0.15s, PD-OFF: 0.920.11s, PD-ON: 0.860.13s, p>0.05. Individual response times can be found in the Appendix, Section III.2). Total error incidence –and hence removal of trials from analysis- was less than 6% in all groups, although more pronounced in PD-OFF subjects (Controls: 1.60%, PD-OFF: 5.59%, PD ON: 5.29%). However, since error incidence was low, and error trials were excluded from the analysis, behavioural results were not likely to be confounded by group-dependent error patterns.  53  Temporal Dynamics- Spectrograms In Normals, spectrograms showed a brief decrease in power relative to baseline levels in the 10-25 Hz range following the initial non-predictive cue (Figure 3-3). Power suppression also occurred approximately 1 s before the imperative "Go" cue, particularly in the FCentral, L Sm and Occipital regions. Following the onset of the imperative cue, a large power suppression in the 13-30Hz range was then observed across all regions. This power suppression lasted until approximately 0.5s following the average response time for each group (red vertical dashed line in Figure 3-3), presumably due to the fact that subjects were instructed to return the joystick to the center position following target selection. A power rebound in the 20-30 Hz range was then observed in all regions approximately 1s after response time, although its magnitude was less pronounced in the Occipital region. A power increase at lower frequencies (1-10Hz, and up to 15Hz for R Sm, Central and Occipital regions) was instead observed in all regions following the non predictive cue. Although the power in these frequency ranges renormalized after about 1s from the beginning of the task, a subsequent  re- increase was observed during  movement in all regions. In contrast, PD patients OFF medication showed a reduced power suppression in the 10-25 Hz range following the non-predictive cue and immediately before the imperative cue. Similarly, power suppression in the 13-30 Hz range was blunted during movement. Lower-band (1-10 Hz) power modulation was also blunted in PD OFF patients both following the non-predictive cue and during movement, particularly in the Central region.  54  Levodopa medication (PD ON) partially restored beta-band suppression before and during movement, albeit not to the levels observed in controls. Power suppression in the 10-25 Hz range following the non-predictive cue was also partially restored. In contrast, PD ON subjects demonstrated blunted power modulation in the 1-10 Hz both following the non-predictive cue and in correspondence to movement.  Temporal Dynamics- PDCograms Overall, smAR-PDC based analysis revealed several asymmetries in the bidirectional patterns of connectivity between the 5 regions of interest, as well as a differential involvement of each region in the temporal dynamics. Connections that appeared to be particularly disrupted in PD OFF subjects were Central FCentral, L Sm  Occipital (Figure 3-4), as well as R Sm  Occipital and Central  Occipital (Figure 3-5; additional figures for the remaining connections can be found in Appendix III.3. Overall, PDCograms for PD OFF subjects showed dramatic changes in connectivity as compared to controls, while PDCograms for PD ON subjects revealed that levodopa medication failed to completely restore normal patterns of modulation, and that improvement was connection-dependent. In particular, greater medication-dependent improvements were observed for L Sm  Occipital (Figure 3-4, bottom panel), and partially for Central FCentral (Figure 3-4, top panel).  55  Quantitative Analysis - Power Spectra Quantitative analysis of the PRE (0-2.75 s) and MOV (2.75-3.59 s) intervals revealed task-dependent differences, while only modest region-dependent differences were observed for all groups (Figure 3-6). Overall, smaller power differences with respect to rest were observed in the PRE phase as opposed to MOV. The biggest relative difference between PD OFF and Normals were observed in L Sm (PRE: 10-20 Hz, MOV: 10-30 Hz), R Sm (MOV: 12-30 Hz) and Occipital (MOV: 12-35 Hz). In the PRE phase, power was increased in normal subjects below 10 Hz, but not in PD OFF or PD ON for all regions (except L Sm where PD OFF > Normals between 5-10 Hz). In the MOV phase, power was also increased in Normals < 8 Hz, while a marked decrease in power was observed between 10-35 Hz. As observed in the temporal dynamics, although PD OFF also showed a down-regulation of power at 10-35 Hz, modulation was attenuated as compared to controls. In addition, PD ON subjects in general did not show a return to normal levels of modulation, and levodopa appeared to exacerbate the decrease in power < 8 Hz that was observed in PD OFF.  Quantitative Analysis - PDC Spectra and Correlation with UPDRS The PDC-based quantitative analysis of the PRE and MOV intervals aided further identification of several abnormalities in the connectivity modulation by PD subjects. In the PRE phase (Figure 3-7), PD OFF subjects showed disrupted connectivity with respect to rest in several connection pairs. As already observed in the PDCograms-based analysis, information flow was most disrupted in the following connections: Central FCentral, OccipitalL Sm, OccipitalR Sm, CentralOccipital. Of these, only the 56  CentralFCentral connection showed an almost complete return to normal modulation patterns following levodopa administration (PD ON). Levodopa however further deteriorated connectivity in L SmFCentral, L SmCentral and Occipital  Central. Correlation of OFF UPDRS Motor Scores with PRE PDC values for each of the PD OFF patients (excluding P016, an outlier with UPRS III Score=41) revealed that PDC values in the Central  FCentral direction of information flow were positively correlated with motor clinical scores in the β (13-30 Hz) and α-β (8-30 Hz) frequency ranges (α: r=0.66, p=0.038; β: r=0.807, p=0.009; α-β: r=0.78, p=0.011 with a FDR-corrected alpha level of 0.025. For scatterplots of these correlations and a table of all remaining correlations, please cf. Appendix III.4.1-2). In the MOV phase (Figure 3-8), information flow for PD OFF was again most disrupted in the Central  FCentral, Occipital  L Sm, Occipital  R Sm and Central  Occipital. In addition, PD OFF subjects also showed deteriorated modulation of connectivity for the L Sm FCentral and L Sm  R Sm connections, as compared to the PRE phase. Levodopa medication  restored  connectivity in the Occipital  L Sm direction to levels comparable to those of controls, as well as in the Occipital  R Sm (above 10 Hz only) and Central  FCentral directions (above 20 Hz only). As for the PRE phase, levodopa further deteriorated connectivity L SmFCentral, L Sm Central and Occipital  Central. Correlation of UPDRS OFF Motor Scores with MOV PDC values again revealed a positive correlation between connectivity values for the Central  FCentral connection. Correlations were significant in the α (8-12 Hz), β (13-30 Hz) and α-β (8-30 Hz) frequency ranges (α: r=0.749, p=0.022; β: r=0.951, p=0.007; α-β: r=0.928, p=0.008,  57  FDR-corrected alpha=0.025. For scatterplots of these correlations and a table of all remaining correlations, please cf. Appendix III.4.3-4). In addition, since the gender of our PD subjects was asymmetrically distributed (7 males, 2 females out of the 9 subjects included in the correlations), we performed a posthoc correlation of PD OFF PDC-values for each connection and each frequency band with subject gender. No significant correlations were found in either PRE or MOV, with an FDR-corrected alpha of 0.025 (A table of correlation values can be found in Appendix III.5).  58  Figure 3-3. Spectrograms for all groups. The five electrode regions (FCentral, L Sm, R Sm, Central and Occipital) are shown on the left. Indicated are power values with respect to rest, in dB. The vertical red bars indicate average response times for each group. “Cue” and “Go” refer to the non-predictive cue and the imperative cue shown in Figure 3-1.  59  Centr  FCent  ___________________________________________________________________  L Sm  Occip.  Figure 3-4. PDCograms for all groups, a). Shown are the FCentral  Central (top) and L Sm  Occipital (bottom) connections . The top row of each figure shows left to right direction of connectivity and the bottom row shows the reverse direction. Colorbars indicate connectivity values with respect to rest (arbitrary units, (a.u.)).  60  Occip.  R Sm  ____________________________________________________________________  Occip.  Centr  Figure 3-5. PDCograms for all groups, b). Shown are the R Sm  Occipital connections (top) and Central  Occipital connections (bottom). The top row of each figure shows left to right direction of connectivity and the bottom row shows the reverse direction. Colorbars indicate connectivity values with respect to rest (a.u.).  61  FCentral  L Sm  R Sm  Central  Occipital  Figure 3-6. Power Spectra for all groups. Shown on the left column are values for the PRE phase of the task, and on the right column values for the MOV phase. All power values(dB) are values with respect to rest. Red lines indicate the PD OFF group, blue lines the Normal group, and black lines the PD ON group.  62  Figure 3-7. PDC Spectra for the PRE phase of the task. In the analysis, each column element is the source of connectivity flow, while each row element is the sink (i.e. where the information flows into) of the connectivity flow. The regions of interest are indicated in the diagonal elements. Normals are indicated in blue, PD OFF in red and PD ON in black. PDC values are connectivity values with respect to rest (a.u.).  63  Figure 3-8. PDC Spectra for the MOV phase of the task. In the analysis, each column element is the source of connectivity flow, while each row element is the sink of the connectivity flow. The regions of interest are indicated in the diagonal elements. Normals are indicated in blue, PD OFF in red and PD ON in black. PDC values are connectivity values with respect to rest (a.u.)  64  3.5  Discussion  We have shown that cortical activity in PD subjects performing a visually guided task is characterized not only by abnormal changes in power dynamics over time, but also by altered directional connectivity, as revealed by the smAR-PDC analysis. We found complex alterations in connectivity were only partially restored - and often exacerbated by levodopa, and might thus underlie remaining deficits in PD subjects ON medication. Most importantly, PDC-based, task-specific measures of connectivity were found to correlate with UPDRS motor scores in PD OFF subjects.  Temporal Dynamics A vast majority of seminal studies involving abnormal oscillatory activity in PD have involved patients only, largely because LFP recordings are only available in PD subjects undergoing DBS surgery. In contrast, our study aimed to assess task-dependent modulation of cortical activity across regions of interest, and to directly compare PD subjects and normal controls using a paradigm that has been previously shown to modulate LFP activity in PD. Simply looking at the power spectrum in the different regions of interest was capable of distinguishing between control and PD subjects. In control subjects, spectrograms showed reduced spectral power relative to baseline levels in the 10-25 Hz range following the initial non-predictive cue (Figure 3-3), as well as approximately 1s before the imperative "Go Cue" and during movement at 10-30 Hz, in agreement with findings investigating EEG event-related (de)synchronization in visually guided motor tasks and choice reaction tasks (Doyle et al. , 2005, Kranczioch et al. , 2008). In contrast,  65  PD OFF subjects demonstrated impaired spectral power modulation in the 10-25 Hz range both following the non-predictive cue and during movement preparation, and reduced power suppression in the 13-30 Hz band in correspondence to movement. Attenuated preparatory desynchronization in the α and β bands in PD subjects ON performing a choice reaction task ON medication has been previously reported (Praamstra and Pope, 2007), and studies reporting abnormal activity in the β band in PD subjects are numerous, both using LFP recordings (Brown et al. , 2001, Brown and Williams, 2005, Cassidy et al. , 2002, Levy et al. , 2002) and MEG (Stoffers et al. , 2008a, Stoffers et al. , 2008b). In our study, PD ON subjects had only partially restored normal modulation of power in the 10-30 Hz range, and blunted modulation in the lower frequencies (< 10 Hz). Spectral analysis revealed that inter-region differences in each group tend to be relatively modest (Figure 3-3), and only the connectivity analysis was able to reveal distinctive regional differences. A previous EEG study showed increased cortico-cortical coherence over the 10-35 Hz range in PD patients at rest (Silberstein et al. , 2005), but coherence analysis is restricted to forcing the connectivity between electrodes/regions to be symmetric. In contrast, PDC-based analysis revealed complex task and regiondependent changes in connectivity that were clearly asymmetric (Figure 3-4 and Figure 3-5). In PD subjects OFF medication, PDC-based analysis revealed extensive connectivity disruptions that were often not restored or even exacerbated by levodopa medication (Figure 3-4, top panel, and Figure 3-5, bottom panel). While it is difficult to associate specific EEG electrodes with particular brain areas, we note that a PDC-based method will tend to be robust to volume conduction,  66  thus allowing some broad interpretations to be made. We found significant connectivity abnormalities in PD OFF subjects in the OccipitalL Sm and OccipitalR Sm connections. The occipital electrodes O1 and O2 may relate to Brodmann area 17, the primary visual cortex (Homan et al. , 1987), while electrodes P3, P4 (in the L and R Sm regions) overlie Brodmann area 7, involved in visuomotor coordination (Homan et al. , 1987). Indeed, PD patients have been known to show impairments in the visual system (Rodnitzky, 1998, Silva et al. , 2005) as well as visuo-cognitive deficits (Antal et al. , 1998). Recent work revealed occipital and posterior parietal hypoperfusion in PD patients without dementia that correlated with impaired cortical visual processing (Abe et al. , 2003), while other work has shown a correlation of poor performance in visuospatial and visuoperceptual tests with gray matter atrophy in posterior temporal, parietal and occipital regions (Pereira et al. , 2009). Thus, the impairments of posterior connectivity we observed might in part be related to abnormal visual processing during task performance in PD. Further work will be required to determine if alterations in connectivity correspond to psychometric indices in PD.  Quantitative Analysis of PRE and MOV phases A comparison of the preparatory (PRE) and movement (MOV) phases of the task revealed that differences in power spectra with respect to rest were greatest between PD OFF subjects and healthy controls during the MOV phase (Figure 3-6, right panel). In particular, the largest MOV differences were found in the 10-30 Hz range in the L and R Sm as well as Occipital regions, in agreement with the visuo-motor nature of the task,  67  and thus pointing to potential visuo-motor deficits in PD that were not completely restored by medication. While visual inspection of the PDCograms suggested that the PRE and MOV phases had similar disruptions in connectivity, the quantitative analysis demonstrated distinct differences between the two phases. During movement preparation, the Occipital  L Sm and Central  FCentral connections were particularly disrupted in PD OFF as compared to controls, suggesting a disruption in visual processing during movement preparation in PD. During movement, differences in connectivity between PD OFF and controls emerged in the L Sm  FCentral connection in the 18-40 Hz range. Levodopa had differing effects on connectivity in the preparatory and movement phases. In the preparatory phase, levodopa medication fully restored connectivity Central  FCentral, while it disrupted or exacerbated connectivity in L Sm  FCentral, L Sm  Central and Occipital  Central (Figure 3-7). In the movement phase, it best restored connectivity Occipital  L Sm and R SmOccipital while it again disrupted modulation L Sm  FCentral, L Sm  Central and Occipital  Central (Figure 3-8). These results suggest that exacerbation of altered connectivity by levodopa might underlie remaining deficits in PD subjects ON medication. The Central FCentral PDC-based connection correlated positively with motor UPDRS scores for PD OFF subjects. Correlations were significant in the β, and α-β range for PRE, and in the α, β, and α-β ranges for MOV. As expected with a positive correlation between UPDRS and connectivity strength in PD subjects, this connection was decreased in normal controls in the preparatory phase (Figure 3-7). In contrast, during the movement phase, the relative connectivity differences between PD OFF and  68  normal controls varied depending on the frequency band examined (Figure 3-8). For frequencies > 12 Hz, this connection was again decreased in normal controls as opposed to PD OFF. However, healthy subjects showed increased connectivity as compared to PD subjects for frequencies <12 Hz. The paradoxical finding of a positive correlation between UPDRS score and connectivity strength in the α band in a connection that is higher in normals may reflect a compensatory mechanism in PD. The positive correlation of Central  FCentral connectivity in PD OFF with motor clinical scores might again underlie PD deficits in visuo-motor processing along a parietal to centro-frontal stream, given that 'Central' electrodes (Cz, Pz) are located over the central midline in parietal (Pz) and central areas (Cz is located in front of the central sulcus, Steinmetz et al. , 1989).  Limitations and Technical Considerations Our choice of electrode regions of interest was limited by some technical considerations. First, while we desired to maintain high spatial resolution, region selection was influenced by the empirical observation that PDC-based analysis becomes less robust as the number of nodes increases. In addition, a higher number of nodes increases computational demands from the mAR calculations, and thus, our choice of 5 regions of interest was a tradeoff between minimizing the number of smAR-PDC nodes while still maintaining spatial resolution. Last, we chose to average the electrodes in each region of interest, but further work is needed to better determine the optimal methods for identifying the contribution of each electrode to the regions of interest. We explored an Independent Component Analysis (ICA)-based method to determine optimal weighting of individual electrodes  69  (data not shown), but the results obtained were not significantly different from those obtained here by the simpler averaging method.  Overall, our results suggest that the use of smAR-PDC-based analysis might be ideally suited to detect connectivity modulation between cortical regions as a function of frequency. The use of a sparsity constraint on the mAR matrices reduces the risk of overfitting the data, and reduces computational intensity while being appropriate for the EEG, which has been shown to possess "small world" network properties (Ferri et al. , 2008, Micheloyannis et al. , 2006). In addition, smAR-PDC analysis is less affected by volume conduction effects that are known to plague traditional connectivity measures such as coherence (Nunez et al. , 1997, Srinivasan et al. , 2007, Winter et al. , 2007) since it takes into account the influence of other nodes when assessing the connectivity between two regions of interest. While EEG-based methods do not provide the level of spatial resolution offered by fMRI techniques, they allow for great temporal resolution and provide information on oscillatory activity in frequency bands of interest. Thus, the use of an EEG analysis method that also provides directional connectivity information -as offered by some fMRIbased techniques (Li et al. , 2008) - can be beneficial when studying neurological conditions such as PD, that are characterized by abnormal oscillatory activity in cortical and subcortical systems. In the case of PD, information on the modulation of cortical connectivity is particularly valuable given recent results pointing to a more important role of the cortex  70  in the modulation of BG oscillatory activity than previously thought (Arbuthnott and Garcia-Munoz, 2009, Dejean et al. , 2009, Gradinaru et al. , 2009, Li et al. , 2007).  3.6  Conclusions  Our results suggest that the EEG, coupled with the appropriate analysis, demonstrates complex abnormalities in connectivity in PD subjects performing paradigms previously shown to modulate LFP β band oscillations. This suggests that the role of the EEG to monitor PD may need to be expanded. In particular, we have shown that PD subjects demonstrate impairments not only in the intrinsic modulation of activity in each region of interest - as shown by spectral analysis- but also profound and complex alterations in modulation of connectivity that might underlie deficits in visuo-motor processing. In addition, our work shows that levodopa medication does not fully restore connectivity, and in some cases further exacerbates the deficits shown by unmedicated subjects. Moreover, changes in connectivity that might underlie visuo-motor processing correlate with motor UPDRS scores in unmedicated patients. Although further work is needed to further elucidate the role of each connection, given the complexity of the impairments that we have found, our work shows promise for a non-invasive method to assess changes in frequency-dependent connectivity in PD, and potentially in other conditions known to be affected by abnormal oscillatory activity, such as schizophrenia and epilepsy.  71  3.7  References  Abe Y, Kachi T, Kato T, Arahata Y, Yamada T, Washimi Y, et al. Occipital hypoperfusion in Parkinson's disease without dementia: correlation to impaired cortical visual processing. J Neurol Neurosurg Psychiatry. 2003 Apr;74(4):419-22. Amirnovin R, Williams ZM, Cosgrove GR, Eskandar EN. Visually guided movements suppress subthalamic oscillations in Parkinson's disease patients. J Neurosci. 2004 Dec 15;24(50):11302-6. Antal A, Bandini F, Keri S, Bodis-Wollner I. Visuo-cognitive dysfunctions in Parkinson's disease. Clin Neurosci. 1998;5(2):147-52. Arbuthnott G, Garcia-Munoz M. Dealing with the devil in the detail - some thoughts about the next model of the basal ganglia. Parkinsonism Relat Disord. 2009 Dec 1;15 Suppl 3:S139-42. Astolfi L, De Vico Fallani F, Cincotti F, Mattia D, Marciani MG, Bufalari S, et al. Imaging functional brain connectivity patterns from high-resolution EEG and fMRI via graph theory. Psychophysiology. 2007 Nov 1;44(6):880-93. Baccalá LA, Sameshima K. Partial directed coherence: a new concept in neural structure determination. Biol Cybern. 2001 Jun 1;84(6):463-74. Brown P, Oliviero A, Mazzone P, Insola A, Tonali P, Di Lazzaro V. Dopamine dependency of oscillations between subthalamic nucleus and pallidum in Parkinson's disease. J Neurosci. 2001 Feb 1;21(3):1033-8. Brown P, Williams D. Basal ganglia local field potential activity: character and functional significance in the human. Clinical Neurophysiology. 2005 Nov 1;116(11):2510-9. Cassidy M, Brown P. Task-related EEG-EEG coherence depends on dopaminergic activity in Parkinson's disease. Neuroreport. 2001 Mar 26;12(4):703-7. Cassidy M, Mazzone P, Oliviero A, Insola A, Tonali P, Di Lazzaro V, et al. Movement-related changes in synchronization in the human basal ganglia. Brain. 2002 Jun 1;125(Pt 6):1235-46. Chen CC, Litvak V, Gilbertson T, Kühn A, Lu CS, Lee ST, et al. Excessive synchronization of basal ganglia neurons at 20 Hz slows movement in Parkinson's disease. Experimental Neurology. 2007 May 1;205(1):214-21. Dejean C, Hyland B, Arbuthnott G. Cortical effects of subthalamic stimulation correlate with behavioral recovery from dopamine antagonist induced akinesia. Cereb Cortex. 2009 May 1;19(5):1055-63. Delorme A, Makeig S. EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. J Neurosci Methods. 2004 Mar 15;134(1):9-21. Doyle Gaynor LMF, Kühn AA, Dileone M, Litvak V, Eusebio A, Pogosyan A, et al. Suppression of beta oscillations in the subthalamic nucleus following cortical stimulation in humans. European Journal of Neuroscience. 2008 Oct 1;28(8):1686-95. Doyle LMF, Yarrow K, Brown P. Lateralization of event-related beta desynchronization in the EEG during pre-cued reaction time tasks. Clin Neurophysiol. 2005 Aug 1;116(8):1879-88. Ferri R, Rundo F, Bruni O, Terzano MG, Stam CJ. The functional connectivity of different EEG bands moves towards small-world network organization during sleep. Clin Neurophysiol. 2008 Sep;119(9):202636.  72  Fogelson N, Williams D, Tijssen M, van Bruggen G, Speelman H, Brown P. Different functional loops between cerebral cortex and the subthalmic area in Parkinson's disease. Cereb Cortex. 2006 Jan 1;16(1):6475. Gatev P, Darbin O, Wichmann T. Oscillations in the basal ganglia under normal conditions and in movement disorders. Mov Disord. 2006 Oct 1;21(10):1566-77. Gradinaru V, Mogri M, Thompson KR, Henderson JM, Deisseroth K. Optical deconstruction of parkinsonian neural circuitry. Science. 2009 Apr 17;324(5925):354-9. Homan RW, Herman J, Purdy P. Cerebral location of international 10-20 system electrode placement. Electroencephalogr Clin Neurophysiol. 1987 Apr;66(4):376-82. Klostermann F, Nikulin VV, Kühn AA, Marzinzik F, Wahl M, Pogosyan A, et al. Task-related differential dynamics of EEG alpha- and beta-band synchronization in cortico-basal motor structures. Eur J Neurosci. 2007 Mar 1;25(5):1604-15. Kranczioch C, Athanassiou S, Shen S, Gao G, Sterr A. Short-term learning of a visually guided power-grip task is associated with dynamic changes in EEG oscillatory activity. Clin Neurophysiol. 2008 Jun 1;119(6):1419-30. Kühn AA, Williams D, Kupsch A, Limousin P, Hariz M, Schneider G-H, et al. Event-related beta desynchronization in human subthalamic nucleus correlates with motor performance. Brain. 2004 Apr 1;127(Pt 4):735-46. Kühn AA, Kempf F, Brücke C, Gaynor Doyle L, Martinez-Torres I, Pogosyan A, et al. High-frequency stimulation of the subthalamic nucleus suppresses oscillatory beta activity in patients with Parkinson's disease in parallel with improvement in motor performance. J Neurosci. 2008 Jun 11;28(24):6165-73. Kühn AA, Tsui A, Aziz T, Ray N, Brücke C, Kupsch A, et al. Pathological synchronisation in the subthalamic nucleus of patients with Parkinson's disease relates to both bradykinesia and rigidity. Experimental Neurology. 2009 Jan 14;215(2):380-7. Lalo E, Thobois S, Sharott A, Polo G, Mertens P, Pogosyan A, et al. Patterns of bidirectional communication between cortex and basal ganglia during movement in patients with Parkinson disease. J Neurosci. 2008 Mar 19;28(12):3008-16. Levy R, Ashby P, Hutchison WD, Lang AE, Lozano AM, Dostrovsky JO. Dependence of subthalamic nucleus oscillations on movement and dopamine in Parkinson's disease. Brain. 2002 Jun 1;125(Pt 6):1196209. Li J, Wang ZJ, Palmer SJ, McKeown MJ. Dynamic Bayesian network modeling of fMRI: a comparison of group-analysis methods. Neuroimage. 2008 Jun;41(2):398-407. Li S, Arbuthnott GW, Jutras MJ, Goldberg JA, Jaeger D. Resonant antidromic cortical circuit activation as a consequence of high-frequency subthalamic deep-brain stimulation. J Neurophysiol. 2007 Dec 1;98(6):3525-37. Marsden JF, Limousin-Dowsey P, Ashby P, Pollak P, Brown P. Subthalamic nucleus, sensorimotor cortex and muscle interrelationships in Parkinson's disease. Brain. 2001 Feb 1;124(Pt 2):378-88. Micheloyannis S, Pachou E, Stam CJ, Breakspear M, Bitsios P, Vourkas M, et al. Small-world networks and disturbed functional connectivity in schizophrenia. Schizophr Res. 2006 Oct;87(1-3):60-6.  73  Nunez PL, Srinivasan R, Westdorp AF, Wijesinghe RS, Tucker DM, Silberstein RB, et al. EEG coherency. I: Statistics, reference electrode, volume conduction, Laplacians, cortical imaging, and interpretation at multiple scales. Electroencephalogr Clin Neurophysiol. 1997 Nov 1;103(5):499-515. Pereira JB, Junque C, Marti MJ, Ramirez-Ruiz B, Bargallo N, Tolosa E. Neuroanatomical substrate of visuospatial and visuoperceptual impairment in Parkinson's disease. Mov Disord. 2009 Jun 15;24(8):11939. Palmer SJ, Eigenraam L, Hoque T, McCaig RG, Troiano A, McKeown MJ. Levodopa-sensitive, dynamic changes in effective connectivity during simultaneous movements in Parkinson's disease. Neuroscience. 2009a Jan 23;158(2):693-704. Palmer SJ, Ng B, Abugharbieh R, Eigenraam L, McKeown MJ. Motor reserve and novel area recruitment: amplitude and spatial characteristics of compensation in Parkinson's disease. Eur J Neurosci. 2009b Jun;29(11):2187-96. Palmer SJ, Wen-Hsin Lee P, Wang ZJ, Au WL, McKeown MJ. theta, beta But not alpha-band EEG connectivity has implications for dual task performance in Parkinson's disease. Parkinsonism Relat Disord. 2010 May 1. Praamstra P, Pope P. Slow brain potential and oscillatory EEG manifestations of impaired temporal preparation in Parkinson's disease. J Neurophysiol. 2007 Nov 1;98(5):2848-57. Rodnitzky RL. Visual dysfunction in Parkinson's disease. Clin Neurosci. 1998;5(2):102-6. Silberstein P, Pogosyan A, Kühn AA, Hotton G, Tisch S, Kupsch A, et al. Cortico-cortical coupling in Parkinson's disease and its modulation by therapy. Brain. 2005 Jun 1;128(Pt 6):1277-91. Silva MF, Faria P, Regateiro FS, Forjaz V, Januario C, Freire A, et al. Independent patterns of damage within magno-, parvo- and koniocellular pathways in Parkinson's disease. Brain. 2005 Oct;128(Pt 10):2260-71. Srinivasan R, Winter WR, Ding J, Nunez PL. EEG and MEG coherence: measures of functional connectivity at distinct spatial scales of neocortical dynamics. J Neurosci Methods. 2007 Oct 15;166(1):4152. Steinmetz H, Furst G, Meyer BU. Craniocerebral topography within the international 10-20 system. Electroencephalogr Clin Neurophysiol. 1989 Jun;72(6):499-506. Stoffers D, Bosboom JLW, Deijen JB, Wolters EC, Stam CJ, Berendse HW. Increased cortico-cortical functional connectivity in early-stage Parkinson's disease: an MEG study. Neuroimage. 2008a Jun 1;41(2):212-22. Stoffers D, Bosboom JLW, Wolters EC, Stam CJ, Berendse HW. Dopaminergic modulation of corticocortical functional connectivity in Parkinson's disease: an MEG study. Experimental Neurology. 2008b Sep 1;213(1):191-5. Sun Y, Zhang H, Feng T, Qiu Y, Zhu Y, Tong S. Early cortical connective network relating to audiovisual stimulation by partial directed coherence analysis. IEEE Trans Biomed Eng. 2009 Nov 1;56(11 Pt 2):27214. Tropini G, Chiang J, Wang Z, McKeown MJ. Partial directed coherence-based information flow in Parkinson's disease patients performing a visually-guided motor task. Conf Proc IEEE Eng Med Biol Soc. 2009 Jan 1;2009:1873-8.  74  Williams D, Tijssen M, Van Bruggen G, Bosch A, Insola A, Di Lazzaro V, et al. Dopamine-dependent changes in the functional connectivity between basal ganglia and cerebral cortex in humans. Brain. 2002 Jul 1;125(Pt 7):1558-69. Williams D, Kühn A, Kupsch A, Tijssen M, van Bruggen G, Speelman H, et al. Behavioural cues are associated with modulations of synchronous oscillations in the human subthalamic nucleus. Brain. 2003 Sep 1;126(Pt 9):1975-85. Winter WR, Nunez PL, Ding J, Srinivasan R. Comparison of the effect of volume conduction on EEG coherence with the effect of field spread on MEG coherence. Stat Med. 2007 Sep 20;26(21):3946-57. Witte H, Ungureanu M, Ligges C, Hemmelmann D, Wüstenberg T, Reichenbach J, et al. Signal informatics as an advanced integrative concept in the framework of medical informatics. New trends demonstrated by examples derived from neuroscience. Methods Inf Med. 2009 Jan 1;48(1):18-28.  75  CHAPTER 4:  GENERAL DISCUSSION AND CONCLUSIONS  4.1  Introduction  The purpose of this thesis was to investigate the modulation of frequencydependent cortical connectivity in PD patients, and to elucidate the effects of medication on overall activity patterns. We found widespread alterations in EEG connectivity that were only partially restored and often exacerbated by levodopa medication. Our results are consistent with the notion that the cortex might play a greater role in the modulation of abnormal rhythms in PD than previously recognized. We utilized a novel sparse mAR-based PDC method, developed by our collaborators, to assess frequency-dependent connectivity changes, and tested it on preliminary data from PD patients OFF medication and healthy controls. We subsequently further optimized the analysis and applied it to a full EEG data set from PD patients ON and OFF medication, as well as healthy controls performing a visually guided task that had previously been shown to modulate β-band activity in STN LFP recordings. This chapter will summarize and discuss the main findings of the thesis, address the limitations of the experiments performed and provide recommendations for future investigations.  76  4.2  Summary of Findings  In Chapter 2, we showed that the proposed sparse mAR- based PDC method (smAR-PDC) was able to reveal distinctly altered patterns of EEG-based connectivity between PD patients off medication and healthy subjects. PDC-based analysis was also more sensitive to region-specific changes than spectral analysis, and because of its ability to provide directional (and hence causal) information we suggested that it could be a suitable technique to investigate altered cortical connectivity in PD. In Chapter 3, we optimized the smAR-PDC analysis and utilized this technique to analyze a full EEG data set from PD subjects ON and OFF medication -as well as healthy controls- performing a visually guided choice reaction task. Our analysis was improved with respect to Chapter 2 methods by providing an overall measure of connectivity and power modulation with respect to baseline levels, by including an additional region (the Occipital region) in the analysis, as well as by dividing the analysis into two components: (1) an overall assessment of temporal dynamics, obtained via the calculation of spectrograms and PDC-ograms, and (2) a quantitative analysis of the preparatory phase of the task interval (PRE) and the movement phase of the task (MOV). The latter type of analysis included both power spectra and PDC-based connectivity spectra, as well as a correlation of average PDC-based connectivity values and UPDRS Motor scores for PD patients OFF medication in different frequency bands of interest (θ, α, β and α-β). Our results showed that PD patients presented impaired power modulation in both the α and β frequency bands during both the PRE and MOV phases of the task, although differences between PD OFF subjects and healthy controls were greatest during the MOV phase. Inter-region differences between the 5 cortical areas of interest (FCentral, L and R 77  Sm, Central and Occipital) within each subject group were relatively modest, although larger differences between PD OFF and controls were found in the 10-30 Hz range in the L and R Sm as well as Occipital regions. In addition, levodopa medication only partially restored normal modulation of power in the 10-30 Hz range, while it further blunted modulation at lower frequencies (<10 Hz). In contrast, complex and region-specific alterations in directional connectivity were revealed by the smAR-PDC analysis. Connectivity was most altered in PD OFF patients  in  the  CentralFCentral,  OccipitalLSm,  OccipitalRSm,  CentralOccipital as revealed by PDCograms and confirmed by quantitative analysis of the PRE and MOV intervals. Quantitative analysis also revealed that specific connections (OccipitalL Sm, and CentralFCentral) were more disrupted in PD OFF during the PRE phase, while previously not apparent disruptions in L SmFCentral emerged in the 18-40 Hz range during the MOV phase. Akin to the spectral analysis findings, PDC-based analysis also revealed that while levodopa was able to partially restore connectivity in some instances (Central FCentral in the PRE phase, as well as OccipitalL Sm in the MOV phase), it disrupted or exacerbated already altered connectivity in several instances (in particular, L SmFCentral, L SmCentral during both PRE and MOV). Last, a very important finding of the quantitative PDC-based analysis was the correlation of motor UPDRS scores with PDC-based connectivity for PD OFF subjects. For both PRE and MOV, a positive correlation was found between the clinical scores and the average connectivity with respect to rest in the CentralFCentral connection.  78  Correlations were significant in the β (13-30 Hz), α-β(8-30 Hz) ranges for PRE, and in the α (8-12 Hz), β, and α-β ranges for MOV.  4.3  Study Significance  Our results indicate that PD patients show not only altered power spectral modulation in the α and β frequency bands in each region of interest (particularly during movement), but also complex alterations in the modulation of connectivity between regions that correlate with motor clinical measures during both movement preparation and execution. Independent studies have assessed spectral changes in the EEG in healthy controls or PD subjects, but very few studies have directly compared power spectral modulation during motor tasks in both groups or included the effect of medication. For example, some EEG studies of choice reaction tasks in healthy individuals have found power modulation patterns that are comparable to our findings (Doyle et al. , 2005, Kranczioch et al. , 2008). Similarly, several LFP studies have revealed abnormal modulation of activity in the β band in PD patients that mirror our findings of blunted power suppression in the 10-25 Hz range both before the imperative cue and during movement (Brown et al. , 2001, Brown and Williams, 2005, Cassidy et al. , 2002, Levy et al. , 2002). A previous EEG study (Praamstra and Pope, 2007) showed attenuated desynchronization, as compared to controls, in the α and β bands in PD patients ON medication performing a choice reaction task, but did not investigate the OFF medication state. A recent MEG study directly compared oscillatory brain activity in PD patients ON and OFF medication to that of age-matched controls, but only investigated an eyes-closed 79  resting state condition (Stoffers et al. , 2007). In contrast, our study directly compared power spectral modulation during a visually-guided motor task in PD patients ON and OFF levodopa, as well as in age-matched healthy controls, thus further clarifying the differences among the three groups and the effects of medication. In particular, our results showed that levodopa medication did not completely restore patterns seen in normals, and that it also further blunted modulation at frequencies <10 Hz, particularly during movement. Our findings also revealed complex alterations in directional connectivity, particularly in posterior connections that might underlie visuo-motor processing. Indeed, PD patients have been known to show several impairments in the visual system. In particular, recent studies have revealed deficits such as occipital and posterior hypoperfusion that correlate with impaired performance during a visuocognitive test (Abe et al. , 2003), as well as a correlation of gray matter atrophy in posterior temporal, parietal and occipital regions with deteriorated performance in visuospatial and visuoperceptive tests (Pereira et al. , 2009). Although further work is needed to elucidate the role of each examined PDC-based connection in specific visuo-cognitive and visuomotor tasks, our results show promise for future ways to examine visual impairments in PD in the context of the BG-thalamocortical oscillatory model of the disease. In addition, we showed that levodopa medication only partially restored impaired connectivity, and sometimes exacerbated or deteriorated existing connections, perhaps explaining remaining deficits in medicated PD patients. It is interesting that medication had a particularly negative effect on connectivity that might underlie motor processing (LSmFCentral, and LSmCentral, Figure 3-7 and Figure 3-8), as side effects of  80  levodopa therapy include motor fluctuations and dyskinesias (Damier, 2009, Schapira et al. , 2009). Our results showed that altered connectivity in the CentralFCentral connection in PD subjects OFF medication correlated with motor UPDRS scores in the β and α-β frequency ranges during both movement preparation and execution. Altered modulation in CentralFCentral might again underlie deficits in visuo-motor processing along a parietal to centro-frontal stream, although more work is needed to elucidate the role of this connection via specific visuomotor tests, as well as purely motor or visual paradigms. The finding that PDC values for this connection consistently correlated with clinical measures during both phases of the task also raises the possibility that altered connectivity CentralFCentral might be utilized in the future as a potential biomarker for PD deficits. Overall, our findings support the BG oscillatory model of PD, and suggest that the EEG might be a suitable method to investigate altered frequency-dependent cortical connectivity. Very few studies have investigated functional cortical connectivity in the context of abnormal oscillatory activity in PD using EEG (Cassidy and Brown, 2001, Silberstein et al. , 2005) or MEG (Stoffers et al. , 2008a, Stoffers et al. , 2008b), and utilized methods such as coherence or synchronization likelihood (SL) that do not provide directional connectivity information. In contrast, our method was able to reveal several asymmetries in connectivity, not only in PD subjects but also in control subjects, thus indicating a differential contribution of each region and connection to task demands in both the healthy and diseased states. PDC-based methods are also more robust to  81  volume conduction issues that are known to affect EEG analysis, since the PDC takes into account the effect of all remaining nodes when assessing pairwise connectivity. Last, an important strength of EEG-based investigation of altered oscillatory activity in PD is the non-invasive nature of the recordings. While a vast majority of studies in this area have involved invasive LFP recordings, our investigation allowed us to assess both normal controls and PD subjects that were not candidates for DBS surgery. EEG-based cortical connectivity is an indirect measure of abnormal oscillatory activity in BG-thalamocortical circuits in PD, but several recent studies have revealed that enhanced coupling of activity exists between the STN and cortex in PD (Klostermann et al. , 2007, Marsden et al. , 2001). In addition, the cortex appears to have a greater role than previously recognized in the modulation of abnormal rhythms in PD, given that cortical areas, and not STN, have been shown to be the primary site of STN DBS action (Dejean et al. , 2009, Gradinaru et al. , 2009, Li et al. , 2007). Overall, our results suggest that the role of the EEG to monitor abnormal oscillatory activity in PD might need to be expanded. Non-invasive findings from the EEG could also potentially benefit existing therapies for PD or open new therapeutic avenues. The effects of DBS on cortical connectivity could be for example assessed via EEG measures, and stimulator settings could be subsequently optimized to reduce altered connectivity. Similarly, given the preliminary result that non-invasive TMS can reduce abnormal rhythms in the BG (Doyle Gaynor et al. , 2008), results from the EEG could be potentially used to direct stimulation and optimize restoration of connectivity. In addition, our techniques could be applied to other conditions known to be characterized by altered oscillatory activity, such as epilepsy or schizophrenia.  82  Findings from this thesis might also increase our understanding of the pathophysiology of PD, particularly in the context of the oscillatory model of the basal ganglia. Frequency-dependent abnormal modulation of connections associated with visual and visuo-motor processing suggests that altered processing in these pathways in PD could be due not only to structural abnormalities or changes in the firing rate of the fibers involved, but potentially also to altered synchronization and desynchronization patterns, although further work is needed to address these speculations. Similarly, the finding that levodopa results in deteriorated modulation of connections that might be associated with motor processing raises the possibility that specific pathways might be responsible for the motor side effects of levodopa. For example, it has recently been shown that poorly-understood levodopa-induced dyskinesias (LID) cannot be considered a purely random movement, and that specific -although yet unidentified- circuits might be involved (Gour et al. , 2007). Although further investigation is needed in dyskinetic patients, and using paradigms that could better allow to distinguish motor and sensory components, our findings could lead to an increased understanding of this, and other, important aspects of the parkinsonian state.  4.4  Study Limitations  LFP Recordings We initially planned to record STN LFP activity while a small subset (n=5) of PD subjects were performing the same visually-guided task that they had performed during EEG recordings, so that we could better compare information derived from deep brain recordings to that inferred via EEG techniques. However, due to several technical and 83  recruitment issues (Please cf. Appendix I for a detailed discussion), we were only able to obtain LFP data from 1 PD subject. In addition, the patient had significant difficulties completing the behavioural task during surgery due to fatigue, and thus was able to correctly perform only a very small subset of trials (7 out of 68). Other technical issues included the presence of noise artifacts in the data, which made analysis particularly difficult. Therefore, we were not able to compare our results from EEG measurements to those obtained via LFP recordings, thus limiting the opportunity to assess whether the EEG might allow to indirectly infer BG activity. However, previous studies have shown mirroring of STN activity in the cortex (Klostermann et al. , 2007, Marsden et al. , 2001), and it is known that abnormal BG oscillations spread throughout BG-thalamocortical networks (Brown et al. , 2001, Brown and Williams, 2005, Cassidy et al. , 2002, Gatev et al. , 2006). Thus, although future LFP recordings might be desirable to directly compare EEG measures to STN activity, our results might still allow us to infer activity that has likely synchronized over large scales in BG-thalamocortical circuits.  EEG Considerations Other study limitations include the choice of electrode regions for EEG data collection and analysis. While we desired to maintain high data spatial resolution, electrode and region selection were influenced by the empirical observation that PDCbased analysis becomes less robust as the number of nodes increases. In addition, a higher number of nodes increases computational demands from the mAR calculations, and thus, our choice of 5 regions of interest was a tradeoff between minimizing the number of smAR-PDC nodes while still maintaining spatial resolution. Further work is  84  also needed to better determine the optimal methods for identifying the contribution of each electrode to the regions of interest. We have previously explored an Independent Component Analysis (ICA)-based method to determine optimal weighting of individual electrodes, but the results were not significantly different from those obtained via the simpler averaging method. In addition, while each electrode in the 10-20 System has previously been referenced to specific brain landmarks or areas (Homan et al. , 1987, Steinmetz et al. , 1989), it can be argued that the EEG can only provide a very rough measure of cortical activity and thus it can be difficult to assess the underlying neurobiology. We chose to utilize the EEG because it might be biased to detect widespread activity that has synchronized over large scales (Volkmann, 1998), and because of its high temporal resolution that would be appropriate to investigate abnormal oscillatory activity in PD. However, a future improvement of our study could include the use of simultaneous high resolution EEG with fMRI (Mantini et al. , 2010), to further increase our understanding of the biological mechanisms underlying the observed connectivity changes.  Subjects Due to recruitment limitations, the majority of our PD subjects (8 out of 10) were male, while our control subjects were more equally distributed (5 males and 5 females). In order to assess gender-specific differences in connectivity we performed a post-hoc correlation of gender with PDC-connectivity values for each connection, frequency band, and task interval (PRE and MOV). While no correlations were significant, it might be desirable in future studies to have a more balanced PD population, especially as there  85  have been anecdotal reports of gender-dependent differences in STN LFP measurements (Alberto Priori, "Basal Ganglia Oscillations as a Window into the Characteristics of Parkinson's Disease", Symposium in the Movement Disorders Society 13th International Congress, Paris June 7-11, 2009). In addition, again due to recruitment limitations, one PD subject was left-handed. We chose to have subjects perform the task with their dominant hand, as the joystick task requires significant dexterity, and would have thus created difficulties in movement execution if subjects were to use their non-dominant hand instead. While subject groups were still directly comparable, as one control subject was also left-handed, it would have been desirable to have greater handedness consistency in our subjects. However, the use of leave-one-out validation techniques did not reveal significant changes in our results when excluding the left-handed subjects from the analysis.  4.5  Future Directions  In light of the issues discussed in section 4.4, an important future direction of this study should be a direct comparison of EEG measures of abnormal oscillatory activity in PD with LFP recordings. In order to facilitate data collection, it might be desirable to perform LFP recordings in PD patients that have had their macroelectrode leads externalized prior to implantation of the subcutaneous stimulator, and thus collect the data in the days following the initial surgery and prior to the final implantation. This would likely reduce patient fatigue, and would thus allow for more extensive testing to be performed, while also allowing potential simultaneous EEG and LFP recordings.  86  In addition, it might be of interest to perform EEG recordings while the stimulator is ON, so that some of the positive effects of DBS on cortical connectivity could be investigated. Additional testing could include a combination of the ON/OFF medication and ON/OFF conditions, as to more accurately define the individual and combined effects of high frequency stimulation and pharmacological therapy on connectivity. Another investigative avenue that we are currently pursuing is the analysis of EEG data from other joystick-based visually-guided motor tasks. Specifically, we collected data from two other types of paradigms that have been shown to modulate LFP activity in PD: (1) A choice reaction task that included warning cues, based on a task by Williams et al. (Williams et al. , 2003); (2) A Go-No Go task based on a task by Kuhn et al. that had shown early β-power increase following presentation of a no go signal (Kühn et al. , 2004). We also collected data from a Stop-signal task based on a paradigm that had shown the involvement of the STN in the inhibition of Go response execution, potentially via a hyperdirect pathway between the STN and the inferior frontal cortex (Aron and Poldrack, 2006). In addition, we collected data from our subjects while they were resting, both with their eyes open (and fixating on a stationary cross), and with their eyes closed. The investigation of EEG-based modulation of connectivity in a variety of tasks and conditions will allow us to extend the interpretation of our results to more behavioural paradigms and situations, thus increasing our understanding of abnormal oscillatory activity in PD. Last, given that our findings suggest PD abnormalities in posterior connectivity that might underlie visuo-motor processing, it would be of interest to investigate specific visual tasks that have been shown to correlate with anatomical and functional alterations  87  in the visual system in PD (Abe et al. , 2003, Pereira et al. , 2009). We chose to investigate a paradigm that had previously been shown to modulate BG β activity (Amirnovin et al. , 2004) so that we could better compare our results with previous reports on abnormal oscillatory activity in PD. However, our results did not allow us to differentiate between visual and motor components of the task. Thus, an investigation of specific tasks that are purely motor or visual in nature might be particularly useful to increase our understanding of motor and sensory connectivity impairments in PD.  88  4.6  References  Abe Y, Kachi T, Kato T, Arahata Y, Yamada T, Washimi Y, et al. Occipital hypoperfusion in Parkinson's disease without dementia: correlation to impaired cortical visual processing. J Neurol Neurosurg Psychiatry. 2003 Apr;74(4):419-22. Amirnovin R, Williams ZM, Cosgrove GR, Eskandar EN. Visually guided movements suppress subthalamic oscillations in Parkinson's disease patients. J Neurosci. 2004 Dec 15;24(50):11302-6. Aron AR, Poldrack RA. Cortical and subcortical contributions to Stop signal response inhibition: role of the subthalamic nucleus. J Neurosci. 2006 Mar 1;26(9):2424-33. Brown P, Oliviero A, Mazzone P, Insola A, Tonali P, Di Lazzaro V. Dopamine dependency of oscillations between subthalamic nucleus and pallidum in Parkinson's disease. J Neurosci. 2001 Feb 1;21(3):1033-8. Brown P, Williams D. Basal ganglia local field potential activity: character and functional significance in the human. Clinical Neurophysiology. 2005 Nov 1;116(11):2510-9. Cassidy M, Brown P. Task-related EEG-EEG coherence depends on dopaminergic activity in Parkinson's disease. Neuroreport. 2001 Mar 26;12(4):703-7. Cassidy M, Mazzone P, Oliviero A, Insola A, Tonali P, Di Lazzaro V, et al. Movement-related changes in synchronization in the human basal ganglia. Brain. 2002 Jun 1;125(Pt 6):1235-46. Damier P. Drug-induced dyskinesias. Curr Opin Neurol. 2009 Aug;22(4):394-9. Dejean C, Hyland B, Arbuthnott G. Cortical effects of subthalamic stimulation correlate with behavioral recovery from dopamine antagonist induced akinesia. Cereb Cortex. 2009 May 1;19(5):1055-63. Doyle Gaynor LMF, Kühn AA, Dileone M, Litvak V, Eusebio A, Pogosyan A, et al. Suppression of beta oscillations in the subthalamic nucleus following cortical stimulation in humans. European Journal of Neuroscience. 2008 Oct 1;28(8):1686-95. Doyle LMF, Yarrow K, Brown P. Lateralization of event-related beta desynchronization in the EEG during pre-cued reaction time tasks. Clin Neurophysiol. 2005 Aug 1;116(8):1879-88. Gatev P, Darbin O, Wichmann T. Oscillations in the basal ganglia under normal conditions and in movement disorders. Mov Disord. 2006 Oct 1;21(10):1566-77. Gour J, Edwards R, Lemieux S, Ghassemi M, Jog M, Duval C. Movement patterns of peak-dose levodopainduced dyskinesias in patients with Parkinson's disease. Brain Res Bull. 2007 Sep 14;74(1-3):66-74. Gradinaru V, Mogri M, Thompson KR, Henderson JM, Deisseroth K. Optical deconstruction of parkinsonian neural circuitry. Science. 2009 Apr 17;324(5925):354-9. Homan RW, Herman J, Purdy P. Cerebral location of international 10-20 system electrode placement. Electroencephalogr Clin Neurophysiol. 1987 Apr;66(4):376-82. Klostermann F, Nikulin VV, Kühn AA, Marzinzik F, Wahl M, Pogosyan A, et al. Task-related differential dynamics of EEG alpha- and beta-band synchronization in cortico-basal motor structures. Eur J Neurosci. 2007 Mar 1;25(5):1604-15.  89  Kranczioch C, Athanassiou S, Shen S, Gao G, Sterr A. Short-term learning of a visually guided power-grip task is associated with dynamic changes in EEG oscillatory activity. Clin Neurophysiol. 2008 Jun 1;119(6):1419-30. Kühn AA, Williams D, Kupsch A, Limousin P, Hariz M, Schneider G-H, et al. Event-related beta desynchronization in human subthalamic nucleus correlates with motor performance. Brain. 2004 Apr 1;127(Pt 4):735-46. Levy R, Ashby P, Hutchison WD, Lang AE, Lozano AM, Dostrovsky JO. Dependence of subthalamic nucleus oscillations on movement and dopamine in Parkinson's disease. Brain. 2002 Jun 1;125(Pt 6):1196209. Li S, Arbuthnott GW, Jutras MJ, Goldberg JA, Jaeger D. Resonant antidromic cortical circuit activation as a consequence of high-frequency subthalamic deep-brain stimulation. J Neurophysiol. 2007 Dec 1;98(6):3525-37. Mantini D, Marzetti L, Corbetta M, Romani GL, Del Gratta C. Multimodal integration of fMRI and EEG data for high spatial and temporal resolution analysis of brain networks. Brain Topogr. 2010 Jun;23(2):1508. Marsden JF, Limousin-Dowsey P, Ashby P, Pollak P, Brown P. Subthalamic nucleus, sensorimotor cortex and muscle interrelationships in Parkinson's disease. Brain. 2001 Feb 1;124(Pt 2):378-88. Pereira JB, Junque C, Marti MJ, Ramirez-Ruiz B, Bargallo N, Tolosa E. Neuroanatomical substrate of visuospatial and visuoperceptual impairment in Parkinson's disease. Mov Disord. 2009 Jun 15;24(8):11939. Praamstra P, Pope P. Slow brain potential and oscillatory EEG manifestations of impaired temporal preparation in Parkinson's disease. J Neurophysiol. 2007 Nov 1;98(5):2848-57. Schapira AH, Emre M, Jenner P, Poewe W. Levodopa in the treatment of Parkinson's disease. Eur J Neurol. 2009 Sep;16(9):982-9. Silberstein P, Pogosyan A, Kühn AA, Hotton G, Tisch S, Kupsch A, et al. Cortico-cortical coupling in Parkinson's disease and its modulation by therapy. Brain. 2005 Jun 1;128(Pt 6):1277-91. Steinmetz H, Furst G, Meyer BU. Craniocerebral topography within the international 10-20 system. Electroencephalogr Clin Neurophysiol. 1989 Jun;72(6):499-506. Stoffers D, Bosboom JL, Deijen JB, Wolters EC, Berendse HW, Stam CJ. Slowing of oscillatory brain activity is a stable characteristic of Parkinson's disease without dementia. Brain. 2007 Jul;130(Pt 7):184760. Stoffers D, Bosboom JLW, Deijen JB, Wolters EC, Stam CJ, Berendse HW. Increased cortico-cortical functional connectivity in early-stage Parkinson's disease: an MEG study. Neuroimage. 2008a Jun 1;41(2):212-22. Stoffers D, Bosboom JLW, Wolters EC, Stam CJ, Berendse HW. Dopaminergic modulation of corticocortical functional connectivity in Parkinson's disease: an MEG study. Experimental Neurology. 2008b Sep 1;213(1):191-5. Volkmann J. Oscillations of the human sensorimotor system as revealed by magnetoencephalography. Mov Disord. 1998;13 Suppl 3:73-6.  90  Williams D, Kühn A, Kupsch A, Tijssen M, van Bruggen G, Speelman H, et al. Behavioural cues are associated with modulations of synchronous oscillations in the human subthalamic nucleus. Brain. 2003 Sep 1;126(Pt 9):1975-85.  91  APPENDICES Appendix I: STN LFP Data Collection  I.1 Overview In order to better compare the results obtained from EEG data with traditional methods of assessment of abnormal oscillatory activity in PD, we initiated a collaboration with Dr. Chris Honey (VGH Neurosurgery) to collect local field potential (LFP) data during STN DBS surgery. We aimed to collect data from n=5 patients. However, due to several difficulties in patient recruitment as well as technical difficulties during data collection, we ultimately collected LFP data from 1 subject only.  I.2 Methods Subjects Due to difficulties in finding suitable study participants, we were only able to recruit 3 candidates (P015, P016, P017) for bilateral STN DBS neurosurgery from the Movement Disorders Clinic at the University of British Columbia (Their clinical details are summarized in Table A-2 in Section III.1 Details for all subjects). Exclusion criteria included atypical Parkinsonism, presence of other neurological conditions, and visual conditions requiring correction beyond the use of eye glasses or contact lenses. All patients were on levodopa medication, but they were off medication during the surgery. The study was approved by the University of British Columbia Clinical Research Ethics Board (CREB), and all subjects gave written and oral consent prior to participation.  92  EEG and LFP data Collection Details Patients were recruited to perform the EEG portion of the study (Please refer to Section 2.4, for full details) approximately 1 week prior to their surgery. Although simultaneous recording of EEG and DBS data would have been most desirable, EEG data collection during surgery was impeded by the presence of sterile surgical dressings. Moreover, patients were typically asked by the neurosurgeon to withdraw from their medication for a few days before the surgery, and thus were often not comfortable participating in the EEG experiment in the days immediately preceding the surgery, after having been off their medication for extended periods of time. Thus, we chose to collect data as close as possible to the day of the surgery, while still maintaining patient comfort. In addition, while typically the leads to the electrodes are externalized for a few days before implantation of the stimulator, thus allowing for more extended testing to be performed (Cassidy et al. , 2002, Fogelson et al. , 2006, Klostermann et al. , 2007, Kühn et al. , 2004, Lalo et al. , 2008, Williams et al. , 2003) DBS surgery at VGH is performed in one day only (i.e. the stimulator is implanted immediately following implantation of STN electrodes). This reduces the LFP data collection window to a short 30 minute intraoperative interval, and also makes it more likely that patients might not be able to complete the study due to fatigue or technical difficulties during the surgery. Ultimately, we were able to collect LFP data from one subject only (P017). One other patient (P015) withdrew from the study prior to completion of the EEG data collection due to discomfort while off medication, while patient P016 performed the EEG portion of the study but ultimately DBS surgery was deemed unnecessary for them.  93  Subject P017 was implanted with bilateral STN macroelectrodes after the STN had been identified by pre-operative MRI. The macroelectrodes used were platinumiridium electrodes (Medtronic Neurological Division Minneapolis, Minneapolis, NM) and each electrode had 4 contacts with a separation of 2 mm each. LFP collection was performed in the left STN (since the patient performed the task with their right hand), via bipolar recordings from contacts # 2 and #4. Unfortunately, no ground electrode was provided. LFP data were collected intraoperatively, while the subject performed the visually guided joystick task described in section 0, as well as during a 1 minute rest period while the subject had their eyes closed, and a 1 minute rest period while they were fixating on a large cross on the computer screen. Data were recorded at a sampling rate of 2000 Hz and stored via the Synamps2 amplifier system and Scan4.3 software (Neuroscan, Compumedics, Texas USA). Alignment of data to task-related events was performed by sending Matlab-coded TTL pulses via parallel port from the stimulus computer to the Neuroscan Synamps2 system. TTL pulses were then recorded as event-specific timestamps by the Scan 4.3 software. Data were then down-sampled to 250 Hz, and band-passed 1-100 Hz for subsequent examination and analysis.  94  I.3 Outcomes and Discussion Due to fatigue during the surgery, the subject had considerable difficulty performing the task, only correctly selecting 7 out of 68 total targets. The average response time for successful trials was 1.27s. In addition, LFP data contained several noise artifacts, likely due to a lack of a ground electrode, and to the presence of several noise sources in the OR in the proximity of the recording apparatus. As very few trials were successful, and as artifacts often occurred right in the middle of the trial, it was very difficult to properly reject artifact regions while still salvaging intact trial data. Thus, it was not possible to make any inferences on task-dependent modulation of LFP activity in our subject. In order to improve future data collection, it might be desirable to collect data via the externalized macroelectrode leads that in some neurosurgery centers (for example, Hotchkiss Brain Institute, Calgary; Western Hospital, Toronto) are connected to the stimulator a few days after the surgery, thus allowing to perform recordings in the perioperative interval, and minimizing patient fatigue and discomfort. Data collection in the perioperative interval can also allow for more controlled experimental conditions, such as a minimization of electrical noise sources that contributed to artifacts in our recordings.  95  I.4 References Cassidy M, Mazzone P, Oliviero A, Insola A, Tonali P, Di Lazzaro V, et al. Movement-related changes in synchronization in the human basal ganglia. Brain. 2002 Jun 1;125(Pt 6):1235-46. Fogelson N, Williams D, Tijssen M, van Bruggen G, Speelman H, Brown P. Different functional loops between cerebral cortex and the subthalmic area in Parkinson's disease. Cereb Cortex. 2006 Jan;16(1):6475. Klostermann F, Nikulin VV, Kühn AA, Marzinzik F, Wahl M, Pogosyan A, et al. Task-related differential dynamics of EEG alpha- and beta-band synchronization in cortico-basal motor structures. Eur J Neurosci. 2007 Mar 1;25(5):1604-15. Kühn AA, Williams D, Kupsch A, Limousin P, Hariz M, Schneider G-H, et al. Event-related beta desynchronization in human subthalamic nucleus correlates with motor performance. Brain. 2004 Apr 1;127(Pt 4):735-46. Lalo E, Thobois S, Sharott A, Polo G, Mertens P, Pogosyan A, et al. Patterns of bidirectional communication between cortex and basal ganglia during movement in patients with Parkinson disease. J Neurosci. 2008 Mar 19;28(12):3008-16. Williams D, Kühn A, Kupsch A, Tijssen M, van Bruggen G, Speelman H, et al. Behavioural cues are associated with modulations of synchronous oscillations in the human subthalamic nucleus. Brain. 2003 Sep 1;126(Pt 9):1975-85.  96  Appendix II: Ethics Certificates II.1 Ethics Certificates for EEG Data Collection  97  98  99  100  II.2 Ethics Certificates for DBS LFP Data Collection  101  102  Appendix III: Additional Materials for Chapter 3 III.1 Details for all subjects  Table A-1. Details for all healthy controls. Indicated in grey are subjects that were not included in the final analysis, as well as the reasons for exclusion (Cf. the 'Notes' column in the table ).  103  Table A-2. Details for all PD subjects. Indicated in grey are subjects that were not included in the final analysis, as well as the reasons for exclusion (Cf. the 'Notes' column). Patients that were selected for the DBS study are indicated in the appropriate column. 'DBS-incompl' indicates subjects for whom DBS recordings could not be collected.  104  III.2 Response times for all included subjects  Normals  PDOFF  PDON  Subject ID N002 N003 N004 N005 N006 N007 N008 N009 N010 N011  Response time (s) 0.624 1.159 0.686 0.674 0.771 0.703 0.765 0.720 0.734 0.813  Avg Stdev  0.765 0.149  P004 P005 P006 P007 P008 P010 P011 P012 P014 P016  0.967 0.686 0.931 1.011 0.911 0.970 0.998 1.003 0.732 0.955  Avg Stdev  0.916 0.114  P004 P005 P006 P007 P008 P010 P011 P012 P014 P016  0.928 0.660 0.789 0.904 1.038 0.942 0.879 0.927 0.640 0.858  Avg Stdev  0.856 0.126  105  III.3 Temporal Dynamics - Additional PDCograms  Figure A-1. Additional PDCograms for all groups, a). Shown are the L Sm  R Sm connections (top) and FCentral  L Sm (bottom). The top row of each figure shows left to right direction of connectivity and the bottom row shows the reverse direction. Colorbars indicate connectivity values with respect to rest (a.u.).  106  Figure A-2. Additional PDCograms for all groups, b). Shown are the FCentral  R Sm connections (top) and FCentral  Occipital connections (bottom). The top row of each figure shows left to right direction of connectivity and the bottom row shows the reverse direction. Colorbars indicate connectivity values with respect to rest (a.u.).  107  Figure A-3. Additional PDCograms for all groups, c). Shown are the R Sm  Central connections (top) and L Sm  Central connections (bottom). The top row of each figure shows left to right direction of connectivity and the bottom row shows the reverse direction. Colorbars indicate connectivity values with respect to rest (a.u.).  108  III.4 Quantitative Analysis-Correlations of PDC connectivity with OFF UPDRS III.4.1 PRE Central to FCentral Correlation Graphs  Figure A-4. PRE Central to FCentral Correlation Graph, 8-30 Hz.  Figure A-5. PRE Central to FCentral Correlation Graph, 13-30 Hz.  109  III.4.2 PRE Correlations Band 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta'  Connections L Sm-->Fcentral' R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm' FCentral-->R Sm' L Sm-->R Sm' Central-->R Sm' Occipital-->R Sm' FCentral-->Central' L Sm-->Central' R Sm-->Central' Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital' Central-->Occipital'  R 0.1580 0.6802 0.4822 0.8492 0.3725 0.1139 0.1081 0.2308 0.1073 0.0093 0.6287 0.1365 0.4139 0.2300 0.3164 0.0463 0.1211 0.3593 0.1816 0.2602  p 0.6886 0.1961 0.1573 0.3769 0.3553 0.7932 0.7404 0.5254 0.7329 0.9668 0.0557 0.7145 0.2375 0.6075 0.5320 0.8912 0.7154 0.3625 0.6235 0.4437  Band 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha'  Connections L Sm-->Fcentral' R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm' FCentral-->R Sm' L Sm-->R Sm' Central-->R Sm' Occipital-->R Sm' FCentral-->Central' L Sm-->Central' R Sm-->Central'  R 0.0892 0.2994 0.6597 0.2524 0.2590 0.1411 0.1672 0.0359 0.2472 0.0600 0.6351 0.0105 0.1825 0.3275 0.3195  p 0.8430 0.6557 0.0384 0.5104 0.4996 0.7269 0.6270 0.9346 0.4728 0.8841 0.0469 0.9906 0.5971 0.4464 0.4984  110  'alpha' 'alpha' 'alpha' 'alpha' 'alpha'  Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital' Central-->Occipital'  0.1671 0.1816 0.4213 0.0282 0.2983  0.6466 0.5913 0.2369 0.9288 0.3687  Band 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta'  Connections L Sm-->Fcentral' R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm' FCentral-->R Sm' L Sm-->R Sm' Central-->R Sm' Occipital-->R Sm' FCentral-->Central' L Sm-->Central' R Sm-->Central' Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital' Central-->Occipital'  R 0.0761 0.3382 0.8067 0.2313 0.0365 0.1978 0.3161 0.1718 0.3195 0.1701 0.5011 0.0727 0.2290 0.2598 0.1610 0.1867 0.1595 0.5304 0.0570 0.4117  p 0.7897 0.3938 0.0086 0.5769 0.9249 0.5820 0.3632 0.6460 0.3594 0.6406 0.1494 0.8162 0.5565 0.4877 0.6754 0.6075 0.6345 0.1235 0.8902 0.2301  Band 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta'  Connections L Sm-->Fcentral' R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm' FCentral-->R Sm' L Sm-->R Sm' Central-->R Sm'  R 0.0178 0.1638 0.7801 0.1381 0.0903 0.1872 0.2861 0.1255 0.3094 0.1456 0.5566  p 0.8893 0.6773 0.0109 0.7541 0.8101 0.6083 0.4137 0.7370 0.3750 0.6940 0.1040  111  'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta'  Occipital-->R Sm' FCentral-->Central' L Sm-->Central' R Sm-->Central' Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital' Central-->Occipital'  0.0586 0.1708 0.2762 0.2027 0.1848 0.1650 0.5298 0.0389 0.3845  0.8509 0.6745 0.4660 0.6129 0.6129 0.6224 0.1282 0.9311 0.2560  III.4.3 MOV Central to FCentral Correlation Graphs  Figure A-6. MOV Central to FCentral Correlation Graph, 8-12 Hz.  112  Figure A-7. MOV Central to FCentral Correlation Graph, 8-30 Hz.  Figure A-8. MOV Central to FCentral Correlation Graph, 13-30 Hz.  113  III.4.4 MOV Correlations Band 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta'  Connections L Sm-->Fcentral' R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm' FCentral-->R Sm' L Sm-->R Sm' Central-->R Sm' Occipital-->R Sm' FCentral-->Central' L Sm-->Central' R Sm-->Central' Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital' Central-->Occipital'  R 0.1069 0.5179 0.5157 0.8611 0.2133 0.0319 0.1757 0.0918 0.0437 0.1096 0.4495 0.0804 0.2637 0.6103 0.3223 0.0752 0.1182 0.5369 0.1304 0.2297  p 0.7884 0.2512 0.1336 0.2789 0.5810 0.9883 0.5924 0.8385 0.8551 0.7540 0.1952 0.8343 0.4571 0.2664 0.5073 0.8300 0.7187 0.3454 0.7280 0.4993  Band 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha'  Connections L Sm-->Fcentral' R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm' FCentral-->R Sm' L Sm-->R Sm' Central-->R Sm' Occipital-->R Sm' FCentral-->Central' L Sm-->Central'  R 0.1344 0.1228 0.7486 0.3841 0.0995 0.0146 0.2510 0.0313 0.1785 0.0632 0.4886 0.0555 0.1423 0.7271  p 0.7317 0.7540 0.0215 0.3150 0.7962 0.9089 0.4638 0.9073 0.5941 0.8367 0.1516 0.8658 0.6702 0.2637  114  'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha'  R Sm-->Central' Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital' Central-->Occipital'  0.3415 0.1708 0.2122 0.4360 0.0318 0.2522  0.5180 0.6356 0.5292 0.3383 0.9242 0.4518  Band 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta'  Connections L Sm-->Fcentral' R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm' FCentral-->R Sm' L Sm-->R Sm' Central-->R Sm' Occipital-->R Sm' FCentral-->Central' L Sm-->Central' R Sm-->Central' Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital' Central-->Occipital'  R 0.0760 0.2553 0.9508 0.1115 0.0129 0.0247 0.3883 0.1644 0.1599 0.0665 0.4642 0.0592 0.3885 0.3679 0.0750 0.1611 0.1826 0.2802 0.0443 0.3271  p 0.8787 0.5604 0.0070 0.8129 0.9593 0.9130 0.2649 0.6443 0.6314 0.8910 0.1893 0.8413 0.4758 0.3464 0.8767 0.6529 0.5859 0.4624 0.8996 0.3438  Band 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta'  Connections L Sm-->Fcentral' R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm' FCentral-->R Sm' L Sm-->R Sm'  R 0.1292 0.1698 0.9282 0.0009 0.0137 0.0235 0.3615 0.1366 0.1696 0.0358  p 0.7849 0.7176 0.0083 0.9608 0.9793 0.9101 0.2955 0.7029 0.6127 0.9592  115  'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta'  Central-->R Sm' Occipital-->R Sm' FCentral-->Central' L Sm-->Central' R Sm-->Central' Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital' Central-->Occipital'  0.4826 0.0635 0.2550 0.4105 0.1442 0.1647 0.1892 0.3171 0.0413 0.3064  0.1688 0.8369 0.6194 0.3113 0.7573 0.6472 0.5703 0.4155 0.9037 0.3688  III.5 Quantitative Analysis- Correlation of PDC connectivity with PD subject gender III.5.1 PRE correlations Band 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta'  Connections L Sm-->Fcentral' R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm' FCentral-->R Sm' L Sm-->R Sm' Central-->R Sm' Occipital-->R Sm' FCentral-->Central' L Sm-->Central' R Sm-->Central' Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital' Central-->Occipital'  R 0.431 0.304 0.356 0.328 0.294 0.380 0.304 0.330 0.304 0.680 0.349 0.365 0.328 0.273 0.330 0.366 0.302 0.380 0.304 0.304  p 0.161 0.460 0.231 0.681 0.674 0.899 0.985 0.715 0.573 0.043 0.899 0.396 0.517 0.354 0.502 0.440 0.619 0.771 0.879 0.518  Band 'alpha'  Connections L Sm-->Fcentral'  R 0.322  p 0.299  116  'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha'  R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm' FCentral-->R Sm' L Sm-->R Sm' Central-->R Sm' Occipital-->R Sm' FCentral-->Central' L Sm-->Central' R Sm-->Central' Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital' Central-->Occipital'  0.299 0.378 0.299 0.299 0.315 0.304 0.304 0.304 0.655 0.371 0.304 0.522 0.304 0.304 0.364 0.304 0.396 0.304 0.304  0.360 0.208 0.959 0.734 0.895 0.933 0.777 0.438 0.074 0.582 0.758 0.103 0.525 0.855 0.364 0.314 0.309 0.910 0.448  Band 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta'  Connections L Sm-->Fcentral' R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm' FCentral-->R Sm' L Sm-->R Sm' Central-->R Sm' Occipital-->R Sm' FCentral-->Central' L Sm-->Central' R Sm-->Central' Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital'  R 0.431 0.379 0.346 0.304 0.304 0.288 0.304 0.304 0.436 0.578 0.512 0.365 0.680 0.299 0.304 0.405 0.426 0.285 0.330  p 0.744 0.640 0.373 0.469 0.533 0.712 0.695 0.429 0.168 0.113 0.131 0.303 0.032 0.654 0.885 0.197 0.161 0.513 0.941  117  'beta'  Central-->Occipital'  0.304  0.547  Band 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta'  Connections L Sm-->Fcentral' R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm' FCentral-->R Sm' L Sm-->R Sm' Central-->R Sm' Occipital-->R Sm' FCentral-->Central' L Sm-->Central' R Sm-->Central' Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital' Central-->Occipital'  R 0.391 0.368 0.316 0.304 0.304 0.315 0.304 0.304 0.401 0.623 0.442 0.345 0.694 0.299 0.304 0.379 0.372 0.330 0.330 0.304  p 0.906 0.510 0.331 0.555 0.574 0.803 0.748 0.498 0.211 0.094 0.176 0.447 0.028 0.616 0.873 0.226 0.182 0.429 0.975 0.509  III.5.2 MOV Correlations Band 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta'  Connections L Sm-->Fcentral' R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm' FCentral-->R Sm' L Sm-->R Sm' Central-->R Sm' Occipital-->R Sm'  R 0.355 0.304 0.304 0.349 0.274 0.354 0.304 0.304 0.304 0.481 0.349 0.365  118  p 0.214 0.464 0.343 0.488 0.465 0.841 0.727 0.985 0.916 0.129 0.956 0.235  'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta' 'theta'  FCentral-->Central' L Sm-->Central' R Sm-->Central' Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital' Central-->Occipital'  0.311 0.330 0.304 0.499 0.302 0.328 0.294 0.304  0.595 0.502 0.437 0.111 0.574 0.468 0.877 0.496  Band 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha' 'alpha'  Connections L Sm-->Fcentral' R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm' FCentral-->R Sm' L Sm-->R Sm' Central-->R Sm' Occipital-->R Sm' FCentral-->Central' L Sm-->Central' R Sm-->Central' Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital' Central-->Occipital'  R 0.377 0.369 0.299 0.304 0.282 0.282 0.304 0.304 0.304 0.304 0.337 0.304 0.349 0.304 0.299 0.477 0.304 0.489 0.304 0.304  p 0.321 0.428 0.532 0.662 0.359 0.894 0.845 0.942 0.552 0.198 0.470 0.574 0.399 0.738 0.780 0.131 0.305 0.162 0.951 0.416  Band 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta'  Connections L Sm-->Fcentral' R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm'  R 0.429 0.436 0.263 0.304 0.302 0.304 0.299 0.299  p 0.446 0.981 0.653 0.682 0.293 0.668 0.695 0.433  119  'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta' 'beta'  FCentral-->R Sm' L Sm-->R Sm' Central-->R Sm' Occipital-->R Sm' FCentral-->Central' L Sm-->Central' R Sm-->Central' Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital' Central-->Occipital'  0.490 0.304 0.508 0.419 0.330 0.304 0.304 0.432 0.397 0.673 0.330 0.304  0.135 0.253 0.174 0.268 0.317 0.918 0.902 0.161 0.193 0.142 0.971 0.526  Band 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta' 'alpha-beta'  Connections L Sm-->Fcentral' R Sm-->FCentral' Central-->FCentral' Occipital-->FCentral' FCentral-->L Sm' R Sm-->L Sm' Central-->L Sm' Occipital-->L Sm' FCentral-->R Sm' L Sm-->R Sm' Central-->R Sm' Occipital-->R Sm' FCentral-->Central' L Sm-->Central' R Sm-->Central' Occipital-->Central' FCentral-->Occipital' L Sm-->Occipital' R Sm-->Occipital' Central-->Occipital'  R 0.429 0.436 0.263 0.304 0.302 0.304 0.299 0.299 0.490 0.304 0.508 0.419 0.330 0.304 0.304 0.432 0.397 0.673 0.330 0.304  p 0.446 0.981 0.653 0.682 0.293 0.668 0.695 0.433 0.135 0.253 0.174 0.268 0.317 0.918 0.902 0.161 0.193 0.142 0.971 0.526  120  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0071053/manifest

Comment

Related Items