ROCKEFELLER MEDICAL URRARY INSTITUTE OF N . ‘ROLGG .’. THE NATIONAL HOS 'iTAL. QUEEN SQUARE, I.ONDOM WCiN

POSITRON EMISSION TOMOGRAPHY INVESTIGATION OF CORTICAL MALFORMATIONS CAUSING EPILEPSY

Mark Philip Richardson MA BM BCh MRCP INSTITUTE OF NEUROLOGY

Thesis for examination of Doctor of Philopsophy UNIVERSITY OF LONDON

1 9 93 ProQuest Number: 10017451

All rights reserved

INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. uest.

ProQuest 10017451

Published by ProQuest LLC(2016). Copyright of the Dissertation is held by the Author.

All rights reserved. This work is protected against unauthorized copying under Title 17, United States Code. Microform Edition © ProQuest LLC.

ProQuest LLC 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 48106-1346 For Emma and Ben Preface

This thesis represents the culmination of four years work; three as a full-time Research Fellow of the Institute of Neurology and a further year completing this work while undertaking the duties of Specialist Registrar in Neurology at the Hammersmith Hospital, London.

The work was funded by a Project Grant from the Medical Research Council, UK, held by Dr. John Duncan and Dr. David Fish of the Epilepsy Research Group of the Institute of Neurology, Queen Square, London and Professor David Brooks of the MRC Cyclotron Unit, Hammersmith Hospital, London. Their foresight enabled me to undertake a rewarding project in the best possible environment. Dr. John Duncan and Professor David Brooks acted as co-supervisors. John Duncan was enormously generous with his time and encouragement; his inspiration provided the driving force behind my thesis. I will always be deeply grateful for his introduction to an approach to research work which is both patient-centred and truly rigorous. His humour and friendship have sustained me through challenging times. David Brooks provided leadership and direction; his views, coming from a perspective outside the world of clinical epilepsy, have been important. A measure of my respect for him has been my desire to work for him as his Registrar after completing my full-time research work.

The experimental work was based at the MRC Cyclotron Unit. PET research is very much a team effort; at the MRC Cyclotron Unit I have had the privilege to work with the best methods group and technical support staff available. In particular, the Radiochemistry Section, under Mr. Ian Watson, has virtually unfailingly provided a faultless supply of radiotracers; the Blood Analysis Laboratory, under Dr. Safiye Osman, has provided essential data; the Methods Section, particularly Dr. Vincent Cunningham, has provided data analysis techniques which have enabled our investigations to proceed; and the Radiographers under Mr. Andrew Blyth have provided good-humoured and excellent technical support and patient protection within the scanning environment. Dr. John Stevens, Consultant Radiologist at the National Hospital for Neurology, Queen Square, reported the vast majority of the MRI studies through which our patients were identified and assigned to study groups; Dr. David Fish, Professor Simon Shorvon, Dr. Ley Sander and Dr. Sheelagh Smith of the National Hospital for Neurology generously referred patients for PET studies; Dr. Jenny Coull and Dr. Paul Grasby of the MRC Cyclotron Unit provided the normal control data and the software for the Rapid Visual Information Processing paradigm described in Chapter 11 ; Dr. Harri Jenkins of the MRC Cyclotron Unit supplied a copy of the software for the Motor Sequence Learning paradigm described in Chapter 11 ; Dr. Sanjay Sisodiya and Dr. Samantha Free of the Epilepsy Research Group of the Institute of Neurology undertook the MRI segmentation described in Chapter 9; finally I would like to thank Dr. Karl Friston and his colleagues for their encouragement and stimulation in the evolution of newer analysis techniques. To all of these individuals I owe the deepest gratitude.

The Epilepsy Research Group at the MRC Cyclotron Unit has had changing personnel over the last few years. Prior to my arrival, Dr. Peter Bartenstein and Dr. Martin Prevett undertook ground-breaking work which provided a momentum allowing my investigations to proceed rapidly. Dr. Matthias Koepp worked alongside me on parallel projects throughout my entire time at the MRC Cyclotron Unit. His contribution, through shared problems and challenges, lengthy discussions, mutual support and just simply being there when help was needed, was possibly the most valuable.

The volunteers for the studies were patients of the epilepsy clinics at Queen Square and normal subjects recruited from friends, relatives and acquaintances of the investigators and patients. I am profoundly grateful and honoured to have had the privilege of working with these selfless and generous volunteers. Their contribution was the most important of all.

Finally, my wife Emma and son Benjamin (born half way through this work) have endured the absence and unavailability of their husband and father during many evenings and weekends. The completion of this thesis marks both an end and a beginning; and also an opportunity to record my love for them. Thank you for your support. Abstract

Positron emission tomography (PET) was employed to investigate the central benzodiazepine receptor (cBZR) system and cerebral blood flow in patients with malformations of cortical development (MCD) and other syndromes associated with epilepsy.

Abnormalities of cortical cBZR binding were shown in 11 of 14 patients with MCD; 7 had abnormalities beyond the lesions seen with high-resolution MRI. These abnormalities consisted of both increases and decreases in cBZR binding.

All six patients with acquired non-progressive cerebral injury had reduced cBZR binding in the lesion seen on MRI; in 2 cases the region of reduced cBZR was more extensive. All six patients with dysembryoplastic neuroepithelial tumour had reduced cBZR density in the region of brain affected on MRI. One subject had a single remote region of increased cBZR.

Thirteen of 18 patients with extratemporal localisation-related epilepsy and normal MRI had regions of abnormal cortical cBZR; 10 had regions of increased cBZR, 6 had regions of decreased cBZR and 3 had both regions of increased and decreased cBZR. Six of 10 patients with unilateral frontal seizure onset had an abnormality in the implicated frontal lobe.

Five of 10 patients with MCD had abnormalities of MRI volumetry; only these same five patients had abnormalities of regional grey matter volume using a voxel based automated technique.

Six of 10 patients with MCD showed regions of cerebral cortex with disproportionate cBZR binding compared with local grey matter volume by integrating PET and MRI data at the voxel level.

Changes in cBZR were seen in 10 patients with unilateral RLE and normal MRI in the cortical-basal ganglia motor system. Some regions showed abnormalities of cBZR correlating with seizure frequency.

Eight of 10 patients with MCD showed activation of affected brain regions with appropriate tasks. In addition, there was a significant alteration in the overall activation pattern in five patients, compared with a normal control group. Contents

P reface 3

A b stract 5

Contents 6

Tables 13

Illustrations 14

Abbreviations 16

C hapter 1. Background 18 1.1. MALFORMATIONS OF CORTICAL DEVELOPMENT 1 8 1.1.1 Normal brain development 18 1.1.2. Abnormal brain development 21 1.1.3. Specific syndromes of MCD 24 1.1.3.1. LIssencephaly and pachygyria 24 1.1.3.2. Subependymal nodular heterotopia 25 1.1.3.3. Band heterotopia (“double cortex syndrome”) 26 1.1.3.4. Hemimegalencephaly 26 1.1.3.5. Polymicrogyria 27 1.1.3.6. Schizencephaly 28 1.1.3.7. Focal grey matter heterotopia 28 1.1.3.8. Focal cortical dysplasia 28 1.1.3.9. Dysembryoplastic neuroepithelial tumour 32 1.1.4. Treatment of malformations of cortical development 33 1.1.4.1. Drug treatment 3 3 1.1.4.2. Surgery 33 1.2. STRUCTURE OF THE NORMAL CEREBRAL CORTEX 35 1.2.1. Neocortex 35 1.2.2. Hippocampus 36 1.3. THE GABAERGIC SYSTEM 37 1.3.1. GABA, GABA receptors and benzodiazepine receptors 37 1.3.2. GABAergic neurons 40 1.4. MECHANISMS OF EPILEPTOGENESIS 41 1.4.1. General principles and terminology 41 1.4.2. Specific models 42 1.4.2.1. Normal hippocampus 43 1.4.2.2. Kindling 43 1.4.2.3. Intraventricular kainic acid 45 1.4.2.4. Previous limbic status epilepticus 46 1.4.2.5. Normal neocortex 47 1.4.2.6. Penicillin and bicucculline in rat neocortex 48 1.4.2.7. GABA withdrawal syndrome, rat neocortex 48 1.4.2.8. Cobalt focus 49 1.4.2.9. Chronic neocortical injury, rat 50 1.4.2.10. Alumina gel, monkey neocortex 51 1.4.2.11. GABA withdrawal syndrome in the baboon 52 1.4.2.12. Subcortical networks 52 1.4.3. Evidence from man 54 1.4.3.1. Temporal lobe epilepsy 55 1.4.3.2. Neocortical epilepsy 56 1.4.4. Conclusions 57

C hapter 2. Rationale for using PET to study localisation-related epilepsy 67 2.1. SPECT 67 2.1.1. Blood flow 67 2.1.2 Benzodiazepine receptors 70 2.2. PET 71 2.2.1. Cerebral metabolic rate 71 2.2.2. Opioid receptors 75 2.2.3. Other ligands 75 2.2.4. Benzodiazepine receptors 75 2.2.4.1. Alcohol 75 2.2.4.2. Neurodegenerative conditions 76 2.2.4.3. Idiopathic recurrent stupor 76 2.2.4.4. Epilepsy 76 2.3. OTHER FUNCTIONAL STUDIES IN MCD 78 2.4. AIMS 80

Chapter 3. Methods 1: Principles of positron emission tomography 82 3.1. PRINCIPLES OF PET 82 3.2. nC-FLUMAZENIL PET 85 3.2.1. Ligands for benzodiazepine receptor imaging 86 3.2.2. Parameters 86 3.2.3. Parameter modelling 87 3.2.3.1. Single scan studies 87 3.2.3.2. Multiple scan studies 89 3.2.4. Input function 91 3.2.5. Partial volume effect 92 3.2.6. Scanner field of view 92 3.2.7. Identification of structures and comparison of scans 93 3.2.7.1. Region of interest methods 93 3.2.7.2. Voxel-based methods 93 3.3. HgTSo PET 95

Chapter 4. Methods II 104 4.1. Subjects 104 4.2. Scanning protocol 104 4.3. T^C-flumazenil scans 105 4.3.1. Derivation of plasma input function 105 4.3.2. Data analysis 106 4.3.3. Production of parametric maps 106 4.4. Flg^^O scans 106 4.5. Coregistration with MRI 106

Chapter 5. Benzodiazepine receptors in malformations of cortical development 108 5.1. INTRODUCTION 108 5.2. METHODS 110 5.2.1. Subjects 110 5.2.2. Scanning protocol 110 5.2.3. Derivation of plasma input function 110 5.2.4. Data analysis 110 5.2.5. Production of parametric maps 110 5.2.6. Coregistration with MRI 111 5.2.7. Statistical Parametric Mapping 111 5.3. RESULTS 112 5.4. DISCUSSION 114 C hapter 6. Benzodiazepine receptors in localisation-related epilepsy due to acquired cerebral insult 126 6.1. INTRODUCTION 126 6.2. METHODS 128 6.2.1. Subjects 128 6.2.2. Scanning protocol 128 6.2.3. Derivation of plasma input function 128 6.2.4. Data analysis 128 6.2.5. Production of parametric maps 128 6.2.6. Coregistration with MRI 128 6.2.7. Statistical Parametric Mapping 128 6.3. RESULTS 129 6.4. DISCUSSION 130

C hapter 7. Benzodiazepine receptors in dysembryoplastic neuroepithelial tumour 134 7.1. INTRODUCTION 134 7.2. METHODS 135 7.2.1. Subjects 135 7.2.2. Scanning protocol 135 7.2.3. Derivation of plasma input function 135 7.2.4. Data analysis 135 7.2.5. Production of parametric maps 135 7.2.6. Coregistration with MRI 135 7.2.7. Statistical Parametric Mapping 135 7.3. RESULTS 136 7.4. DISCUSSION 137

C hapter 8. T ^C-flumazenil PET in extratemporal epilepsy with normal MRI 141 8.1. INTRODUCTION 141 8.2. METHODS 1 43 8.2.1. Subjects 143 8.2.2. Scanning protocol 143 8.2.3. Derivation of plasma input function 143 8.2.4. Data analysis 143 10

8.2.5. Production of parametric maps 143 8.2.6. Coregistration with MRI 143 8.2.7. Statistical Parametric Mapping 143 8.3. RESULTS 144 8.3.1. Patients with unilateral frontal seizure onset 144 8.3.2. Patients without clear evidence of unilateralfrontal seizure onset 144 8.4. DISCUSSION 146

C hapter 9. A voxel-level method for the comparison of structural and functional imaging data 151 9.1. INTRODUCTION 151 9.2. METHODS 153 9.2.1. Subjects 153 9.2.2. Scanning protocol - PET 1 53 9.2.3. Production of parametric maps of ^^C-flumazenil binding 1 53 9.2.4. Scanning protocol - MRI 1 53 9.2.5. Segmentation of MRI and MRI volumetry 1 53 9.2.6. Data analysis 1 54 9.2.7. Spatial normalisation of PET and MRI 1 54 9.2.8. Statistical Parametric Mapping 1 54 9.2.9. Analysis of structural data using SPM 1 55 9.2.10. Analysis of ^^C-flumazenil PET data 1 55 9.2.11. Comparison of MRI and PET images 1 55 9.3. RESULTS 1 57 9.3.1. SPM analysis of MRI structural data and comparison with region-based morphometries 157 9.3.2. Comparison of structural and functional images 157 9.4. DISCUSSION 1 60 9.4.1. Voxel-based analysis of MRI using SPM 160 9.4.2. Voxel-based comparison of MRI and PET in MCD using SPM 160 9.4.3. Voxel-based comparison of MRI and PET - general methodological considerations 161

Chapter 10. The cortical-subcortical GABAy^-BZR system infrontal lobe seizures 169 10.1. INTRODUCTION 169 10.2. METHODS 170 10.2.1. Subjects 170 11

10.2.2. Scanning protocol 170 10.2.3. Derivation of plasma input function 170 10.2.4. Data analysis 170 10.2.5. Production of parametric maps 170 10.2.6. Statistical Parametric Mapping 170 10.3. RESULTS 172 10.4. DISCUSSION 173

Chapter 11. Cerebral activation in malformations of cortical development 179 11.1. INTRODUCTION 179 11.2. METHODS 181 11.2.1. Subjects 181 11.2.2. PET scanning 181 11.2.3. Experimental designs 182 11.2.3.1. Occipitalcortex activation 182 11.2..3.2. Dominant left frontal cortex activation 182 11.2.4. Data analysis 183 11.2.5. Statistical Parametric Mapping 183 11.2.5.1. Significant changes in rCBF in the whole brain of each patient 184 11.2.5.2. Comparison of individual activation pattern with the control group 184 11.3. RESULTS 186 11.3.1. Occipital cortex activation 186 11.3.1.1. Significant changes in rCBF in the whole brain of each patient 186 11.3.1.2. Comparison of individual activation patterns with the control group 186 11.3.2. Left frontal cortex activation 187 11.3.2.1. Significant changes in rCBF in the whole brain of each patient 187 11.3.2.2. Comparison of individual activation pattern with the control group 187 11.4. DISCUSSION 189

Chapter 1 2. Discussion 203 12.1. SUMMARY OF THE MAIN FINDINGS 203 12.2. NEUROBIOLOGICAL IMPLICATIONS 204 12.3. FUTURE STUDIES 206 12

References 208

List of publications 256

Independent contribution

PET cannot be undertaken by a single individual. Many have contributed to the data contained in this thesis. Specifically: - T^C-flumazenil was produced by the Radiochemistry section, MRC Cyclotron Unit - radiotracer purity was checked by Quality Control, MRC Cyclotron Unit - PET scans were supervised by Radiographers of the MRC Cyclotron Unit - images were made available in digital format courtesy of the Computing Unit, MRC Cyclotron Unit - analysis of blood for radioactivity and metabolites was undertaken by the Blood Laboratory, MRC Cyclotron Unit - the spectral analysis software was written by Dr. Vincent Cunningham, MRC Cyclotron Unit - SPM software was used courtesy of the Wellcome Department of Cognitive Neurology, Institute of Neurology - Dr. John Stevens, Consultant Radiologist at the National Hospital for Neurology, Queen Square, reported the vast majority of the MRI studies through which the patients were identified and assigned to study groups - Dr. David Fish, Professor Simon Shorvon, Dr. Ley Sander and Dr. Sheelagh Smith of the National Hospital for Neurology generously referred patients for PET studies - Dr. Jenny Coull and Dr. Paul Grasby of the MRC Cyclotron Unit provided the normal control data and the software for the Rapid Visual Information Processing paradigm described in Chapter 11 - Dr. Harri Jenkins of the MRC Cyclotron Unit supplied a copy of the software for the Motor Sequence Learning paradigm described in Chapter 11 - Dr. Sanjay Sisodiya and Dr. Samantha Free of the Epilepsy Research Group of the Institute of Neurology undertook the MRI segmentation described in Chapter 9 13 Tables

Table 5.1. Clinical details of the 14 patients with MCD 118 Table 5.2. Results of SPM analysis 119 Table 6.1. Clinical features and results of SPM analysis 132 Table 7.1. Clinical features and results of SPM analysis 139 Table 8.1. Clinical features of 10 patients with probably unilateral onset frontal lobe seizures and normal MRI 148 Table 8.2. Clinical features of 8 patients without clear evidence of unilateral onset frontal lobe seizures and normal MRI 149 Table 9.1. Clinical features of the 10 patients with MCD 165 Table 11.1. Clinical details of patients with MCD 193 Table 11.2. Atypical cortical organisation in the patients with MCD 194 Table 11.3. Summary of the cerebral activation findings in10 patients with MCD 195 14 ustrations

Figure 1.1. Cartoon showing a simplified scheme of embryonic neuronal migration 59 igure 1.2. Lissencephaly 60 igure 1.3. Subependymal nodular heterotopia 60 igure 1.4. Band heterotopia 60 igure 1.5. Bilateral perisylvian polymicrogyria 61 igure 1.6. Schizencephaly 61 igure 1.7. Isolated heterotopic grey matter nodule 61 igure 1.8. Focal malformation of cortical development 62 igure 1.9. Dysembryoplastic neuroepithelial tumour (DNT) 62 igure 1.10. A simplified scheme of intracortical circuits, layers and cell-types in eocortex 63 igure 1.11. Schematic diagram of intrahippocampal connections 64 igure 1.12. Principal GABAergic and excitatory interactions between the frontal eocortex and the basal ganglia in man 65 igure 1.13. (a) Schematic representation of normal cortical connections (b) Schematic epresentation of cortical connections following cell loss 66 igure 3.1. Annihilation of a positron 97 igure 3.2. The two y-rays are emitted from the brain 97 igure 3.3. The PET camera 98 igure 3.4. Imaging in 2D and 3D 99 igure 3.5. Modelling of PET data 100 igure 3.6. Three dimensional representation of the 11 C-flumazenil molecule 101 igure 3.7. Two possible models of flumazenil distribution in brain 102 igure 3.8. Outline scheme of the method of spectralanalysis 103 igure 5.1. Subject TJ 120 igure 5.2. Subject AF 121 igure 5.3. Subject JW 122 igure 5.4. Subject RH 123 igure 5.5. Subject TP 124 igure 5.6. Average transformed MRI 125 igure 6.1. Subject SM 133 igure 7.1. Subject RIH 140 igure 8.1. Subject HD 150 igure 9.1. Cartoon illustrating schematically the method of voxel-by-voxel comparison 15 of Structural and functional data 166 Figure 9.2. Comparison of block volume and SPM analysis of MRIGM images 167 Figure 9.3. Results of the comparison of MRIGM and FMZVD images using SPM 168 Figure 10.1. Results of group analyses of LRE patients 176 Figure 10.2. Scattergraphs of seizure frequency versus FMZVD 177 Figure 10.3. A model of the GABAergic system and its influence on seizures 178 Figure 11.1. Summary of RVIP results 196 Figure 11.2. Summary of MSL results 197 Figure 11.3. Subject DL 198 Figure 11.4. Subject EM and subject SN 199 Figure 11.5. Subject TP and subject MW 200 Figure 11.6. Subject RH and subject KS 201 Figure 11.7. Subject KT 202 16 Abbreviations

Explanation First use

MCD malformation(s) of cortical development 1.1 MRI magnetic resonance imaging 1.1 EEC electroencephalogram 1.1 HS hippocampal sclerosis 1.1 NJPA1 neuron-glial junctional polypeptide 1.1.1 GABA Y-aminobutyric acid 1.1.2 PAF platelet-activating factor 1.1.2 EPSP excitatory postsynaptic potential 1.1.2 IPSP inhibitory postsynaptic potential 1.1.2 GluR glutamate receptor 1.1.2 SSEP somatosensory evoked potential 1.1.3.1 IQ intelligence quotient 1.1.3.4 CMV cytomegalovirus 1.1.3.5 CT computed tomography 1.1.3.8 GFAP glial fibrillary acidic protein 1.1.3.8 PGP protein gene product 1.1.3.8 ECoG electrocorticography 1.1.3.8 DNT dysembryoplastic neuroepithelial tumour 1.1.3.9 CA cornu Ammonis 1.2.2 GAD glutamic acid decarboxylase 1.3.1 Ca^+ calcium ion 1.3.1 K+ potassium ion 1.3.1 ci- chloride ion 1.3.1 cBZR "central" benzodiazepine receptor 1.3.1 PET positron emission tomography 1.3.1 NMDA N-methyl-D-aspartate 1.4.2.1 mRNA messenger ribonucleic acid 1.4.2.2 AMPA a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid 1.4.2.2

^max available receptor density 1.4.2.8 Kd receptor affinity 1.4.3.2 rCBF regional cerebral blood flow 2.1.1 SPECT single photon emission computed tomography 2.1.1 HIPDM N,N,N'-trimethyl-N'-(2-hydroxy-3-methyl-5- iodobenzyl)-! ,3,propanediamine 2.1.1 17

HMPAO hexamethylpropyleneamine-oxime 2.1.1 ECD 99mTc-ethyl cysteinate dimer 2.1.1 TLE temporal lobe epilepsy 2.1.1 T8FDG ’ ^F-deoxyglucose 2.1.1 rCMRgluc regional cerebral metabolic rate for glucose 2.2.1 MAO monoamineoxidase 2.2.3

Vd volume of distribution 2.2.4.2 HPLC high pressure liquid chromatography 2.2.4.3 FWHM full width at half maximum 3.1 2D two-dimensional 3.1 3D three-dimensional 3.1 ROI region of interest 3.2.7.1 SPM statistical parametric mapping 3.2.7.2 SPGR spoiled gradient recalled 4.1 ARSAC Administration of Radioactive Substances Advisory Committee 4.1 AC-PC anterior commissure - posterior commissure 4.2 IRF impulse response function 4.3.3 FMZVD TT C-flumazenil volume of distribution 4.3.3 ANOOVA analysis of covariance 5.2.7 CSF cerebrospinal fluid 6.4 RVIP rapid visual information processing 11.2.3.1 MSL motor sequence learning 11.2.3.2

EPC epilepsia partialis continua 2.2.1 MRIGM MRI-defined cortical grey matter 9.2.5 FLE frontal lobe epilepsy 10.1 18 Chapter 1.

Background.

1.1. MALFORMATIONS OF CORTICAL DEVELOPMENT

Malformations of cortical development (MCD) originate during intrauterine development as a consequence of a variety of inherited and acquired insults, resulting in a wide spectrum of structural cortical anomalies and clinical syndromes. These disorders are rare but of importance in epilepsy practice. MCD is seen in 4.3-12.6% of adult seizure patients referred for magnetic resonance imaging (MRI) [1, 2]. MCD may underlie the important paediatric epilepsy syndromes West Syndrome and Lennox- Gastaut syndrome [3, 4]. The detection of localised slow activity on an electroencephalogram (EEC) should also prompt a search for MCD [5]. Furthermore, MCD may be present as a second lesion in 15% of patients with the commonest lesion underlying treatment-resistant localisation-related epilepsy, hippocampal sclerosis (HS) [6]; this second lesion may in some cases be related to failure to respond to surgery for HS. Although epilepsy, mental retardation and focal neurological and cognitive deficits are all well-recognised in patients with malformations of cortical development, the pathophysiological and neurochemical correlates of this group of disorders are little studied.

1.1.1 Normal brain development

The sequence of events in the ontogenesis of the brain in man has been extensively described [7-9]. The neuroectoderm forming the neural plate appears along the dorsal midline of the embryo on the 16th day post conception in man. The neural plate folds to form a tube which is complete by about the 26th day. The rostral portion of the neural tube differentiates extensively to form the brain. Three primary brain vesicles appear initially at around day 28 - the prosencephalon, mesencephalon and rhombencephalon. By about day 35 the first and third vesicles have each matured into two vesicles. The five secondary brain vesicles comprise the telencephalon (which will become the cortex, striatum and olfactory system), diencephalon (which will become the thalamus, hypothalamus and subthalamus), mesencephalon (which will become the midbrain), metencephalon (which will become the pons and cerebellum) and myelencephalon 19

(which will became the medulla oblongata).

The vast majority of neurons and a large number of glial cells are derived by proliferation of the neuroepithelial cells in the germinal matrix at the ventricular surface. The rate of cell division in the forebrain ventricular surface germinal zone is greatest during the second trimester. A great excess of neuroblasts is formed, greatly more in number than the final number of neurons in the mature brain. Programmed cell death occurs in the brain as in other tissues during embryonic and foetal life. A major stimulus to undergo cell death is the failure to form appropriate synaptic contacts [10]. Four anatomically distinct migrations of neuroblasts into their final anatomical positions have been identified [7]: via the corpus pontobulbare to form the pontine and medullary nuclei; via the corpus gangliothalamicus to form the basal ganglia and thalamus; migration in the cerebellum; and migration from the periventricular germinal matrix to the cerebral cortex. Malformations of cortical development involve errors in this fourth migration path.

Building on work in the fly Drosophila, a number of proliferation and cell-fate determination factors are beginning to emerge in man, such as Notch 1, inherited variably during neuroblast division depending on the orientation of the line of cell cleavage [11] and Numb, distributed towards the apical area of mitotic cells [12]. The precise role of these factors in MCD in man remains to be elucidated.

The future cortical neurons migrate in an organised series of waves such that the cells destined to be in the deepest layers of the cortex migrate first, with more superficial layers migrating subsequently and passing through the layers of cells already arrived in order to reach the cortical surface (figure 1.1)[13]. The neurons are guided to their cortical targets by radial glial fibres which span the entire hemisphere from the ventricle to the pia [7]. This radial migration has been long recognised [14] but more recently extensive lateral dispersion during migration has been demonstrated [15, 16]. Multiple factors may be responsible for migration. In particular, the mechanisms of initial attachment of neurons to radial glia and the signals which lead to detachment and appropriate positioning within the cortical layers have been shown to be crucial in animal models. In the reefer mouse, which shows a histological pattern somewhat reminiscent of lissencephaly, failure of detachment of neurons from glia has been demonstrated [17]. In contrast, in the weaver mutant mouse (which has a cerebellar migration fault) granule cells fail to attach to radial glia [18]. The processes of adhesion during migration are apparently robust and similar in different brain areas; hippocampal granule cells migrate normally along cerebellar radial glial fibres in vitro, and cerebellar granule cells migrate along hippocampal glia, in both cases the 20

neuronally expressed surface marker astrotactin appearing critical [19]. Neurons appear to be committed to specific cortical layers before migration commences [20]. If attachment to glia has occurred and the cell is already committed to a destination before migration, the critical signals for ensuring appropriate lamination should be cortically-derived “stop” signals [21]. A monoclonal antibody to neuron-glial junctional polypeptide (NJPAl) identified the presence of this protein along radial glial fibres of developing rat cortex. NJPAl immunoreactivity disappeared where radial glial fibres entered the marginal zone at the distal side of the developing cortex. In the presence of the antibody, migrating neurons were greatly slowed and tended to detach from the radial glia, suggesting NJPAl provides part of the mechanism for maintaining contact between migrating neurons and radial glia. The absence of this factor at the distal end of radial glial fibres may be part of the “stop” signal [22].

The establishment of the pial-glial barrier is one of the earliest histogenetic events in neurogenesis. This is accomplished by coordinated interaction among the processes of radial glia, various extracellular matrix components, and mesenchymal cells at the pial surface, with the formation of a basement membrane. The pial-glial barrier and the basement membrane appear to be critical to the migration and final positioning of neurons and to the differentiation of the laminar cortical pattern within the developing cortex [23]. Before the migrating future cortical neurons reach the outside of the developing brain, the outer layers consist of a so-called marginal zone, inside this the cortical preplate, below this the subplate. Migrating cells penetrate the subplate and come to rest between the layers of older cells, forming the cortical plate (figure 1.1). Prominent in the marginal zone are very large neurons called Cajal-Retzius cells. Cajal-Retzius cells were found evenly distributed throughout the murine cerebral cortex during development. Neuronal migration has been shown to be dependent on glutamate receptors; it has been proposed that Cajal-Retzius cells releasing glutamate may direct migrating neuroblasts toward the marginal lamina, therefore creating the "inside-out" sequence of cortical development [24]. A secreted factor structurally related to other known adhesion molecules in the central nervous system, which has been termed reelin, has been shown to be defective in the reeler mouse; this factor appears to be confined to Cajal-Retzius cells [25]. Migrating neurons from mice with this defect fail to penetrate the subplate and accumulate below it producing a lissencephalic-like cortex. A mouse with similar phenotype, scrambler, has normal reelin; it has been suggested the defect may be at the putative receptor for reelin [26]. A further defect in the reeler/scrambler system affects the gene for cdkS, a kinase involved in cell-cycle regulation [27].

In humans migration commences around weeks 6-8 [28, 29] and continues prolifically 21

until week 16, though marked migration occurs until week 25 and is not complete until 5 months postnatally; migration in the medulla oblongata is complete by 8 weeks of intrauterine life, but cerebellar migration continues until at least the end of the first postnatal year. During the first half of gestation most cells of the germinal matrix become neurons; in the second half of gestation most of the germinal cells become glia. Differentiation and migration of glia is characterised by retraction of the radial process of radial astroglial cells; these cells differentiate into astrocytes, oligodendrocytes and ependymal cells [30]. However, most astrocytes destined for the cortical layers are produced in the periventricular germinal zone after the migration of neurons and themselves migrate into the cortex [31 ]. This process is prolonged compared to neuronal differentiation and migration; astrocytes of the frontal lobe migrate centrifugally from 28 weeks gestation until 6 years of age [32].

1.1.2. Abnormal brain development

The wide range of cerebral malformations can be explained in terms of the stage of gestation at which the putative error in embryogenesis occurred [28, 29]. Disorders of neurulation, the formation of the neural tube and major brain vesicles occur earliest, at approximately 1 -4 weeks. These abnormalities include anencephaly, encephalocoele and meningomyelocoele. Midline malformations of the forebrain occur later at 4-8 weeks and include holoprosencephaly, arhinencephaly, septo-optic dysplasia, agenesis of the corpus callosum and colpocephaly. Errors of neuronal proliferation, especially at 8-20 weeks, may be responsible for hemimegelencephaly and hydranencephaly. The disorders of neuronal migration are due to errors occurring largely at 6-25 weeks. These disorders, which will be described in detail below, include the generalised disorders lissencephaly, band heterotopia and subependymal heterotopia, and focal disorders pachygyria, polymicrogyria, schizencephaly, focal grey matter heterotopia and focal cortical dysplasia.

Malformations of cortical development may be seen in association with a variety of clinical syndromes [29]. These include metabolic disorders such as Zellweger syndrome, Menke kinky-hair disease, Gm2 gangliosiosis and glutaric aciduria type II; chromosomal abnormalities including Miller-Dieker syndrome, 17p13.3 mutations, trisomy 13, 18 and 21, and deletion 4p; neurocutaneous disorders such as incontinentia pigmenti, neurofibromatosis I, hypomelanosis of Ito, tuberous sclerosis and epidermal naevus; neuromuscular disorders including Walker-Warburg syndrome, Fukuyama muscular dystrophy and dystrophia myotonica; intrauterine infection with cytomegalovirus, toxoplasma or syphilis; and environmental agents such as toxins or radiation. Despite 22 these numerous associations, the nature of the aetiological process in malformations of cortical development frequently remains elusive.

In some human cases, and in many animal models, specific aetiological factors may be demonstrated. The neuropathological examination of the brains of children prenatally exposed to methyl mercury revealed dysplasia of the cerebral and cerebellar cortex [33]. Using MRI, 11 patients with congenital cytomegalovirus infections involving the brain were retrospectively reviewed. Lissencephaly was found in four patients. Focal areas of dysplastic cortex, presumably poly microgyria, were found in five patients [34].

In the rat, methylmercury induces mitochondrial degeneration in the endothelium of the cerebral capillaries with subsequent haemorrhages which interfere with neuronal migration [33]. A wide variety of types of cortical malformation were produced following focal cortical freezing, electrocoagulation, focal cortical aspiration or gentle brushing of uncovered meninges, in newborn or 1- to 3-day-old rats [35]. Similar procedures from postnatal day 4 onwards, at a time when a reactive astrogliosis is possible, produced cavitating infarcts and tissue scars rather than malformative lesions. Lissencephaly was produced in ferrets by toxic exposure to methylazoxymethanol acetate given to the pregnant female [36]. These lissencephalic ferrets were shown to have lower seizure threshold and abnormal interictal EEG.

Exposure to a strong alkylating agent on a specific gestational day in the rat gives rise to histological abnormalities in the cerebral cortex of the rat pups [37]. These abnormalities are various, and are accompanied by cognitive defects and a lowered seizure threshold [38]. Redirected perforant and commissural connections of neurons in the hippocampus of such rats has been noted [39] in addition to deformed dendrites and reduced spine numbers on ectopic neurons in the hippocampus [40]. Furthermore, decreased expression of the GluR2 receptor gene and decreased GABA^ a1 subunit gene have been detected in this model [41], all of which may be related to increased seizure susceptibility. Fetal rats exposed to irradiation developed a variety of patterns of MCD. Experimental rats had prolonged ictal activity on EEG; none of the control rats demonstrated ictal activity [42].

The mouse Small eye (Sey) manifests a broad spectrum of neuronal migration disorders. The gene locus is situated on chromosome 2 and is a point mutation in the paired box (Pax)-6 gene, which contains paired-like and homoeobox domains and is a developmental control gene. Embryogénie studies have shown neurons with swollen and vacuolated cytoplasm, reduced formation of axons, and vacuolar degeneration in existing 23 axons, glial cells and radial glial fibres. Consequently, there is an impairment of the peripheral migration of putative neurons [43]. Using a neocortical culture, ionizing radiation in very low doses was shown to slow migration and reduce expression of the neural cell adhesion molecule N-CAM [44]. LIST, the gene defective in Miller-Dieker lissencephaly syndrome, encodes a subunit of brain platelet-activating factor (PAF) acetylhydrolase which inactivates PAF, a neuroregulatory molecule. LIST is strongly expressed in developing human cortex. The murine homologue, Lisi, is strongly expressed in developing and adult cortex [45].

Despite considerable evidence of functional abnormalities in experimental models of MCD, the cortical neurons themselves may be able to function normally despite malpositioning. In the reeler mouse, the normal intrinsic firing patterns, action- potential shapes, resting membrane potentials, input resistances, and evoked excitatory postsynaptic potentials (EPSPs) and inhibitory postsynaptic potentials (IPSPs) imply that cortical neurons develop the membrane and synaptic properties appropriate to their function, despite being malformed and mispositioned [46].

The functional anomalies in MCD may therefore be due to abnormal connectivity rather than abnormal cellular properties. During neuronal maturation, axons are produced by neurons before dendrites. If a neuron is oriented in the wrong direction following abnormal migration, the axon direction can shift orientation up to 180°. Dendrites cannot reorient, and may be greatly abnormal in preterm infants, trisomy 13, other “unclassified” mental retardation and metabolic encephalopathies, even in the presence of normal neuroimaging [47, 48]. Transplanted embryonic putative neurons migrate into areas of neuronal degeneration in adult mouse cortex, subsequently differentiating into large neurons with correct orientation of dendrites and axons [49]. ^H-thymidine birthdating studies showed that layer 5 callosal and subcortically projecting neurons were generated at the same stage of corticogenesis, implying that from early stages of axon extension, callosal and subcortically projecting cells are distinct classes of neurons and initiate growth toward class-specific and non-overlapping sets of targets [50]. Malpositioned neurons unable to send an axon to a predetermined target are therefore likely to produce abnormal connections. Prenatal injury in foetal animals produces non­ contiguous distant gyral anomalies including some in the opposite hemisphere [51] and abnormal connectivity in rat neocortex following prenatal injury has been demonstrated [52]. Further, the connections demonstrated between heterotopic nodules and other cortical areas [53] are highly likely to be most abnormal. Additional indirect evidence of abnormal connectivity is provided by the observation on MRI of distant, even contralateral, abnormalities of cortical grey matter volume in the majority of patients with localised MCD [54]. These lines of evidence point to the likelihood of functional 24 abnormality in both affected and apparently normal regions of brain in addition to any localised structural abnormality.

1.1.3. Specific syndromes of MCD

1.1.3.1. Lissencephaly and pachygyria

Lissencephaly describes the macroscopic appearance of a brain with absent cortical sulcation, though preserved major fissures (figure 1.2). Classically, lissencephaly consists of two types [55]: Type I, with histologically a four layered cortex, widened ventricles and failure of opercolation; Type II, in which no definite layers of cortex can be distinguished due to the entirely haphazard cortical architecture. Barkovich has recently proposed three further types of lissencephaly - microcephalia vera (type III), radial microbrain (type IV) and diffuse polymicrogyria (type V) [56]. Type I includes Miller-Dieker syndrome; type II includes Walker-Warburg syndrome. Pachygyria refers to a brain with diffusely thick cerebral cortex with poor, but not absent, sulcation. A deletion in chromosome IZplS.S has been demonstrated both in cases of Miller-Dieker syndrome [57] and isolated lissencephaly [58]. The gene product is a subunit of brain platelet-activating factor acetylhydrolase [59]. Persistence of vimentin immunoreactivity has been detected in lissencephalic cortex [60]; vimentin expression occurs in normal development only in cells before differentiation into neuronal and glial lineages, suggesting the error responsible for lissencephaly occurs before such differentiation and allows persistence of undifferentiated cells in mature lissencephalic cortex.

A number of features are typical of type I lissencephaly [28, 55, 61 ]. The four-layered cortex consists of a molecular layer, a disorganised outer cellular layer, a cell-sparse layer (possibly the result of necrosis) and a thick inner cellular layer. It is likely that the disruption of migration is at 3-4 months gestation. Other malformations may be associated, such as cerebellar or callosal agenesis, ventricular enlargement or heterotopias. Miller-Deiker or Norman-Roberts syndromes may be seen. Patients who do not fulfil the criteria for these syndromes often have microcephaly, infantile hypotonia, developmental delay, cognitive impairment and seizures from an early age, usually infantile spasms. EEG has been variously described as showing “delta-theta” rhythms [62], “rapid rhythms” [62] or “slow spike-wave pattern” [63]. Other electrophysiological measures may also be abnormal - somatosensory evoked potentials (SSEPs) recorded in ten patients after stimulation of the median nerve were abnormal in all [63]. MRI shows an 8-shaped brain with vertical sylvian fissures, absent or 25 minimal sulcation, thick cortex, reduced white matter volume with large (foetal-type rather than hydrocephalic) ventricles and failure of complete inversion of the hippocampi [64].

Type II lissencephaly is not an intracortical malformation but is the result of massive glial and neuronal ectopia in the leptomeninges [65]. This results from a failure of arrest of neuronal migration due to defects in the integrity of the pial/glial barrier. The resulting cerebral cortex appears entirely disorganised. Type II lissencephaly may be associated with hydrocephalus, infantile hypotonia, eye malformations, developmental delay, severe cognitive impairment and seizures from an early age [62]; EEG shows fast high-amplitude rhythms in 80% and MRI may show thick cortex, hydrocephalus and hypomyelination [55, 66].

Clinically, the syndrome of pachygyria is more heterogeneous. Patients with extensive pachygyria who did not have typical lissencephaly were found to have variable clinical and EEG features, but this group were almost all severely impaired neurologically, cognitively and had severe intractable epilepsy. Histology of a cortical biopsy showed cortical layers 5 and 6 could not be differentiated and that the white matter was poorly myelinated and contained clusters of heterotopic neurons. This syndrome, a congenital disorder of neuronal migration, with prolonged survival, may represent a mild form of lissencephaly; lissencephalic and pachygyric areas may coexist in the same patient [28, 66].

7 .7 .3 .2 . Subependymal nodular heterotopia

Subependymal nodular heterotopia consists of clusters of neurons around the ventricles, representing cells which have failed even to begin migration from the germinal zone (figure 1.3). Three described series [67-69] include 8, 13 and 7 patients with this condition respectively. Between 60 and 100% are bilateral; those few unilateral cases are predominantly right-sided. The great majority, 66-92%, are female and the majority have seizure onset after 10yrs. Other associated MCDs are exceptional. EEG is usually diffusely abnormal or even shows a pattern compatible with generalised epileptiform activity. MRI reveals material with cortical grey matter imaging characteristics surrounding the ventricles. Intelligence is usually normal with no neurological deficits. Familial cases have been described; in all the pedigrees only females were affected and in some cases did not have epilepsy [69]. Genetic linkage to markers in distal Xq28 has been shown [70]. The X-linkage of this disorder partly explains the phenotype: through the process of Lyonisation, some neurons will be normal and some abnormal. The normal cells migrate fully to the cortex; the abnormal cells 26 remain in the ventricular zone.

Histological studies have shown the presence of highly differentiated neurons with neuronal-glial relationships indistinguishable from normal grey matter and rich afferent innervation, though the origin of this innervation remains unclear [70].

1.1.3.3. Band heterotopia (*"double cortex syndrome”)

Band heterotopia, or “double cortex syndrome”, refers to the occurrence of a continuous, usually generalised or sometimes predominantly frontal, layer of material with macroscopic and MRI characteristics of cortex lying below the real cortex within the white matter (figure 1.4). Occurrence in males is exceptional [71, 72], though familial association with lissencephaly in males has been noted [73]. This X-linked lissencephaly has tentatively been mapped to chromosome Xq22 based on the observation of a single X-autosomal translocation in a girl [74]. It has been postulated that lissencephaly results in males because, as in the reeler mouse, no cells penetrate the subplate to form normal cortex, whereas in females the phenomenon of Lyonisation will give rise to some normal neurons which form a normal cortex and a population of affected neurons which accumulate beneath the subplate forming a heterotopic band.

The associated epileptic disorder varies in nature and degree of severity. Patients may present with infantile spasms, a Lennox-Gastaut syndrome, or other forms of secondary generalised or multifocal epilepsy [71]; partial seizures are most frequent [72]. Response to medical treatment is variable. Mental retardation is usually mild or moderate, and only rarely severe. It correlates with the type of epileptic syndrome, and is greater in patients with more disorganised cortex overlying the heterotopia [71 ]. EEG may be normal, show widespread theta and focal spikes [72], or even generalised spike- wave [28]. Other MRI anomalies were frequent, particularly loss of sulcation, ventricular enlargement and prolongation of T2 signal. Thicker bands and more pachygyria correlated with poorer development and greater likelihood of clinical presentation with symptomatic generalised epilepsy [72]. Histology showed normal layers 1-4, but layer 5 was not seen and layer 6 merged with the U-fibres of white matter; clusters of ganglion cells were seen beneath. Cortical thickness was usually normal [28].

1.1.3.4. Hemimegalencephaly

Hemimegalencephaly is a brain malformation characterized by gross cerebral asymmetry and MOD. Infants with the condition present with early seizures and severe 27 encephalopathy. MRI demonstrates unilateral hemispheric hypertrophy with lateral ventricle dilatation, abnormal gyral pattern, and a thick cortex on the enlarged side. There is an inverse relationship between the severity of abnormalities and the size of the hemisphere [28]. The usual presentation is with seizures in infancy; these may be unilateral with unilateral neurological signs. Some individuals with normal IQ and minimal neurologic dysfunction have been reported [28].

The pathology has been described in detail [60]. Tissues from three cases of hemimegalencephaly causing intractable seizures treated by cortical resection were studied using immunohistochemical, ultrastructural, and morphometric techniques. Severe cortical dysplasia was seen in all cases and included lesions best characterized as hemilissencephaly and polymicrogyria. Blurring of the cortex-white matter junction, the presence of large neuronal heterotopias, and neuronal cytomegaly were frequent observations. Immunohistochemical analysis demonstrated cellular colocalization of astrocytic markers, glial fibrillary acidic protein and vimentin, suggesting the presence of poorly differentiated cells with multiple phenotypic features.

7.7.3.5. Polymicrogyria

Polymicrogyria consists of a region of cortex with a disordered gyral pattern with multiple very small gyri; MRI shows thick cortex resembling pachygyria - the multiple small gyri often cannot be resolved (figure 1.5). Presentation is very variable and is often severe if diffuse. The usual histology is probably ischaemic midcortical laminar necrosis of layer 5; superficial to this are normal layers 2,3 and 4 [28]. Polymicrogyria can be unlayered and may be associated with other anomalies [75]; indeed, polymicrogyria may be the histology underlying hemimegalencephaly. Perfusion failure due to CMV, toxoplasma, syphilis or maternal hypoxia is suggested as a possible mechanism [76], though a familial predisposition has been noted [77]. The mechanism of production of excessive folding has been suggested to be that, with the middle layer of cortex destroyed, the superficial and deep layers of cortex are opposed to one another and have markedly different predetermined growth rates [78]. Alternatively, microgyria formation may be the consequence of brain repair mechanisms occurring during neuronal migration to the neocortex. Placement of a freezing probe on the skull of neonatal rats produced four-layered microgyria. Reactive astrocytes and macrophages arrived in the damaged area within 24 hours of the injury, and repair of the damaged tissue peaked within the first week. Damaged radial glial fibres regrew, and supragranular neurons migrated through this damaged area, also within the first week. The damaged cortex then began to assume its adult-like microgyric appearance [79]. 28

One clearly recognisable syndrome is the so-called congenital bilateral perisylvian syndrome [80] first described in adult twins [81]. This consists of congenital pseudobulbar palsy, intellectual delay and bilateral opercular or perisylvian poly microgyria. 80% have generalised slow spike-wave and 20% focal EEG discharges [28]. Only a small minority do not have epilepsy [82]. Patients with similar unilateral imaging findings have been described [83], one of whom at postmortem examination had polymicrogyria. Epilepsy and mental retardation were prominent features, with focal neurological signs in a minority. All had diffuse, multifocal or generalised EEG abnormality.

1.1.3.6. Schizencephaly

Schizencephaly refers to a cleft in the cortical mantle lined with grey matter in which the edges of the cleft may (“closed lip”) or may not (“open lip”) be in apposition to each other (figure 1.6). The cleft may extend to the underlying ventricle. From an MRI series, approximately one third were bilateral and one third were open-lip [84]. From this series, patients with bilateral schizencephalies or large or medium open lip schizencephalies had significantly worse neurological impairments. Strong similarities were noted in the patient outcomes and the locations of cortical anomalies of patients with schizencephaly and those with nonschizencephaly focal cortical dysplasias. Cortex lining the cleft is usually polymicrogyric [85]; a developmental vascular aetiology of schizencephaly has been suggested [28, 86]. Some patients with severe schizencephaly have been shown to have mutations at the EMX-2 locus, a homeobox gene involved in early embryogenesis of the head [87].

7 .1.3.7. Focal grey matter heterotopia

Single regions of subependymal heterotopia may be seen, particularly in the right hemisphere and around the trigones of the lateral ventricles (figure 1.7)[68]. Subcortical nodules may be associated with focal signs and focal seizures with appropriately positioned lesions [67]. The white matter and overlying cortex may be abnormal in some individuals. The ventricle may appear compressed by a mass-like aggregate of neurons.

1.1.3.8. Focal cortical dysplasia

The most extensively described and studied malformation of cortical development is focal cortical dysplasia (figure 1.8), first described in association with epilepsy [88]. Focal cortical dysplasia may be associated with hemiplegic dystonia [89] and may underlie 29 some cases of autism [90]. In addition, focal temporal neocortical dysgenesis may rarely be associated with hippocampal malformation [91].

The clinical features of 31 patients with focal cortical gyral anomalies, assumed to be dysplasia, have been described [92]. Generalised developmental delay was very frequent in patients with bilateral focal dysplasia, accompanied by a high incidence of bilateral motor dysfunction and delayed speech. When the dysplasia was unilateral, contralateral spastic hemiplegia or monoplegia was present in the majority, but dysphasia was uncommon, even in patients with dysplasia in the frontal lobe of the dominant hemisphere. Patients with cortical dysplasia were younger at onset of seizures and had a lower IQ compared with patients with mesial temporal sclerosis [93]. A full-scale intelligence quotient of less than 80 was found in 13 of a series of 30 patients [94]. A wide variety of seizure types, usually partial and frequently refractory to drug treatment, was seen, though some authors emphasise the occurrence of particular seizure types, notably motor seizures and hemiparesis [92] and epilepsia partialis continua [94-97]. The epileptic syndrome may, in some cases be relatively benign; 3 patients have been described with complex partial seizures only, a unifocal EEG abnormality and good response to drug treatment.

Computed tomography (CT) is greatly inferior to MRI for the detection of focal cortical dysplasia [93, 94]; the first MRI description of this condition was ten years ago [98]. The MRI features of 31 patients with focal cortical gyral anomalies, assumed to be dysplasia, have been described [92]. Focal areas of cortical dysplasia were most common in the frontal lobes, but were seen in all areas of the brain. The most common MRI appearances were a thickened, irregularly bumpy cortex with shallow, wide sulci, and a deep infolding of thickened cortex. High T2 signal intensity may be seen [99] which was much more frequent in histopathologically more severe lesions [100]. MRI may appear normal in surrounding areas of cortex which histology shows still to be abnormal [100]. In a series of 30 patients, some identified by postoperative pathology, magnetic resonance imaging was superior to computed tomography for identification of the dysplastic cortical lesions, usually showing cortical thickening with blurring of grey- white interface. In one third, MRI showed only subcortical abnormalities [94]. MRI did not allow distinction between a number of pathological entities, namely true pachygyria, focal cortical dysplasia, or the forme fruste of tuberous sclerosis [94]. Compared with patients with intractable epilepsy due to tumour and mesial temporal sclerosis, patients with cortical dysplasia showed significantly more frequent extratemporal lesions [93]. In the largest series described, 52 patients with cortical dysplasia who underwent partial lobectomy for medically intractable seizures [101], the temporal lobe was involved in 34 patients, frontal lobe in 18, parietal lobe in four, and occipital lobe in 30 three. In three patients multiple lobes showed dysplasia. Dysplasia was right-sided in 29 patients and left-sided in 23 patients. Dysplasia was focal in 23 patients, multifocal in four patients, and diffuse in 25 patients[101 ]. Focal cortical dysplasia is often located in the central and insular regions in 50-70% of patients [94, 97, 102, 103].

In an extensive retrospective review of 94 cases [104], the EEG features of cases with localised cortical dysplasia were very variable. The EEG could be normal even when the cortical dysplasia was extensive. An EEG with very high amplitude rhythmic activity was found to have high specificity for severe cortical dysplasia but low sensitivity. Abnormal fast activity was not specific and was seen with very diverse pathologies. The EEG showed interictal abnormalities extending beyond the boundary of the lesion on MRI in 17 of 30 patients, indeed sometimes into the opposite hemisphere of patients with unilateral lesions [94]. Compared with patients with intractable epilepsy due to tumour and mesial temporal sclerosis, patients with cortical dysplasia showed significantly more frequent non-epileptiform EEG abnormalities [93]. An ictal or “ictal-like” pattern of continuous spikes recorded at electrocorticography (ECoG) in 21 of 32 patients with dysplasia but no patients with cortical tumours has been described [100]. Moreover, these abnormalities were recorded directly over the dysplastic region; abnormalities recorded in the tumour patients were in surrounding cortex. This provides evidence for the intrinsic epileptogenicity of focal cortical dysplasia. The occurrence of rhythmic epileptiform discharges on scalp EEG, in association with continuous electrical discharges on ECoG, is highly specific for cortical dysplasia [105].

Three main histologic patterns of cortical dysplasia were observed in the largest described series [101]: (1) a cortical laminar architectural disorganisation and/or malalignment of neurons, (2) clusters of atypical neurons and glia within the cortex, and (3) a hypercellular molecular layer with increased numbers of neurons and glia; these patterns occurred with similar frequency. In many patients more than one pattern of dysplasia was identified. Coexistent tumours were present in a quarter of the patients, including ganglioglioma, dysembryoplastic neuroepithelial tumour, and low-grade astrocytoma. Tuberous sclerosis was very unusual, present in less than 10%. However, an earlier series described the “forme fruste” of tuberous sclerosis occurring as frequently as typical cortical dysplasia [94]; the former may show more marked cytoarchitectural anomalies and subpial clusters of giant astrocytes. An emphasis has been placed on the presence of eosinophilic “balloon cells” in deep grey and subcortical white matter. These “balloon cells” were positive for glial fibrillary acidic protein (GFAP), neurofilaments, and protein gene product (PGP) 9.5 [99]. The presence of such cells has been incorporated into one of a variety of grading systems [100]: Grade I - area of cortical dyslamination; Grade II - dyslamination plus giant polyploid 31

“dysplastic” neurons; Grade III - grade II plus “balloon cells” or other cells showing characteristics intermediate between glia and neurons. The similarity of these cells to the gemistocytic astrocytes of tuberous sclerosis prompted the coining of the phrase “forme fruste of tuberous sclerosis”. Other more complex scoring systems for histopathological diagnosis exist [106].

More sophisticated in vitro examinations have revealed further features of dysplastic cortex. Using GFAP and synaptophysin as markers of glial and neuronal differentiation respectively in resected specimens from children with infantile spasms, both glial and neuronal features of the balloon cells in cortical dysplasia have been shown [107]. A strong abnormal increase in immunoreactivity to the high and medium molecular mass neurofilament epitopes was seen in hypertrophic neurons of cortical dysplasia [108]. These neurofibrillary accumulations in cerebral cortical dysplasia share some common antigens with the neurofibrillary tangles of Alzheimer’s disease, namely tau and ubiquitin, but do not demonstrate immunoreactivity to paired helical filaments antiserum.

A patient suffering from epilepsia partialis continua resistant to antiepileptic drugs underwent focal surgical resection of a lesion [95]. The histological examination revealed focal cortical dysplasia. Surrounding areas showed decreased numbers of neurons, astrocytosis and proliferation of capillaries, compatible with chronic tissue necrosis. Subpopulations of local-circuit neurons were examined with parvalbumin, calbindin D-28k and somatostatin immunocytochemistry. Focal areas of cortical dysplasia contained abnormal immunoreactive neurons. Huge parvalbumin- immunoreactive cells were distributed at random and resembled axo-axonic (chandelier) and basket neurons. Abnormal calbindin D-28k-immunoreactive cells were reminiscent of double-bouquet neurons and multipolar cells. Very large somatostatin-immunoreactive cells were seldom observed in the dysplastic foci. On the other hand, areas of tissue necrosis displayed massive reduction of immunoreactive cells and fibres. Abnormalities in the morphology and distribution of local-circuit (inhibitory) neurons observed uniquely in this study of focal cortical dysplasia may have a pivotal role in the appearance and prolongation of electrical discharges and continuous motor signs in human focal epilepsy. Additionally, this study demonstrates that regions of increased inhibitory interneurons and regions with a decreased density of such cells may be seen in patients with focal cortical dysplasia.

Application of 4-aminopyridine (a convulsant drug) gave rise to spontaneous seizure­ like discharges in resected human dysplastic neocortex, but not in normal neocortex in temporal lobectomy specimens from patients with MS, probably due to abnormal local 32 connectivity allowing neuronal recruitment and synchronisation [109]. In resected cortical dysplasia both laminar distribution and density of catecholaminergic tyrosine- hydroxylase immunoreactive neurons was altered in the dysplastic area and fibre density increased in surrounding areas [110]. Catecholamines may contribute towards limiting seizure propagation [111, 112].

1.1.3.9. Dysembryoplastic neuroepithelial tumour

Dysembryoplastic neuroepithelial tumour (DNT) is a recently described brain mass lesion with distinctive pathological features and a favourable prognosis (figure 1.9). It shares some features with dysplastic lesions and may have a developmental origin. The usual presentation is with treatment-resistant complex partial seizures of long standing [113-116]. Onset of epilepsy is usually in childhood [113, 114, 116] although presentation in the fifth decade has been described [115]. Normal IQ is usually found, but EEG is usually abnormal: localised slow activity and interictal spiking were seen in the substantial majority, being usually more extensive than or distant to the lesion [113].

CT may be normal, show calcification [113], or show cystic hypodense lesions with slight contrast enhancement [116]. MRI is superior, allowing other characteristics to be defined: DNT was predominantly intracortical, although there may also be white matter involvement [113]. In temporal lobe cases, MRI showed that the lesion involved, or was in close proximity to, mesial temporal structures in the great majority [113]. Other MRI features included circumscribed hyperintensity on T2 weighted images, hypointensity on T1 weighted images, and cyst formation [113, 116]. Deformities of the overlying cranium have also been observed [116].

Local surgical resection is usually extremely effective, bringing seizure freedom to 12 of 14 patients in one series [113], although the outcome is less good on longer follow-up of temporal lobe lesions, especially if the hippocampus remains. Histopathological characteristics included a heterogeneous composition in all cases with calcification and dysplastic features being particularly prominent [113]. Typical focal cortical dysplasia may be seen in areas of cortex distant to the DNT [117]. Differential diagnosis includes oligodendrogliomas, mixed gliomas, and gangliogliomas [114]. Features of the DNT that are useful in making the distinction include a multinodular and multicystic appearance, the presence of both neuronal and glial (oligodendrocytic and astrocytic) components with little if any cytologic atypia, the presence of accompanying cortical dysplasia, and the lack of an arcuate vascular pattern [114]. Two multifocal cases showed periventricularly located tumour with nodular extension to the periphery, suggesting an 33

origin from subependymal germinal matrix with nests of primitive neuroblasts arrested in their embryonal migration [115], although such multifocal cases are unusual. In one series of 14 cases, six tumours were frontal, six were temporal, one was parietal, and one was occipitoparietal [116]. In a larger series of 16, the lesion involved the temporal lobe in all but one patient where it was in the cingulate gyrus [113]; one of these patients did not have epilepsy.

1.1.4. Treatment of malformations of cortical development

7 .7 .4 .7 . Drug treatment

Very little data exist- describing outcome in patients with MCD with regard to antiepileptic drugs. Response to medical treatment in ten patients with band heterotopia was variable, although one patient became seizure-free with valproate alone [71]. 60% of patients with bilateral perisylvian dysplasia were refractory to drug treatment [80]. Four patients with hemimegalencephaly were also noted to be intractable [118]. The largest series included 79 patients [119]; however, the numbers in any one diagnostic group were small. In generalised disorders, partial seizures were occasionally controlled with a variety of drugs and infantile spasms were also occasionally controlled, but generalised seizures were intractable. In localised disorders, spasms were often well controlled but partial seizures were more resistant.

7 .7 .4 .2 . Surgery

Surgical treatment in epilepsy may take one of three possible strategies: local resection of a putatively epileptogenic lesion; corpus callosotomy, to prevent seizure generalisation, which is used as a palliative treatment in patients with severe drop attacks; and hemispherectomy, used to remove very extensive lesions. All of these methods have been used in the treatment of MCD, though the accounts are largely anecdotal except in patients with focal cortical dysplasia.

Treatment of pachygyria by anterior callosotomy has led to “considerable improvement” of the epilepsy [120]. Callosotomy in band heterotopia may lead to considerable reduction of drop attacks [71 ].

Hemimegalencephaly causing intractable seizures has been treated by resective hemispherectomy [60] though functional hemispherectomy may be of more value in children with dysgenesis [121]. One series described 10 of 11 children becoming 34 seizure-free after hemispherectomy for intractable seizures; cognitive and motor problems also improved [122].

Six patients with grey-matter heterotopias have been reported undergoing surgery [123]. Good outcome correlated only with resection of coincident HS in one patient and resection of a localised heterotopion in another; the remainder fared badly.

Surgical resection has been undertaken in schizencephaly with improvement in seizure frequency noted [124], although this is generally only feasible if the patient already has a hemiparesis or is very young.

Local resection of focal cortical dysplasia is much more widely described. Of 24 post- surgical resection patients only 10 achieved good or excellent outcome; the variable most strongly correlated with surgical outcome was the amount of lesion removed [125]. Specifically there was no significant correlation between the amount of excision of the epileptogenic area as judged by scalp electroencephalography and electrocorticography studies, and surgical outcome. Other series note better outcomes: the oldest series [88] included 50% of patients with lesions in the temporal lobe, some with histology at the mild end of the spectrum and noted 75% improved by surgery: a more recent series [126] included only patients with temporal lobe lesions, none of whom had severe histology and found 90% were improved by surgery. In a separate series of 17 patients with cortical dysplasia who had surgical resection for medically intractable partial epilepsy, compared with two groups of surgically treated patients with intractable epilepsy due to tumour and HS, patients with cortical dysplasia showed significantly less favourable surgical outcome for seizure control [93]. Surgical outcome related to the extent of pathology but not to the histological abnormality. Lesions outside the temporal and frontal lobes were correlated with poor surgical outcome, as were generalised interictal EEG abnormalities, which may reflect extensive or multiple lesions. Ictal intracranial recordings were not useful for presurgical evaluation of cortical dysplasia [100].

Multiple subpial transection, employed in patients where some or all of the epileptogenic zone cannot be resected because it lies in a vital cortical area, has been reported to show worthwhile improvement in seizure frequency in a small number of patients with focal cortical dysplasia [127]. 35

1.2. STRUCTURE OF THE NORMAL CEREBRAL CORTEX

1.2.1. Neocortex

The adult human cerebral cortex consists almost entirely of neocortex, phylogenetically the most recently evolved cortical area. More ancient cortical regions are the paleocortex (the lateral pallium in amphibians, lateral olfactory area of the uncus in man) and archicortex (medial pallium in amphibians, hippocampus in man) [128]. Neocortex contains neurons which may be divided into three categories: projection neurons, which connect directly to subcortical centres; association neurons which connect with other neurons in the cortex of the same hemisphere; and commissural neurons which project to the opposite hemisphere, mostly through the corpus callosum.

Histologically, typical neocortex contains six layers (figure 1.10), although there are considerable variations in relative cells numbers and cortical thickness in different cortical areas. [128, 129]. The most superficial layer, layer 1 or the molecular layer, is largely a synaptic field containing dendrites of pyramidal cells and cells of Martinotti (having the principal characteristic of an axon projecting towards the superficial cortical surface) from other layers and axons from ipsilateral cortex, contralateral cortex and the thalamus. A few horizontal cells of Cajal and stellate cells (interneurons) are seen. The external granular layer or layer 2 contains many small pyramidal cells and stellate cells. The dendrites of these cells largely run into layer 1 and the axons into deeper cortical or subcortical areas. Layer 3, the external pyramidal layer, largely contains typical pyramidal cells which project into the white matter as association, projection or commissural fibres. The internal granular layer, layer 4, contains a large number of stellate cells whose input comes substantially from the thalamus. These cells in turn synapse with dendrites of layer 5 and 6 cells, with cells of Martinotti and also with each other. Layer 5 contains pyramidal cells (such as Betz cells in the motor cortex) and some stellate cells and cells of Martinotti. The lowest layer, layer 6 or the multiform layer, contains mostly fusiform cells, mostly projecting into the white matter.

The functional organisation of the neocortex (figure 1.10) at the level of cellular circuits is beyond the scope of the thesis. In brief, the cortex receives afferents largely from other cortical regions and from the thalamus. Thalamic afferents synapse in all cortical layers, but especially layers 1 to 4, in particular layer 4 where the profuse afferents may be seen as the outer line of Baillarger. This is particularly prominent in striate cortex and termed the line of Gennari. Axons of the larger pyramidal and fusiform 36 cells form efferents which enter the white matter. Large numbers of intracortical interneurons are present, particularly the small pyramidal cells of layer 2 and the stellate cells of layer 4. The overall cortical organisation is in the form of minute vertical units or columns, particularly well-demonstrated in sensory, especially visual, areas [130]. The localisation of different functions to different cortical areas has been demonstrated through the effects of lesions in man, experimental animal studies involving lesioning and stimulation and through functional imaging in man.

1.2.2. Hippocampus

The hippocampal formation includes the dentate and parahippocampal gyri, up to the entorhinal cortex which is outside the hippocampal formation in the lateral olfactory area, in addition to the hippocampus (figure 1.11). This structure is sometimes called Ammon's horn and has been divided histologically into four subregions (Cornu Ammonis (CA) 1 -4). The ventricular surface of the hippocampus is a layer of white matter called the alveus which is continuous with the fornix posteriorly and contains efferent and afferent axons. The hippocampal archicortex contains three layers; the molecular layer, continuous with the outer layer of neocortex; the pyramidal cell layer, continuous with neocortical layer 5, containing pyramidal cells with dendrites extending into the molecular layer and axons passing through the alveus to the fornix; and the polymorphic cell layer, similar to neocortical layer 6, containing neurons projecting through the fornix and others with axons projecting into the molecular layer [128].

The major input is via the perforant path to the dentate gyrus. In the dentate gyrus the pyramidal cells of the hippocampal pyramidal cell layer are replaced by granule cells. Efferent “mossy” fibres from these cells terminate mostly in CAS of the hippocampus; pyramidal cells of CAS project via Schaffer collaterals to CA1. The neurotransmitter of these pathways is glutamate. The principal inhibitory cell is the basket cell, providing both feed-forward and feedback inhibition via GABA. Pyramidal cells project from the hippocampus via the alveus to end in the subiculum and via the fornix to the septum and mamillary bodies. The subiculum projects, amongst other areas, to the entorhinal cortex, creating a loop involving the hippocampus. 37

1.3. THE GABAERGIC SYSTEM

1.3.1. GABA, GABA receptors and benzodiazepine receptors

GABA is the dominant inhibitory neurotransmitter in the mammalian brain, and exerts its actions through GABA^, GABAg and the more recently described GABA^ receptors. Small changes in GABA-mediated inhibition have profound changes in neuronal excitability, and a wide range of pharmacological drugs that modify GABA receptors, particularly GABA^^ receptors, are used clinically as anaesthetic agents, anxiolytics and anticonvulsants. It has been estimated that approximately 17% of the synapses in the mammalian brain are GABAergic [131]. GABA is formed by the conversion of glutamate to GABA by glutamic acid decarboxylase (GAD). Following release into the synaptic cleft, GABA is removed by uptake into neurons and glia. GABA is metabolised by GABA transaminase to succinic semialdehyde [132].

The GABAg receptor [133] is part of the seven-transmembrane spanning, G-protein- coupled receptor superfamily. The GABAg receptors are coupled to calcium (Ca2+) and potassium (K+) channels via G-proteins and second messenger systems [134]. The drug baclofen, used to reduce spasticity, acts via the GABAg receptor [133]. GABAg receptors on postsynaptic membranes activate a slow outward K+ current, resulting in a relatively slow, weak and prolonged inhibition. Presynaptic GABAg receptors activate slow Ca^+ currents which impair release of neurotransmitters of all kinds from vesicles.

GABAq receptors are stimulated by cis-4-aminocrotonic acid but not by the ligands of the GABAa and GABAg receptors. Following molecular studies they have been localised to subpopulations of retinal neurons, and are composed of the p-subunits, pi and p2 [135]. The presence of the p2 subunit outside the retina indicates that GABA^ receptors are also present in the spinal cord, optic tectum, cerebellum and hippocampus. GABA^ receptors are integral membrane channels that stabilise the resting potential of the cell by increasing the membrane conductance to chloride (Cl ). Inhibition mediated at the GABA^ receptor occurs with lower concentrations of GABA and is longer lasting then via the GABA^ receptor [136].

GABAy^ receptors are membrane-spanning, ligand-gated Cl- ion channels [137], which are targeted by many pharmacological agents in addition to GABA, including several clinically important drugs [138]. These agents interact with several distinct binding sites on the GABA^ receptor to allosterically modulate the GABA-induced Cl- ion flux. Drugs that act in this manner include benzodiazepines, barbiturates and some anaesthetics (including etomidate and propofol), although interaction with the GABAy^ 38 receptor may not be responsible for the full pharmacological effect of these substances.

When GABA binds to the receptor it increases the neuronal membrane conductance for Cl- which results in the membrane becoming hyperpolarized and therefore less excitable [138]. Both high and low-affinity GABA^^ binding sites exist, responding to nanomolar or micromolar concentrations of GABA or its antagonists respectively, and may represent different conformational states of the same receptor [139]. GABA has its physiological action by working on the low-affinity sites and therefore it requires micromolar concentrations of GABA or its analogues to activate the GABA^ binding site. The GABA^ receptor is composed of multiple subunits. Benzodiazepines such as diazepam and flunitrazepam act at specific high-affinity benzodiazepine binding sites on brain membranes closely associated with GABA^ receptors [140]. Stimulation of these receptors increases the frequency of the Cl- channel opening [141]. Binding of benzodiazepines to these bindings sites shows an excellent correlation between the clinical effects and the affinity for the binding sites, and these are therefore assumed to be the "central" benzodiazepine receptors (cBZR) from which the clinically important affects of benzodiazepine exert their action [142]. "Peripheral" BZR sites are found on the outer mitochondrial membrane of many tissues, but are pharmacologically distinct from the GABA^^ receptor-associated benzodiazepine binding sites and do not bind many typical benzodiazepines including flumazenil [143]. There are at least two distinct cBZR binding sites related to the GABA^ receptors, known as BZ1 and BZ2 [144]. Another group of compounds is believed to act by interacting with the cBZR to reduce the GABA- induced Cl- ion flux, and are referred to as "inverse" BZR agonists. BZR antagonists, such as flumazenil, interact with the cBZR and do not influence the GABA induced Cf ion flux, but antagonise the actions of cBZR receptor agonists and inverse agonists. In addition to the cBZR, other distinct subunits of the GABA^ receptor include the picrotoxin binding site, the barbiturate binding site, the steroid binding site, the avermectin Bla binding site and the zinc binding site among others.

Molecular cloning of the GABAy^ receptor has shown at least eighteen genetically distinct subunit subtypes including six a, four p, four y, one ô, one e and two p-subunits [135, 137, 146, 147]. The molecular architecture of the GABA^ receptor consists of a variety of combinations of these subunits. Five subunits are necessary for the formation of the Cl- ion channels; potentially several thousand different structural combinations are possible, although it appears that far fewer are found as native GABAy^ receptors in the brain [148]. A functional GABAy^ receptor depends upon the presence of 2a, 2p and either a Y or a Ô [149]. Certain preferred subunits a 1 , p2 and y 2 are widely distributed in the brain, while others are expressed in specific regions, such as aS and pi in the hippocampus and cerebellar cortex [150-152]. The extensive structural heterogeneity 39 of GABAy^ receptors may point to a functional heterogeneity, which might in the future be exploited pharmacologically. The potential diversity of GABAy^ receptor function is presently being analyzed using recombinant GABAy^-receptors, which consist of various subunit combinations. These studies point not only to variations in the affinity of GABA, depending on the type of subunit combination, but also to differences in the affinities and intrinsic efficacies of benzodiazepine receptor ligands [153].

Although the action of benzodiazepines as psychotropic drugs had been recognised previously, the presence of a specific BZR was not described until 1977, using homogenised membranes from rat brain cells, [154]. Closely following on from this, benzodiazepine receptors were also shown to be present in high densities in the cerebral and cerebellar cortical regions of the human brain [155]. Monoclonal antibodies to a and p subunits of the GABAy^ receptor were shown to have a very similar distribution to labelling with ^H-flumazenil, a neutral antagonist at the cBZR, in the rat olfactory bulbs, cerebral cortex, ventral pallidum, globus pallidus, hippocampus, dentate gyrus, substantia nigra, geniculate nuclei, inferior colliculus, cerebellum, reticular formation, spinal cord, and retina. In contrast, no receptors could be detected in white matter, pineal, pituitary, adrenals, and superior cervical ganglia. Sex and age had no effect on the distribution of the cBZR. Only among the cerebellar layers was there a conspicuous difference between the staining intensity for GABAy^ subunits and the radiolabeling for cBZR [156]. In postmortem human brain the same close correlation between the immunohistochemical and radiohistochemical findings was observed [156]. At the electron microscopic level, the immune reaction product was localized to pre- and postsynaptic membranes of axodendritic and axosomatic synapses [156].

Benzodiazepines interact with components of neuronal membranes to modify excitability in three different ways; a high affinity central receptor (dissociation constant, K^, of 3 nM) linked to the GABAy^ receptor enhances the inhibitory action of GABA by increasing the number of openings of Ch channels produced by a given concentration of GABA, which correlates with anticonvulsant and anxiolytic activity; a lower affinity membrane site

(K d 100 nM to 1 piM) limits repetitive firing as observed in isolated neurons (in a manner similar to the action of phenytoin or carbamazepine), probably involving an increase in the population of sodium channels in the inactive state; a lower affinity site

(K q 45 jiM) in presynaptic terminals decreases voltage sensitive Ca^+ conductance and, by limiting Ca^+ entry, decreases neurotransmitter release [157]. The physiological and pharmacological relevance of the two lower affinity benzodiazepine binding systems is unclear. 40

Autoradiographic studies have added more information to the understanding of the distribution of cBZR within the central nervous system. The striking absence of cBZR density in the subcortical white matter was noted and, in addition, the variation within the Brodmann layers, with the highest densities in layers 3 and 4 of the cerebral cortex was demonstrated [158]. In an extensive autoradiographic study using ^H-flunitrazepam [159] the distribution of benzodiazepine receptors was mapped within the human brain. The highest density of receptors was found in the cerebral cortex, hippocampus, nucleus accumbens, amygdala and mamillary bodies. Intermediate levels were found in the basal ganglia, cerebellum and thalamus. Low levels were found in the brainstem nuclei. Sex and age had no effect on the distribution of the benzodiazepine receptors. This distribution correlates well with that found by positron emission tomography (PET). PET studies have also shown the binding of ^^C-flumazenil did not change significantly with age [160].

1.3.2. GABAergic neurons

The cerebral cortex contains four different types of GABAergic cells, which represent a subpopulation of interneurons. Basket cells project to the soma and proximal dendrites of the pyramidal cells, and are predominantly seen in layers 3 and 5 of the cortex. Chandelier cells which synapse on the initial axon segment of the pyramidal cells, are most numerous in layer 3. The double bouquet cells are chiefly found in layers 2 and 3 and to a lesser extent layer 5 and synapse onto the spines and fine dendritic branches of the pyramidal cells. The fourth type of interneuron, the clutch cell, does not synapse on pyramidal cells but onto the spiny stellate cell of layer 4 [161].

The interneurons contain various peptides, including parvalbumin (basket and chandelier cells) and calbindin D-28k (double bouquet cells). These are calcium- binding proteins and, although the exact role these proteins play is uncertain, it has been suggested that they are involved in intracellular calcium buffering, preventing the intracellular calcium rise and therefore may provide resistance to degeneration. Excitotoxicity may be due to calcium entry into cells, which then activates enzyme systems and indirectly causes the production of free radicals. Therefore, if cells contain calcium buffering proteins, it may be a factor in making them less vulnerable to excitotoxic damage [162]. Parvalbumin immunoreactive interneurons have been shown to be present in all areas of the neocortex, apart from layer I and in the cerebellar Purkinje cells and hippocampus in the rat brain [163]. In addition parvalbumin is virtually only found in GABA containing cells, and can therefore be considered to be a protein marker for these cells [163]. The role of GABAergic interneurons is undoubtedly complex and not yet fully elucidated. The basic set of cortical neurons described above are probably connected as a basic circuit which is repeated innumerable times in each cortical layer with many modifications appropriate to specific output neuron properties [506]. Recent work, reviewed in detail [507], suggests that GABAergic inhibitory interneurons have a major role in the synchronisation of neuronal acticity and are involved in the generation of large-scale network oscillations. In essence, GABAergic interneurons probably dictate the excitability of groups of principal cells of similar function, permitting both controlled normal activity and plastic reorganisation. One potential neurochemical mechanism involved in plastic reorganisation was proposed by Hendry and Jones [508,509]. They suggested that downregulation of GABA and GABA^ receptors led to the unmasking of latent intracortical horizontal connections [510]. In this way, for example, latent intracortical connections within a cortical locus deprived of afferent input, such as might occur following a localised retinal lesion producing a scotoma [509], would be released from inhibition and allow adjacent cortical loci to drive the de-afferented cortical locus. This might account, for example, for the perceptual filling in of scotomas [509] i.e. small blind spots in the visual field are not perceived as such, but are perceptually “filled in”. GABAergic neurones have also been suggested to underlie the maintenance of motor maps in the primary motor cortex (M l) [511]. 41

1.4. MECHANISMS OF EPILEPTOGENESIS

In vivo and in vitro models of epileptic disorders in animals have provided many insights into the mechanisms by which seizures may arise. The largest amount of data is derived from studies of limbic, particularly hippocampal, epilepsy in rodents. It must be prudent to apply caution in extrapolating from rodent limbic epilepsy to extratemporal localisation related epilepsy in man, although data from neocortical epilepsy models in rodents demonstrates that many limbic mechanisms also apply in the neocortex. The most pertinent models are those from neocortical epilepsies in primates. Moreover, particularly relevant information may be derived from in vitro studies of resected surgical specimens and from in vivo investigations in man. The subsequent section will follow this logic, initially describing some general principles and terminology, then describing epileptogenesis in the normal rodent hippocampus and other rodent limbic epilepsy models. Following this, a description of epileptogenesis in normal rodent neocortex is set out. Subsequently, neocortical epilepsy models are described, culminating with primate models. These models primarily address local cortical mechanisms; a growing attention to cortical-subcortical networks in epileptogenesis is next described. Finally, data derived from man is discussed.

1.4.1. General principles and terminology

The cerebral cortex is composed of local circuits which have both inhibitory and excitatory connections, and inputs from distant areas which also may be inhibitory or excitatory. In simple terms, anything disturbing the balance of excitation and inhibition can give rise to a seizure [164]. In order to interpret animal models it is necessary to describe the correlation between in vitro phenomena and seizures. The clinical investigative hallmarks of epilepsy are the ictal seizure discharge (which accompanies a seizure) and the interictal spike discharge recorded by EEG. An early clue to the mechanism of focal epileptic seizures was that neurons in the hippocampus can show paroxysmal depolarisations; many neurons in a local region can show this phenomenon simultaneously due to recurrent excitatory connections between pyramidal cells; this activity is synchronised with an EEG spike [165]. The interictal discharge represents the synchronous, large depolarisaton (the depolarising shift), accompanied by bursts of action potentials, of thousands of neurons in the locality of the spike recorded on the scalp surface EEG; this is followed by a hyperpolarising potential [166]. Neurons surrounding the focus are inhibited [167]. If a seizure is to develop, the post- depolarising shift hyperpolarisation gradually disappears, to be replaced by a depolarisation with superimposed small depolarisations resembling small depolarising 42 shifts occurring synchronously in many neurons; the EEG shows afterdischarges [166]. The afterdischarges become longer wih each interictal discharge until a seizure appears. Many insults can, in animal models, lead to this characteristic sequence of epileptic events. It is also noteworthy that the majority of neurons involved as a seizure develops and spreads are normal.

The depolarising shift is regarded to be predominantly a synaptic phenomenon [166] resulting from an amplification of the normal excitatory post-synaptic potentials. This amplification could have many origins. It could result from loss of inhibition, potentiation of the excitatory post-synaptic potential [168, 169], or a variety of other mechanisms modulating the efficacy of receptor function [166]. There may be abnormal intrinsic properties of the neuron: significant differences in the ratios of sodium channel subtype mRNAs have been found between normal and epileptic human brains [170].

The development of synchronous firing of many neurons is essential for the further development of seizure activity. Synaptic excitation through recurrent collaterals is possible both in hippocampus and neocortex [171]. Excitatory projections to other brain areas may also spread seizure activity. Further, it is possible that extracellular ion changes may lead to increased excitability of some neurons and ephaptic interactions [172]. Recruitment may also occur or be facilitated by lateral diffusion of neurotransmitter to influence neighbouring neurons [173]. Finally, there is some evidence for the direct electrical coupling of some cortical neurons [174].

The transition to seizures in acute focal models involves a loss of the post depolarising shift hyperpolarisation, possibly by frequency-dependent desensitisation which may arise through presynaptic feedback by GABA acting via GABAg receptors [176] and/or via increases in intracellular chloride and extracellular potassium which affect ion channels in a manner leading to reduction of inhibitory postsynaptic potentials [176]. Thus epileptic activity is permitted to spread into surrounding normal brain. Subcortical networks may act as seizure gating areas.

1.4.2. Specific models

Many models of localisation-related epilepsy and partial seizures exist; their appropriateness in relation to differing syndromes of human epilepsy is difficult to establish. Moreover, the findings from different models (and even from different studies in the same model) may be contradictory in terms of mechanism. There follows a 43 description of the most influential models, starting with rodent limbic models, progressing through rodent neocortical models to primate neocortical models, models of cortical-subcortical networks, and finally to evidence from man.

7 .4 .2 .7 . Normal hippocampus

The development of a hyperexcitable hippocampus able to support epileptiform bursting requires an imbalance between GABA^^-mediated inhibition and N-methyl-D-apartate (NMDA)-mediated excitation. Once bursting is established, hippocampal plasticity allows bursting to be maintained even if the excitation is withdrawn, both in the rat dentate gyrus [177, 178] and the CA3 region of the guinea-pig hippocampus [179]

The GABA,^ mechanisms shown by these experiments to be crucial to preventing abnormal excitability are mediated via inhibitory interneurons. These interneurons may themselves be inhibited by GABAg mechanisms, resulting in an overall increase in excitation. In rat hippocampal slices, granule cell firing was induced by perforant path stimulation; recurrent inhibition was induced by pairing the perforant path stimulus with a preceding mossy fibre stimulus and was mediated through GABA^ receptors [178]. Baclofen strongly suppressed recurrent inhibition and converted the perforant path-evoked response to an epileptiform response because its disinhibitory effect on the interneurons was substantially greater than its suppressive effect on synaptic excitation - in other words, inhibition of the inhibitory mechanism was excitatory.

In summary, these experiments demonstrate that normal tissue can support epileptiform activity if the balance between GABA^^-mediated inhibition and NMDA- mediated excitation is disrupted. These results are pertinent not only to the mechanisms by which focal epileptiform activity arises but also to the mechanism by which normal cortex surrounding an epileptic focus may become involved in epileptic discharges.

1.4.2.2. Kindling

Kindling refers to the process by which repeated subconvulsive electrical stimulation to a region of the brain in a living, freely-moving animal gives rise to a sustained increase in excitability such that progressively more severe seizures arise with repeated stimulation; ultimately, spontaneous seizures may occur. Many brain areas may be kindled in a variety of species [180]. In animals kindled at various sites of the limbic system the pattern of development of kindling and the nature of the seizures is very similar, regarded as analagous to complex partial seizures in man [180]; neocortical stimulation produces a rather different pattern with motor seizures. Kindling models 44 generally show a repense to conventional antiepileptic drugs such as phénobarbital, phenytoin, carbamazepine, valproic acid, diazepam, and clonazepam [181].

The dependence of kindling on GABA^ mechanisms can be demonstrated by drug treatments in such animals. Progabide, acting at GABA^ receptors, [182, 183], SKF89976A, a GABA uptake inhibitor, [182] and vigabatrin, an inhibitor of the GABA catabolic enzyme GABA-transaminase, [182, 184] all suppress kindling. In contrast, baclofen, a selective GABAg receptor agonist, does not show anticonvulsant effects [182, 183].

Attempts have been made to clarify the nature of the GABA^-dependent mechanisms in kindling. In the rat kindling model of epilepsy, quantitative radiohistochemistry has shown increased GABA^ and BZR on somata and dendritic trees of dentate granule cells at approximately 24hrs after the last seizure [185]. Further, certain GABA^ receptor subunit messenger RNAs (mRNAs) were increased in all hippocampal areas following kindling [186]. In a rat olfactory bulb kindling model, ^H-diazepam revealed a bilateral increase of benzodiazepine receptors in the olfactory bulb and an ipsilateral increase in the hippocampus were observed [187]. It could be hypothesised that in the hyperexcitable hippocampus produced by kindling, impaired GABAy^-mediated inhibition is reflected by an up-regulation of GABA^ receptors. It has recently been argued [188] that central cortical postsynaptic neuroreceptors are always fully saturated whenever transmitter is released into the synapse; hence, an increase in inhibition (increased magnitude of IPSP) could only be acheived by increasing the number of postsynaptic receptors. Evidence for such an increase on dentate granule cells following kindling has been obtained [188].

Excitatory mechanisms have also been investigated. An antagonist of NMDA receptors (MK-801, non-competitive channel blocker) suppressed the development of kindling [189]. NMDA receptor function is enhanced in hippocampal dentate granule cells during kindling [190]. Also, CA3 pyramidal cells from kindled rats were more sensitive to NMDA than control cells; this was argued to be due to an increased number of NMDA receptors [191]. An increase in the amplitude of both the slow NMDA and fast non-NMDA receptor-mediated components of glutamatergic transmission accompanied by a decrease in GABA^-mediated inhibition has been found using whole-cell current-clamp techniques [192]. Further, a long-lasting decrease in the inhibitory effect of GABA on glutamate excitatory responses has been found in hippocampal pyramidal cells from a kindling model [193]. In vivo microdialysis found progressive increases in glutamate release as kindling progressed [194]. In amygdala kindled rats the GluR2 subunit of the a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) receptor, which is 45 uniquely responsible for controlling calcium permeability, was significantly reduced in the amygdala and piriform cortex following kindling, potentially allowing calcium- related excitotoxicity [195].

In summary, the epileptogenic effect of kindling can be simplified to a process of enhancement of at least partly NMDA-dependent excitatory mechanisms coupled with a reduction in inhibitory mechanisms via GABA^ receptors. The finding of elevated GABA^/BZR density is fairly consistent in granule cells of the dentate gyrus, though GABAergic interneurons (the presynaptic source of GABA) may be reduced in number. Consistent changes in NMDA receptor numbers have not been observed.

1.4.2.3. Intraventricular kainic acid

In adult rats, intraperitoneal or intraventricular administration of kainic acid, a glutamic acid analogue and potent neurotoxin, induces persistent seizure activity that results in electrographic alterations and neuropathology that closely resemble human temporal lobe epilepsy.

Intracellular recordings of evoked bursts of action potentials made from rat hippocampal CA1 pyramidal cells in slices where the CA3/CA4 region had been lesioned using kainic acid showed that the bursting activity is largely due to a loss of inhibition mediated by GABA^ receptors [196]. Using GAD immunocytochemistry, no acute or chronic decrease in numbers of CAT interneurons or in qualitative characteristics of the pericellular distribution of their terminals in CAl was found. In vitro autoradiography showed an increase in GABA;^ receptor density in CAl [197]. lontophoretic application of GABA to the soma and dendrites of CAl pyramidal cells indicated that there had been no change in the efficacy of the postsynaptic GABA receptors on these cells [198]. These data indicate that the elements necessary for inhibition are still present in CAl, despite lesion- induced hyperexcitability, but a transient disconnection between inhibitory and excitatory elements in CAl, or a loss of normal afferent drive from CA3 onto some CAl interneurons, or both, has occurred.

In the CA3/CA4 area, the hippocampal region most vulnerable to neurodegeneration after kainate acid treatment, expression of GluR2 (the AMPA/kainate receptor subunit that limits Ca^+ permeability) and GluR3 was decreased markedly at 12 and 24 hr, such that increased formation of Ca^+-permeable AMPA/kainate receptors in the CA3/CA4 area might lead to glutamate excitotoxicity. Also, GABAy^ a l subunit expression was decreased. In the dentate gyrus, concomitant increases in GluR2 and GluR3 expression were observed; GABAy^ a l subunit expression was not detectably altered. These findings 46 indicate that kainate seizures modify hippocampal glutamate and GABA^^ receptor expression in a cell-specific manner and that these changes could result in excitotoxicity [199].

In addition to functional changes, this model reveals some important structural changes. In rats treated with kainate, collateral sprouting of mossy fibres (the axons of granule cells, a major input to hippocampal CA3) has been observed [200-202]. Epileptogenic lesions causing partial deafferentiation of the dentate gyrus or other hippocampal structures are followed by a regenerative process of the mossy fibres, which may provide anatomically abnormal excitatory connections. This phenomenon has been seen in animal models and in human temporal lobe [203]. These structural changes presumably result from changes in gene expression; certainly increased expression of the proto-oncogene c-fos is observed [204], though the role of this gene is not established. In rats with two copies of a null mutation for c-fos there was marked protection against mossy fibre sprouting and seizure development [205]. The mossy fibre sprouting is associated with an early (within 90 mins of kainic acid exposure) expression of mRNA for the growth-associated pepetide GAP-43 in granule cells; this gene expression, mossy fibre sprouting and subsequent seizures are all blocked by administration of MK-801 or pentobarbital within 55 mins of kainic acid exposure [206]. Intraventricular infusion of a synthetic peptide mimic of a nerve growth factor domain that interferes with the binding of neurotrophins to their receptors resulted in significant inhibition of mossy fiber sprouting [207]. Other cells also are damaged; Immunohistological studies showed a decrease in the number of somatostatin positive non-pyramidal cells in the CAl area [198]. Dramatic protection against the cell losses seen in this model has been conferred by knockout of both copies of the p53 tumour supressor gene [208].

In summary, the kainic acid-treated rat model reveals the important mechanism of synaptic reorganisation via mossy fibre sprouting, such that abnormal recurrent excitatory circuits are produced resulting in abnormal hyperexcitability. This is accompanied by an increase of GABA^^ density. A loss of inhbitory action is seen however, perhaps due to a functional disconnection of GABAergic interneurons, possibly related to the loss of somatostatin-positive cells.

1.4.2.4. Previous limbic status epilepticus

Prolonged electrical stimulation of the perforant path may give rise in the rat to a period of limbic status epilepticus and subsequent chronic spontaneous seizures. In this model, extended electrographic monitoring in the hippocampus revealed spontaneous 47 recurrent paroxysmal discharges, persistent interictal spiking and fully developed electrographic seizures [209].

Hippocampal slices demonstrated an increased excitability relative to slices from control animals as evidenced by epileptiform bursting in increased extracellular potassium and decreased extracellular calcium [209]. Changes in paired pulse inhibition elicited in CAT pyramidal cells was studied in this model with a protocol known to reflect the potency of inhibition mediated by GABA^ receptors, showing a profound and enduring disturbance of GABA-mediated inhibition [209-211]. The possibility that CAl inhibitory neurons (basket cells) were hypofunctional or "dormant" due to a loss of excitatory input to inhibitory cells from damaged CA3 pyramidal cells was tested by stimulating the contralateral perforant path in order to activate the same CAl basket cells via different inputs. Contralateral stimulation evoked CAl pyramidal cell paired- pulse inhibition immediately in the previously stimulated hippocampus [211]. To demonstrate the same effect more directly, inhibitory postsynaptic potentials were elicited in CAl pyramidal cells by activation of basket cells; responses from rats previously undergoing status were compared to those from control animals not electrically stimulated. Inhibitory postsynaptic potentials evoked indirectly by activation of terminals that excited inhibitory interneurons were reduced in the epileptic tissue, whereas inhibitory postsynaptic potentials evoked by direct activation of basket cells, when excitatory neurotransmission was blocked, were not different from controls [212]. These results provide support for the "dormant basket cell" hypothesis.

In further studies, granule cell pathophysiology was seen only in animals that exhibited a loss of adjacent dentate hilar mossy cells and hilar somatostatin/neuropeptide Y- immunoreactive neurons. GABA-immunoreactive dentate basket cells (the local inhibitory interneurons) survived despite the extensive loss of adjacent hilar neurons. However, parvalbumin immunoreactivity, present normally in a subpopulation of GABA-immunoreactive dentate basket cells, was absent on the stimulated side, which could represent decreased parvalbumin synthesis in surviving basket cells or a loss of a specific subset of inhibitory cells [211, 213].

In summary, rats undergoing electrical stimulation leading to prolonged limbic status epilepticus show a loss of functional inhibition clearly due to functional underactivity of inhibitory interneurons, possibly due to a loss of specific somatostatin-positive cells.

1.4.2.5. Normal neocortex

As in the hippocampus, epileptiform excitatory activity is constrained in the neocortex 48 by GABAergic inhibitory interneurons acting on pyramidal cells via GABA^j^ receptors [214, 215].

A series of meticulous studies of the induction of hyperexcitability in the normal cat striate cortex by convulsant drugs has recently been reviewed [216]. Through minute attention to the laminar positioning of microinjections and recording electrodes, it was shown that the principle excitatory interneuron of neocortex (the spiny stellate neuron of layer 4) was the crucial initial cell involved in the development of excitability after the application of penicillin or bicucculine and that the action of this cell is modulated by GABA. These cells receive a heavy afferent projection form the thalamus. Blockade of glycine by stryzhnine was effective in layers 2 and 3 via local cortical networks [217].

In summary, GABA^ mechanisms and glycine are responsible for inhibitory influences in neocortex; GABAergic mechanisms are more generally implicated. This mechanism does not need to be completely lost in order for seizures to arise. This parallels the finding from hippocampus in which epileptiform activity can be produced despite preserved inhibitory activity.

1.4.2.6. Penicillin and bicuc ulline in rat neocortex

At low concentrations, both penicillin and bicuc ulline specifically block GABAergic inhibitory synaptic transmission in brain tissue acting via GABA^ receptors. Direct application of these agents to neocortex evokes an acute epileptic focus.

Afterpotentials were analyzed in penicillin foci of the rat motor cortex in vivo using intracellular recording techniques [218]. Part of the afterpotential is chloride- dependent, probably representing a GABA^ response. Surprisingly, the Ch-dependent potential can be depolarising, possibly due to intracellular Ch accumulation, which might favor transition to ictal discharges [218].

In summary, partial blockade of GABA^^ receptors may give rise to an acute epileptic focus in this model. A mechanism by which continued opening of GABA^-dependent Ch channels may paradoxically contribute to the sustained depolarisations leading to a seizure has been revealed.

1.4.2.7. GABA withdrawal syndrome, rat neocortex

A few days of localized application of GABA into the somatomotor cortex of rats induces, upon withdrawal, the appearance of epileptogenic activity with maximal electrographic 49

expression circumscribed to the infused site. This GABA-withdrawal syndrome lasts for several days [219]. The paroxysmal discharges induced by withdrawal from unilateral GABA application are accompanied by myoclonus of the body territory corresponding to the infused area.

The effect of an NMDA receptor blockade was evaluated in this model [220]. Bursting properties and morphology of neurons were also analyzed in slices. The noncompetitive NMDA antagonist phencyclidine was administered systemically before discontinuation of the GABA infusion. Phencyclidine prevented the development of seizures; if phencyclidine was injected after GABA discontinuation it failed to suppress seizures. Most neurons studied showed paroxysmal depolarization shifts to white matter stimulation or intracellular depolarizing current injection. Morphologically, these cells were large, spiny pyramidal neurons localized in layer 5 of the sensorimotor cortex. Application to the slices of a selective antagonist of NMDA receptors reduced the amplitude and the duration of paroxysmal depolarization shifts [220].

In summary, this model contains many features familiar from hippocampal models, particularly the opposing roles of GABA* and NMDA receptors acting on pyramidal cells.

1.4.2.8. Cobalt focus

Cortical implantation of cobalt metal in the rat, cat, monkey, rabbit and gerbil gives rise to intermittent and paroxysmal seizures [221].

Of note in this model is strong evidence for the development of a secondary or “mirror” focus in the homotopic area of the contralateral hemisphere. This focus shows vigorous independent spike activity [222].

GABA and its synthetic enzyme GAD, are both at normal levels prior to the development of seizures, are significantly decreased during the period of seizures, and return toward control values at a time when seizures are no longer apparent. There is no change in postsynaptic GABA^ receptor density (Bm^J prior to seizures, a significant increase in B^ax during seizure activity, and a return toward normal when seizure activity has terminated [223]. Changes in putative inhibitory interneuron numbers are inversely related to these receptor binding changes [224].

However, contradictory results have been obtained: Benzodiazepine, GABA, and picrotoxinin receptor bindings were measured in different brain areas using ^h - flunitrazepam, ^H-muscimol and ^ss-t-butylbicyclophosphorothionate respectively as 50 ligands, and the values were compared with glass-implanted controls. In the focal area, all the specific receptor bindings decreased and Scatchard analysis showed a decrease in the number of binding sites without any effect on binding affinity. No change was seen in the binding characteristics of the other areas studied [225].

Abnormalities of the cholinergic system have been demonstrated in this model. Significant reductions of both the acetylcholine synthetic enzyme choline acetyltransferase and in the degradative enzyme acetylcholinesterase were seen in both primary and secondary foci during the period of peak epileptic activity. Moreover, the cholinesterase inhibitors physostigmine and diisopropylfluourophosphate were highly effective anticonvulsants.

Alterations of the cortical noradrenergic system in chronic cobalt epileptogenic foci in the rat have been documented [111]. These changes include a decreased density of noradrenergic terminals in the focus and a decreased concentration of noradrenaline. The homotopic area showed no changes, however. Facilitation of focal cobalt-induced epilepsy after lesions of the noradrenergic locus coeruleus system [112] has also been noted.

Cobalt implantation also induces epilepsy in the cat. Changes in the strength of recurrent inhibition in the feline cortex were observed. The strength of inhibition was analyzed in terms of paired-pulse depression of the amplitude of somatosensory evoked potentials elicited by stimulation of the ventral posterolateral thalamic nucleus. An enhancement in recurrent inhibition was observed shortly after cobalt application. The reduction of inhibition that appeared later was associated with afterdischarges evoked by ventral posterolateral thalamic nucleus stimulation. These afterdischarges frequently extended to epileptic discharges. These results suggest that the reduction in recurrent inhibition induced by cobalt application plays an important role in the spread of seizure activity [226].

In summary, the cobalt implantation model demonstrates functional impairment of GABA^-mediated inhibition, accompanied by a loss of GABA-positive interneurons and a reduction in GABA and GAD levels. Some studies have found an increase in GABA^ receptor density during the period of seizure activity, but this has not been a constant finding. Intriguing abnormalities of cholinergic and adrenergic transmitter systems have been noted.

1.4.2.9. Chronic neocortical injury, rat

Neocortical lesions were surgically inflicted in rat neocortex and the epileptogenic 51 cortex subsequently resected and their axonal arborisation reconstructed. The results suggested that a significant degree of axonal reorganization had taken place in the chronically injured cortex, presumably due to axonal sprouting. It was speculated that this might be an adaptive mechanism for recovery of function after injury, or might be maladaptive and play an important role in the generation of epileptiform events by increasing the numbers and density of synaptic contacts between neurons [227].

1.4.2.10. Alumina gel, monkey neocortex

Subpial injections of aluminum hydroxide gel into the sensorimotor cortex of the Macaca mulatta monkey produce chronic epilepsy. The structural and functional consequences of this primate model are extremely clearly characterised.

The epileptic cortex demonstrated markedly reduced GABA^^ receptor binding and diminished tissue GABA concentration and glutamic acid decarboxylase activity [228]. GABA^ receptor loss was more marked than receptor losses for other neurotransmitters and was more widespread. Scatchard plot analysis demonstrated that the diminished GABA^ receptors within the focus were due to receptor loss and not affinity changes. Spearman rank correlations showed a significant correlation between the degree of GABA^ receptor loss and decrease in GAD activity and the seizure frequency [228].

A series of detailed studies from one group over more than a decade has clarified the GABAergic abnormalities of this syndrome [229-232]. Using an immunocytochemical method for the localization of GAD, GABAergic nerve terminals were seen throughout all layers of normal monkey sensorimotor cortex. These terminals displayed ultrastructural characteristics that suggested that they arose from aspinous and sparsely spinous stellate neurons. In monkeys made epileptic by cortical application of alumina gel, a highly significant numerical decrease of GAD-positive nerve terminals occurred at sites of seizure foci indicating a functional loss of GABAergic inhibitory synapses [229, 231]. Monkeys were studied at progressive intervals following unilateral application of alumina gel to sensorimotor cerebral cortex. Using an immunocytochemical method, the amounts of GAD-positive terminal-like structures were determined on control and experimental sides of motor cortex (layer 5). This showed reductions of GAD-positive terminals on the experimental cortical side, greater the longer the time since implantation. Electron microscopy at an early stage revealed degenerating axon terminals in layer 5 of motor cortex, as well as phagocytosis of degenerating material and astrogliosis. Similar findings were obtained from a chronically epileptic specimen, except that degenerating terminals were observed less often and fibrous astrocytic scarring was more prevalent, especially surrounding the 52 somata of pyramidal neurons [230, 233]. Further, a selective loss of GAD-positive neuronal somata occurs in monkeys with alumina gel implants even before seizures begin [232]. Electron microscopy showed that chandelier cell axons appeared to degenerate in epileptic cortex. The highly strategic site of GABAergic inhibitory synapses on axon initial segments suggests that they exert a strong influence on the output of pyramidal cells. The near absence of these chandelier cell axons in epileptic foci most likely contributes to the hyperexcitability of neurons [234].

In summary, in this model, perhaps analagous to a localised injury in man, there is a specific loss of a class of GABAergic interneurons, accompanied by a loss of GABA* receptors.

7.4.2.7 7. GABA withdrawal syndrome in the baboon

After the chronic infusion of GABA into the cortex of baboon Papio papio, a sudden cessation of the infusion provoked cortical epileptic discharges around the infused area; the baboon exhibited partial epilepsy for about 3 days. The anticipated mechanism would be the same as for GABA withdrawal sydrome in the rat [235].

1.4.2.12. Subcortical networks

Cortical excitability and the spread of seizure activity are both strongly influenced by subcortical structures.

The substantia nigra may be a crucial component of the network underlying limbic kindled seizures and possibly kindling itself [180]. Ipsilateral substantia nigra stimulation delivered prior to each amygdala kindling stimulation significantly retarded the appearance of seizures. Bilateral substantia nigra prestimulation blocked seizure generalization. These effects were only partially antagonized by haloperidol, but were completely abolished by picrotoxin, suggesting that GABAergic rather than dopaminergic pathways were involved [236]. In animals pretreated with microinjections of isoniazid, an inhibitor of activity of GAD, into the substantia nigra pars reticulata bilaterally, non-convulsant doses of pilocarpine resulted in severe motor limbic seizures and status epilepticus [237].

Further evidence strengthens the hypothesis of a GABA-mediated effect in the substantia nigra. In kindled rats abnormalities in GABA synthesis have been detected - reduced glutamic acid decarboxylase has been detected in synaptosomes prepared from the substantia nigra of kindled rats [238]. Bilateral microinjection of a GABA agonist. 53 muscimol, into the area of the substantia nigra markedly suppressed both motor and limbic seizures induced by stimulation of amygdala, olfactory structures, or lateral entorhinal cortex [239]. Microinjection of saline did not suppress seizures. Muscimol injections 1 to 2 mm dorsal to the substantia nigra or into neocortex did not suppress the seizures. Microinjection of an irreversible inhibitor of GABA transaminase (vigabatrin) into the area of the substantia nigra also suppressed kindled seizures. Seizures were markedly suppressed in animals with bilateral destruction of the substantia nigra but not in animals in which the substantia nigra was spared bilaterally. It is noteworthy that partial as well as generalised seizures were suppress ed by manipulations of the substantia nigra [239]. The substantia nigra is not simply a part of a network allowing seizures to become generalised; it plays a part in partial seizure activity. Afterdischarges can be detected in the substantia nigra during partial limbic seizures and metabolism of this area is increased [240].

The physiologically relevant inhibitory input comes from the striatum. Susceptibility to seizures induced by systemic administration of pilocarpine to rats is increased by a lesion of the GABAergic striatonigral projection [241]. These experiments show that inhibition of the substantia nigra by GABA acting at GABA^ receptors is protective against seizure development and that destruction of a GABAergic striatonigral projection is facilitatory in seizure development. GABAergic basal ganglia outflow may play an important role in controlling the spread of seizures by influencing the activity of the substantia nigra.

Some contradictory evidence exists, however: The role of the substantia nigra in seizures induced by low frequency premotor cortical stimulation in the rat has been investigated [242]. Unilateral intranigral injection of muscimol was found, paradoxically, to facilitate seizure development; the facilitatory effect of intranigral muscimol was not modified by haloperidol pretreatment [242]. Other studies [243] show no effect of bilateral substantia nigra lesions on kindled seizures.

The influence of the globus pallidus pars interna and the substantia nigra pars compacta on hippocampal electrical activity has been studied [244]. Injection of intravenous sodium penicillin produces steady interictal spikes in the hippocampus. Substantia nigra stimulation induces regular theta rhythm and inhibits the spikes. Pallidal stimulation, on the contrary, appears to strongly enhance epileptiform activity, proceeding to generalised seizure activity [244].

Blockade of GABA receptors in the rat superior colliculus has been shown to protect against maximal electroshock-induced tonic convulsions and spontaneous generalized 54 non-convulsive seizures [245, 246], Other nuclei closely involved with limbic circuits and potentially relevant to an interaction between the pallidum and hippocampus have been studied. The roles of the habenular nucleus, raphe, nucleus accumbens, septum and caudate have been studied in a cat model involving direct penicillin application to the hippocampus [247, 248], Habenular electrical stimulation caused an increase in spike frequency and amplitude. Intraperitoneal methysergide bimaleate (5-HT antagonist) suppressed the effects of habenular stimulation. In contrast to the effects of pallidal and habenular stimulation, raphe electrical stimulation inhibited hippocampal spiking and intra-raphal muscimol (a GABA receptor agonist) enhanced hippocampal-based epilepsy. After muscimol, raphe stimulation at the same threshold parameters failed to affect hippocampal activity. In cats with habenular lesions hippocampal spike frequency and amplitude were reduced and intra-raphal muscimol did not affect the hippocampus [247].

The relevance of these findings to hippocampal epilepsy in man is not known; equally, the role of basal ganglia influences in neocortical epilepsy in man is not established, although a role for the substantia nigra seems likely on the basis of its influence in several animal models. A summary of the interactions between the basal ganglia and frontal neocortex in man is shown in figure 1.12 [249]. In a single human subject with a frontal seizure focus causing simple partial motor seizures of the face, signal changes in the left ventrolateral thalamus showed a high degree of temporal correlation with signal changes in the left frontal cortical seizure focus, demonstrating close corticothalamic coupling of metabolism [250].

1.4.3. Evidence from man

For obvious reasons, many of the experiments described in animal models cannot be carried out in man. However, surgical treatment of epilepsy presents several opportunities for investigations in human epilepsy. Resected specimens can be subjected to many of the pharmacological and neurophysiological manipulaions used in in vitro preparations from animals. In addition, in patients in whom depth electrodes are placed an opportunity is presented to correlate electrical activity with dialysates of extracellular fluids obtained through miniaturised dialysis probes placed into the brain alongside the electrodes. The largest amount of data is available from temporal lobe epilepsy. 55

1.4 .3 . 1. Temporal lobe epilepsy

The frequently-noted observation in animal models that although inhibitory function may be impaired the cellular components required for inhibition are retained has parallels in man. In resected sclerotic hippocampi, extracellular unit activities of human hippocampal neurons and their responses to single pulse stimulation were recorded in patients in vivo during interictal periods. It was observed that prolonged inhibition in synchronously firing epileptic neurons could occur with little or no prior excitation, suggesting that this hyperpolarization was likely to be synaptic rather than dependent on an intrinsic membrane property. Although this evidence was indirect, it could suggest that recurrent inhibition may remain functional in human temporal lobe epilepsy [251].

Binding to excitatory and inhibitory receptors was examined in frozen hippocampal sections from surgical specimens resected from patients with medically refractory temporal lobe epilepsy and compared with age-comparable autopsy control subjects. Binding to phencyclidine receptors associated with the NMDA channel was reduced by 35 to 70% in all regions in the hippocampi of the patients. In contrast, binding to the NMDA recognition site and its associated glycine modulatory site was elevated by 20 to 110% in the CA1 area and dentate gyrus of the hippocampus of the patients. Binding to these sites was unaffected in area CA4. Binding to the quisqualate-type excitatory amino acid receptor was unchanged in all regions except the stratum lacunosum moleculare CA1, where it was increased by 63%. GABA^ and BZR binding was reduced by 20 to 60% in CA1 and CA4, but unchanged in dentate gyrus [252]. Patients in this study had evidence of neuronal loss but not advanced hippocampal sclerosis. However, other studies in hippocampal sclerosis showed a reduction of NMDA receptor binding in CA1, CA3 and CA4 [253]. A comparison of cell counts and autoradiography of benzodiazepine receptors using ^25|_jomazenil in resected hippocampus from patients with hippocampal sclerosis showed a significant reduction of BZR and neuronal density in CA1, CA2, CA3, CA4 and dentate gyrus. All regions showed a tendency for the loss of receptors to be over and above the loss of cells, but this reached significance only in CA1 and CA3 [254]. Additionally, this study showed the “peripheral” benzodiazepine receptor was increased in the same regions.

Synaptic rearrangements due to mossy fibre sprouting appear also to occur in man [255]. Epileptogenic lesions causing partial deafferentiation of the dentate gyrus or other hippocampal structures are followed by a regenerative process of the mossy fibres, which may provide anatomically abnormal excitatory connections. This phenomenon has been seen in animal models and in human temporal lobe [203]. 56

Quantitation of mossy fibre sprouting in patients with HS compared with patients with extrahippocamapi temporal mass lesions showed significantly more mossy fibre sprouting in the HS patients, although a few extra-hippocampal subjects also showed mossy fibre sprouting [256]. An enhanced and abnormal NMDA receptor response has been detected on surviving dentate granule cells in resected hippocampi of patients with chronic temporal lobe epilepsy [257]; it was suggested that excitotoxic responses consequent to this might be responsible for the abnormal dendritic morphology seen in these cells.

A role for excessive glutamate-mediated excitation in seizures in man has been proposed. Bilateral intrahippocampal microdialysis was used to test the hypothesis that an increase in extracellular glutamate may trigger spontaneous seizures. The concentrations of glutamate and GABA were measured in microdialysates before and during seizures in 6 patients with complex partial epilepsy investigated before surgery. Before seizures, concentrations of glutamate were higher in the epileptogenic hippocampus, whereas GABA concentrations were lower. During seizures, there was a sustained increase in extracellular glutamate to potentially neurotoxic concentrations in the epileptogenic hippocampus. Moreover, the increase preceded the seizures. GABA concentrations were unchanged before seizures, but increased during them, with a greater rise in the non-epileptogenic hippocampus. These data suggest that a rise in extracellular glutamate may precipitate seizures and that the concentrations reached may cause cell death. Also, potentially compensatory inhibition, as measured by elevated GABA levels, is more pronounced in the non-epileptogenic hippocampus [258]. In addition to elevated glutamate, other potentially excitatory amino acids have been assayed - elevations of aspartate, glycine and serine in association with seizures have also been noted with similar methods [259].

1.4.3.2. Neocortical epilepsy

Areas of focal spiking were compared with samples from surrounding nonspiking neocortex in surgical resection specimens from patients with partial seizures. In areas of spiking cortex, GAD and tyrosine hydroxylase were significantly increased; glutamic acid, GABA-aminotransferase and glutamine synthetase were not significantly different [260]. Excitatory amino acids were raised in human epileptic foci [261]. Comparison of tissue from an area of temporal neocortex shown at operation to be spiking showed decreased glutamic acid decarboxylase compared to non-spiking tissue [262]. No differences in B^gx o*" affinity (K^) of BZR were observed when epileptic foci were compared with nonspiking cortex [263], although a relative decrease in the number of a l postsynaptic receptor sites was seen [264]. 57

The role of specific neuronal populations in epileptic foci was studied by comparing epileptic and non-epileptic cortex removed from patients with low-grade gliomas [265]. Epileptic and nearby (within 1 to 2 cm) non-epileptic temporal lobe neocortex was identified using electrocorticography and shown histologically to be free of tumour. Although there was no significant difference in the overall cell count, a significant decrease in both somatostatin- and GABAergic-immunoreactive neurons was found (74% and 51%, respectively) in the epileptic cortex compared to that in nonepileptic cortex from the same patient [265]. Cortex from neocortical resection in patients suffering from drug-resistant epilepsy were examined for parvalbumin and calbindin-D28k immunocytochemistry to determine local circuit neuron populations [266]. Around neoplasms a reduced percentage of local circuit neurons was seen. Abnormal morphology and distribution of local circuit neurons was seen in patients with focal cortical dysplasia and in other focal migrational disorders, including neuronal nests in the white matter. Increased percentages of immunoreactive local circuit neurons and fibers in focal neocortical necrosis (cavernous angiomas), and diffuse hypoxic encephalopathy were detected. There was a lack of consistent morphologic abnormalities in the neocortex of patients with temporal lobe epilepsy, and in patients with cryptogenetic frontal lobe epilepsy [266].

Bursting activity was demonstrated from epileptogenic neocortex in vitro. This activity was suppressed by NMDA blockers [267]. Preliminary data has also been reported showing similar bursting activity suppressed by NMDA blockers in resected focal cortical dysplastic tissue [268].

1.4.4. Conclusions

The processes underlying abnormal cortical development give rise to disorganisation of cortical architecture with the potential for aberrant conections and abnormal synaptic activity. The mechnisms of epilefogenesis in localisation-related epilepsy depend centrally, in most models, on abnormalites of GABAergic function subserved at GABA^ receptors.

A simple model of epileptogenicity in both hippocampal and neocortical partial seizure disorders in which cell loss has occurred has been proposed by Engel [269] and is summarised in figure 1.13. This model may be of limited relevance to malformations of cortical development, since cell loss is not usually a prominent feature. Nonetheless, the bringing together of evidence pointing to both excessive excitation and preserved inhibition to create a model of hypersynchronisation may prove to be applicable in MCD. 58

Full causative explanations of epileptogenesis in localisation-related epilepsy in man are still lacking. Profound and unexpected insights are likely to emerge from genetic studies of such disorders, with the growing identification of single-gene disorders causing localisation-related epilepsies. Familial temporal lobe epilepsy is an autosomal dominant disorder in some kindreds [270]; another family with localisation-related epilepsy were shown to have a susceptibility gene mapping to lOq [271]. Localization of a gene for autosomal dominant nocturnal frontal lobe epilepsy to chromosome 20q 13.2 [272] was followed by the identification of a missense mutation in the neuronal nicotinic acetylcholine receptor a4 subunit [273]. Postulated mechanisms of epileptogenesis described above do not involve this receptor, suggesting that many present concepts may need to be re-examined. 59

CP

SP

IZ

vz ego

Figure 1.1. Diagram showing a simplified scheme o f embryonic neuronal migration in the neocortex. Cells are numbered in the order of arrival at the cortical plate.

MZ - marginal zone; CP - cortical plate;

SP - subplate; IZ - intermediate zone; VZ

- ventricular zone; CR - Cajal-Retzius cell; RG - radial glial fibre 60

Figure 1.2. Lissencephaly.

Coronal section through the T1-weighted MR! of a patient with lissencephaly. The smooth cortex, marked in black on the cartoon above, appears greatly thickened and the ventricles enlarged.

Figure 1.3. Subependymal nodular heterotopia.

Transverse section from T1-weighted MRI of subject KS, studied in Chapters 5, 9 and 11. A continuous nodular layer is seen lining the lateral ventricles, shown in black on the cartoon above.

Figure 1.4. Band heterotopia.

Transverse section through T1-weighted MRI of subject EG, studied in Chapter 11. A continuous band of grey matter is present bilaterally, shown on the cartoon above in black, giving the appearance of a “double cortex”. 61 1

Figure 1.5. Bilateral perisylvian polymicrogyria.

Transverse section through a 71-weighted MRI. The perisylvian cortex, marked in black on the cartoon above, appears greatly thickened. The polymicrogyric nature of this abnormality is only seen clearly in a few areas. A

Figure 1.6. Schizencephaly.

Transverse section through T l- weighted MRI of subject RH, studied in Chapters 5, 9 and 11. This subject has two clefts, one in the right central area and one in the left frontal lobe. The right central cleft is illustrated and indicated in black on the cartoon above; the upper part o f the left frontal cleft is also partly shown. The cortex lining the clefts is likely to be polymicrogyric.

■ Figure 1.7. Isolated heterotopic grey matter nodule.

Sagittal section through T1-weighted MRI of subject SN, studied in Chapter 5, 9 and 11. The grey matter heterotopion is above the trigone of the right lateral ventricle, marked in black on the cartoon above. 62

Figure 1.8. Focal malformation of cortical development, probably focal cortical dysplasia.

Transverse section through 71-weighted MRI of subject MW, studied in Chapters 5 and 11. The region of right lateral temporal cortex, shown in black on the cartoon above, has a thickened appearance, with blurring of the grey- white boundary and a macrogyric pattern. Cortical dysplasia is the likely explanation.

Figure 1.9. Dysembryoplastic neuroepithelial tumour (DNT).

Coronal section of a T l- weighted MRI of a patient with a left amygdala DNT, indicated in black on the cartoon above. 63

Meninges

1 Molecular layer

2 External granular layer

3 External pyramidal layer

^ # - 4 Internal granular layer

5 Internal pyramidal layer

6 Multiform layer

White matter

Figure 1.10. A simplified scheme o f intracortical circuits, layers and cell-types in neocortex. Modified from Barr, LB. and Kiernan, J.A. The Human Nervous System: an Anatomical Viewpoint. Philadelphia: Lippincott, 5th Edition, 1988. 64

Alveus

Pyramidal cell of CAT Basket cell

Schaffer collateral Temporal horn of lateral ventricle

Subiculum

Perforant path Pyramidal cell of CA3 Mossy fibre Dentate gyrus

Figure 1.11. Schematic diagram of intrahippocampal connections. Coronal section through right hippocampus. Modified from O'Keefe, J.O. and Nadel, L. The Hippocampus as a Cognitive Map. Oxford: Clarendon Press, 1978. 65

GABA Glutamate

Cortex

STN

GPi GPe / Striatum

SNr Thalamus VA, VL

Figure 1.12. Principal GABAergic and excitatory interactions between the frontal neocortex and the basal ganglia in man, after Flaherty and Graybiel

1994. STN - subthalamic nucleus; GPe - globus pallidus pars externa; GPi - globus pallidus pars interna; SNr - substantia nigra pars reticulata; VA - ventral anterior nucleus of thalamus; VL - ventrolateral nucleus of thalamus. 66

/ / / / / / / / / / 1 ????? / \ (a) (b)

Figure 1.13. (a) Schematic representation of normal cortical connections. Reciprocal innervation of a hypothetical neuronal system is shown, typical of synaptic organisation in both neocortex and hippocampus. Excitatory afferents terminate on the dendrites of principal neurons. Axon collaterals from these principal cells terminate on inhibitory interneurons which in turn make inhibitory contacts on the soma o f the same and adjacent (not shown) principal cells. (b) Schematic representation of cortical connections following cell loss and axon sprouting. Fewer afferent inputs sprout to innervate more principal neurons, predisposing to hypersynchronisation. Excitability is further increased by the development of monosynaptic recurrent excitatory collaterals. Although not demonstrated, it is possible inhibitory interneurons also show sprouting, further encouraging hypersynchronisa tion. 67 Chapter 2.

Rationale for using PET to study localisation-related epilepsy.

Although abnormalities of the GABAergic system in localisation-related epilepsy in man are very likely, investigation of this system is highly challenging. Invasive in vivo experimentation in man is not possible except in some circumstances asociated with surgery; similarly, in vitro study of human brain slices is severely limited. Study of post-mortem material is confounded by rapid post-mortem changes in the central nervous system. Therefore, in vivo methods of visualising normal and pathological cerebral functions are required. The radiotracer techniques employed in single photon emission computed tomography (SPECT) and PET permit measurement of regional blood flow, metabolic rate for glucose and parameters of binding to certain neuroreceptors.

2 .1 . SPECT

2.1.1. Blood flow

SPECT may be used to indicate relative regional cerebral blood flow (rCBF) using a variety of radiotracers, including ^^Sxenon, N ,N ,N '-trim ethyl-N '-(2-hydroxy-3- methyl-5-iodobenzyl)-1,3,propanediamine (HIPDM), hexamethylpropyleneamine- oxime (HMPAO), i23|_n_isopropyl-p-iodoamphetamine and s^n^Tc-ethyl cysteinate dimer (ECD). Investigations using these methods began before the era of high-resolution MRI, when a considerably higher proportion of patients would have been found normal with structural imaging; comparisons between imaging modalities in this substantial older literature now need cautious interpretation. In essence, interictal SPECT shows reduced regional rCBF, ictal and postictal studies show increased rCBF. Older studies, comparing SPECT with inferior technologies in other imaging modalities, are enthusiastic; more recent studies have defined a more limited role for blood flow SPECT. The literature below is reviewed chronologically.

One of the earliest studies revealed the essential phenomena of increased local blood flow in the ictal state and reduced flow interictally [274]. The particular localising value of regional hyperperfusion during ictal studies was soon amply confirmed [275], both in 68 secondary generalised, as well as partial, seizures [276, 277]. In the early MRI era, interictal SPECT localisation correlated with EEC in 75% and with MRI in only 61%, leading to the conclusion that, with these technologies, SPECT was superior to MRI [278]. With rapid progress in MRI technology, interictal SPECT was recognised to be inferior to MRI, though ’ ®F-deoxyglucose (i^FDG) PET was at this stage still regarded as the most sensitive modality [279]. However, further studies revealed the paradoxical and clinically limiting observations of interictal diffuse, multifocal and bilateral hypoperfusion, or focal hyperperfusion, as well as focal hypoperfusion [280] in addition to limited concordance with EEC - only 69% in one study [281]. Some regional abnormalities were possibly responsible for neuropsychological defects [282, 283].

MRI, with limited resolution, continued to be surpassed by ictal and even by interictal SPECT [283-290]. Further investigations confirmed the greater sensitivity of ictal SPECT: interictal studies were positive in 46-76% compared to 75-97% of ictal studies [291-298]. In a review of 80 selected patients who had ictal SPECT and EEC studies and temporal lobe localisation, diagnostic sensitivity of ictal SPECT judged by EEC was 90% [299]. The technical difficulties of ictal studies were somewhat assuaged by the finding that early postictal SPECT was more sensitive than interictal SPECT [300].

Patterns of blood flow change were clarified. Serial SPECT studies in the interictal, ictal and immediate postictal states were performed in 12 patients with refractory temporal lobe epilepsy to define the patterns and duration of peri-ictal cerebral blood flow changes. Visual and quantitative analysis showed a constant pattern of unilateral global increases in temporal lobe perfusion during seizures which suddenly switched to a pattern of relative mesial temporal (hippocampal) hyperperfusion and lateral temporal hypoperfusion in the immediate postictal period [295, 301]. “Reversed” crossed cerebellar diaschisis was observed in ictal studies, with cerebellar hyperperfusion contralateral to the cerebral focus [302]. Focal hyperperfusion was demonstrated in epilepsia partialis continua with normal EEC [303]. In temporal lobe epilepsy, ictal dystonia was associated with a relative increase in perfusion of the basal ganglia opposite the dystonie limb. Additionally, slight increases in cortical blood flow on the side opposite the direction of version were associated with head-turning, irrespective of the side of seizure focus [304].

Frequent discordance with EEG findings was noted in interictal studies [305-308]. The continued identification of multiple and bilateral abnomalities with interictal studies was ascribed to diaschisis, projection sites and mirror foci [309]. Although recognising the growing superiority of MRI over interictal SPECT, particularly of hippocampal T2 69 signal and hippocampal volume [310-314], it was found that preoperative interictal HMPAO SPECT could contribute to the prediction of postoperative verbal memory function [315] and appeared to be more helpful in children where the rate of detection of abnormality with MRI was markedly lower than in adults [316]. Nonetheless, studies continued to be produced in which interictal SPECT, now recognised to be of limited sensitivity, was compared with the inferior imaging by other modalities, such as X-ray CT, or only compared with EEG [317-320]; the value of such studies is, retrospectively, limited. Some later studies continued to report superiority of SPECT over MRI [321, 322]; the relative sophistication of the acquisition and analysis methods may be responsible for these conflicting studies.

In conclusion, interictal SPECT is of little value in the current era; ictal SPECT may be helpful in patients with normal imaging, especially in temporal lobe epilepsy (TLE). 539 patients reported in the literature with interictal SPECT have been reviewed [299]. 291 patients were localised by SPECT to the temporal lobe, 65 to extra­ temporal locations and 183 were unlocalised. 79% showed concordance of the interictal SPECT and EEG localisation. In a recent overview of the extensive experience from the Austin Hospital, ictal SPECT achieved 97% correct localisation in unilateral temporal lobe epilepsy, compared with 71% for postictal SPECT and 48% for interictal scans. In extratemporal seizures ictal SPECT studies localized the focus in 92%, compared to 46% for postictal studies and interictal SPECT was of little value [323]. Ictal ^^"^Tc- HMPAO scans may be especially useful in the evaluation of patients with extra-temporal seizures with normal MRI, localizing the seizure focus in 10 of 12 patients in one series and in 20 of 22 in another [324, 325]. The latter study revealed asymmetric tonic posturing, contralateral head and eye deviation and unilateral clonic jerking were associated with an ictal increase in CBF in the frontocentral, medial frontal or dorsolateral areas. Focal ictal increases in CBF have also been found in patients with parietal lobe epilepsy, some with normal MRI [326]. The coregistration of interictal and ictal SPECT images, to result in an "ictal difference image" that may be coregistered with an individual's MRI enhances the accuracy of data interpretation [327].

Blood flow SPECT has been used in a small number of studies specifically investigating developmental disorders. HMPAO SPECT demonstrated cerebral blood flow similar to cortex in two patients with heterotopia, in the areas of their heterotopic lesions [328]. Profound hemispheric hypoperfusion in the affected hemisphere in 6 children with hemimegalencephaly has been noted [329]. During ipsilateral EEG discharges in a child with hemimegalencephaly, rCBF was 40% higher in all regions of the affected hemisphere; interictally, rCBF was 45% lower [330]. Interictal ^^f^Tc-nMPAO SPECT imaging has recently revealed identical or increased cerebral perfusion of laminar 70 heterotopia in comparison to the overlying neocortical gray matter [331]. Postictal SPECT has shown a focal increase in rCBF in areas of focal cortical dysplasia [332] and a focal ictal increase in rCBF rise has been demonstrated in patients with malformations of cortical development and non-localizing ictal scalp EEG [333].

2.1.2 Benzodiazepine receptors

^23|-Ro 1 6-01 54 ( ^23|-jomazenil) was shown to have intracerebral retention parallel to the known distribution of BZD receptors in man [334]. A multi-centre European trial of i23|-jomazenil SPECT was reported. Examining the predictive value for identifying the seizure focus in localisation-related epilepsy, a 100% positive predictive value for iomazenil compared to 92% for flow images was found. Negative predictive values were calculated as 81% for ^23|_jomazenil and 54% for flow images [335]. ’ 23|.iomazenil SPECT in 17 patients who were considered candidates for surgery because of medically intractable complex partial seizures showed asymmetry in over 80%, in agreement with EEG findings [336].

In a comparative study of 12 patients with localisation-related epilepsy and normal MRI, 10 patients had focal reduction of HMPAO uptake and 9 patients had focal reduction in ^23|_jomazenil uptake. In 8 these abnormalities were coincident [337]. In 10 patients with normal MRI but a region of interictally reduced blood flow with HMPAO SPECT, all showed a concordant reduction in i23|_iomazenil binding [338]. 14 patients with reduced interictal HMPAO binding regions showed reduced i23|_jomazenil binding in 11, the reduction with iomazenil being more marked. In this study clobazam was shown to have no effect on the interpretation of iomazenil SPECT images [339]. These studies suggest little advantage in imaging BZR over blood flow. However, in a patient with focal epilepsy due to MCD a marked focal decrease in ^23|.jomazenil uptake was seen in this area but normal HMPAO uptake [340]. Some animal evidence suggests BZR imaging should be clinically useful. In hippocampal kindled rabbits, in vivo double tracer autoradiography using i^sj.iomazenil and ^^f^Tc-HMPAO, ’ 25|_iomazenil accumulation was more markedly and extensively decreased than ^^f^Tc-HMPAO accumulation [341].

The reason for the disappointing results of clinical application of SPECT BZR imaging may be methodological. Acute occlusion of the middle cerebral artery in rats has been shown to cause no acute change in the number of BZR. However, in an experiment in which dual tracer studies were carried out simultaneously with occlusion, reduction of HMPAO binding was seen which correlated very strongly with reduction in iomazenil binding, indicating that a single image of iomazenil binding in man, without any kinetic 71 modelling, may be seriously confounded by changes in regional blood flow [342]. Tracer kinetic and equilibrium methods have been shown in man to give good measures of BZR parameters without the confound of blood flow; however, dynamic imaging and blood sampling are required [343]. In the baboon, very good agreement with ex vivo data has been shown for regional B^gx and values using dynamic SPECT [344]. These experiments also demonstrated the inadequacy of semi-quantitative methods commonly used in clinical practice in comparison with rigorous kinetic and equilibrium techniques [345].

More recent studies using higher resolution equipment and optimal scan orientation have suggested that the area of reduced specific binding of i23|.|omazenil is more restricted than the defect of blood flow and of greater sensitivity for the localization of an epileptogenic focus [346, 347].

2 .2 . PET

2.2.1. Cerebral metabolic rate

T8FDG has been used with PET to measure regional cerebral metabolic rate for glucose (rCMRgluc). The underlying hypothesis permitting interpretation of ^®FDG PET data is that a local increase in neuronal activity will be reflected by an increase in rCMRgluc, and a local decrease in activity by a decrease in rCMRgluc. Localised ictal activity in animals was shown 25 years ago to result in a localised increase in rCMRgluc: In a baboon, intraamygdaloid injection of kainic acid resulted in repetitive limbic status epilepticus and susequent localisation-related epilepsy. During the interictal state, T8FDG uptake was normal. During the ictal state ^®FDG uptake was increased [348].

However, most human studies have been of the interictal state; the hallmark in localisation-related epilepsy has been a regional decrease in rCMRgluc interictally, although a study of drug-naive patients found interictal increased rCMRgluc [349]. Several studies of ^ ®FDG PET have found a 60-90% incidence of hypometabolism in the temporal lobe interictally in adults and children with temporal lobe epilepsy [279, 3 5 0 -3 5 8 ]. ^®FDG PET revealed a hypometabolic region in 25 of 31 patients with localisation-related epilepsy; MRI detected abnormality in only 10, though with the now obsolete technology appropriate to the study era [355]. Surgical ECoG findings confirmed the epileptogenic nature of regions of interictal hypometabolism [359]. The hypometabolic zone is frequently extensive and, particularly in TLE, may involve regions outside the ictal onset area; in patients with a unilateral temporal lobe focus the 72 asymmetry of metabolism was more pronounced in the lateral temporal cortex than in the mesial part of the temporal lobe. In most cortical and subcortical regions on the side of the epileptic focus, glucose consumption rate was lower than in the contralateral region or than in controls [358, 360]. Hypometabolism of basal ganglia and thalamus has been further confirmed in patients with TLE of both limbic and neocortical origin [361]. A well-localised region of decreased glucose metabolism is found less frequently in extratemporal seizure disorders than in TLE [362].

Although regional hypometabolism may be detected in patients with lesions or normal MRI, the clinical utility of such a finding is controversial. In patients judged to have temporal lobe epilepsy, the presence of temporal lobe hypometabolism was associated with a better outcome after ipsilateral temporal lobectomy than in patients without hypometabolism [363]. Conversely, in 31 of 40 TLE patient with normal MRI who had unilateral temporal hypometabolism, unilateral EEG spikes were a better predictor of surgical outcome than ^^FDC PET findings [364]. Further, in a comparison of CT, MRI, T8FDG PET and depth EEG in 22 patients, MRI correlated with depth EEG in 50%; regional abnormalities were found with quantified PET in 95% of the patients but were concordant with depth EEG for side of onset in only 77% of the patients and for lobe of onset in only 59%. A possibly false localising PET result for lobe of onset was obtained in 36%, limiting the usefulness of such data [365]. However, of 37 patients with concordant findings with non-invasive EEG and ^®FDG PET, depth studies did not reveal discordant findings in any case, suggesting ^®FDG PET might obviate the need for depth recording [366].

In a comparative study of ^^FDG PET, HMPAO SPECT and MRI in 20 patients with TLE, PET showed an abnormality in 80% with normal MRI and 100% with MRI lesions; SPECT revealed abnormality in 20% and 80% respectively [313]. With increasing sophistication of MRI technology, the place of ’ ®FDG PET in detecting a lesion has been questioned. Of 24 patients with partial epilepsy of neocortical origin, two-thirds of patients had regions of hypometabolism, but these were usually associated with structural imaging abnormalities [367]. More profound reduction of metabolism was noted in patients with increased T2 signal in the mesial temporal lobe (presumably indicating HS) compared with those without increased T2 signal [357]. Similarly, 16 of 18 patients with TLE had temporal lobe hypometabolism on ^®FDG PET; 11 had clear concordant MRI abnormalities, suggesting ^®FDG PET did not provide clinically useful data if the MRI findings were definite, but had some additional sensitivity [368]. Although in one study hippocampal neurone loss was not correlated with the degree of temporal lobe hypometabolism [369], recent comparisons of hippocampal atrophy and glucose metabolism in patients with TLE found that the degree of hippocampal atrophy 73 correlated with the degree of hypometabolism in the hippocampus [368, 370].

The superiority of quantitative over visual interpertation of ^^FDG PET in TLE has been demonstrated [371]. It may be possible to improve the delineation of the hypometabolic zone in TLE by performing activation tasks during the scan, such as “emotional speech”, which increase the contrast between normal and abnormal regions in the temporal lobe [372].

In paediatric practice ispDG PET may have a more valuable role. ^®FDG PET has enabled a characterisation of the functional abnormalities in children with severe epilepsies, in some cases enabling remarkably effective surgical intervention in children with normal MRI [373]. In eight children with intractable neonatal onset seizures, interictal PET revealed unilateral diffuse hypometabolism in three, leading to highly effective hemispherectomy. In one child, ictal PET showed hypermetabolism in the left frontal cortex, left striatum, and right cerebellum; again, a large resection was immensely successful [374]. In 13 children with infantile spasms of undetermined cause, unilateral hypometabolism involving the parieto-occipito-temporaI region was found in 5; MRI was normal in 4 of these. 4 of these 5 infants underwent surgical removal of the cortical focus with seizure-free outcome; pathological examination of resected tissue in each showed microscopic cortical dysplasia [375].

Despite reservations about the current place of ispDG PET in clinical investigation, this technique has yielded interesting information relating to the neurobiology of seizure disorders. Some differences exist in the metabolic patterns associated with different aetiologies of TLE: patients with HS have the lowest ’ ®FDG uptake in the entire temporal lobe, followed by patients with lateral temporal seizure origin. Patients with tumours in the mesiobasal temporal lobe show only a slight decrease of glucose metabolism [376]. The comparison of metabolic patterns on ^®FDG PET with video EEG telemetry in 48 patients revealed patients with frontal hypometabolism alone had shorter ictal and postictal durations; auras were more likely to be present in patients with temporal hypometabolism alone; other metabolic patterns did not predict specific ictal clinical features [377]. Relative hypometabolism of the left hemisphere correlated with lower cognitive performance in 13 unilateral TLE patients. Hypometabolism of the left lateral temporal lobe and thalamus independently correlated with verbal memory difficulties [378]. Patients with localisation-related epilepsy had bilateral cerebellar hypometabolism that was not fully explained by treatment with phenytoin [371]. The metabolic consequences of successful surgery have been investigated. After selective amygdalo-hippocampectomy in patients with HS there was an increase of rCMRgluc in the ipsilateral and also the contralateral hemisphere [379]. 74

^ ®FDG PET has been employed to examine patients with MCD. Glucose metabolism similar to normal cortex has been detected using ’ ^EDG PET in the ectopic neurons in band heterotopia [380, 381] and in heterotopic nodules [382, 383]. Of 17 patients with MCD, fifteen had abnormal PET findings, consisting of focal hypometabolism in 9 patients and displaced metabolic activity of normal gray matter, reflecting underlying grey matter heterotopia, in 6. All 15 patients had MRI abnormalities [384]. 4 of 8 children with hemimegalencephaly had regions of reduced metabolism in the unaffected hemisphere [385]. A study including both patients with acquired lesions and congenital abnormalities such as tuberous sclerosis and agenesis of the corpus callosum showed regions of decreased metabolism which were more extensive than the MRI lesions in some cases [386]. Attempts have been made to determine the functional capacity of dysgenetic regions. A medical student with extensive unilateral heterotopic grey matter underwent a resting ^ ®FDG PET in which metabolic activity similar to normal cortex was seen in the heterotopia; during a second scan the patient undertook a verbal fluency task which resulted in a slight increase in the metabolic rate in the heterotopic region [387].

A small number of studies have examined ictal metabolic patterns. By chance, 18 children undergoing ^®FDG PET had seizures during the ^®FDG uptake period. Three major metabolic patterns were determined based on degree and type of subcortical involvement: Nine children had asymmetric glucose metabolism of striatum and thalamus, usually involving unilateral cortical hypermetabolism and crossed cerebellar hypermetabolism. Five children had symmetric metabolic abnormalities of striatum and thalamus; this pattern was accompanied by hippocampal or insular cortex hypermetabolism, diffuse neocortical hypometabolism, and absence of any cerebellar abnormality. Four children had hypermetabolism restricted to cerebral cortex [388]. Focal hypermetabolism has been noted in EPC [389]. Hypometabolism is accentuated after a seizure and may not return to the interictal state for more than 24 hours [390].

In summary, although ^^FDG PET has been useful in the past for detecting a seizure focus possibly amenable to surgery, the improvements in MRI technology have rendered such routine clinical investigation superfluous in the majority of patients. However, in selected MRI normal patients, particularly children, a place remains for evaluation with 18FDG PET, with the caveat that the hypometabolic region may be found to be nx)re extensive than the ictal onset zone. However, i®FDG PET has provided valuable information about metabolic abnormalities in an extensive cortical-subcortical network and an insight into patterns of ictal metabolism. 75

2.2.2. Opioid receptors

The rationale for employing opioid receptor ligands is the demonstration that endogenous opioids are released following partial and generalized tonic-clonic seizures and contribute to the post-ictal rise in seizure threshold [391]. An increase in ^i-agonist T^C-carfentanil binding to ^-receptors in lateral temporal neocortex in areas which also showed reduced glucose metabolism has been shown [392]. This may be a tonic antiepileptic system that serves to limit the spread of electrical activity. The increase in i^C-carfentanil binding to lateral temporal neocortex was confirmed and reduced binding to the amygdala was noted [393], though this may be an artefact of partial

volume effect. Using ’ ^C-diprenorphine, which binds with similar affinities to n , k and Ô subtypes of opioid receptors, there were no abnormalities of binding of ’ ^C- diprenorphine in the temporal lobe or elsewhere [393, 394]. Also there was no overall

asymmetry of binding of ^^F-cycloFOXY, which binds to p, and k receptors in patients with TIE [395]. The explanation of these findings is not clear, although either an

upregulation of ja receptors and reduction of number or affinity of k receptors, or

upregulation of receptors and occupation of k receptors by an endogenous opioid ligand would fit the evidence.

2.2.3. Other ligands

Deprenyl binds to monoamineoxidase (MAO)-B receptors, which are mainly located on astrocytes. In 9 patients with TLE ^ ^ C-deuterium deprenyl binding was increased in the epileptogenic temporal lobe, possibly reflecting gliosis [396]. ^^C-doxepin was increased in epileptic foci also showing reduced interictal glucose metabolism [397], although the explanation for this finding requires clarification.

2.2.4. Benzodiazepine receptors

2.2 .4 . 7. Alcohol

In mice it has been demonstrated that acute ingestion of high doses of alcohol causes an increase in ^H-flumazenil binding [398]. Using a saturation procedure with two PET experiments for each subject, no differences in B^^^ or were found following either acute alcohol injestion in previously healthy men, or in men with chronic alcohol dependency [399, 400], though the variance in B^^ax may be greater among alcohol dependent subjects than in normals, possibly implying multiple effects of alcohol on the benzodiazepine receptor [401 ]. 76

2.2A.Z. Neurodegenerative conditions

No abnormality of ^ ^C-flumazenil binding was found in a group of 9 patients with Friedreich's ataxia [402], using a multi-injection protocol. Since the pathology may be primarily in regions of low i^C-flumazenil binding this is perhaps not surprising.

T^C-flumazenil volume of distribution (V^) was significantly decreased by 26% in the caudate nucleus in 6 untreated patients with Huntington's disease [403]. This is commensurate with data obtained in postmortem autoradiographic studies of receptor density. The postmortem changes known to occur in putamen and thalamus were not found, possibly because the patients studied with PET had relatively early disease. The T ^C-flumazenil PET findings may prove to be a useful tool for early diagnosis and for monitoring the effects of treatment.

Extensive changes in ^^C-flumazenil in the motor cortex, dorsolateral and medial prefrontal regions have been found in 10 patients with motor neuron disease compared to normals [404]. Though this may simply reflect neuronal loss in these areas, it may be a consequence of altered function in GABA^ mediated inhibitory circuits. Again, this raises the tantalising possibility of a method for monitoring response to treatment, which may be available in the future.

2 2 .4 .3 . Idiopathic recurrent stupor

A recent case report describes a 48 year old woman with idiopathic recurrent stupor whose T^C-flumazenil PET was normal interictally, but during a period of stupor flumazenil binding was greatly reduced. Drug screens and mass spectrometry for benzodiazepines were negative, but during the stupor endozepines, purified by high pressure liquid chromatography (HPLC) and measured by radioreceptor assay, were detected suggesting strongly that PET can detect endozepine release in this condition [405].

2.2.4.4. Epilepsy

The epilepsies have been more extensively studied with ’ ^C-flumazenil PET than other conditions. Using PET and ^’C-flumazenil Savic et al. found a significant 29% decrease in in the region of the epileptic focus in 10 subjects, six of whom had medial temporal epileptic foci [406]. A further 8 patients, 6 with normal MRI, showed focal decreases in BZR density which were more closely localised than reduction in glucose metabolism as indicated by ^^FDG PET [407]. The reduction in BZR density in the 77 epileptogenic hippocampus has been further confirmed in other studies [408-410]. The study of Koepp et al. addressed in detail the question whether changes in BZR density are seen elsewhere in the brain in hippocampal epilepsy: no differences in binding were found outside the affected hippocampus, which had a reduction in of 23% [410],

Although the finding of reduced benzodiazepine density in the hippocampus (and not elsewhere) of patients with hippocampal epilepsy seems clear, one important issue is inadequately addressed: to what extent is the apparent reduction of tracer binding actually an artefact resulting from atrophy of the affected hippocampus? The dimensions of the hippocampus are of the same order as PET resolution, hence a reduction of volume of the hippocampus leads to increased partial volume effect and so an apparent reduction in intensity at that point. Methods to “correct” PET images for partial volume effect are required, A recent study using such a method showed reduced cBZR over and above loss of hippocampal volume in hippocampal sclerosis [411].

The pathological basis underlying reduced ’ ’ C-flumazenil binding in hippocampal sclerosis is not yet fully elucidated. Autoradiographic and histopathological studies of surgically removed sclerotic hippocampi have shown reduced neuron counts and cBZR density, with a reduction of cBZR density over and above the neuronal loss [254]. A highly statistically significant correlation between neuronal loss and reduced cBZR binding on autoradiographic analysis of hippocampi from patients treated surgically for hippocampal sclerosis has been observed [412]. This study concluded that neurone loss was the sole basis of reduced ^’C-flumazenil binding in vivo. However, the possibility of cBZR loss over and above neuronal loss was not excuded by this study.

Epileptic foci outside the hippocampus have not been systematically studied, although small numbers of cases have been described with focal decrease in receptor density [406].

The place of ’ ’ C-flumazenil PET in the clinical investigation of patients with refractory localisation-related and unremarkable high quality MRI has been little investigated. In six patients with frontal lobe epilepsy, reduction in cBZR binding was demonstrated with T^C-flumazenil PET, that was consistent with clinical and EEC data [413]. In 2 of 4 patients who also had ^^FDG scans the region of reduced metabolism was more extensive than the reduction in cBZR. The MRI was unremarkable in five, implying greater sensitivity of PET. Comparisons of this nature, however, crucially depend on the relative sophistication of the instruments and methodologies used. Greater reduction of 1’ C-flumazenil binding in the putative seizure focus has been reported in patients with frequent seizures compared to those with less frequent seizures [414]. This finding has 78 not been confirmed by other studies.

Primary generalised epilepsy hes been much less studied than focal epilepsy. Diagnostic classification is often difficult; this may explain the contradictory results noted in different studies. There is Increasing evidence that the fundamental abnormality underlying primary generalised epilepsy appears to lie at the GABAg receptor [415], so abnormalities of ’ ^ C-flumazenil binding might not be expected. Childhood and juvenile absence epilepsy patients were shown to have normal interictal ^^C-flumazenil Vy [416]; other primary generalised epilepsy patients have been shown to be not significantly different from normals [417] or to have increased receptor density in the cerebellar nuclei and decreased receptor density in the thalamus [418]. This latter finding is intriguing, pointing to abnormalities outside the thalamocortical loop usually regarded as the substrate of generalised seizure discharges [419]; the finding has not yet been replicated in man or confirmed in animal studies. Further work is needed to clarify the findings in well-defined diagnostic groups.

The effects of drugs on benzodiazepine receptor binding in man have not been widely examined using ^’C-flumazenil PET. Treatment with sodium valproate was associated with a significant reduction in in a cross-sectional study of patients with childhood and Juvenile absence epilepsy [416], suggesting a possible mechanism of therapeutic action.

Absence seizures were induced by hyperventilation in susceptible volunteers during the course of a study. Comparing the ^^C-flumazenil binding with that in a scan without seizure provocation in the patients, with scans of normals and with computer simulations of the effects of changes in blood flow or receptor binding due to seizures, alterations of benzodiazepine receptor characteristics during absences were not detected [420].

2.3. OTHER FUNCTIONAL STUDIES IN MCD

Language representation was mapped by electrical stimulation of chronically implanted subdural electrodes in 34 patients being evaluated for epilepsy surgery, 28 of whom had magnetic resonance imaging or histological evidence of MCD. Developmental lesions were generally associated with a normal location of language cortex even if the lesion extended into the language area, whereas lesions acquired before the age of 5 years were in some cases associated with relocation of language representation to the opposite hemisphere [421]. SSEPs were recorded in patients with MCD affecting eloquent sensory areas. 79

Approximately half of the patients demonstrated normal SSEPs recorded over regions of MCD [422]. Both of these studies suggest the possibility of functional capacity of regions of MCD. 80

2 .4 . AIMS

In summary, methods of functional imaging may be used to elucidate the metabolic, blood flow and neuroreceptor consequences of localisation-related epilepsy. SPECT is limited by poor spatial resolution and, usually, by acquisition techniques which do not permit kinetic or equilibrium modelling of tracer parameters and by analysis techniques which do not permit quantitation. The requirement for these techniques, in order to eliminate the confounding effects of blood flow and to allow the greater sensitivity of quantitative methods, has been amply demonstrated. Using PET, kinetic or equilibrium approaches and quantitative analysis, ’ ^C-flumazenil binding has consistently been shown to be abnormal in a more restricted pattern than abnormalities of metabolism, suggesting that T^C-flumazenil binding abnormality may be a more faithful reflection of the epileptogenic zone, without the confounding effect of diaschisis. No prior study has examined ^^C-flumazenil binding in MCD. Further, no prior study has attempted to clarify the aetiological correlates of abnormalities of ^^C-flumazenil binding, or any subcortical ^^C-flumazenil binding abnormalities. Recent studies have begun to suggest a role for iiQ-flumazenil PET in patients with normal MRI. The capacity of regions of MCD to support normal cortical functions has been examined with PET in case reports only and using questionable methodology; attempts have been made to examine this question using electrophysiological methods, but these do not permit anatomical visualisation of the findings.

The aims of the investigations carried out were:

1) To investigate benzodiazepine receptor binding in MCD using ’ ^C-flumazenil PET.

2) To investigate the functional capacity of regions of MCD using PET and to seek evidence of reorganisation of cortical functions

3) To investigate the characteristics of any abnormalities of ^^C-flumazenil binding in patients with extratemporal localisation-related epilepsy and acquired cerebral injury.

4) To investigate the characteristics of any abnormalities of ^^C-flumazenil binding in patients with localisation-related epilepsy and dysembryoplastic neuroepithelial tumour.

5) To investigate the characteristics of any abnormalities of ^’C-flumazenil binding in patients with extratemporal localisation-related epilepsy and unremarkable high- resolution MRI and then to attempt to interpret any abnormalities seen using the results 81 found in patients with known aetiologies.

6) To investigate the characteristics of any abnormalities of ^^C-flumazenil binding in the subcortical structures of patients with extratemporal localisation-related epilepsy.

7) To develop a technique for combining structural MRI data and functional PET data at the voxel level, in order to determine if observed abnormalites of ^^C-flumazenil PET are due to underlying anatomical, rather than functional, abnormality.

The hypotheses which were tested were:

1) MCD is associated with abnormalities of ^^C-flumazenil binding which are more extensive than the MRI abnormality.

2) Regions of MCD will fail to be activated by cognitive and motor tasks and cortical reorganisation occurs in this context.

3) Different aetiologies of localisation-related epilepsy are associated with characteristic ^^C-flumazenil binding changes.

4) Extratemporal localisation-related epilepsy is associated with abnormalities of the GABAergic system in subcortical structures which will be visualised with ^^C- flumazenil PET.

5) Abnomalities of structure do not account for all the abnormalities of function seen in localisation-related epilepsy with ^’C-flumazenil PET. 82 Chapter 3.

Methods 1 : Principles of positron emission tomography.

In this chapter, the first of two methodology chapters, the general principles of PET methods are described. The details of the experimental methods used are described in the subsequent chapter.

3.1. PRINCIPLES OF PET

PET provides a means for the in vivo measurement of regional cerebral blood flow, metabolic rate or binding parameters of a variety of neuroreceptor ligands and enzyme substrates in man. These measures are obtained as tomographic images. PET can be performed with the subject in the resting state or during, or following, a variety of potentially perturbing stimuli, such as the performance of a cognitive or motor task, the administration of a drug or the occurence of a seizure. Positron-emitting isotopes with short half lives are used in PET, commonly and ^®F, with half-lives of 2, 20 and n o minutes respectively. Because of the short half-lives of and it is necessary to produce many PET radiotracers with an on-site cyclotron, severely limiting the availability of the technique. Although incorporated into a tracer molecule by substituting for hydrogen or a hydroxyl group, hence potentially altering the properties of the molecule, and replace the naturally occurring element and so do not alter the properties of the molecule. Molecules with specific properties (e.g. receptor binding) can be labelled with positron emitting chemical nuclei and then injected into a subject [423].

The positron is emitted from the radioligand and, after travelling a short distance, is annihilated following collision with an electron. Through Einstein’s principle of the conservation of mass-energy (e=mc^), two photons (y rays) are emitted. The mass and velocity of the positron and electron are both relatively small, thus the momentum of the collision is almost zero. Momentum is conserved in a collision, therefore the two photons travelling at the speed of light travel in diametrically opposite directions (180° to each other) in order to conserve virtually zero momentum (figure 3.1). The PET camera consists of a ring of detectors arranged around the organ(s) of interest. The 'coincident' photons are absorbed in diametrically opposing detector crystals at approximately the 83 same time (figure 3.2). The advantage of imaging a positron-emitting ligand is that the detection of coincident photons establishes the origin of those photons to be somewhere along a line linking the two detectors (line of response) (figure 3.3). Additionally, the probability of two photons arriving simultaneously at diametrically opposing detectors by chance is very small, hence the signal to noise ratio is very favourable. In contrast, SPECT cameras, which detect single emitted photons, cannot distinguish between a single photon emitted from the radiotracer and a photon arriving at the detector by chance, hence the signal to noise ratio is very much less favourable.

The data collected by the PET scanner may be mathematically reconstructed to produce tomographic images of tissue radioactivity concentration [424]. The resolution of a PET camera is defined by the apparent width of a point source of radioactivity positioned in the field of view at half the maximum value (full width at half maximum, FWHM). Two point sources cannot be resolved unless separated by at least one FWHM. The resolution of PET is limited by the short distance that the emitted positron must travel in order to encounter an electron and be annih/feted. This distance varies with the isotope used, since positrons of different energies are emitted by different isotopes. For this theoretical best resolution is approximately 2mm.

In the past, images were obtained with lead septa separating each ring of detectors collecting each transaxial slice of data (two-dimensional or 2D imaging). The number of possible lines of response in 2D is small, hence the electronic linking of detectors to detect coincidences and mathematics of reconstruction are relatively simple. However, many photons will be absorbed by the lead septa and will not reach a detector, hence reducing the sensitivity of the camera. By removing the septa, the absorbtion of photons no longer occurs, resulting in a very substantial gain in sensitivity (three-dimensional or 3D imaging, figure 3.4) [425]. However, two problems arise: Firstly, at each detector a very large number of lines of response is possible, requiring complex electronics and reconstruction algorithms; secondly, the problem of scatter arises. With very many photons being emitted from the organ of interest, some of these photons inevitably collide with subatomic particles and are deflected from their original trajectory, or scattered. Although such photons will no longer be able to arrive coincidentally with the “partner” photon travelling in the opposite direction, by chance some scattered photons will arrive at a detector at the same moment as another photon. Because the number of photons being emitted is so great, the proportion of photons which become scattered is not insignificant. In 2D, the scattered photons usually deviate from the possible lines of response of paired detectors and are absorbed by the septa, so scatter is much les of a problem in 2D. The scatter fraction varies across the field of view of the scanner due to its geometry, and also will vary depending on the local level of 84

radioactivity in the organ of interest. Hence, scatter needs to be subtracted from the data during reconstruction in order to enable accurate radioactivity measures. This can be done by modelling scatter according to the known geometry of the camera [426] or by measuring the scatter simultaneously with the unscattered data [427]. The former method is easier but cannot account for scatter from photons emitted from parts of the subject outside the field of view of the camera, which may be significant; the latter method has the advantage of allowing removal of the real scatter, rather than merely a model of it, which allows for out of field of view scatter. Measuring scatter depends upon the observation that scattered photons have lost some of their energy through a collision with another particle; it is possible to adapt the detector properties such that photons can be collected at two different “energy windows” simultaneously, one set at the level of the emitted photons (the “data window”) and the other at a lower energy level appropiate for scattered photons (the “scatter window”). This is known as dual-window scatter correction [427]. One possible scatter modelling technique is termed convolution-subtraction scatter correction [426].

Over the time period of the scan the distribution of the radioligand in the organ of interest is likely to change. In the case of a neuroreceptor ligand, the inital period of the experiment will show a radioactivity distribution reflecting blood flow and ligand delivery. As the PET scan continues the ligand will become bound and the radioactivity distribution will then reflect the pharmacokinetics of the ligand at the receptor. In addition, the radioactivity measurements will be affected by decay of the radioisotope. Depending on the ligand in question, the time course of these changes is likely to be of the order of minutes to hours, hence the acquisition of the PET data will need to continue through this time. In order that the changing radioactivity distribution may be recorded, the scanning period is subdivided into a number of discrete intervals known as “frames”. In turn, the data for each frame are subdivided spatially into a very large number of very small subvolumes, usually roughly cubic, known as “voxels”. All of the voxels for each frame can be assembled as an image.

The data are recorded in such a way that at each voxel there is a different voxel value (radioactivity counts per unit time) for each frame of the experimental period, hence there is an image corresponding to each frame. The change in voxel value for a particular voxel over time is, in essence, a time-activity curve for that region of the organ of interest. The set of images corresponding to all the frames is, taken together, a very large set of time activity curves for all regions within the scanner field of view. This process of collecting data subdivided over time in order to detect changes in radiotracer distribution is termed “dynamic imaging”. Although a conventional technique with PET, it is possible, though rarely done, to perform dynamic imaging with SPECT. 85

Frequently, data are also collected from the subject’s blood. A variety of parameters of the ligand will be affected by the rate of delivery and wash-out of the ligand, which are in turn affected by the concentration of radioligand in the blood and the partitioning of the ligand between plasma, blood cells and plasma protein binding. The concentration of radioligand in the blood is also affected by the rate of decay of the isotope and any metabolism of the ligand. Although venous blood sampling is easy, many of these measures will differ between venous and arterial blood. Arterial blood sampling is usually essential in order to determine the availability of the radiotracer in question to enter from the blood into the organ being scanned. In general, a continuous measure is collected of total radioactivity in arterial blood over time. In order to determine how much of this total radioactivity is unmetabolised available radioligand, metabolites are assayed and the partitioning between the plasma, intracellular and protein bound pools is measured in discrete blood samples taken throughout the scanning period. These data are used to correct the continuous measure of a rterial blood radioactivity, in order to produce a time-activity curve for unchanged, available radiotracer in arterial plasma. This is termed an “input function”.

These data, the dynamic image set and the input function, can then be used to derive regional parameters of ligand binding, such as available receptor density (B^gx)» receptor affinity (K^) and binding potential or volume of distribution between blood and the organ in queston (V^). These parameters are derived via modelling techniques (figure 3.5). A model is a set of internally consistent hypotheses to explain the behaviour of complex systems, based on available data from that system, which is used to interpret novel observations from that system. The essential underlying assumption is that the tissue time-activity curve at each voxel is the convolution of the input function and the tissue response function which describes the physiological process under study at each voxel.

3.2. 1 ^C-FLUMAZENIL PET

PET has been used to image benzodiazepine receptors for over ten years [428]. Through this time a number of methodological issues have been addressed and clarified, allowing the selection of optimal methods [423]. 86

3.2.1. Ligands for benzodiazepine receptor imaging

A number of non-selective benzodiazepine site agonists have been used as labelled ligands to investigate benzodiazepine receptors in man and animals: ^^C-flunitrazepam [429], T ^C-suriclone [430] and ’ ^C-triazolam [431]. All of these agonists have significantly more non-specific binding than the neutral antagonist flumazenil. ^ ^C-flumazenil has been used for the vast majority of the studies carried out to date.

The characteristics of ^ ^C-flumazenil are that it is specifically bound exclusively in the brain, it crosses the blood brain barrier easily, it is not metabolised in brain, its non­ specific cerebral binding represents about 10% of the specific binding (small but not negligible), it has polar metabolites (one of which is radiolabelled) which do not cross the blood brain barrier, it dissociates from the benzodiazepine receptor rapidly (tissue half life estimates vary between 12 and 22 minutes), it reaches peak concentrations in the head by about 10 minutes and it is localised principally in cortex with some binding in hippocampus, basal ganglia, thalamus and cerebellum from which it washes out at a differential rate than from the cortex.

T^C-Ro 15-4513 [432], a partial inverse agonist, has also been employed; it binds to receptor subtypes which incorporate the a6 subunit, which is not labelled by flumazenil [433].

The discussion that follows will focus exclusively on PET scanning using ^C-flumazenil.

3.2.2. Parameters

Generally, an image of radioactivity distribution is not sufficient; in most settings, an attempt is made to derive quantified parameters from the dynamic image series, either for regions of interest or for every voxel (“voxel-by-voxel”). Possible neuroreceptor parameters of interest which might be measured are ^nd binding potential, related to the ratio of and K^. Of note is that in order to separate and B^g^ at least two scans are needed [434] while one suffices for a ratio measure. In clinical studies this is an important decision since some patients may find it difficult to tolerate more than one scan. In addition, the longer the time separating different scans in the same subject the more likely it is that extraneous factors may change the desired measures. Further, it is unlikely that a person could have more than three scans in one year as the maximum absorbed dose of radioactivity allowed for volunteers (at present in the UK) would be exceeded. 87

Since it has been demonstrated that ’ ^C-flumazenil dissociates very rapidly from the benzodiazepine receptor, the measure obtained with only one scan is a volume of distribution (V^) [435] rather than a binding potential. This may suffice in many situations unless one postulates that B^gx and change by the same proportion and in the same direction in which, case the ratio of the two would remain unchanged. A compromise might be to undertake a small number of studies in order to establish that Kj is not different between groups of subjects under comparison, and then to collect larger numbers of subjects using the more practicable single scan technique.

3.2.3. Parameter modelling

Compartmental models which describe the pharmacokinetics of a receptor ligand postulate the existence of three possible tissue 'compartments' which represent free, specifically bound and non-specifically bound ligand, in addition to the the vascular compartment. Rate constants are used to describe the rate of passage of ligand between one compartment and another (figure 3.7). Non-compartmental modelling (e.g. spectral analysis [436]) does not make specific compartment predictions but assumes a partition between blood and tissue and a number of functions which describe the tissue behaviour of the ligand.

This large number of variables, many of which cannot be measured directly, is difficult to model. The general principle which governs the analysis of these data is that a balance has to be achieved between the number of assumptions and the number of unknown parameters. This is because an increase in unknown parameters results in a greater variability in their estimate [437] and in some cases achieves no greater accuracy for the 'fit' to the time activity curve [435]. On the other hand, assumptions made in order to simplify the model and thus reduce the number of parameters have often been inaccurate. Examples of the latter are the assumption of negligible ^^C-flumazenil binding in the pons and of a pseudoequilibrium being achieved at some time during an essentially non-equilibrium scanning procedure.

3.2.3.1. Single scan studies

For single scan studies, two methods are in current use to determine flumazenil V^: kinetic modelling [435] and spectral analysis [436]. Spectral analysis has the advantage that it makes no assumptions about the number of compartments which are needed to model the data. The single extravascular compartmental assumptions of the kinetic method result in the model producing less accurate parameters in areas which 88 have less specific binding such as the thalamus or cerebellum [435]; it has been demonstrated, however, that a better fit to the data is not obtained using a model with an additional extravascular compartment, though this was not shown for all brain regions [438].

Spectral analysis provides a way of deconvolving the plasma input function out of the tissue time-activity curve to produce a tissue response functon, in a manner which is independent of any kinetic or compartmental model and from which voxel-by-voxel parametric images may be produced. No a priori assumptions about the number of compartments in the kinetics of the tracer need be made. Instead, a large but finite number of basis functions which span the expected kinetic behaviour of the tracer in the tissue is defined a priori. Flumazenil is eliminated according to first order kinetics; hence, the basis functions are exponentials of the form:

Equation!.

a e ~ ^ ^ where a and p are constants. These functions are convolved with the measured plasma input function. A computer algorithm is used to select, scale and combine a small subset of these functions which best fit the measured voxel time-activity curve. The general model takes the form:

Equation 2. j- n -^jt C tis À t) = Cm(0® y a j e

where C^jssCt) denotes tissue concentration at time t, C^(t) the measured arterial input function, ® convolution, and aj the contribution (a weight or scaling factor) of the basis function exp(- pjt). Variables aj are constrained to be zero or positive. A least squares technique is used to obtain best fit solutions for each voxel of data. Time-activity curves are analysed without decay correction - this allows convenient limits to be placed on values of p. The slowest possible tissue response corresponds to irreversible trapping of the label and subsequent isotopic decay. Rapid transients due to label in the vasculature can be accommodated by setting the upper limit of the range p =10° s-i. In the model employed n=64 basis functions are used with the values of p distributed logarithmically. Typically, most aj are returned as zero, with only two to four positive values returned for each pixel time-activity curve. This method is sufficiently rapid for routine analysis of entire image sets on a voxel-by-voxel basis (60-90 minutes on a Sun SPARCclassic workstation). 89

For each voxel a characteristic unit impulse response function (IRF) is produced, which is the tissue response function scaled to a unit concentration value of tracer arriving at the brain instantaneously at time=zerof Essentially, this IRF is produced by combining the same weighted basis functions as selected by the least-squares fit above, but without convolution of these basis functions with the plasma input function:

Equation 3. j - P - ( f i ; — A ) r IRF(t)= 2 7 = 1

where X is the decay constant for the isotope and p the number of positive values of a.

Having calculated the IRF for each voxel, a variety of parametric images can be produced: clearance of tracer from blood to tissue (K^ ), of tracer in tissue relative to blood, and images constructed from the IRF at specific time points. Because the images are all constructed on the basis of response to unit impulse activity, the equivalent parametric images are directly comparable between different subjects, or in the same subject in different conditions. The method is summarised in figure 3.8.

In practice, the intercept is poorly defined; an index of clearance from blood to tissue is obtained by producing an image corresponding to tissue response at 2 minutes. An index of specific retention of the tracer is produced by imaging tissue response at 25 minutes. For some tracers, such images are less noisy than Vy images and may be used to identify regions of interest, which can then be mapped back onto the data set for calculation of regional V^. In the case of ’ ^C-flumazenil, images have an acceptably low level of noise, particularly if patient movement is minimised. Images of ^ C - flumazenil have been termed FMZVD images in this thesis. In the case of flumazenil, which has minimal non-specific binding, volume of distribution is closely correlated with receptor density [435].

3.2.3.2. Multiple scan studies

Various methods have been employed to model ’ ^C-flumazenil studies where a separate Ky and are required: these can be broadly divided into equilibrium and non­ equilibrium methods.

a) equilibrium method

* The high frequency spectral components, corresponding to ligand in the vasculature, are omitted in deriving the impulse

response function (IRF), such that the IRF represents the response of the peripheral (specific binding) compartment only. 90

A recent study applied one of the possible equilibrium experimental designs to ’ ^C- flumazenil PET [438]. In brief, the radioligand (which was administered at tracer doses and was not in equilibrium) was injected on two occasions: one with no competing ligand, one when the competing ligand had already been administered and was in an equilibrium state. The main advantage of the method is that the system under study is truly in equilibrium and thus this method should be applied whenever robust parameter quantitation is essential. b) non-equilibrium methods

These rely on the injection of cold ligand either mid-study, to displace the radioligand, or as a coadministration with the radioligand. The former design is known as a displacement study, the latter as a low specific activity study. After the initial descriptive studies [439, 440] which did not attempt to separate out the contribution to the signal made by radioligand in the vasculature, two types of modelling have been applied: pseudo-equilibrium and kinetic.

In pseudo-equilibrium modelling, an equilibrium between the region of interest and a region assumed to contain only 'free' ligand is assumed for particular time-points in the scan. Scatchard plots are applied to determine the desired indices. Different strategies have been employed to determine the time of occurrence of pseudo-equilibrium: (a) the period in which the calculated 'specific binding' time activity curve (for example having subtracted the pons time activity curve from a cortical curve) has zero gradient [401 ] (b) the period in which the time activity curves from two different brain areas (one being the receptor rich area, and the other the receptor 'poor' and thus the 'control' area) have a constant ratio [441] c) the time activity curve from a 'saturating' (radiolabel administered with high doses of cold ligand) experiment is subtracted for the same region from a tracer alone ('high specific activity') study. The first assumption (a) is more inaccurate than the others, since it has the additional complication that the time taken to reach maximum activity is different in different brain areas depending on the local blood flow [435] and permeability. All these methods can produce misleading results because they are essentially non-equilibrium methods which overestimate the binding potential [442] and the assumptions about reference values are often invalid [438, 443, 444]. These methods have been extensively used because of their ease of application [406, 441, 445-448].

Kinetic analyses do not make any assumptions about equilibrium and employ kinetic compartmental models fitting linearly [437] or non-linearly [449] for a number of unknowns. Measured input functions are used in some [435, 437] but not all [448] 91 studies. Protocols vary from dual injections of ligand (one at high specific activity and one at low specific activity) to multi-injection protocols; some of these would only be possible in primates because of the doses of radiation used. While this approach is more rigorous than a pseudoequilibrium method, it still has the disadvantage that it examines a non-equilibrium system using more than one study.

3.2.4. Input function

In order to dissect out the contribution to total radioactivity in each voxel of vascular, free, non-specifically bound and specifically bound compartments, an accurate quantitation of the unmetabolised radiotracer in the vascular (cerebral capillary) compartment should be made. A number of possible approaches have been taken; obviously, none measure the cerebral capillary radioligand concentration directly.

1) A 'reference’ area in the brain which has no specific binding, the same vascular delivery and the same non specific binding as the cortex could be used to provide a measure of the sum of non-specific, free and vascular compartments, which can then be subtracted from other brain areas to leave specific binding only. The area most often employed has been the pons although some early studies employed the white matter [441]. A formal comparison of corpus callosum, white matter and pons showed that the pons was the most stable comparison area for ’ ^C-flumazenil [447].

2) The vascular time activity curve may be measured using venous samples, arterial discrete or continuous samples or continuous 'arterialised' venous samples. Arterial blood has the advantage of similar pH and protein binding to capillary blood and also reflects delivery, whereas venous blood has had the ligand extracted. Also, venous data produces an input function which is 'smoothed' in time as the various ligand molecules from the same cardiac ejection will travel at greatly different speeds through the peripheral tissues.

3) A variant of method (1) is to use the same brain region in a separate second study as the reference region. In this second study a very high concentration of cold ligand is administered to completely block specific binding, so that a sum of vascular, free and non-specfic remains, which can be subtracted from the total radioactivity of the first study to leave only specific binding [441 ].

4) Low specific activity injections may be administered with the assumption that when the dose of cold ligand coadministered is greater than a threshold (e.g 10 jug/kg [445]), all time activity curves from the region of interest follow the same course which is 92 indicative of signal due to nonspecific binding and free ligand.

3.2.5. Partial volume effect

The information collected by a PET camera is recorded as counts per unit volume. These brain unit volumes are of the order of lOmm^ and can contain millions of neurons and glia, and different proportions of gray matter, white matter, blood vessels and cerebrospinal fluid. As the resolution of the camera improves, partial volume effects decrease in magnitude. However, even at theoretical best, a PET voxel cannot be below 2mm3, still a large volume in terms of neuroanatomical structures. Thus the signal from any voxel will always be 'mixed' in terms of the cytological constituents producing it.

This fact has a number of implications for the interpretation of data: a) There will be an apparent spread of voxel radiotracer parameter values for uniform anatomical structures since the proportion of signal derived from other tissues varies, especially at the boundaries of a structure. b) A decrease in signal from an area of atrophy may only reflect reduced number of neurons, rather than a change in receptor numbers per neuron. c) Conversely, a small structure with low activity in a volume of high radioactivity may have decreased binding but may be identified as having normal binding because of “spill- in" from surrounding tissue.

Hence it has been recognised that methods of partial volume correction are necessary for a full characterisation of PET data [450-454]. These methods all rely on structural imaging with MRI and subsequent segmentation of the MRI into grey matter, white matter and, in some methods, a single region of interest. These structural data are coregistered with the PET images in order to correct radioactivi^ measures and parameter estimates for mixed-tissue sampling and partial volume effect.

3.2.6. Scanner field of view

Many of the initial studies conducted using PET could only produce inforrmation from a number of separate non contiguous slices (4 or 6) of the head. Current technology (ECAT-953B, Siemens/CTI, Knoxville, Tennesee, USA) now permits 3D imaging with contiguous slices, but the field of view in the axial direction is of the order of 10cm, 93 preventing imaging of the entire brain simultaneously. New cameras (eg. EXACT-966, Siemens/CTI, Knoxville, Tennesee, USA) will however have a larger axial field of view of approximately 24 cm which will allow the simultaneous imaging of both cortical and brainstem structures.

3.2.7. Identification of anatomical structures and comparison of different scans

3 .2 .7 .7. Region of interest methods

In general, previous PET investigations in epilepsy have employed region of interest (ROI) methods to detect changes in metabolism or receptor density. The regions of the brain to be sampled were defined visually or by a template atlas in the patients and compared to ROIs placed in corresponding regions of normal subjects. The placement of regions was guided by EEC foci, projection areas, the findings of MRI and presumed underlying pathology. The ROI may be of a predetermined geometrical shape (eg. ellipse) placed over the structure of interest, which results in mixed tissue sampling as well as partial volume effect, but permits easy region placement and parameter calculation. Alternatively, the ROI may be carefully anatomically delineated on a structural image such as an MRI, which reduces mixed tissue sampling but does not avoid partial volume effect. ROI methods permit a priori hypotheses concerning regional changes in metabolism or receptor binding to be addressed with accurate quantitation; drawbacks of such methods include potential observer bias and subjectivity affecting region delineation and placement. Furthermore, areas of brain outside the ROIs are not examined. One approach to this problem is to develop techniques for the automated placement of ROIs [455, 456]. Analysis is still limited to the ROIs, however.

3.2.7.2. Voxel-based methods

In order to avoid the drawbacks associated with placement of ROIs, the conventional analysis of Hg^^O PET has employed analysis techniques at the voxel level. The most widely-used method is Statistical Parametric Mapping (SPM)[457]. The principles underlying SPM are straightforward. In order to compare sets of images, the analysis proceeds through three stages, all automated, entirely objective and “hands off”.

Firstly, all the images are automatically “warped” to fit a standard brain size and shape, in order to permit comparison of images from different subjects without the confounding effects of normal anatomical variability. As a result of this process, all of the images are identical in terms of overall morphology. Localised regional 94 characteristics of each individual image remain, however. This normalizing spatial transformation matches each image to a reference or template image that already conforms to the standard space, without requiring user-defined landmarks [458]. An initial transformation employs translation, rotation, scaling and quadratic transformations; following this a slice-by-slice transformation employing predetermined basis functions is used. The matching to the template is non-iterative and employs least-square methods to reduce the difference between the subject and template image. This process is termed volumetric normalisation.

Secondly, image noise is reduced. PET and SPECT are inherently noisy imaging techniques; this noise is reduced by filtering or smoothing the images. A simple Gaussian filter is used [457].

Thirdly, an automated statistical test is performed. This employs the General Linear Model. Each voxel is examined independently. The voxel intensities at one particular voxel from one group of images are compared with the voxel intensities at the same voxel in another group of images using the General Linear Model, in which any number of covariates may be introduced to model the data. By examination of the residual error terms, which follow a t-distribution, the fit of the data to the model can be examined. With the null hypothesis of no difference between the two groups of images and via a “Gaussianisation” of the t-distribution to a Z-distribution, a p-value is calculated for that voxel. This process is repeated at each and every voxel, with all of these thousands of voxel-based tests performed absolutely independently in a univariate sense. The result is a voxel-by-voxel map of p-values. By thresholding at a suitable p-value (eg. p<0.001 ) a map of anatomical areas which differ between the groups of images at that statistical threshold is produced [457, 459].

However, because the analysis regards the voxels as independent, this map of uncorrected p-values cannot be regarded as the final result - because of partial volume effects, smoothing and physiological relationships between different brain areas, voxels from the same image do not all have entirely independent voxel values. Two approaches can be taken to the interpretation of the result: firstly, an a priori hypothesis can be used stating the region(s) which will be examined and all other brain regions will be ignored, whatever the p-values in those other areas; secondly, a correction can be applied to the p-values to correct for the multiple non-independent comparisons inherent in examining all of the voxels together. This second approach allows the whole brain volume to be interrogated, without the necessity for a priori hypotheses regarding which regions should be examined. Indeed, these two approaches can be used together, the first to address questions regarding a specific region or regions, the second to examine 95 the entire brain. The method for correction of p-values draws on the Theory of Gaussian fields. Foci of supra-threshold (uncorrected) voxels are characterized in terms of spatial extent ( k ) and peak height ( j i ) [459]. This characterization is in terms of the probability that a region of the observed number of voxels (or bigger) could have occurred by chance (P(nmax > k ) ) , or that the peak height observed (or higher) could have occurred by chance (P(Zmax > ^)) over the entire volume analyzed (i.e a corrected p-value). Appropriate corrected p-values are selected a priori.

SPM has been widely used to analyse activation PET studies. Typically, data have been pooled from a number of subjects. Recently, more sensitive PET scanners [460] allow data to be analysed from Just one subject, rather than requiring a group to be analysed. Such approaches have been highly successful and illuminating in the fields of cognitive neuropsychology and motor control [461] and are fast becoming the standard analysis techniques for functional MRI studies. Limitations of SPM are the requirement for smoothing, which reduces the effective spatial resolution of the imaging data, and the expression of results in terms of Z-scores at “peak” voxels rather than as quantitative parameters.

3 .3 . HgTSQ p e t

Measurement of regional cerebral blood flow (rCBF) using and PET is now a well- established technique. In order to measure rCBF, a single scan of 1 -2 minutes duration with arterial sampling can be used. rCBF (0 can be calculated from the integral of the tissue counts(C-r(t)) for the scan period and the arterial input function (Cp(t)):

Equation 4: rc,(t)dt = r/.c/o®

where p is the partition coefficient, X is the decay constant for and 0 denotes convolution. Using a fixed value for p the equation can be solved for f [462, 463]. rCBF is sensitive to changes in local neuronal activity and under normal conditions is coupled with local cerebral metabolic rate [464]. The short half-life of means that repeated measures can be made in the same subject within a short period of time, typically 12 measures at 10 minute intervals. For the comparison of rCBF between different conditions, measurement of absolute rCBF values is not necessary, since in man local tissue radioactivity and rCBF are nearly linearly related and changes in relative rCBF (relative to a global CBF value) provide good estimates of true changes in rCBF, hence 96 obviating the need for arterial sampling [465].

The issues of scatter, scanner field of view, partial volume effect and comparison of images have been discussed above and are equally applicable to 97

unstable nucleus

511 keV

511 keV

Figure 3.1. Annihilation of a positron (p+) by collision with an electron (p-) results in the creation of two y-rays with energy of 511 keV.

Figure 3.2. The two y-rays are emitted from the brain and detected simultaneously by a pair of detectors. The PET camera detectors are arranged in a series of rings around the subject’s head. D3 98

D6

D5 □

D2

D4

Figure 3.3. The PET camera consists of a series of rings of detectors, of which a small selection are shown (D1-D6). A source of activity, S, emits pairs of photons which are detected simultaneously by pairs of detectors, in this example D I/D 2, D3/D4, D5/D6. The lines of response, indicated by lines linking the detector pairs, can be reconstructed, allowing the point S to be imaged. 99

a) 2D imaging Septa

b) 3D imaging

Figure 3.4. (a) When imaging in 2D, lead septa are positioned inside the ring of detectors, D. “In plane” photon pairs (PI) are detected, but “out of plane” pairs (PS) strike the lead septa and are not detected. Photon pairs in which one photon is scattered (P2) also strike the septa and are not detected, (b) In order to image in 30, the septa are removed. As a result, no photons are absorbed by the septa, resulting in a very substantial gain in sensitivity. However, account needs to be taken of “out of plane” pairs (P3) in the image reconstruction process. In addition, scattered photons can be detected (P2), giving rise to an apparent, but incorrect, line of response (dotted line). 100 Dynamic image series

Tissue time-activity curves

thalamus

occipital

Time

60 mins

Input function

MODEL

Time

Parameter estimates: - regions of interest - voxel-by-voxel images

Figure 3.5. The dynamic image series consists of a set of time-activity curves, one for each voxel of the image. Two examples are shown, one for an occipital voxel with little specific binding, one for a thalamic voxel with marked specific binding. The blood and tissue data are combined with a model which allows parameter estimated to be made, either on a region of interest basis, or as voxel-by-voxel images. This example is of the non-selective opioid ligand, 11 C-diprenorphine. 101

Figure 3.6. Three dimensional representation of the 71 C-flumazenil molecule (ethyl 8-fluoro-5,6-dihydro-5-( 71 C)methyl~6-oxo-4H-imidazo( l,5-a)( 1,4)benzodiazepine-3- carboxylate). The main structure is formed of carbon; other elements are indicated. 102

C t

C ns

Intravascular Extravascular

Blood-brain barrier

Figure 3.7. Two possible models of flumazenil distribution in brain. The inner compartments illustrate a model of greater complexity, accounting for ligand in arterial plasma (Cp), free extravascular ligand (CF), non-specificalty bound (CNS) and specifically bound ligand (CS) in the brain. The kinetic rate constants (kn) describe exchange between these compartments. The outer compartments demonstrate the relationships between these ligands in a simplified two compartment model, used by Frey et al. (1991). 103

(b)

(a) (c)

K1

Vd

2 25 (d)

Figure 3.8. Outline scheme of the method of spectral analysis; y-axis represents radioactivity, x-axis time in all graphs. A range of exponential basis functions (a) is defined; 3 are illustrated, 64 are used in the method. These are convolved with the measured plasma input function (b) to produce a set of convolved basis functions (c). These functions are selected, weighted and combined by a computer algorithm to obtain a best-fit to the measured voxel time-activity curve (d). The equivalent weighted basis functions without convolution are then combined to produce the impulse response function (IRF), (e). This curve can be used to derive a variety of images: K l, specific time points such as 2 mins or 25 mins, or VD, the area under the curve extrapolated to infinity.

It should be pointed out that in the IRF illustrated in (e) the y-axis has been normalised to unit clearance. 104 Chapter 4

Methods II.

This chapter details the methods which were used in common in many of the studies described in subsequent chapters. To prevent unnecessary repetition, each subsequent chapter refers back to this section with regard to methods, except where a particular technique was used only in that study, in which case it is described in detail in the relevant chapter.

4.1. SUBJECTS

The patients studied were recruited from the epilepsy clinics of the National Hospital for Neurology and Neurosurgery, Queen Square, London, UK and the National Society for Epilepsy - Chalfont Centre, Chalfont St. Peter, UK. All patients suffered from partial seizures and were taking antiepileptic drugs. All had undergone high-resolution MRI including a T1-weighted spoiled gradient recalled (SPGR) volume acquisition with 1.5mm thick slices as well as T2-weighted and proton-weighted sequences with contiguous 5mm thick partitions (Signa 1.5T, General Electric, Milwaukee, USA). All MRI images were reported by an experienced consultant neuroradiologist. Dr. J.M. Stevens, aware of the clinical information. Patients were excluded from the study if taking benzodiazepines. Normal control subjects were all volunteers recruited from friends and colleagues of both patients and investigators. All normal controls had no evidence of neurological disorder, no family history of epilepsy or malformation of cortical development and were on no medication. Consumption of alcohol was not allowed for 48 hours preceding the scan. Written informed consent was obtained in all cases and approval of local ethical committees at both the National Hospital for Neurology and Neurosurgery and the Hammersmith Hospital and of the UK Administration of Radioactive Substances Advisory Committee (ARSAC) were obtained.

4.2. SCANNING PROTOCOL

PET scans were carried out at the Medical Research Council Cyclotron Unit, Hammersmith Hospital, UK using an ECAT-953B PET scanner (CTI/Siemens, Knoxville, Tennesee, USA) with performance characteristics as described previously [460]. Data were acquired in 3D mode, with the septa retracted to improve sensitivity by a factor of 105

6.4 [425]. Dual energy window scatter correction was employed in reconstruction to produce images with a resolution of 4.8 x 4.8 x 5.2mm, in air in the centre of the scanner field of view. Images containing 31 contiguous slices were produced with voxel dimensions 2.09 x 2.09 x 3.43mm. Subjects heads were positioned with the glabella- inion line parallel to the detector rings so that the transaxial planes were parallel to the intercommissural (AC-PC) line. During scanning subjects heads were rested in individualized head moulds, and continuous direct observation maintained to minimise movement. A transmission scan using three rotating ®®GaA®Ge-rotatory line sources was performed to enable emission scans to be corrected for attenuation.

4.3. 1 ^C-FLUMAZENIL SCANS

An eight channel EEG was recorded throughout patient scans to ensure the scan was truly interictal. A 22 gauge cannula was inserted into the radial artery after infiltration with 0.5% bupivacaine. High specific activity C-flumazenil tracer prepared by a modification of the method of Maziere [466] was injected intravenously. Twenty frames were acquired over 90 minutes. The effective dose equivalent of radioactivity per subject was 2.6mSv.

4.3.1. Derivation of plasma input function

Radioactivity in arterial blood was assayed continuously on-line [467], and intermittent blood samples were taken for calibration purposes and for assay of radiolabelled metabolites in plasma [468]. Typically for C-flumazenil, radiolabelled metabolites appear in plasma after a short delay following intravenous injection of the parent tracer, and the proportion of the total radioactivity in blood which is present in the plasma increases with time, because a radiolabelled polar metabolite does not enter red cells. The derivation of a plasma input function from the data obtained from arterial blood was carried out as follows, as described previously [469, 470]. First, the ratio of the concentration of total radioactivity in plasma to that in whole blood, as measured in the discrete blood samples, was fitted to an equation describing an exponential approach to a constant value (greater than one) as a function of the time of sampling. Second, the proportion of radioactivity in plasma recovered in the parent (non­ metabolised) fraction was also fitted to a function incorporating a delay followed by an exponential approach to a constant fraction. These two fitted functions were then applied to the continuous on-line arterial data, which, after appropriate cross-calibration with the PET camera, gave a plasma parent (unchanged labelled C-flumazenil) input 106 function for subsequent kinetic analyses.

4.3.2. Data analysis

Data were analysed on a Sun SPARCclassic workstation (Sun Microsystems, Mountain View, CA, USA) using Analyze version 7.0 (Mayo Foundation, Rochester, Minnesota, USA) [471], MATLAB (The MathWorks Inc., Natick, Massachusets, USA) and Statistical Parametric Mapping (Wellcome Dept, of Cognitive Neurology, Institute of Neurology, London) [457].

4.3.3. Production of parametric maps

The 20 frames of the dynamic image were initially realigned with one another by an automated least squares technique using software from the Cyclotron Unit designed for this purpose. Movement artifact during the scan is substantially reduced by this means [458]. Voxel-by-voxel parametric images of C-flumazenil volume of distribution (FMZVD) consisting of 31 contiguous slices with voxel dimensions 2.09 x 2.09 x 3.43mm were produced using the technique of spectral analysis [436, 472].

4 .4 . Hz 150 SCANS

Radioactivity was administered as an Hgi^O bolus infused over 20 seconds followed by a 20 second saline flush. 12 PET emission scans were collected over 2 hours for each subject, with a 10 minute interval between the start of each scan. The effective dose equivalent of radioactivity per subject was 5.0mSv. Integrated radioactivity counts were collected over a 90 second acquisition period, commencing at the rising phase of radioactivity counts in the head; these data were used as an index of rCBF [473]. All active experimental conditions began 60 seconds before the start of the data acquisition period and continued through to the end of this time. The single frame image, effectively an integral of radioactivity counts per voxel over the scanning period, was used for subsequent analyses without any parameter modelling. The 12 scans from each individual experiment were realigned to remove movement artefact.

4.5. COREGISTRATION WITH MRI

Parametric images of FMZVD or mean images of Hg^^O were precisely coregistered with high-resolution volume acquisition MRI scans [474]. In brief, this method employs an iterative technique to rotate, translate and scale the two images to be coregistered. Segmentation of the images into homogeneous tissue types based on voxel intensity allows 107 calculation of a "ratio image"; this ratio is successively minimised by the iterative process. 108 Chapter 5.

Benzodiazepine receptors in malformations of cortical development.

5.1. INTRODUCTION

The neutral amino acid y-aminobutyric acid (GABA) is the major inhibitory neurotransmitter in mammalian brain. The GABAy^ receptor functions as a chloride ion- selctive channel and has multiple allosteric modulatory sites, one of which binds benzodiazepines [138]. Flumazenil is a neutral antagonist at the benzodiazepine receptor which can be labelled with and used as a PET tracer [466].

Animal models of localisation-related epilepsy implicate abnormalities of GABAergic function. In a primate model of neocortical epilepsy [230] a marked decrease in inhibitory GABAergic synapses was seen at the seizure focus. However, in hippocampal kindling models [185, 187, 188], an intraventricular kainic acid model in the rat [197] and in the rat cobalt focus seizure model [223] an increase in GABAy^ receptor density at the seizure focus has been observed. In resected human epileptogenic hippocampus reduced GABAy^ and benzodiazepine receptor binding have been demonstrated [252, 254, 412], though no change in GABAy^ receptor binding in spiking temporal neocortex has been reported [263]. These findings, combined with the findings from animal models described above, make it difficult to predict the nature of any GABAy^-BZR receptor changes that might be observed in human neocortical epilepsy when investigated with C-flumazenil PET.

PET and TT C-flumazenil have been used to examine patients with localisation-related epilepsy. No previous study has examined BZR binding exclusively in patients with malformations of cortical development, although using SPECT, i23|_jomazenil to label benzodiazepine receptors and s^mTc-HMPAO to indicate blood flow, reduced BZR binding with normal blood flow has been reported in a single patient with focal cortical dysgenesis [340]. Previous investigations have employed region of interest (ROI) methods to detect changes in BZR binding. ROI methods permit a priori hypotheses concerning regional changes in receptor binding to be addressed with accurate quantitation; drawbacks of such methods include potential observer bias and subjectivity 109 affecting region delineation and placement. Furthermore, areas of brain outside the ROIs are not examined. Voxel-based methods have the advantage of producing an image of a receptor binding parameter at the voxel level, which may then be compared with other individuals, again at the voxel level. In this way the bias and subjectivity which may result from an ROI method is eliminated and regions of significant abnormality in all parts of the brain may be detected. This is particularly relevant for a pathology such as cortical dysgenesis which is heterogenous between subjects in terms of location, severity and extent. In this study spectral analysis was used to produce voxel by voxel images of BZR parameters and Statistical Parametric Mapping was used to undertake a voxel by voxel statistical comparison between each patient and a group of normal subjects. This was the first study employing voxel-level comparison between ^^C- flumazenil PET images and the first study to compare ligand PET images from single individuals with a normal control group to be published [475].

It was hypothesised that abnormalities of BZR would be demonstrated using ’ ^C- flumazenil PET in vivo in malformations of cortical development and that these abnormalities would be more extensive than the area of abnormality seen with MRI. To specifically address this question a method was employed which measured BZR independently of blood flow and allowed an entirely objective assessment of any BZR changes throughout the entire brain on a voxel-by-voxel basis. n o

5.2. METHODS

5.2.1. Subjects

Fourteen patients were studied, all of whom had undergone high resolution volumetric MRI (GE Signa 1.5T, SPGR) which had identified MCD. Patients were excluded from the study if taking benzodiazepines. Twenty-four normal subjects of similar age were studied.

5.2.2. Scanning protocol

Scans were performed using the ECAT-953B PET scanner with performance characteristics as described previously [460]. Data were acquired in 3D mode, with the septa retracted to improve sensitivity by a factor of 6.4 [425]. Dual energy window scatter correction was employed in reconstruction. High specific activity ^ ’C-flumazenil tracer, approximately 370MBq, was injected intravenously. Twenty frames were acquired over 90 minutes.

5.2.3. Derivation of plasma input function

Radioactivity in arterial blood was assayed continuously on-line and intermittent blood samples were taken for calibration purposes and for assay of radiolabelled metabolites in plasma. The derivation of a plasma input function from the data obtained from arterial blood was carried out as described in Chapter 4.

5.2.4. Data analysis

Data were analysed on a Sun SPARCclassic workstation (Sun Microsystems, Mountain View, CA) using Analyze version 7.0 (Mayo Foundation), MATLAB (The MathWorks Inc., Natick, Mass.) and Statistical Parametric Mapping (SPM95, Wellcome Dept, of Cognitive Neurology, Institute of Neurology, London).

5.2.5. Production of parametric maps

The 20 frames of the dynamic image were initially realigned. Voxel-by-voxel parametric images of flumazenil volume of distribution (FMZVD) consisting of 31 contiguous slices with voxel dimensions 2.09 x 2.09 x 3.43mm were produced using the technique of spectral analysis [436]. m

5.2.6. Coregistration with MRI

Parametric images of FMZVD were precisely coregistered with high-resolution volume acquisition MRI scans using an automated technique [474].

5.2.7. Statistical Parametric Mapping

Statistical parametric maps are statistical processes that are used to characterise regionally specific effects in imaging data. All images were volumetrically normalised to a TT C-flumazenil template as described in Chapter 4. Subsequently all images were smoothed using a 10x10x6mm FWHM Gaussian kernel.

The automated volumetric normalisation was achieved using a standard template constructed from a normal FMZVD scan with dimensions very close to those of the atlas of Talairach and Tournoux. Eight of the normal subjects underwent MRI scanning as a quality control of the normalisation process: the MRI scan for each individual was coregistered with the FMZVD PET scan for that individual; the coregistered MRI was then transformed into the standard space using the transformation parameters calculated for the volumetric normalisation of that individual's FMZVD PET scan. These eight coregistered and volumetrically normalised MRI scans were then combined on a voxel- by-voxel basis to produce an "average transformed MRI" for these eight normals, as a quality control of the normalisation process.

Effects were estimated according to the general linear model at each and every voxel [457]. Global activity was included as a confounding covariate and this analysis can therefore be regarded as an ANCOVA [476]. To test hypotheses about regionally specific effects the estimates were compared using linear contrasts, comparing each individual patient with the normal control group, examining both for regions of significantly increased and decreased FMZVD. The resulting set of voxel values for each contrast constitutes a statistical parametric map of the t statistic SPMjt}. The SPMjt} were transformed to the unit normal distribution (SPM{Z}) and thresholded at 3.09 (or p = 0.001 uncorrected). The resulting foci were then corrected by characterising in terms

of spatial extent ( k ) and peak height ( ja ) [459]. The threshold chosen for both P(nmax >

k ) and P(Zmax > ^i) was P<0.05. 112

5.3. RESULTS

The normal subjects ranged in age from 26 to 71 years, median 41 years. The patient group ranged from 18 to 50 years, median 30 years. The MRI findings, seizure pattern and EEG findings, drug treatment, time since last seizure and C-flumazenil PET findings for the patient group are summarised in Table 5.1.

Each of the 24 normal subjects was compared individually with the remaining 23 using SPM. 23 of the 24 showed no regions of significant alteration of FMZVD; one 60 year old female showed a region of decreased FMZVD in the right frontal cortex.

Using voxel by voxel ANCOVA with age as a covariate, no effect of age on FMZVD was seen either globally or regionally in the normal group.

Of the 14 patients with cortical dysgenesis studied, 11 showed at least one region of significant change in FMZVD (Table 5.2). Three individuals showed normal distribution of FMZVD; one subject with bilateral band heterotopia (SJ), one subject with a small heterotopic nodule in the white matter (SH) and one subject with a small region of medial frontal MCD (MH).

MRI of subjects TJ, MS and SJ revealed bilateral band heterotopia with apparently normal overlying cortex. All the heterotopic bands were narrow, approximately 1mm. Being below the resolution of the PET method, the bands themselves were not detected. However, TJ showed multiple areas of both increased and decreased FMZVD in cortex of normal MRI appearance (figure 5.1), MS showed a single region of decreased FMZVD in cerebral cortex. SJ had a normal FMZVD PET scan.

Subjects KS and AF had extensive bilateral subependymal heterotopia. The heterotopia was identified as a region of increased FMZVD in KS; in AF bilateral changes were seen in the cortex consisting of reduced FMZVD in medial frontal and occipital lobes, and extensive areas of increased FMZVD in the lateral frontal lobes (figure 5.2). These areas of cortex appeared normal on high resolution MRI.

Subjects JW (figure 5.3), RH (figure 5.4), TP (figure 5.5), MC, MW, IB and MH, all with focal cortical dysgenesis and, in some cases, other malformations of cortical development (RH, TP, MC) demonstrated by MRI, showed regions of increased FMZVD in the areas of MCD and also more widespread areas of increased FMZVD in regions which had a normal MRI appearance. There were also extensive areas of decreased FMZVD in other areas of cortex with normal appearance on MRI, considerably beyond the MRI- 113 defined boundaries of the lesion.

The very small single heterotopic nodule of subject SH was not identified on a FMZVD PET scan; the single large heterotopic nodule of subject SN was identified as a region of increased FMZVD; the FMZVD PET was otherwise normal. 114

5.4. DISCUSSION

Abnormalities of cortical BZR binding have been shown in a variety of dysgenetic syndromes. Eleven of the 14 patients with malformations of cortical development had FMZVD PET abnormalities and 7 had abnormalities beyond the lesions seen with high- resolution volumetric MRI. These abnormalities consisted of both increases and decreases of FMZVD, in close spatial relation to one another. Non-contiguous regions were detected remote from the abnormality seen on MRI. In the three individuals with band heterotopia, FMZVD binding was abnormal in two, with multiple abnormalities in one, in areas of cortex that appeared normal on MRI. Both individuals with extensive bilateral subependymal heterotopia had abnormal FMZVD; in one the heterotopic nodules were identified, in the other extensive changes were seen in cortex that was normal on MRI, with areas of both decreased and increased FMZVD. One of two individuals with single heterotopic nodules had a normal FMZVD PET; in the other the nodule was clearly identified. Six of the 7 patients with focal cortical dysgenesis had markedly abnormal FMZVD PET. In particular, a pattern of increased FMZVD was seen frequently in the abnormal regions of cortex, with extensive areas of reduced FMZVD in the nearby normal cortex, in addition to remote lesions in cortex of normal MRI appearance.

The finding of increased FMZVD in FMZ PET studies of patients with epilepsy is new; this study has also demonstrated the anatomical extent of areas of abnormal FMZVD throughout the brain volume in a way that has not previously been possible. The methods have a number of features which allowed these findings to be made. In particular, the adverse effects of patient movement were minimised by continuous observation and by automated realignment of the dynamic image frames; accurate volumetric normalisation was used to allow precise structure-to-structure coregistration between individuals, confirmed by inspection of the "average transformed MRI" that was used as a quality control, in which considerable anatomical detail is preserved (figure 5.6); the statistical method applied was highly rigorous; the statistical results were superimposed onto the volumetrically normalised MRI for each patient to allow a detailed correlation between structure and function to be made.

There are two clear limitations to this method. Firstly, the resolution of the SPM is approximately equivalent to a FWHM of 12 x 13 x 9mm because of the smoothing process, hence small regions of significant abnormality were not detected. Increasing the spatial resolution of the SPM results in an increase in noise. Secondly, the method of volumetric normalisation will not produce satisfactory results with brains of gross anatomical abnormality. Careful inspection of the volumetrically normalised images showed that this was not a problem in the patients in this series. In addition, failure of 115 adequate normalisation would result in significant abnormality being seen in areas where there is no overlap between the patient's brain and the brains of the normal group as a result of artefact; this was not observed.

SPM permits a voxel-by-voxel analysis of all brain regions; the result of these analyses is a map of regions of significant difference between the patient and the normal group. In contrast, an ROI based method would permit these differences to be characterised in terms of absolute values of FMZVD and also would allow corrections for partial volume effect to be made, which is important in all PET studies [450-454]. SPM and ROI-based analyses should be regarded as complementary rather than as mutually exclusive alternative methods.

The observation of an alteration in regional FMZVD could have several different explanations:

(1)lt could simply be the result of partial volume effects: regions of increased FMZVD resulting from more grey matter and less white matter and CSF being included in the PET voxel; regions of decreased FMZVD resulting from the opposite. This is likely to be the explanation for some of the findings in areas of anatomically abnormal grey matter - for example, increased FMZVD found in a subependymal nodule, which was compared statistically with voxels of white matter and CSF in the normal group. It seems probable that this method has particular sensitivity for the detection of heterotopic grey matter. However, careful inspection of the MRI data alongside the SPM showed that areas of altered FMZVD were seen in regions of cortex which had entirely normal thickness and morphology, so partial volume effect does not seem a likely explanation for the FMZVD abnormalities in these cases.

(Z)Altered FMZVD might be observed as a result of a change in the density of neurons per volume of cortex, or from the presence of ectopic BZR-bearing neurons in subcortical white matter.

(3)Changes in FMZVD would also be observed as a result of a change in the density of BZR per neuron or by the occupation of the benzodiazepine receptor by an irreversibly bound endogenous ligand.

(4)A change in FMZVD could also result from a change in receptor characteristics affecting affinity (KJ. The occupation of the benzodiazepine receptor by a competitive endogenous ligand might also alter the apparent affinity. 116

It is not possible to distinguish between these possibilities using this method; in vitro quantitative correlational studies of neuronal and receptor numbers are needed for this distinction. However, our results are entirely consistent with observed alterations in BZR and GABA^s^ receptor density identified by previous investigations. ^^C-flumazenil PET studies [254, 406-409, 413, 414, 477] and surgical pathology [254] found regional decreases in BZR-GABA^ receptors at epileptogenic foci in hippocampus and, in some of the PET studies, in neocortex. In the rat cobalt focus seizure model of spontaneous neocortical epilepsy, in hippocampal kindling models [185, 187, 188], and in an intraventricular kainic acid model in the rat [197] an increase in GABA^ receptor density was observed in the seizure focus; these findings demonstrate that an increase in

GABAa receptors can occur in epileptic foci and predict that BZR increases might be found in some patients with neocortical partial epilepsy. Our results suggest that, in some cases, the epileptic focus may be characterised by a region of increased BZR-GABA^ receptor density (perhaps resulting from impaired GABA release and postsynaptic upregulation) surrounded by a region of reduced BZR-GABA^^^ receptor density (possibly resulting from increased GABA release in a region of increased inhibition as a response to seizure activity). These possibilities, however, are speculative at present and require further elucidation.

We found the FMZVD changes to be more extensive than the MRI changes in seven cases and regions of altered FMZVD were found in unanticipated areas such as the hippocampi bilaterally (subject TP) or the cingulate cortex (subject JW).

In general, the PET findings are not contrary to the EEG findings. The EEG data were collected during conventional patient investigation and were mostly interictal. These patients had not had systematic intracranial ictal recordings and EEG data were not intended to act as a “gold standard” for comparison with the PET data. It is recognised that MRI and EEG may not be well-correlated in MCD [104]. The data in the tables shows the EEG and PET findings are correlated, though not perfectly. It is not possible without surgical or autopsy pathological material to explain fully the finding of alterations in FMZVD outside the MRI lesion. It seems likely that the MRI identifies the “tip of the iceberg”; FMZVD PET identifies a greater extent of dysgenesis than MRI. A strength of the SPM method is the ability to detect abnormalities beyond the expected areas. It remains to be determined whether these areas are indeed occult dysplasia.

These findings may also have clinical relevance. In seven cases the abnormalities on FMZVD PET were more extensive than the MRI changes. This may affect surgical planning and have implications for prognosis. This is not yet proven; further studies of the impact of FMZVD findings, particularly multiple abnormal regions, on surgical 117 outcome need to be performed. Poor outcome from resective surgery in patients with MCD most likely reflects the extensive nature of the abnormality, that has been incompletely identified with MRI. It is well-recognised that interictal ’ SpDG PET shows widespread hypometabolism in both hippocampal sclerosis and cortical dysgenesis [375, 478]. Widespread metabolic effects are likely secondary to diaschisis and do not imply abnormalities in the neurons in the hypometabolic region [478]. The finding of abnormalities of FMZVD, in contrast, implies neuronal abnormalities.

In conclusion, FMZVD PET analyzed using SPM showed, in most patients with MCD, multiple and extensive regions of both increased and decreased ^^C-flumazenil binding that extended beyond the lesions defined by high-resolution volumetric MRI. This has implications both for understanding the mechanisms of epileptogenesis in these individuals and for their clinical management. n e

MRI finding Age Duration of Gender Seizures Interictal EEG Drugs (daily dose Time since

epilepsy features in mg) last seizure;

(years) (* - ictal) frequency (per month)

TJ Bilateral band heterotopia 33 31 Male CPS + 2° gen. Irregular generalised ESM 1000; PKT 350; 36hrs

spikes LTG 100 1

MS Bilateral band heterotopia 30 18 Male R arm focal L hemisphere slow CBZ 600 4 days motor + 2° gen. 30

SJ Bilateral temporal band 23 5 Female CPS Normal CBZ 400 2 weeks

heterotopia 30

KS Subependymal heterotopia 30 6 Female R arm focal Normal VPA 600; PHT 300 36hrs sensory + 2° gen. 8

AF Subependymal heterotopia 47 21 Female CPS and 2° gen. Bilateral temporal fod PHT375 2 months tonic 1

JW L frontal and R temporal 37 13 Female 2° gen. tonic Bilateral independent GBP 1200; CBZ 1300 6 hrs

cortical dysgenesis foci 30

RH Bilateral cortical 34 33 Female L arm focal No definite abnormality CBZ 800 2 days dysgenesis and schizencephaly motor + 2° gen. 4

TP R temporoparietal cortical 29 27 Female CPS + 2° gen. R posterior hemisphere VPA 2500; CBZ 600; 8 days

dysgenesis and heterotopias focus LTG 100 30

MC Bilateral cortical 18 17 Female L arm focal Bilateral independent LTG 400; CBZ 1000 2 days

dysgenesis; heterotopic nodules motor + 2° gen. fbd * 30

MW R post temporal 18 13 Male CPS R anterior focus CBZ 1000 2 weeks

cortical dysgenesis 1 IB Bilateral perisylvian 28 14 Male SPS + 2° gen. No definite abnormality CBZ 900; GBP 1200 24 hours polymicrogyria 30

MH L medial frontal focal 50 21 Male CPS + 2° gen. Bilateral slow PHT 475 1 month cortical dysgenesis 1

SN Single R posterior parietal 26 7 Female SPS No definite abnormality CBZ 700 4 years heterotopic nodule 0

SH Small nodule in post L 29 14 Female L arm focal Generalised spike wave CBZ 1000 3 days parietal white matter motor + 2° gen. 4

Table 5.1. Clinical details of the 14 patients with MCD. R - right; L = left; post - posterior; P H I = phenytoin; CBZ= carbamazepine; ESM - ethosuximide; LTG = lamotrigine; VP A = sodium valproate; GBP = gabapentin; CPS = complex partial seizures, BPS = simple partial seizures, 2° gen = secondary generalised. 119

MRI finding PET findings: PET findings:

Increased FMZVD decreased FMZVD

TJ Bilateral band tieterotopla L post, par.; posterior Bllat. post temp;

cingulate cortex; R frontal

medial occipital

MS Bilateral band fieterotopla None L motor strip

SJ Bilateral temporal band None None fieterotopla

KS Subependymal fieterotopla Fleterotopic nodule None

AF Subependymal fieterotopla Extensive bllat. lat. Bllat med front; front; L lat temp bllat occipital

JW L frontal and R temporal L front dysgenesis; Bllat. front;

cortical dysgenesis bll. dngulate L ocdpital

RFI Bilateral cortical dysgenesis L front and R par. defts; L front;

and scfilzencepfialy L front and R par. dysgenesis “ R parietal

TP R temporoparietal cortical Fteterotopias; tilppocampi; L parieto-ocdpltal;

dysgenesis and tieterotopias posterior dngulate R occipital

MC Bilateral cortical dysgenesis Bll Inferior frontal; Fleterotopic nodules; fleterotopic nodules bll post par/ocdpltal R temp cortex

MW R post, temporal cortical None R post. temp,

dysgenesis and occipital

IB Bilateral perisylvian Bilateral perisylvian None

polymlcrogyrla

MFI L medial frontal focal None None

cortical dysgenesis

SN Single R posterior parietal Fleterotopic nodule None fleterotopic nodule

SFI Small nodule In post. L None None parietal wfilte matter

Table 5.2. Results of SPM analysis. FMZVD=flumazenil volume of distribution; R=right; L=left; par=parietal; temp=temporal; front=frontal; lat=lateral; bil=bilateral; post=posterior. 120

» $

Figure 5.1. Subject TJ. A) Coregistered, normalised 71-weighted MRI B) Volumetrically normalised FMZVD PET 0) Regions of significantly abnormal FMZVD superimposed on the MRI. Red, increased FMZVD, blue, decreased FMZVD (p<0.001, correctedp<0.05). 121

&

P

A

Figure 5.2. Subject A F. A) Coregistered, normalised T1-weighted MRI B) Volumetrically normalised FMZVD PET C) Regions of significantly abnormal FMZVD superimposed on the MRI. Red, increased FMZVD, blue, decreased FMZVD (p<0.001, corrected p<0.05). 122

*

2

Figure 5.3. Subject JW. A) Coregistered, normalised T1-weighted MRI B) Volumetrically normalised FMZVD PET 0) Regions of significantly abnormal FMZVD superimposed on the MRI. Red, increased FMZVD, blue, decreased FMZVD (p<0.001, corrected p<0.05). 123

a

a 7%

Figure 5.4. Subject RH. A) Coregistered, normalised T1-weighted MRI B) Volumetrically normalised FMZVD PET C) Regions of significantly abnormal FMZVD superimposed on the MRI. Red, increased FMZVD, blue, decreased FMZVD (p<0.001, corrected p<0.05). 124

n W 4 y

)*«

K § i

Figure 5.5. Subject TP. A) Coregistered, normalised T1-weighted MPI B) Volumetrically normalised FMZVD PET C) Regions of significantly abnormal FMZVD superimposed on the MR I. Red, increased FMZVD, blue, decreased FMZVD (p<0.001, corrected p<0.05). 125

M

Figure 5.6. “Average transformed MR I". 126 Chapter 6.

Benzodiazepine receptors in localisation- related epilepsy due to acquired cerebral

insult.

6.1. INTRODUCTION

Abnormalities of GABA^^-mediated inhibition are implicated in a number of animal models of localisation related epilepsy [213, 230-232, 239, 479, 480], including models of localised neocortical injury. Similar mechanisms may underlie similar seizure syndromes in man. ^’C-flumazenil (FMZ) positron emission tomography (PET) provides a means to investigate this hypothesis in man in vivo. Regions of decreased flumazenil binding may be due to decreased neuronal density, decreased receptors per neuron or a decrease in receptor affinity; conversely, increased flumazenil binding may be due to the opposite phenomena.

Using FMZ PET, reduction of FMZVD has been detected in the hippocampus of patients with hippocampal sclerosis (HS) and mesial temporal lobe epilepsy [408-410] with no abnormalities elsewhere in the neocortex [410]. Moreover, the loss of cBZR binding is of a greater degree than the loss of hippocampal volume [477]. The study described in Chapter 5 found multiple regions of both increased and decreased FMZVD in patients with malformations of cortical development (MCD), possibly reflecting abnormal neuronal density or abnormal receptor density per neuron [475]. In many cases the cBZR abnormalities were more extensive than the abnormality detected with MRI. It is proposed that the finding of a region of increased FMZVD may be indicative of a dysgenetic lesion.

Acquired cerebral injury is an important aetiology of localisation-related epilepsy. No study has previously specifically investigated the FMZ PET characteristics of this patient group; the results of such a study would be needed to interpret studies in patients with normal MRI.

The identification of regions of increased FMZVD in MCD raises two questions: Is increased FMZVD a feature only of MCD, or could it be a response to local seizure 127

activity which might be seen in other aetiologies? Is regionally increased FMZVD an artefact of the analytical method used?

A study of FMZ PET in patients with acquired cerebral injuries would identify the occurrence and nature of FMZVD abnormalities in such patients and in addition clarify an understanding of the nature of FMZVD increases observed in MCD. Statistical Parametric Mapping (SPM) involves comparison of scans from different individuals and normalisation of the global mean, which effectively adjusts the voxel values in each individual scan such that the mean activity across the entire image is the same for each subject. Thus, a large area of decreased activity may reduce the global mean for that subject to such an extent that when adjusted to match the other subjects, the activity in other areas of the brain is artificially increased. A study of patients with acquired cerebral injury would determine whether this artefact occurred in practice. It is expected that patients with acquired cerebral injury might have regions of reduced FMZVD, possibly of large extent. If a normalisation artefact occurred in such cases, regions of increased FMZVD would be detected. Further, this method relies on a stereotaxic transformation of all images to a standard brain size and shape. Patients with MCD may have abnormalities of overall brain morphology; imperfect stereotaxic transformation might also result in artefacts such as regions of increased FMZVD and abnormalities of FMZVD beyond the MRI-defined lesion. Patients with acquired cerebral injury may also have extensive abnormalities of cerebral morphology; study of this group will assess the consequences of stereotaxic transformation. If no increases of FMZVD were found it would suggest that increased FMZVD in MCD is not an artefact of the method and not a general feature of the response to a neocortical epileptic focus, but a specific marker of MCD.

Using a method identical to that of Chapter 5, the study described in the present chapter aimed to investigate the nature of abnormalities of FMZVD using FMZ PET in patients with acquired cerebral injury leading to localisation-related epilepsy. In addition, this study sought to clarify whether regions of increased FMZVD would be observed in such patients and whether abnormalities would be seen distant to the MRI-defined lesion. 128

6.2. METHODS

6.2.1. Subjects

Six patients were studied who were recruited from the epilepsy clinics of the National Hospital for Neurology and Neurosurgery, Queen Square, London, UK. All had undergone high-resolution MRI including a 11-weighted spoiled gradient recalled (SPGR) volume acquisition with 1.5mm thick slices as well as T2-weighted and proton-weighted sequences with contiguous 5mm thick slices and T2-weighted images (GE Signa 1.57). All had a clear history and imaging findings compatible with a past acute and non­ progressive cerebral injury. The same normal control group as described in Chapter 5 was used.

The hypothesis was that no regions of increased FMZ binding would be observed in patients with acquired cerebral injury. Six patients were studied in order to demonstrate that the observation of no regions of increased FMZ binding in any patient in the present study was significant at p<0.05, given the observation in our previous study of 16 regions of increased FMZVD in 14 patients with malformations of cortical development (regions of increased FMZVD due to heterotopia were excluded for this purpose). Using Fisher's exact test, if 16 increased FMZVD regions were seen in 14 patients with malformations and no regions of increased FMZVD were seen in 6 patients with acquired lesions, Fisher's exact p-va lue = 0.02.

6.2.2. Scanning protocol 6.2.3. Derivation of plasma input function 6.2.4. Data analysis 6.2.5. Production of parametric maps 6.2.6. Coregistration with MRI

These aspects of the method were identical to that used in Chapter 5.

6.2.7. Statistical Parametric Mapping

Each subject was compared with the normal control group. The threshold chosen for both

P(nmax > k ) and P(Zmax > \i) was P<0.05. The resulting significant regions were mapped onto the coregistered normalised MRI to compare the anatomical abnormality with the abnormality of cBZR. 129

6.3. RESULTS

The clinical features of the patients and the PET results are set out in table 6.1. The patients all had medically refractory localisatbn-related epilepsy for a median of 25 years, range 6 to 40. All of the patients had regions of significantly decreased FMZVD. In all cases the abnormalities were in the same location as the abnormalities seen on MRI, though in two cases the areas of reduced FMZVD were more extensive: SM had reduced FMZVD extending into the right temporal cortex which was normal on MRI (figure 6.1) and GJ had reduced FMZVD extending into the MRI-normal left frontal cortex. No regions of significantly increased FMZVD were seen, despite the very extensive nature of all of the regions of decreased FMZVD. In addition, no PET abnormalities were seen remote from the MRI lesions. 130

6.4. DISCUSSION

All six patients with acquired non-progressive cerebral injury leading to localisation- related epilepsy had reduced FMZVD in the region of brain seen to be damaged on MRI; in 2 cases the region of reduced FMZVD was more extensive. No regions of increased FMZVD were observed and no abnormalities of FMZVD were seen remote from the MRI lesion. This finding confirms our hypothesis that the effect of cerebral injury on the GABA*- cBZR complex in man would be comparable to that in animal models.

The absence of regions of increased FMZVD despite extensive regions of decreased FMZVD and the absence of remote lesions despite abnormalities of overall brain morphology in these subjects, indicate that analysis of such images using SPM does not introduce artefacts, either through spatial normalisation of each image to a standard template or through normalistion of mean global activity. The absence of artefact due to normalisation of mean global activity is not surprising. The calculation is carried out in two automated steps. The first step is the calculation of mean voxel intensity for each entire image. The image is then thresholded at a chosen value, in these studies 80% of the mean voxel intensity. The mean of all voxels above this threshold is calculated and used as the "global" mean in the ANCOVA. This process excludes cerebrospinal fluid (CSF), white matter and extracranial regions so that the voxel values are adjusted relative to mean intensity of grey matter structures. In addition, this process excludes regions of markedly low cortical FMZVD, such that even large regions of reduced FMZVD do not bias the global mean in such a manner that would artefactually increase areas after ANCOVA. This method is identical to that used in the study of MCD described in Chapter 5. In addition, it should be emphasised that the present study and the prior study of MCD are identical in every methodological respect and the same normal control series was used.

Regions of low FMZVD found in these patients with acquired cerebral injury may be due to loss of cortical neurons in the injured region. Additionally, specific loss of cBZR receptors may play a part. That no increases in FMZVD were found implies that increased cBZR is not a response to a neocortical epileptic focus, but that increased cBZR binding is a specific marker of MCD. This finding is of clinical importance in the interpretation of FMZ PET in patients with localisation-related epilepsy and normal MRI, as the finding of increased cBZR binding suggests MCD as the underlying pathology. The observation of regions of decreased FMZVD in patients with acquired lesions which are more extensive than the MRI lesion implies that FMZ PET may be more sensitive in the detection of such abnormalities and hence be of potential value in patients with normal MRI. 131

In conclusion, on the basis of this study, the study described in Chapter 5 and previous studies [410, 475], isolated HS results in a localised region of loss of GABAyj^-cBZR binding which does not extend beyond the mesial temporal structures and involves no remote cortical areas; acquired cerebral injury results in localised neocortical regions of loss of GABA^-cBZR which may be more extensive than the MRI lesion. MCD may result in both localised and distant regions of increased and decreased GABA^-cBZR binding which may be multiple. Extrapolating from these findings, it may be possible to infer an aetiology in patients with localisation-related epilepsy and normal MRI who are observed to have an abnormality of cerebral cBZR using ^^C-flumazenil PET. This information may be most useful in the evaluation of such patients for possible surgery. 132

Duration Regions with MRI EEG AED Subject Age Sex of epilepsy Aetiology Seizures significantly finding finding mg/day (years) reduced FMZVD

SM 33 Male 31 R occipital Stroke in CPS Widespread CBZ 1200 cortical infancy R lEA PMD 300 atrophy

GJ 47 Male 40 L temporoparietal Qosed CPS L slow; CBZ 1600 cortical atrophy head GTCS multifocal lEA LTG 40 0 injury

MD 26 Male Bifrontal Qosed SPS Bilatera LTG 80 0 brain damage head CPS frontal lEA GVG 300 0 injury GTCS

IJ 32 Male 29 R posterior Bacterial CPS Bilateral PHT 300 temporal meningitis GTCS slow VPA 100 0 cortical damage

GH 28 Male Bilateral frontal Closed CPS Bilateral CBZ 1800 brain damage head frontal slow L > R injury

OM 49 Male 21 Extensive L Qosed CPS L centro- LTG 400 cortical atrophy head GTCS temporal lEA GBP 8 00 injury CBZ 1200

Table 6.1. Clinical features and results of SPM analysis. R=right; L=left; SPS=simple partial seizures; CPS=complex partial seizures; GTCS=generalised tonic-clonic seizures; IEA=interictal epileptiform abnormality; CBZ=carbamazepine; PMD=primidone; LTG=lamotrigine; GVG= vigabatrin; VPA=sodium valproate; GBP=gabapentin 133

A i

*

Figure 6.1. Subject SM. A) Coregistered, normalised T1-weighted MRI B) Volumetrically normalised FMZVD PET 0) Regions of significantly abnormal FMZVD superimposed on the MRI. Red, increased FMZVD, blue, decreased FMZVD (p<0.001, corrected p<0.05). 134 Chapter 7.

Benzodiazepine receptors in dysembryoplastic neuroepithelial tumour.

7.1. INTRODUCTION

Dysembryoplastic neuroepithelial tumour (DNET) is a recently described brain mass lesion with distinctive pathological features and a favourable prognosis. It shares some features with dysplastic lesions and may have a developmental origin. The usual presentation is with treatment-resistant complex partial seizures of long standing [113-116]. Onset of epilepsy is usually in childhood [113, 114, 116]. EEG is usually abnormal: localised slow activity and interictal spiking were seen in the substantial majority, being usually more extensive than or distant to the lesion [113]. Typical focal cortical dysplasia may be seen in areas of cortex distant to the DNET [117].

Using a method identical to that of the previous studies, the aim of this study was to investigate the nature of abnormalities of cBZR using FMZ PET in patients with DNET. 135

7.2. METHODS

7.2.1. Subjects

This study included 5 male and 1 female patients who were recruited from the epilepsy clinics of the National Hospital for Neurology and Neurosurgery, Queen Square, London, UK. All had undergone high-resolution MRI including a T1-weighted spoiled gradient recalled (SPGR) volume acquisition with 1.5mm thick slices as well as T2-weighted and proton-weighted sequences with contiguous 5mm thick slices and T2-weighted images (GE Signa 1.5T) which had identified a probable DNET. The same 24 normal subjects as in Chapter 5 were included.

7.2.2. Scanning protocol 7.2.3. Derivation of plasma input function 7.2.4. Data analysis 7.2.5. Production of parametric maps 7.2.6. Coregistration with MRI

These aspects of the method were identical to that used in Chapter 5.

7.2.7. Statistical Parametric Mapping

Each subject was compared with the normal control group as described in Chapter 5. The threshold chosen for both P(nmax > k ) and P(Zmax > f i ) was P<0.05. In addition, for those subjects in whom no regions were found to be significant when corrected for multiple non-independent comparisons, the region of the DNET was examined in isolation. The threshold chosen in this case was P<0.001, uncorrected. 136

7.3. RESULTS

The clinical features of the patients and the PET results are set out in table 7.1. The patients all had medically refractory localisation-related epilepsy for a median of 15.5 years, range 6 to 33. Pathological confirmation was obtained at resective surgery in 4 cases; the remaining 2 have not undergone surgery. Three of the patients had single regions of significantly decreased FMZVD. In all 3 cases the abnormalities were in the same location as the abnormalities seen on MRI. One of these subjects had a single region of significantly increased FMZVD in the occipital cortex. Two of the remaining 3 subjects had a significant reduction in FMZVD in the DNET examined in isolation, without correction for multiple non-independent comparisons. In these two the small region size did not allow the regions to be significant when a correction for multiple non- independent comparisons was made. Furthermore, in every case the Z-score peak within the DNET represented the greatest decrease in FMZVD compared with the normal controls anywhere in the dataset. 137

7.4. DISCUSSION

All six patients with DNET had reduced cBZR density in the region of brain affected by DNET on MRI. In 3 cases the reduction of cBZR reached significance; in the remaining 3, the DNET contained the most reduced FMZVD, but the small region size did not permit this region to be significant when multiple non-independent comparisons across the whole brain volume were taken into account. One subject had a single region of increased FMZVD in the occipital cortex.

Regions of low FMZVD found in our patients may be due to relative reduction of neuronal elements in the DNET compared with normal grey matter. Histological features of DNET include a multinodular and multicystic appearance and the presence of both neuronal and glial (oligodendrocytic and astrocytic) components [114, 116]; hence a relative reduction of neuronal elements is plausible. Additionally, specific loss of cBZR may play a part.

It remains unclear whether DNET should be classified as a tumour, that is a lesion with prolferative potential, or as a developmental lesion related to malformations of cortical development. The presence of accompanying cortical dysplasia is often prominent [114, 116]. It is this associated dysplasia that has led to the inclusion of DNET with developmental lesions. Furthermore, DNET located in the subependymal region with nodular extension to the periphery has been recorded, suggesting an origin from subependymal germinal matrix with nests of primitive neuroblasts arrested in their embryonal migration [115]. However, the presence of occasional mitoses and immunoreactivity to the proliferating cell nuclear antigen in some cases indicate that these lesions may have cellular proliferative activity [113].

Surgery for DNET may be highly successful [113], somewhat in contrast to the results of surgery in MCD. It is possible this is explained by contrasting neurobiology: MCD may often be a widespread phenomenon despite circumscribed MRI findings; DNET may be a truly (uni)focal disorder permitting successful resfttive strategies. The findings of this study would support this suggestion. The single region of increased cBZR found in one subject may indicate distant MCD; however, this patient has had a good outcome following resective surgery, implying that the region of increased FMZVD may not necessarily indicate underlying dysplasia. Increased FMZVD in the occipital region has been observed in some subjects with MCD in Chapter 5 and also in patients with normal MRI in Chapter 8. The explanation of this finding remains obscure.

In conclusion, DNET results in a localised region of reduced FMZVD which does not extend 138 beyond the lesion; increased FMZVD may be seen exceptionally in distant cortex, but does not ne cessarily imply a poorer surgical outcome. 139

MRI location of Age Duration of Seizures Interictal EEG Drugs Reduced FMZVD Highest Z-score lesion; histology epilepsy features (dose in mg) regions p<0.001 within lesion (years) (* = ictal) corrected p<0.05

NM R lateral temp 19 15 Aura of anxiety; Rhythmic delta R LTG 400 None 3.43 automatisms temporal GBP 1800 TPM 200

RIH R ant temp; 23 6 CPS - deja vu, music R temporal focus * GVG 2000 R ant temp 6.33 DNET Also secondary gen PHT 425

MIH L ant lat temp 50 18 CPS - strange sound, L temporal slow VPA 2000 L ant lat temp 4.79 loss of awareness

KO L med temp 30 16 CPS - deja vu, aphasia, L temporal focus* GBP 900 None 4.28 DNET loss of awareness LTG 100

SP R frontal; 19 6 SPS with motor R frontal focus * CBZ 1300 R frontal 5.01 DNET phenomena L arm PHT 300 VPA 1000

RW R amygdala 42 33 CPS - deja vu, epigastric Rhythmic delta over R CBZ 1600 None 2.76 DNET rising, olfactory temporal. Ictal activity LTG 300 hallucination widespread R *

Table 7.1. Clinical features and results of SPM analysis. R=right; L-left; ant-anterior; lat=lateral; med=medial; temp=temporal; SPS= simple partial seizures; CPS=complex partial seizures; CBZ=carbamazepine; GBP=gabapentin; GVG=vigabatrin; PHT=phenytoin; VPA-sodium valproate 140

I 1

Figure 7.1. Subject RIH. A) Coregistered, normalised T1-weighted MRI B) Volumetrically normalised FMZVD PET 0) Regions of significantly abnormal FMZVD superimposed on the MRI. Blue, decreased FMZVD (p<0.001, corrected p<0.05). 141 Chapter 8.

1 iC-flumazenil PET in extratemporal epilepsy with normal MRI.

8.1. INTRODUCTION.

Advances in magnetic resonance imaging (MRI) have had a major impact on the management of patients with refractory localisation-related epilepsy, allowing a surgical approach to the resection of epileptogenic lesions [481-483]. An excellent surgical outcome can be expected in patients with extratemporal epilepsy in whom a single lesion is identified and resected completely [103, 484-486]. However, approximately 15-20% of potential surgical candidates with localisation-related epilepsy have normal MRI, even with the highest resolution equipment and multiple sequences [487]. Removal of a presumed epileptogenic region of neocortex, in which no lesion is identified on preoperative imaging, often results in a poor surgical outcome [488, 489]. In order to detect an epileptogenic region in a patient with normal imaging, invasive electrode studies are frequently undertaken. Such studies may identify the seizure focus, but with the risk of concomitant morbidity from electrode placement. Furthermore, it is not always apparent, even with high quality ictal surface EEG data, where the electrodes should be placed to gain the greatest information.

Functional imaging may help to identify the source of seizures. There «/e little data on the role of T^C-flumazenil PET in patients with refractory partial seizures and unremarkable high quality MRI. In six patients with frontal lobe epilepsy, 5 of whom had normal MRI, focal reduction in flumazenil binding was demonstrated that was consistent with clinical and EEG data [413]. The validity of a comparison of this nature crucially depends on the relative sophistication of the instruments and methodologies used for the PET and MRI studies.

Two key features of this study should be emphasised: Firstly, high-quality high- resolution volumetric MRI was used; secondly, an entirely objective, automated and statistically rigorous method was used for analysis of the PET data which permitted analysis of the whole brain volume [410, 490].

The aim of the current study was to investigate cBZR in patients with a clear-cut 142 electroclinical syndrome of localisation-related epilepsy and normal high resolution volumetric MRI, in order to detect focal cortical regions of abnormal cBZR binding in individual patients in comparison with normal controls. 143

8.2. METHODS

8.2.1. Subjects

18 patients were studied who were recruited from the epilepsy clinics of the National Hospital for Neurology and Neurosurgery, Queen Square, London, UK. All had undergone a high resolution MRI protocol including volumetric spoiled gradient recalled (SPGR) sequence with 1.5mm slices, and coronal T2-weighted and proton-density sequences with contiguous 5mm slices (GE Signa 1.5T), which had been reported to be normal by an experienced neuroradiologist aware of the clinical information. Patients were excluded from the study if taking drugs which are thought to interact with the cBZR- GABAy^ complex (benzodiazepines, barbiturates, vigabatrin). The control group was the same 24 normal control subjects as in Chapters 5, 6 and 7.

8.2.2. Scanning protocol 8.2.3. Derivation of plasma input function 8.2.4. Data analysis 8.2.5. Production of parametric maps 8.2.6. Coregistration with MRI

These aspects of the method were identical to that used in Chapter 5.

8.2.7. Statistical Parametric Mapping

Each patient was compared with the normal control group using the same method as in

Chapter 5. The threshold chosen for both p(nmax > k ) and p(Zmax > n) was p<0.05. 144

8.3. RESULTS

The clinical characteristics of the patients, drug treatment, seizure charaderistics and EEG features are described in Tables 8.1 and 8.2. According to these characteristics, the patients were divided into two groups: 10 patients with unilateral frontal seizure onset and 8 patients with neocortical onset not clearly from one frontal lobe.

In summary, 13 of the 18 patients had regions of significantly abnormal cortical FMZVD; 10 had regions of increased FMZVD, 6 had regions of decreased FMZVD, and 3 of these had both regions of increased and decreased FMZVD. 7 patients had an abnormality somewhere in the lobe of presumed origin of the seizures; 12 patients had an abnormality in the hemisphere of presumed origin, including 4 with the only abnormality being in the medial occipital cortex, medial frontal cortex or the cingulate gyrus. 9 patients had other remote abnormalities of FMZVD.

8.3.1. Patients with unilateral frontal seizure onset

From clinical and EEG evidence, 10 patients had good evidence for a unilateral frontal seizure onset. Patient AC had a left frontal region of decreased FMZVD with no other abnormalites. Interictal epileptiform activity was seen only in the left frontal area. SW had a left frontal region of increased FMZVD; ictal onset was from the left frontocentral region, A further region of increased FMZVD was seen in the medial occipital cortex. AO had bilateral frontal increased FMZVD, more on the right than the left; seizure semiology and interictal EEG favoured a right frontal onset, though some interictal EEG features were bifrontal. HD had bifrontal decreased FMZVD with also mid-cingulate and right frontal increased FMZVD. In this case interictal EEG showed bifrontal abnormality, but the seizure semiology favoured a right frontal onset (Figure 8.1). PK had bifrontal increased FMZVD, more on the right. Ictal onset was right frontal. ML had a medial frontal decrease in FMZVD, with seizure semiology suggesting a left frontal onset. PW and PS had medial occipital increases in FMZVD. Patients MF and SC had no significant abnormalities of FMZVD. Of these 10 unilateral frontal onset patients, 6 had an abnormality in the lobe of presumed seizure onset.

8.3.2. Patients without clear evidence of unilateral frontal seizure onset

Eight patients did not have good evidence for a unilateral frontal seizure onset. Patient CN had bilateral frontal increases in FMZVD and medial bilateral frontal decreases in FMZVD. Seizure onset was probably in the right hemisphere. Patient NM had a right frontal decrease in FMZVD; ictal onset was in the right posterior temporal region. 145

Patient AG had bilateral parietooccipital increases in FMZVD; localisation of seizure onset was unclear, though possibly frontal. Patient AA had a medial cingulate increase in FMZVD and a medial occipital decrease. Again, localisation of seizure onset was unclear, though possibly frontal. Patient RB, with seizure onset probably from the posterior left hemisphere had a right frontal increase in FMZVD. Patients KIO and RD, with bilateral largely posterior temporal EEG abnormalities, both had normal FMZVD. Patient AIF with seizures of possible frontal lobe origin also had normal FMZVD. 146

8.4. DISCUSSION

Thirteen of the 18 patients with extratemporal localisation-related epilepsy and normal high-resolution MRI had regions of significantly abnormal cortical FMZVD; 10 had regions of increased FMZVD, 6 had regions of decreased FMZVD, and 3 of these had both regions of increased and decreased FMZVD. 7 patients had an abnormality somewhere in the lobe of presumed origin of the seizures; 12 patients had an abnormality in the hemisphere of presumed origin, including 4 with the only abnormality being in the medial occipital cortex, medial frontal cortex or the cingulate gyrus. 9 patients had other remote abnormalities of FMZVD. Of 10 patients with unilateral frontal seizure onset, 6 had an abnormality in the implicated frontal lobe.

The method has a number of advantages over previously used methods. Voxel-by-voxel images of FMZVD were obtained through a single injection, single scan protocol and were then easily amenable to analysis using a voxel-based statistical technique, SPM. This has the advantages of being entirely objective, automated, statistically rigorous and able to analyse the entire brain volume. Region of interest (ROI) analyses, previously the conventional technique for the analysis of ligand PET data, have the disadvantage of being subjective, liable to observer bias and only sampling the areas of brain which were predetermined. The analysis has shown that, even at a highly conservative statistical threshold, abnormalities may be seen remote from the expected areas. This is consistent with our previous study of patients with MCD [475]. The method had good specificity: only one out of 24 normal controls had any regional abnormality of FMZVD. Further, a previous study of patients with HS using the same method showed no abnormalities outside the hippocampus in any case [410]. From this it may be concluded that the remote and discordant abnormalities seen in many of the patients with normal MRI are not “false positives”, but represent real abnormalities

In the group of 10 patients with unilateral frontal seizure onset, 6 of 10 had an abnormality in the lobe of presumed seizure onset, although 4 were bilateral. These results might provide guidance for further neurophysiological studies, particularly by guiding the placement of invasive electrodes. The interpretation of the findings in the 8 patients without good evidence of unilateral frontal onset is less clear. However, the finding of single lesions, particularly in patients NM and RB might again provide guidance for further neurophysiological studies.

The study of Chapter 5 identified regions of increased FMZVD in patients with MCD [475]. Increased FMZVD has been identified in 10 patients in the current study.lt is possible that the finding of increased FMZVD in these patients may represent occult MCD. 147

The studies of Chapters 6 and 7 revealed only regions of decreased FMZVD, confined to the area of the lesion, in patients with extratemporal localisation-related epilepsy due to dysembryoplastic neuroepithelial tumour, and also in patients with acquired lesions. It is possible that regions of decreased FMZVD detected in patients with normal MRI represent neuronal loss.

In the context of an attempt to locate the seizure focus in patients with localisation- related epilepsy and normal MRI, the “gold standard” may be considered to be ictal recordings from invasive electrodes. None of these patients have, as yet, been subjected to such investigations. Intracranial EEG recordings would be the best way to advance an understanding of these findings and place them into context in the clinical evaluation of such patients. In addition, any patients subsequently undergoing resective surgery would provide tissue to confirm or refute the suggestions as to possible aetiological significance of the changes in FMZVD.

In conclusion, using an entirely objective method, regional abnormalities of the cBZR- GABA^ receptor complex in 13 of 18 patients with localisation-related epilepsy and normal MRI have been demonstrated. In a subgroup of patients with unilateral frontal seizure onset, the regional abnormalities of cBZR-GABAy^ receptor density were concordant with other data in the majority of cases. These findings may be very useful in the evaluation of drug-resistant patients for surgery. 148

Subject Age Gender Years Drugs Time since Seizures EEG Probable Increased FMZVD Decreased FMZVD o f (daily last seizure; focus epilepsy dose frequenc y localisation in m g) (per month)

AC 18 Male 7 PHT 3 0 0 2 weeks Frequent nocturnal; head L front frequent interictal L frontal No significant CBZ 1 0 0 0 8 turns to L, L arm rises spikes regions

SW 24 Female 18 GBP 1 8 0 0 3 days Tingling in m o u th, R arm Ictal rfiythmic discharge L frontal L frontal and i 3 significant CBZ 1 2 0 0 12 posturing, evolving L frontocentral regions to 2° gen tonic-clonic

AO 36 Female 29 PHT 600 1 6 hrs L se nsorim otor SPS; Runs of bilateral anterior- Bilateral frontal No significant 3 0 CPS with arm dystonia; dominant sharp and slow. regions 2” gen tonic-clonic Som e sharp w aves in R frontotemporal region No ictal focal features; generalised attenuation only.

18 Female 11 LTG 600 3 weeks R arm dystonie and Bifrontal theta. Bilateral frontal Bilateral frontal 2 rising; head turns to R; 2” gen tonic-clonic

PHT 375 2 hrs Feeling of weakness on R, R front slow and sharp; Bilateral frontal No significant C B Z 1 8 0 0 4 0 loss o f aw areness, ictal onset R frontal fast. regions Rem ac 6 0 0 R arm dystonia, head looks down.

HL 24 Female 18 GBP 1200 36hrs Left face contorts with No clear ictal onset; R frontal No significant Medial frontal CBZ 4 0 0 3 0 left limb movements irregular w idespread regions rhythmical activity

2 2 M ale 21 PHT 5 0 6 weeks Head turns to L; Continuous R mid-front Mejj^l occipital No significant VP A 4 0 0 3 L arm dystonia; slow; occasional sharp regions CBZ 6 0 0 taps R hand activity in same area.

CBZ 8 0 0 4 hrs Head turns to L, Sharp waves phase reverse L frontai Medial occipital No significant PHT 4 0 0 4 0 R arm dystonia, a t F3 in sta g e II sleep. regions fidgetting with L arm Higher amplitude theta L anterior. No clear ictal onset; slow activity bilateral anterior.

MF 21 M ale 1 7 CBZ 1 9 0 0 1 6 hrs Head turning to R and L central sylvian high No significant No significant GBP 1 6 0 0 4 0 R limb jerking: second amplitude fast and spike regions regions pattern arm dystonia activity, may be and facial contortion. accompanied by head turn to R and R limb jerks.

28 Female 17 LTG 400 CPS with head turning to L; Low amplitude slow and No significant No significant wanders around and may sharp in R fronto-temp regions regions run; 2” gen tonic-clonic

Table 8.1. Clinical features of 10 patients with probably unilateral onset frontal lobe seizures and normal MRI. L=left; R=right; PHT=phenytoin; CBZ= carbamazepine; GBP=gabapentin; 2° gen =secondary generalised; occip= occipital; LTG=lamotrigine; Remac=remacemide; VPA=sodium valproate; FMZVD=flumazenil volume of distribution. 149

Subject Age Gender Years Drugs Time since Seizures EEG Probable Increased FMZVD Decreased FMZVD o f (daily last seizure; focus epilepsy dose frequency localisation in m g) (per month)

CN 37 Male 33 PHT 400 2 weeks L leg sensory SPS evolving Theta more marked R R hem isphere Bil fro n ta l Bil m edial fro n ta l Oxcarb 1200 10 to 2° gen tonic-clonic hemisphere. High voltage polyspike/wave during attack; no clear latéralisation

NM 47 Male 37 CBZ 1400 2 days CPS with facial grimacing, R post temp spikes at R post temp No significant Right frontal GBP 1 8 0 0 6 head turning to L;may seizure onset regions evolveto 2“ gen tonic-clonic

AG 26 Male 11 CBZ 2000 1 week Frequent nocturnal jerking Anterior dominant runs of parietal No significantBil PHT 3 0 0 12 of arms. CPS with sharp delta. Ictal events regions no clear latéralisation. preceded by diffuse high voltage sharp, followed by diffuse slow.

39 Male 30 LTG 100 12 hrs Usually from sleep;arms Bilateral anterior slow. Cingulate Bil occipital CBZ 1 6 0 0 16 held rigidly straight in Two events: bilateral central front; may evolve to rhythmic activity and 2“ gen tonic-clonic distorted spikes and sharp. No clear latéralisation.

RB 25 Male 17 CBZ 1000 1 week SPS with R visual field L post temp theta; more L post Right frontal No significant GBP 1 2 0 0 2 phenomena, may evolve extensive L temp/par sharp hem isphere regions to 2° gen tonic-clonic, and slow on hyperventilation. convulsing R>L

KIO 18 Female 10 CBZ 1400 24 hrs SPS (auditory phenomena); First pattern: ictal onset R ant temp No^nificant Rem ac 6 0 0 3 may progress to CPS with R ant temp theta. and regions with head turning to L. Second pattern: ictal onset L post temp Further pattern with R arm with L post temp sharp. dystonia and head turning to R

RD 2 3 Male 1 7 CBZ 1 2 0 0 4 hrs R sensory SPS; CPS Independent L and R post L post temp No significant No significant LTG 1 0 0 4 0 with L arm automatisms; temp spikes interictally. and R post regions regions 2° gen tonic-clonic Ictally; generalised attenuation temp only in most attacks; late R temp theta in one.

40 Male 13 CBZ 2800 2 days CPS; stands, claps hands; Frequent high amplitude No significant No significant LTG 7 0 0 4 rocks back and forward sharp waves over both regions regions if sitting frontal regions

Table 8.2. Clinical features of 8 patients without clear evidence of unilateral onset frontal lobe seizures and normal MRI. L=left; R=right; PHT=phenytoin; CBZ=carbamazepine; GBP=gabapentin; 2° gen = secondary generalised; occip=occipital; LTG=lamotrigine; Remac= remacemide; VPA=sodium valproate; Oxcarb=oxcarbezepine; FMZVD= flumazenil volume of distribution. 150

Figure 8. h Subject 4; regions of abnormal FMZVD rendered onto the coregistered, volumetrically normalised T1-weighted MRI. Right is on the left of the figure. Red: increased FMZVD p<0.001, corrected p<0.05; blueidecreased FMZVD p<0.001, corrected p <0.05. (a) transverse, (b) sagittal, (c) coronal sections; the white lines mark the position of the other two planes orthogonal to each section. 151 Chapter 9.

A voxel-level method for the comparison of structural and functional imaging data.

8.1. INTRODUCTION

Ligand PET studies have been used successfully to investigate many neurological diseases including epilepsy [393, 395, 406, 475, 491, 492], Parkinson's disease [493], Huntington's disease [403, 494, 495], Alzheimer's disease [454], amyotrophic lateral sclerosis [404] and others. Conventionally, ligand parameters were measured in a region of interest (ROI) in a patient, or group of patients, and these data were compared to the homologous region in a control group. Although automated region identification has been proposed using a variety of different methods [455, 456] region placement generally relies on an observer's expertise, is subject to observer bias and is limited in terms of the regions of brain evaluated. A further difficulty with ROIs results from systematic differences between the anatomy of patient and control groups. In a degenerative disease such as Huntington's disease, for example, the reduction of ^^C- flumazenil binding detected in the striatum may be a consequence of volume loss in this structure leading to partial volume effects. The apparently low ligand binding seen in an atrophied structure may actually be normal in terms of ligand binding per unit volume; indeed, partial volume effects leading to an apparent decrease in ligand binding could conceivably mask a real increase in binding per unit volume. Hence it has been recognised that methods of partial volume correction are necessary for a full characterisation of PET data [450-454], These methods share in common the use of high-resolution MRI to define cerebral structure and its coregistration with the PET data.

In order to avoid the restrictions imposed by ROI methods, particularly the subjectivity of region placement and the necegary limitation of the analysis to the selected regions, a method was described in Chapter 5 for the analysis of ’ ’C-flumazenil PET in patients with epilepsy using a voxel-based approach. Statistical Parametric Mapping (SPM)[475]. This study, and others described in this thesis and elsewhere [410, 490], demonstrated that SPM successfully detected significant differences in ligand binding in expected regions in the patient group compared to normal controls, but also detected significant abnormalities in brain regions that were not predicted in advance, with high 152 specificity.

SPM has also been applied to the voxel-based analysis of structural MRI data [496]. The next logical step was to attempt a voxel-based comparison of functional imaging and structural imaging in the same patient. PET data from a patient or patients can be compared with PET data from normal controls and the MRI from a patient or patients can be compared with the MRI of normal controls. The differences between the patients and control subjects within each of the two modalities can then be compared, both in terms of the anatomical postion and relative magnitude. Thus the issue of partial volume correction of the PET data is bypassed. This allows the identification of apparent functional abnormalities in ligand binding that are actually simply proportional to a change in the volume of a grey matter structure at that point (that are not generally interesting) and, conversely, the identification of functional changes that are not related to, or not accounted for by, any structural change (and hence interesting). This represents a group x modality interaction.

As an example of this approach the structural and functional imaging data from patients with partial seizures due to malformations of cortical development (MCD) have been compared. The data were derived from two published studies of the same group of patients: a study of regional cerebral volume measured on MRI in patients with partial seizures due to MCD and in a normal control group [54], and a study of cerebral ^^C- flumazenil binding to the y-aminobutyric acid type A-benzodiazepine receptor complex (GABA^-BZR) in patients with partial seizures due to MCD and in a normal control group [475], described in Chapter 5. The region-based data derived from the MRI study was used as a validation of the voxel-based approach to MRI morphometry. A method was developed that was based entirely on existing tools in the current release of the Statisitical Parametric Mapping software at the time of writing (SPM95). This method not only contributes to the understanding of the specific issue of alterations of benzodiazepine receptor density in malformations of cortical development, but may also be applied generally to compare structural and functional data in the whole range of brain disorders. 153

9.2. METHODS

9.2.1. Subjects

Ten patients were studied, who were recruited from the epilepsy clinics of the National Hospital for Neurology and Neurosurgery, Queen Square, London, UK. All have previously been reported in two previous studies, one of MRI volumetry in malformations of cortical development associated with localisation-related epilepsy [54] and one of benzodiazepine receptor density in this disorder [475], described in Chapter 5. The data from these studies were used. Twenty-four normal subjects of similar age were studied in the PET investigation. A separate group of 32 normal subjects were included in the MRI study. Sixteen normal controls were selected from each study control group in order to obtain the best possible age and sex match between the two control groups.

9.2.2. Scanning protocol - PET 9.2.3. Production of parametric maps of 11 C-fiumazenil binding

These are as described in Chapter 5.

9.2.4. Scanning protocol - MRI

MRI was performed using a 1.5T GE Signa (GE, Milwaukee, USA). A coronal spoiled gradient recalled sequence (SPGR) was performed for morphometric analysis (TE 5ms, TR 35ms, flip angle 35 degrees, acquisition matrix 256x128, 1 NEX, field of view 24cm, producing 124 contiguous slices 1.5mm thick).

9.2.5. Segmentation of MRI and MRI volumetry

MRI images were transferred to an independent imaging analysis workstation (Allegro, ISG Technologies, Toronto, Canada). This permitted segmentation, division into ROIs and subsequent volume measurements, as described previously [54]. In brief, the segmentation was acheived by a combination of selection of a pixel intensity threshold with subsequent region growing from a seed placed by the operator and manual editing to separate other unwanted structures from the ROI. A binarised image of the entire MRI- defined cortical grey matter from both cerebral hemispheres (MRIGM) was produced for each subject and used for analysis with SPM. Cortical grey matter and subcortical matter volumes of interest from each hemisphere were divided into 10 coronal regions of interest, each spanning one tenth of the total anterior-posterior extent of the 154 hemisphere. The volume of each grey matter and subcortical block was measured and normalised for total intracerebral volume as described previously [54]. In this previous study a high degree of intra- and inter-observer reliability was established for the volume measurements. For each coronal volume of interest, measurements were made of the grey matter volume for each hemisphere, the subcortical matter volume for each hemisphere, the ratio of grey matter volumes from opposite hemispheres, the ratio of subcortical matter volumes from opposite hemispheres, and the ratio of grey matter volume and subcortical matter volume for each hemisphere. Thus, a total of 80 measures was produced for each brain. The previous study showed no normal control had more than one abnormal measure; abnormality was defined as the presence of more than one abnormal measure.

9.2.6. Data analysis

Data were analysed on a Sun SPARCclassic workstation (Sun Microsystems, Mountain View, CA) using Analyze version 7.0 (Mayo Foundation) [471], MATLAB (The MathWorks Inc., Natick, Mass.) and Statistical Parametric Mapping (Wellcome Dept, of Cognitive Neurology, Institute of Neurology, London) [457-459].

9.2.7. Spatial normalisation of PET and MRI

All of the images were transformed into a standard three-dimensional space. Because the normal data for MRIGM and FMZVD were from different individuals, two separate but anatomically identical images were required as templates: one for the MRIGM images and one for FMZVD images. To ensure these templates were in exact register, one normal MRI was chosen. This image was segmented as described above to produce an MRIGM image and a second image with the thalami, basal ganglia and cerebellum included. Both of these images were flipped in the transverse plane and then averaged with its unflipped counterpart to produce two symmetrical images. Finally, both images were smoothed with an 8mm isotropic Gaussian kernel, to produce two anatomically identical images, one consisting of only cerebral cortex (the MRIGM template) and the other with cortex and subcortical structures (the FMZVD template).

9.2.8. Statistical Parametric Mapping

For the analysis of each individual modality in isolation, the SPM{t[ were transformed to the unit normal distribution (SPM{Z}) and thresholded at p < 0.001. To correct for the multiple non-independent comparisons inherent in this analysis, the resulting foci were then characterized in terms of spatial extent (k ) [459]. This characterization is in 155 terms of the probability that a region of the observed number of voxels (or bigger) could have occurred by chance (P(nmax > k )) over the entire volume analyzed (i.e a corrected p-value). The corrected p-value chosen was p < 0.05.

9.2.9. Analysis of structural data using SPM

Each individual patient's MRIGM image was compared with the normal group as described previously in Chapter 5 for the comparison of individual and normal group FMZVD images. All images were smoothed with a 13x13x9mm Gaussian kernel, resulting in an effective smoothness identical to FMZVD images smoothed at 10x10x6mm. The smoothed MRIGM images can be regarded as images in which voxel intensity reflects the local grey matter volume at a spatial resolution identical to that of the PET studies. Global mean voxel value was included as a confounding covariate and this analysis can therefore be regarded as an ANCOVA [476]. The occurrence and position of significant regional differences between the patients and the normal controls were compared with the results from the MRI volumetric analysis described above [54], as a validation of the use of SPM in this context.

9.2.10. Analysis of ^ ^C-flumazenil PET data

Each individual patient's FMZVD image was compared with the normal group, described previously in Chapter 5.

9.2.11. Comparison of MRI and PET images

Having spatially normalised the images, the comparison of MRI and PET was greatly facilitated. The transformed, smoothed images were maps of local grey matter volume and local flumazenil V^ at the same resolution. The question addressed at each voxel was: which is the greater effect, change in local grey matter volume, or change in benzodiazepine receptor density? In essence, the analysis was a group by condition interaction within SPM, "group" being individual patient or normal controls, and "condition" being imaging modality, MRI or PET.

In order for this analysis to be possible there must be a consistent relationship between grey matter volume and FMZVD at each voxel over subjects. The measure of FMZVD per unit grey matter volume could be variable across brain regions, however. In order to compare the two sets of images the variance at each voxel should be comparable between the imaging modalities. A consistent correlation between grey matter volume and FMZVD at each voxel was assumed. Proportional scaling was used to acheive global normalisation 156

of voxel values between scans, to render the variance at each voxel, in the two sets of data, the same.

Two contrasts can be specified testing for an interaction between the MRI and PET datasets: Regions showing unexpectedly high FMZVD compared to grey matter volume, and regions showing unexpectedly low FMZVD compared to grey matter volume. These results can be compared with the comparison of individual FMZVD images against the normal controls to reveal those areas of brain in which apparently significant differences in FMZVD were due to abnormalities of local grey matter volume. Regions were defined with a threshold of p < 0.01. Regional abnormalities were retained only if they were significant after a correction for multiple comparisons [459], using p < 0.05 for the analysis of MRIGM and FMZVD images in isolation and and p < 0.5 for the interactions. A low threshold was chosen for the interactions to ensure a low type II error in detecting mismatch between tracer uptake and grey-matter density. Since heterotopic nodules were excluded from the MRIGM images by virtue of the segmentation process, any result showing a significant interaction confined only to heterotopic nodules (present in the FMZVD images but not in MRIGM images) was excluded.

The method is summarised in figure 9.1. 157

9.3. RESULTS

The clinical features of the 10 patients are described in table 9.1.

9.3.1. SPM analysis of MRI structural data and comparison with region- based morphometries

Comparison of each normal control MRIGM image with the remaining 15 normal controls revealed only one significant region in one individual at the statistical threshold of p < 0.001, corrected p < 0.05. Since 32 tests were made (examination for regions of increased grey matter volume and examination for regions of decreased volume in each of 16 subjects) at least one significant region was expected by chance.

Five of the 10 patients had abnormal block volumes according to the region-based morphometries; the approximate position of these abnormal blocks is marked in Figure 9.2. Since the block volume measurements were performed on data which were not spatially normalised, the position of the blocks were approximately indicated by dividing the spatially normalised brain into 10 equal coronal partitions. Several measures were made in each block (grey matter volume, white matter volume and both within-block ratios and ratios between the block volumes and those of the opposite hemisphere homotopic block); the various categories of block abnormality have not been separated since the different measures are not necessarily independent.

Using SPM and a threshold of p < 0.05 (corrected) with ANCOVA, the same five patients were found to have abnormalities of MRIGM. The remaining five patients were found to be normal by both techniques. The SPM results are displayed next to the block volume results for comparison in Figure 9.2. Although some abnormalities are anatomically concordant, particularly in MC, RFI and JW, others are not. Flowever, using these two methods, the concordance in terms of detection of an abnormal brain is striking. SPM permitted the identification of regions of increased or decreased grey matter volume with respect to controls in greater anatomical detail compared to region-based morphometries.

9.3.2. Comparison of structural and functional images

Figure 9.3 shows the SPM results of analysis of the MRIGM data displayed alongside the SPM analysis of the FMZVD data; the results of the analysis of the interaction between these two imaging modalities is also shown. For all of these analyses proportional scaling 158 was used and regions were defined using a threshold of p < 0.01, significant regions are reported at p < 0.05 for the within-modality comparisons and at p < 0.5 for the interaction. In two cases (SH, KS) no abnormalities were found with either modality and no significant interaction was revealed. In two further cases (TJ, and TP) significant abnormalities in one imaging modality were associated with comparable abnormalities in the other modality such that no regions of abnormal FMZVD per grey matter volume were identified: TJ had a region of increased FMZVD in the right central region accounted for by increased grey matter volume in that area; TP had a region of decreased FMZVD in the right posterior temporal/occipital area accounted for by a decrease in grey matter volume in that area. In both of these cases there were other areas of abnormal grey matter volume which were associated with comparable changes in FMZVD in those areas which, while non-significant when examined alone, were sufficient to exclude any significant interaction.

The remaining six cases all showed significant interaction effects, with regions of significantly abnormal FMZVD per grey matter volume. Five of the seven subjects found to have abnormalities of FMZVD examined in isolation (MC, AF, RH, MS, JW) were shown to have a significant interacton between MRIGM and FMZVD, demonstrating that the abnormal FMZVD is not merely a consequence of abnormal anatomy, but reflects abnormal benzodiazepine receptor density in these grey matter regions. However, this was not the case for all areas in these subjects. For example, the cingulate gyrus in JW and left parietal cortex in RH, which were found to show abnormal FMZVD examined in isolation, were found to have a proportionate change in grey matter volume in these areas. Specifically, in the case of JW, increased cingulate FMZVD was accounted for by an unsuspected increase in grey matter volume in this area; in the case of RH, decreased left parietal cortex FMZVD was accounted for by an unsuspected decrease in grey matter volume in this area.

In one subject, SN, interaction analysis showed an abnormality of ^^C-flumazenil that was not evident on analysis of the FMZVD data in isolation: the left prefrontal area showed an increase in grey matter volume, but no corresponding increase in FMZVD. Even in those subjects in whom both abnormality of MRIGM and abnormality of FMZVD were seen when analysed in isolation, the interaction analysis provided “added value”. For example, in RH a schizencephalic cleft is seen in the left central area, reflected by increased MRIGM and increased FMZVD in this area. The interaction analysis shows, unexpectedly, that the posterior lip of the cleft has reduced ^^C-flumazenil binding per grey matter volume and the anterior lip has increased ’ ^C-flumazenil binding. Further, in subject MS there were no regions of decreased FMZVD but several regions of decreased grey matter volume; the interaction analysis revealed that some of these areas of 159 decreased grey matter volume have relatively high ^^C-flumazenil binding, which was not apparent in the analysis of FMZVD images alone.

SN, TP and KS had significantly high FMZVD compared with MRIGM in the region of heterotopic nodules only. These results were artefacts consequent upon the exclusion of heterotopic nodules from the MRIGM images, and are not shown. 160

9.4. DISCUSSION

9.4.1. Voxel-based analysis of MRI using SPM

This study has shown that five of ten patients with MCD and localisation-related epilepsy had abnormalities of regional cerebral volumes as measured in coronal blocks. It has also been shown that these same five patients had abnormalities of regional grey matter volume detected using SPM. The remaining five patients were found to be normal by both techniques. The specificity of SPM in this context was demonstrated by the finding that only one of the sixteen normal controls was found to have a single region of abnormal MRIGM in comparison with the remaining 15 controls. This chance finding was expected at a corrected threshold of p < 0.05. SPM permits an anatomical localisation of regional abnormalities of cortical grey matter volume at a resolution greatly superior to the block volume method. In addition, SPM is entirely objective and is not constrained by predetermination of volumes of interest. The finding of an abnormality of a block volume does not allow the location of this abnormality to be pinpointed. The analysis of structural MRI data using SPM may be a useful technique in the investigation of patients with epilepsy as well as other neurological and psychiatric disease, particularly to delineate the extent of a structural abnormality or when no abnormality is evident by visual inspection.

9.4.2. Voxel-based comparison of MRI and PET in MCD using SPM

This study has also shown that six of 10 patients showed regions of cerebral cortex with disproportionate flumazenil binding compared with local grey matter volume by formally testing for an interaction effect. Five of these had regional abnormalities of FMZVD on examination of the FMZVD images alone. These included both areas of increased FMZVD as well as areas of decreased FMZVD. In addition, it was shown that one patient with abnormalities of MRIGM but no abnormality of FMZVD examined alone showed abnormal FMZVD when compared with MRIGM. Finally, no patient with both MRIGM and FMZVD images found to be normal when examined in isolation was found to have any disproportionate flumazenil binding. The identification of regions of increased flumazenil binding in patients with MCD when abnormalities of cortical volume have been taken into account confirms the previous finding of increased flumazenil binding when FMZVD images were examined in isolation in Chapter 5 [475]. This might be explained by either an increased density of neurons bearing GABA^ receptors, an increased density of these receptors per neuron, or an increased affinity of these receptors. Other potential explanations for the finding of increased flumazenil binding in patients with epilepsy included abnormalities of grey matter structure, or possibly an 161 artefactual consequence of global normalisation of FMZVD data. Since both anatomical abnormality and global normalisation would have the same effect in MRIGM images as in FMZVD, direct comparison of these two modalities allows the exclusion of these effects: it was found, in the majority of cases, that regions of relatively low or high flumazenil binding remained when these factors were taken into account. Therefore, the finding of regions of increased flumazenil binding, as well as decreased flumazenil binding per unit volume of grey matter, is a real phenomenon. In some cases, however, abnormalities of FMZVD images examined in isolation were shown to be accounted for by subtle and previously unnoticed abnormalities of grey matter volume. For example, RH had a region of low FMZVD in the left parietal lobe; in the same area there was a region of reduced MRIGM. As a result, no disproportionate flumazenil binding was seen in that area. Similarly, increased FMZVD was observed in the cingulate of JW; increased MRIGM was also seen in this area, such that no disproportionate flumazenil binding was detected.

No individual with both normal FMZVD and normal MRIGM examined independently had any disproportionate flumazenil binding when the two modalities were compared. The normal control group for MRIGM was composed of different individuals from the normal control group for FMZVD. This study has shown that the detection of abnormality of MRIGM in normals is exceptional. Chapter 5 has already shown that abnormalities of FMZVD in normal controls are also very rare [475]. It is an assumption of this study that both anatomy and function would be normal in all of the normal controls, although only one modality of imaging has been performed in each normal. It is reassuring that no patient with normal MRI and PET imaging, analysed independently, showed any significant abnormalities when the structural and functional data were compared directly.

9.4.3. Voxel-based comparison of MRI and PET - general methodological considerations

Several important general issues are addressed by these findings. It might be thought that the processes of volumetric normalisation and smoothing of images might artefactually create apparent anatomical abnormalities in normal brain images, or remove real abnormalities from abnormal images. It has been demonstrated that neither of these phenomena occur. In total 26 individuals were studied; five were found to be abnormal by a rigorous and meticulous regional volume measurement technique; the same five were designated abnormal by SPM.

The technique allowed for inferences about abnormalities in FMZVD per grey matter volume in a way that avoids using the ratio itself. This was achieved by testing for 162 interactions. This is important because alternative approaches based explicitly on the ratio would not lend themselves to parametric statistical analysis because quotients of this sort are not normally distributed. In order to combine the two modalities of data implicitly in a statistical model, similar distributions under the null hypothesis are required. In this case proportional scaling by the whole brain mean was used to achieve this.

A crucial feature of the method is the use of proportional scaling for global normalisation of voxel values. Assuming a constant relationship between MRIGM and FMZVD in each brain region, global normalisation by proportional scaling also leads to a proportional scaling of the error variance such that homoscedasticity is assured. The interaction effect can only be examined under this condition. The assumption of a constant relationship between MRIGM and FMZVD in each brain region is entirely reasonable given that (under normal circumstances) tracer uptake will be proportional to receptor density, which will in turn be proportional to local grey matter volume. It has been shown that differences between normal controls in each nrradality are hard to detect, therefore there is a predictable distribution of values at each voxel in the two modalities. It is also important to ensure the spatial autocovariance structure of the two data sets is the same. This can be achieved by differential smoothing to the same smoothness; the method of estimating smoothness has been described elsewhere [497].

The feasability of using different templates for different imaging modalities while maintaining exact registration of the two normalised images for the same individual has been shown. This is partly demonstrated by the point above: misregistration would result in artefactual differences between the two imaging modalities in individuals with normal imaging in each modality considered independently. It is also demonstrated by the observation that in a number of cases abnormalities in one modality are precisely counterbalanced by abnormality of the other modality in the same anatomical region. Abnormalities of FMZVD and MRIGM would only be seen to "cancel out" if the coregistration was adequate.

A number of criticisms could be made of this method. Firstly, MRIGM images were assigned a single constant voxel value to all regions of cortex before smoothing; benzodiazepine receptor density is known to be not constant across all grey matter regions in human brain by autoradiographic mapping [159]. Although gross differences in receptor density over the brain may violate the assumption of homoscedasticity above, they are not, in themselves a problem for the technique: This is because a low (or high) receptor density will result in a difference between FMZVD and MRIGM means only ie. a main effect of modality. However, FMZVD per unit grey matter differences represent 163 interactions and are not affected by this main effect. This argument assumes, of course, that the error variances are the same for the FMZVD and MRIGM data after normalisation. If this is not the case more sophisticated normalisation transformations may be required (eg. by weighting every segmented image by the mean FMZVD image over all subjects). Secondly, no account has been taken of the contribution to apparent FMZVD values in cortex by "spill-in" from activity in surrounding white matter. However, it has been shown that benzodiazepine receptor binding in white matter is negligible compared to cortex [159]. For a ligand with significant white-matter binding, the assumption of a non-grey value of zero in the MRIGM images would have to be re-evaluated. A more sophisticated approach would be to combine grey, white and CSF segmented image partitions linearly in a way that best predicted the mean FMZVD image in a least squares sense. Finally, the method outlined here has still required interactive segmentation of the MRI data by a skilled investigator. This is labour-intensive and somewhat subjective, although our data has been shown previously to have been segmented with good inter- and intra-operator reliability [54]. Automated MRI segmentation techniques now available as part of SPM96 (Wellcome Department of Cognitive Neurology, Institute of Neurology, London, UK) should enable this step of the analysis to be automated and rendered entirely objective.

Although voxel-based comparison of MRI and PET using SPM has many advantages, particularly the automated and objective analysis of all brain regions, this method may have some disadvantages. The spatial transformation of images and normalisation to global mean enable relative regional changes to be detected but at the expense of the detection of global change and the measurement of absolute quantitative values. A ROI- based approach, though subjective in terms of region placement and limited to the regions selected, allows absolute quantification. Using SPM to compare structural and functional images might be used to detect regions of significant abnormality and to determine whether any functional abnormality could be explained by structural abnormality in that region. These significant regions could then be analysed with a ROI technique employing partial volume effect correction [450-454] to obtain absolute quantitative values.

In conclusion, the validity of a voxel-based technique for the comparison of individual structural images with a normal group of such images has been demonstrated, which allows subtle structural abnormalities to be located. Previous chapters of this thesis described the application of voxel-based methods to the comparison of individual PET ligand images with a group, which allowed functional abnormalities to be located. Building on these two approaches, a novel technique has been described in this chapter for the direct comparison of structural and functional images using SPM, which allows 164 issues concerning partial volume effects and mixed tissue sampling in PET ligand images to be resolved. This is pertinent to any ligand PET investigation in which either atrophy of a structure or increase in volume, as in MCD, may be encountered. Using this method we have demonstrated that although patients with MCD show changes in cortical volume and changes in FMZVD these changes are often disproportionate; 6 out of 10 of the patients showed abnormalities of FMZVD which were not accounted for by changes in cortical volume. The abnormality of FMZVD per grey matter volume detected by this method demonstrates the existence of a functional abnormality in patients with MCD in addition to structural abnormality. This functional abnormality could be due to abnormal neuronal density or abnormal numbers or affinity of benzodiazepine receptors. The analysis of ligand PET data should always include a comparison with structural data, in the form of high-resolution MRI, which is greatly facilitated by the approach described here. 165

MRI finding Age Duration of Gender Seizures Interictai EEG Drugs (daily dose Time since

epiiepsy features in mg) last seizure

(years) (* - ictal) (before PET)

MC Bilateral cortical dysgenesis 18 17 Female L arm focal Bilateral independent LTG 400; CBZ 1000 2 days

heterotopic nodules motor + 2° gen. foci *

AF Subepenc^al heterotopia 47 21 Female CPS and 2° gen. Bilateral temporal fod PHT 375 2 months

tonic

RH Bilateral cortical dysgenesis 34 33 Female L arm focal No definite abnormality CBZ 800 2 days

and schizencephaly motor + 2° gen.

TJ Bilateral band heterotopia 33 31 Male CPS + 2° gen. Irregular generalised ESM 1000; PHT 350; 36hrs

spikes LTG 100

SN Single R posterior parietal 26 7 Female SPS No definite abnormality CBZ 700 4 years

heterotopic nodule

TP R temporoparietal cortical 29 27 Female CPS + 2° gen. R posterior hemisphere vpa 2500; CBZ 600; 8 days

dysgenesis and heterotopias focus LTG 100

MS Bilateral band heterotopia 30 18 Male R arm focal L hemisphere slow CBZ 600 4 d ^

motor + 2“ gen.

JW L frontal and R temporal 37 13 Female 2° gen. tonic Bilateral independent GBP 1200; CBZ 1300 6 hrs

cortical dysgenesis fod

SH Small nodule in post. L 29 14 Female L arm focal Generalised spike wave CBZ 1000 3 days parietal white matter motor + 2° gen.

KS Subependymal heterotopia 30 6 Female R arm focal Normal VPA 600; PHT 300 36hrs sensory + 2° gen.

Table 9. h Clinical features of the W patients with MCD. R=right; L=left; Z'"gen=secondary generalised; CPS=complex partial seizures; SPS=simple partial seizures; LTG=lamotrigine; CBZ= carbamazepine; PHT=phenytoin; ESM=ethosuximide; VPA=sodium valproate; GBP=gabapentin 166

PET

As acquired FMZVD i.

Segmentation and binarisation

MRIGM

Volumetric Volumetric normalisation normalisation

Normalised Normalised MRIGM FMZVD Comparison of structural and functional data Smoothing Smoothing

Smoothed, Smoothed, normalised normalised MRIGM Voxel-based comparison FMZVD o f MR IG M and F M Z V D with respect to controls h Comparison Comparison with with 'vf controls, controls, within within modality modality J Significant Significant regions regions superimposed superimposed on on normalised normalised MRI MRI

Figure 9 . 7. Cartoon illustrating schematically the method of voxel-by-voxel comparison of structural and functional data. Subject JW is used as an example. 167

Approximate position of Increased grey matter Decreased grey matter Subject block volume abnormality volume (SPM) p<0.001 volume (SPM) p<0.001

■Qflfttol corortd

MC

RH

! SI s'# ^ 4 TJ No significant voxels

■ogittat corond

MS No significant voxels

JW

Figure 9.2. Comparison of block volume and SPM analysis of MRIGM Images. 168

. , . Increased Decreased Increased Decreased Relatively high Relatively low Subject MRIGM p<0.01 MRIGM p<0.01 FMZVD p<0.01 FMZVD p<0.01 FMZVD p<0.01 FMZVD p<0.01 EEG focus

Bilateral widespread MC epilept abnormalityabnorn

Bilateral No significant anterior AF voxels temporal spikes

No definite RH abnormality

Left 4 No significant No significant No significant No significant hemisphere TJ voxels voxels voxels voxels slow

No significant No significant No significant No definite SN voxels voxels voxels abnormality

Right No significant No significant No significant posterior TP voxels voxels voxels spikes

Irregular No significant No significant generalised MS voxels voxels spikes

Bilateral JW fronto- temporal spikes

Figure 9.3. Results of the comparison of MRIGM and FMZVD images using SPM to detect a significant interaction effect. For each subject the significant regions of increased and decreased grey matter volume examined in isolation, the significant regions of increased and decreased flumazenil binding examined in isolation, and the regions of disproportionately increased and decreased flumazenil binding compared to grey matter volume are shown. 169 Chapter 10.

The cortical-subcortical GABA^-BZR system in frontal lobe seizures

10.1. INTRODUCTION

Most hypotheses explaining how seizures arise implicate a relative imbalance between local cortical excitatory and inhibitory influences. Over the past two decades these concepts have been refined, particularly in animal models of epilepsy arising in the limbic system and in epilepsy in man caused by hippocampal sclerosis. Other than in an animal model of chronic neocortical injury in which a sub-population of GABAergic cortical interneurons is lost [230], relatively little is known of the mechanisms of seizures in neocortical epilepsy.

Cortical excitability and the spread of seizure activity are strongly influenced by subcortical structures [180, 236, 241]. The current model of basal ganglia function in relation to the neocortex is complex, but permits some predictions in man concerning the control of cortical excitability by the striatum (figure 1.12)[249]. Neocortical input to the striatum is excitatory, and glutamatergic. The striatum also receives an inhibitory, GABAergic, input from the pallidum. The major outputs from striatum are inhibitory, GABAergic outputs to the internal portion of the globus pallidus and the substantia nigra pars reticulata. These latter areas project via GABAergic inhibitory pathways to the ventral anterior and ventral lateral nuclei of the thalamus, which in turn project, via excitatory pathways, to the cortex. Thus, inhibiting the activity of the striatum leads to increased inhibition of the thalamus and hence decreased cortical excitability.

Potentiation of inhibition is associated with an increased number of GABA^-BZRs [1 8 8 ]. Flumazenil volume of distribution (FMZVD), a parameter measurable at the voxel level with PET, is directly correlated with benzodiazepine receptor density and hence acts as an index of GABA^-BZR density.

Using the technique of Statistical Parametric Mapping (SPM), an automated and objective means of comparing groups of images, a group of 10 patients with unilateral frontal lobe epilepsy (FLE) and normal MRI was investigated in order to detect any abnormalities of FMZVD in the cortex and basal ganglia. 170

10.2. METHODS

1 0.2.1. Subjects

We studied 18 patients who were recruited from the epilepsy clinics of the National Hospital for Neurology and Neurosurgery, Queen Square, London, UK. Clinical details of the patients are summarised in tables 8.1 and 8.2; the individual patient findings were described in Chapter 8. All had partial seizures and had undergone a high resolution MRI protocol including II-weighted volume acquisition with 1.5mm partitions and both coronal T2-weighted and coronal proton density acquisitions covering the whole brain in contiguous 5mm slices (GE Signa 1.5T). The MRI data had been reported to be normal by an experienced neuroradiologist aware of the clinical information. The data from the same 24 normal subjects as in Chapter 5 were used.

10.2.2. Scanning protocol

Scans were performed using an ECAT-953B PET scanner (CTI/Siemens, Knoxville) [460]. Data were acquired in 3D mode, with the septa retracted to improve sensitivity [425]. Dual energy window scatter correction [427] was employed.

10.2.3. Derivation of plasma input function 10.2.4. Data analysis 10.2.5. Production of parametric maps

These aspects of the method were identical to that used in Chapter 5.

10.2.6. Statistical Parametric Mapping

Images from patients with a unilateral onset in the left hemisphere were flipped so that all subjects with unilateral onset had right-sided foci.

For the comparison of the patient and normal groups, the SPMjt} were transformed to the unit normal distribution (SPMjZ}) and thresholded at p<0.01, p<0.001 and p<0.0001 uncorrected. There was no a priori hypothesis defining the regions to be examined;therefore, to permit analysis of the entire brain volume a correction for multiple non-independent comparisons was made. Each focus exceeding the uncorrected threshold was then characterized in terms of spatial extent ( k ) [459]. The corrected threshold chosen was p<0.05. 171

For the analysis of correlation between FMZVD and seizure frequency, seizure frequency (total number of all seizure types per month as determined from patients’ prospectively compiled diaries) was included in the model as a covariate of interest in a parametric analysis. The patient data were examined alone. Firstly, all 18 patients were examined for regions where FMZVD correlated significantly with seizure frequency; this analysis was confined to those regions found significant in the group comparison above. Therefore, the SPM{Z} was thresholded at p<0.001, uncorrected. Secondly, the 10 patients with unilateral frontal seizures were examined alone for regions where FMZVD correlated significantly with seizure frequency; this analysis was confined to the frontal cortex. Therefore, the SPM{Z} was thresholded at p<0.001, uncorrected.

The significant regions were then superimposed on an “average normalised” MRI constructed from a voxel-by-voxel mean of the normalised T1 -weighted MRIs of 8 of the normal control group. 172

10.3. RESULTS

Three separate analyses were performed. Thirteen patients had unilateral seizure foci, 7 on the right and 6 on the left; five had bilateral foci or seizures not clearly lateralised. Images from patients with unilateral left-sided foci were flipped so that all unilateral foci were on the right during the group analysis.

The first analysis compared all 18 patients with the 24 normal controls. This revealed FMZVD significantly increased interictally in a number of cortical and subcortical regions. No regions of decreased FMZVD were found. The main regions of increased FMZVD were the ipsilateral motor cortex (highest Z-score in this region 5.68), ipsilateral putamen (Z=5.21), ipsilateral lateral premotor cortex (Z=4.8), anterior cingulate (Z=5.08), contralateral putamen (Z=4.4), contralateral lateral premotor cortex (Z=4.35) and contralateral motor cortex (Z=3.94) (figure 10.1a).

In the second analysis only the data from the 18 patients was examined, looking for regions where FMZVD correlated significantly with seizure frequency, confining attention only to those regions identified in the first analysis. FMZVD correlated negatively with seizure frequency in a number of regions. Three of these regions coincided with anatomical areas found significant in the first analysis; ipsilateral putamen (Z=3.97), cingulate (Z=3.45) and ipsilateral premotor cortex (Z=3.1) (figure 10.1b and figure 10.2). No significant positive correlations between FMZVD and seizure frequency were observed when examining all 18 patients together.

In the third analysis only the data from a subgroup of 10 patients with unilateral frontal lobe seizure foci was examined for voxels in which FMZVD correlated significantly with seizure frequency. The analysis was confined to frontal cortex. FMZVD correlated positively with seizure frequency in only one area, a region of ipsilateral frontal cortex, Z=3.76 (figure 10.1c). No significant negative correlations between seizure frequency and FMZVD were found for this subgroup. 173

10.4. DISCUSSION

Significantly increased FMZVD in 18 patients with LRE and normal MRI, compared to 24 normal controls, has been shown in the ipsilateral putamen, ipsilateral primary motor cortex, ipsilateral lateral premotor cortex, anterior cingulate, contralateral putamen, contralateral primary motor cortex and contralateral lateral premotor cortex. Furthermore, FMZVD correlated negatively with seizure frequency in ipsilateral putamen, cingulate and ipsilateral premotor cortex. FMZVD correlated positively with seizure frequency in a region of ipsilateral frontal cortex in 10 patients with unilateral frontal lobe seizures.

These changes are complex and suggest a central role of the GABAergic system in either the pathogenesis or adaptive modulation of seizures, or both. It is postulated that these findings indicate an adaptive response. This adaptive response could either be a physiological one, via intrinsic mechanisms, or the result of antiepileptic medication.

The increase in FMZVD (assumed to reflect an increase in GABA^-BZR density) is largely confined to the striatum/pallidum, motor cortex, anterior cingulate and lateral prefrontal cortex, a distributed network known to be involved in simple motor tasks in man [498]. Two of these areas, the ipsilateral premotor cortex and anterior cingulate, also demonstrated a negative correlation between FMZVD and seizure frequency, suggesting that this putative adaptive response serves to reduce seizure frequency. This increase in GABAy^^-BZR density could reflect receptor upregulation because of impaired GABA release, implying impaired inhibition in these regions and hence that these changes are part of the cause of the seizures. Alternatively, the increase in receptors may be an 'attempt" to increase cortical inhibition, an adaptive response reducing the tendency for seizures. The negative correlation between FMZVD and seizure frequency in a subset of these regions suggests that this is an adaptive response.

These data suggest that an increase in (jABA^^-BZR density in the ipsilateral putamen is associated with a reduced seizure frequency in patients with LRE and normal MRI. If increased GABA^-BZR density reflects increased inhibition locally, the increased inhibition of the putamen would result in reduced cortical excitability in the lobe of seizure origin, via a polysynaptic pathway through either globus pallidus pars interna or substantia nigra pars reticulata and the ventral anterior and ventrolateral thalamic nuclei. Thus, one can speculate that the greater the increase in GABAy^-BZR density in the striatum the fewer the seizures, as observed in this study. If the increase in GABA^-BZR density were due to an upregulation following reduced GABA release, then one could hypothesise that a greater increase in striatal GABA^-BZR density (and hence the less 174 the inhibition in this area because of reduced GABA release) would result in more cortical excitation and hence more seizures. This was not observed.

A region of the ipsilateral frontal cortex showed a positive correlation with seizure frequency. We interpret this as receptor upregulation because of impaired GABA release, implying impaired inhibition in this region. Only one other study has examined the correlation between GABA^-BZR density and seizure frequency; the majority of the patients in that study had temporal lobe seizure foci and the GABA^-BZR density at the seizure focus correlated negatively with seizure frequency [414]. The differences in patient group and methods prevent a meaningful comparison of this previous study with the data presented here.

These arguments may be assembled into a model for the role of the GABAy^-BZR system as a feedback modulator of neocortical seizures (figure/^j). The essential elements of this model are: (1) locally impaired GABA^-mediated inhibition in the seizure focus as part of the local epileptogenic mechanism; (2) increased GABAy^-BZR density in a motor network including cortical and basal ganglia regions as a result of seizures, as an adaptive response; (3) reduced cortical excitation as a result of increased striatal inhibition. The model suggests that the increased cortical GABAy^^-BZR density is a relatively ineffective seizure suppressing mechanism, perhaps indicating a relative failure of cortical GABA^-BZR function as part of the underlying epileptogenic mechanism. The increase in inhibition of the striatum is relatively more effective, possibly since it relies on the reduction of excitatory mechanisms rather than in increase in inhibitory mechanisms. Evidence for (2) has been discussed in the introduction: Increased GABA^-BZR density has been observed in the dentate gyrus, hippocampus and olfactory bulb of limbic kindled seizure models [185, 187], in a kainic acid model [197] and in the cobalt focus model [223}. An increased number of GABAA-BZRs is associated with potentiation of inhibition [Otis, 1994 #1069]. The increased GABA^^-BZR density observed in our study could be the result of gene expression or the result of antiepileptic drug action. Alteration in neuroreceptor [199] and other gene expression [203] as a result of the occurence of seizures has been demonstrated.

Although the changes in FZMVD have bee/»interpreted as changes in GABA^-BZR density, it is recognised that FMZVD is a compound measure related both to density of available receptors and receptor affinity. However, studies in which BZ receptor density and affinity have been measured separately in patients with epilepsy [406, 416] have shown no changes in receptor affinity, so this is unlikely to be the explanation. 175

These data demonstrate for the first time in man the existence of a potential GABAergic basal ganglia influence on epileptogenicity. It has previously been suggested that the principal site of action of GABA-modulating antiepileptic drugs in some animal models is in the basal ganglia regions rather than in cortical seizure foci [499]. It is plausible that in man a variety of cortical abnormalities may give rise to partial seizures, but a common pathway in seizure frequency modulation and of the effect of antiepileptic drugs may lie in the GABAergic system of the basal ganglia.

In conclusion, a widespread increase in FMZVD in motor regions in patients with neocortical localisation-related epilepsy and normal MRI has been demonstrated. A subset of these regions show a negative correlation between FMZVD and seizure frequency. In patients with frontal lobe seizure foci, a cortical region in the ipsilateral frontal lobe showed a positive correlation between FMZVD and seizure frequency. A testable model is postulated to explain these findings in which seizures give rise to a widespread increase in GABA^-BZR density; the increase in cortical GABAy^-BZR density does not effectively modulate seizure frequency, but the increase in GABA/^-BZR density in the basal ganglia acts to reduce seizure frequency. 176

A

(a) r

Figure W .U Results of group analyses of LRE patients. (a) Regions of significantly increased FMZVD in 18 patients with LRE compared with 24 normal controls using SPM. Results for three different uncorrected thresholds, each then corrected for multiple non-independent comparisons at p<0.05, are superimposed: p<0.01 - yellow; p

(b) Regions showing a negative correlation of FMZVD and seizure frequency in the group of 18 patients at p<0.01 uncorrected are shown in green.

(c) Region showing a positive correlation between FMZVD and seizure frequency in a subgroup of 10 patients with unilateral frontal lobe seizures at p<0.001 uncorrected is shown in blue. 177 3.51 Ipsilateral putamen

• ♦

FMZVD 2.5 (a)

2 I I I "I I I I I r r I I I I I I I I I' I 0 20 40 60 80 Seizure frequency / month

5.5 Cingulate

5d ♦ 4.5 FMZVD (b) 4

3.5 1—I I I I I I I I I I—1—1—I—I—I—r 0 20 40 60 80 Seizure frequency / month

5- ♦ Ipsilateral premotor 4.5-

4- FMZVD • 3.5- (C)

3-

2.5- "■1 "V 1 ' 20 40 60 80 Seizure frequency / month

4.8 1 If 4.6 i 4.4 ' 4.2 ■ FMZVD 4 - (d) ♦ 3.8 ■ ♦ 3.6 : ♦ 3.4 ■ 0 20 40 60 Seizure frequency / month

Figure 10.2. Scattergraphs of seizure frequency versus FMZVD, corrected for global activity using ANCOVA, at the voxels with maximum Z-score in a parametric analysis of the patient data alone: (a), (b) and (c) include all 18 LRE patients: (d) includes only 10 right frontal focus patients. 178 CORTICAL SEIZURE FOCUS Hyperexcitable via potentially multiple mechanisms

SEIZURES

INCREASED GABAa-BZR

CORTEX BASAL GANGLIA Increased GABA^-BIR density. Increased GABA^-BZR leads Relatively Ineffective seizure to Increased local Inhibition suppression mechanism

DECREASED EXCITATORY OUTPUT FROM THALAMUS Relatively effective seizure suppression mechanism

SEIZURE SUPPRESSION

Figure 10.3. A model of the GABAergic system and its influence on seizure frequency in frontal lobe seizures in man. 179 Chapter 11.

Cerebral activation in malformations of cortical development.

n.1. INTRODUCTION

Malformations of cortical development (MCD) are increasingly recognised as an important aetiology of localisation-related epilepsy [500, SOI]. MRI detects most of these lesions [28] but both EEC [102, 501] and surgically resected specimens from such patients [100] frequently suggest the abnormality is more extensive than the MRI- defined lesion. Using PET and ’ ^C-flumazenil to label benzodiazepine receptors. Chapter 5 has shown that the majority of patients with MCD had regions of significantly abnormal receptor binding remote from the MRI lesion [475]. These lines of evidence suggest that lesions detected by MRI may be merely the "tip of the iceberg".

Impairments of cognitive performance are frequently observed in such individuals [71, 93, 94, 102], although only 9 out of a series of 100 patients with MCD [501] were found to be mentally retarded. This finding is perhaps surprising in view of the extensive nature of the lesions in many of these patients. One possible interpretation of this finding is that the developmentally and structurally abnormal regions of cerebral cortex may be able to participate in normal cognitive tasks to good effect; alternatively, the brain may be a typically organised such that remote and unaffected areas of the brain, which would not normally participate in the performance of a particular task, may have developed the ability to perform a task normally associated with the abnormal area of brain. These two possibilities are not mutually exclusive.

Activation studies using PET can identify areas of cerebral cortex that subserve a variety of tasks, from simple motor to complex psychological paradigms [461]. Comparison of the distribution of blood flow in scans at rest with the distribution of blood flow during performance of a particular task identifies brain regions showing increases or decreases in relative cerebral blood flow (rCBF) during performance of that task. In the past such investigations required the pooling of results from several subjects in order to detect significant change. Recent improvements in scanner technology have resulted in a five-fold improvement in sensitivity, through the implementation of 3D acquisition [425], such that it is now possible to detect with confidence local changes in perfusion of 3% between conditions in a single subject. Furthermore, recent advances in 180 the technique of statistical parametric mapping (SPM) [457-459] allow both the assessment of significant change within an individual and also a comparison of the pattern of activation seen in that individual with that seen in a normal control group.

The aim of the present study was to investigate whether regions of malformed cerebral cortex participate in normal cognitive functions. We also compared the overall pattern of activation in each subject with MCD with a normal control group for the task in question, in order to determine whether there was atypical organisation of cortical function in apparently normal cortex in these subjects, as a consequence of MCD. In order to adopt the strongest possible methodology, we chose activation paradigms which had already been demonstrated to produce robust activation in normal subjects and a widely established analysis technique, SPM, to identify activation in individual subjects and to compare each subject with a normal control group. 181

11.2. METHODS

11.2.1. Subjects

Patients were recruited from the Epilepsy Clinic of the National Hospital for Neurology and Neurosurgery, Queen Square, London, UK and from the National Society for Epilepsy, Chalfont Centre, Chalfont St. Peter, UK. All had undergone MR imaging including an SPGR T1-weighted volume acquisition, proton density and T2-weighted images (GE Signa 1.5T) which had identified MCD. The visually evident extent of the MR images was determined by an experienced consultant neuroradiologist. All patients suffered from partial seizures and were taking antiepileptic drugs. Clinical details of the patients are provided in table 11.1.

Normal controls were volunteers recruited from friends and colleagues of both the investigators and patients. All were neurologically normal, with no personal or family history of neurological illness or family history of MCD.

Ethical committee approval for this project was obtained from both the National Hospital for Neurology and Neurosurgery and the Hammersmith Hospital. Approval to administer the radioisotope was obtained from the Administration of Radioactive Substances Advisory Committee (ARSAC) UK.

1 1.2.2. PET scanning

PET scans were carried out at the Medical Research Council Cyclotron Unit, Hammersmith Hospital, UK, on a Siemens-CTI 953B (CTI Inc., Knoxville, Tenn., USA) in 3D mode with the collimating septa retracted. Radioactivity was administered as an Hg^^O bolus infused over 20 seconds followed by a 20 second saline flush. 12 PET emission scans were collected over 2 hours for each subject, with a 10 minute interval between the start of each scan. The effective dose equivalent of radioactivity per subject was S.OmSv. Integrated radioactivity counts were collected over a 90 second acquisition period, commencing at the rising phase of radioactivity counts in the head; these data were used as an index of rCBF [473]. A transmission scan was collected at the beginning of the scanning period to correct for attenuation of radioactivity by cerebral and extracerebral tissues. All active experimental conditions began 60 seconds before the start of the data acquisition period and continued through to the end of this time. 182

11.2.3. Experimental designs

11.2.3.1. Occipital cortex activation

Five patients with MCD involving the occipital cortex and 7 normal subjects were studied. All undertook a Rapid Visual Information Processing (RVIP) task [473]. This task, primarily a test of visual sustained attention, has been shown by this study to give rise to significant increases in rCBF in bilateral calcarine sulci, fusiform gyri, superior parietal cortex, inferior frontal gyri, supplementary motor areas and right superior frontal gyrus, when compared with a rest condition. Subjects studied were of either left or right handedness; handedness has not been shown to influence the pattern of activation.

Two experimental conditions were used: rest and RVIP. Each condition was repeated 6 times. During the rest condition the subjects remained still and at rest with eyes closed with low ambient noise. During the RVIP condition subjects were asked to look at the centre of a screen on which were presented digits one at a time in an apparently random order at a rate of 100 per minute. Subjects were required to detect any of four specified sequences of three digits (viz. 1-3-5, 2-4-6, 3-5-7, 4-6-8) and to register detection with a button press. Target sequences occurred 8 times per minute; correct responses during the 1.5 seconds after target presentation ("hits") and button presses occurring without the appropriate target stimulus ("false alarms") were recorded automatically. An index of performance was determined as described previously [473].

11.2.3.2. Dominant left frontal cortex activation

Five right handed patients with MCD involving the left frontal cortex and 7 right handed normal subjects were studied. All undertook a Motor Sequence Learning (MSL) task [502]. This task gives rise to significant increases in rCBF in left sensorimotor cortex, bilateral lateral premotor cortices, supplementary motor cortex, dorsal prefrontal cortex, bilateral cerebellar cortex and nuclei, and bilateral putamina.

Two conditions were used: rest and MSL. During the active condition subjects were asked to learn a sequence of 8 key presses with the fingers of their right hand on a four-key keypad. This was performed as follows: A computer-generated pacing tone occurring every two seconds was the cue to press a key. Initially, the subject would guess, and press any key. If the response was incorrect a low-pitched tone was produced, indicating that the subject should guess again at the next pacing tone. If the response was correct a high-pitched tone was produced to indicate the subject had guessed correctly and should 183

proceed to guess the next key press in the sequence at the next pacing tone. In this way the subject proceeded until the sequence of 8 key presses had been successfully guessed. Following this a series of rapid tones indicated the subject should return to the beginning of the process and attempt to recall the sequence. If the sequence was recalled correctly a new sequence was started for the subject to guess. If the subject made errors, an "incorrect" tone was produced in an identical manner to the first attempt at the sequence. The number of errors during sequence learning was recorded. For each of the six active condition scans a different sequence of 8 key presses was presented. The rest condition consisted of lying with eyes closed, fingers resting unmoving on the keys, listening to tones produced by the computer which simulated the tones produced during an active condition.

11.2.4. Data analysis

Data were analysed on a Sun SPARCclassic workstation (Sun Microsystems, Mountain View, CA) using Analyze version 7.0 (Mayo Foundation) [471], MATLAB (The MathWorks Inc., Natick, Mass.) and Statistical Parametric Mapping (Wellcome Dept, of Cognitive Neurology, Institute of Neurology, London) [457-459].

11.2.5. Statistical Parametric Mapping

Each of the twelve rCBF images for each subject were aligned together using an automated PET to PET coregistration. Following realignment all images were transformed into a standard anatomical space. This normalizing spatial transformation matched each scan to a template image that already conformed to the standard space. As a final pre-processing step these images were also smoothed using a 20x20x12mm FWHM Gaussian kernel.

Effects were estimated according to the general linear model at each and every voxel [457]. Global activity was taken to be a confounding covariate and this analysis can therefore be regarded as an ANCOVA [476]. To test hypotheses about regionally specific effects the estimates were compared using linear compounds or contrasts. The resulting set of voxel values for each contrast constituted a statistical parametric map of the t statistic SPMjt}. The SPM{t} were transformed to the unit normal distribution (SPM{Z[) and thresholded at a chosen threshold. To correct for the multiple non- independent contrasts involved in examining many voxels simultaneously, the resulting foci were characterized in terms of spatial extent ( k ) and peak height (^) [459]. This characterization is in terms of the probability that a region of the observed number of voxels (or bigger) could have occurred by chance (P(nmax < k )), or that the peak 184 height observed (or higher) could have occurred by chance (P(Zmax < ^i)) over the entire volume analyzed (i.e a corrected p-value).

The high-resolution SPGR volume acquisition (GE Signa 1.5T) MRI for each patient was coregistered with a PET image for that individual [458]. The PET image used was the mean image of all 12 realigned scans. The coregistered MRI was transformed to the standard space using the identical transformation parameters used for transformation of that individual's PET.

Contrasts were assessed between task and rest for each entire normal control group. The Z-score threshold used was 3 (p < 0.001); the regions reported remained significant after a subsequent correction for multiple non-independent corrprisons at p < 0.05.

For each individual patient, the analysis was carried out in two steps. The first step revealed the brain regions involved in task performance in each individual patient; the second step allowed each individual patient's activation pattern to be compared with the normal control group, which may reveal reorganisation of cortical functions. Finally, the effect of difference in task performance between the individual patient and the normal group was examined, to address the possibility that differences in activation pattern were due to differential task performance.

11.2.5.1. Significant changes in rCBF in the whole brain of each patient

Contrasts were assessed between the active condition and rest for each individual patient in order to demonstrate all of the regions associated with significantly increased rCBF and significantly decreased rCBF during the task versus rest. To identify the overall pattern of activation, a Z-score threshold of 3 (p < 0.001) was used: the regions reported remained significant at p < 0.05 after a subsequent correction for multiple non-independent comparisons. If no significant regions were seen at this threshold, the visually-determined region of MCD was examined alone at a threshold of p < 0.01 uncorrected.

11.2.5.2. Comparison of individual activation pattern with the control group

To demonstrate the significant differences in activation pattern between the findings for each patient and the appropriate normal control group an interaction between group (ie. individual patient or normal controls) and condition (task or rest) was assessed. Two contrasts can be examined, one revealing brain regions showing a relatively greater tendency to increase rCBF with the task in the patient compared to the normal controls. 185 the other revealing brain regions showing a relatively greater tendency to decrease rCBF with the task compared to the normal controls. It was anticipated that the normal subjects would exhibit a certain amount of variability in terms of brain regions activated by the task and the relative "strength" of activation observed in those regions. The aim was to detect reorganisation of cerebral function in the patients which was outwith the variability seen in the normal control group. In order to determine normal variability, each normal subject was compared with the remainder of the control goup; this interaction was also thresholded at p < 0.001 uncorrected, corrected for multiple non-independent comparisons at p < 0.05. The number of brain regions detected in this interaction, the peak Z-scores and the extent of each region were noted for each normal individual. Similarly, each patient was compared with the normal control group and the number of brain regions detected in this interaction, the peak Z-scores and region extent were noted. Post-hoc, a more stringent Z-score threshold and a threshold for region extent were determined, such that no normal individual was found to have any region above this new threshold in comparison with the remainder of the normal group. A patient was regarded as showing atypical organisation of function if any brain regions were detected above these more stringent Z-score and region extent thresholds.

Finally, to test the hypothesis that differences in activation pattern between the individual patients and the control group were the result of differential task performance between the patients and normal subjects, an interaction between patients and normal subjects was examined using task performance as a confounding covariate, in order to identify those areas of brain with a difference in activation in the patients and normal subjects that was not due to differential task performance. 186

11.3. RESULTS

Clinical features of the patients are described in table 11.1.

11.3.1. Occipital cortex activation

11.3.1.1. Significant changes in rCBF in the whole brain of each patient

All 5 patients showed regions of significantly increased blood flow with the task relative to rest at the statistical thresholds selected (figure 11.1). Four of the 5 patients showed significant change in rCBF in the region of MCD: Subject DL, with frontoparietal band heterotopia and occipital subependymal heterotopia showed widespread activation of cortex overlying heterotopia and activation of a region of subependymal heterotopia in the left occiptal lobe; subject EM was found to increase rCBF in a heterotopic nodule; subject SN showed no activation of a heterotopic nodule but a significant decrease in rCBF in overlying cortex; subject TP showed activation of part of the area of right temporo-parieto-occipital dysgenetic cortex. Subject MW showed failure of activation of the abnormal cortex.

11.3.1.2. Comparison of individual activation patterns with the control group

Comparing each normal subject with the remainder of the control group at Z > 3 (p < 0.001, corrected p < 0.05), all the normal subjects showed at least one significant region. No normal subject had any region significantly different from the rest of the control group at a threshold Z > 6 with an extent of greater than 72 voxels; this was chosen as the threshold for examination of the patients. Three of the five patients had brain regions significant at this threshold (table 11.2 and figure 11.1). DL had a left frontal region showing a relatively greater increase in rCBF with the task than in the controls (figure 11.3). This subject also had two regions showing relatively greater decrease in rCBF: one in the right superior temporal region, the other in the left occipital cortex. EM had relatively greater increase in rCBF in the heterotopic nodule in the left occipital lobe and in the overlying cortex, and relatively greater decrease in rCBF in the right primary motor cortex, right superior temporal cortex and in the mid- cingulate (figure 11.4). TP had significantly greater increase of rCBF in the left posterior-inferior temporal area (figure 11.5).

There was no significant difference between the task performance of the patient group and the normal control group. Furthermore, using task performance as a covariate, no differences between the control group and any patient were found to be due to differential 187 task performance.

The results are summarised in table 11.3 and illustrated in figures 11.1, 11.3, 11.4 and 11.5.

11.3.2. Left frontal cortex activation

11.3.2.1. Significant changes in rCBF in the whole brain of each patient

All 5 patients showed regions of significantly increased blood flow with the task relative to rest at p < 0.001, corrected p < 0.05 (figure 11.2). Four patients showed significant activation of the region affected by MCD. In subject AF activation was seen in cortex overlying subependymal nodular heterotopia. In subject EG activation was seen in cortex overlying band heterotopia. In subject RH activation was seen in cortex lining the schizencephalic clefts and in surrounding regions of focal cortical dysgenesis. Subject KS showed vigorous activation of the nodular heterotopia on the left. In addition, AF had a region of significantly increased rCBF in a heterotopic nodule in the right frontal lobe at p < 0.01 and EG had extensive task-related decrease in rCBF in the heterotopic bands at p < 0.01.

11.3.2.2. Comparison of individual activation pattern with the control group

Five of the normal subjects, when compared with the remainder of the control group at Z > 3 (p < 0.001, corrected p < 0.05), showed at least one significantly different region. No normal subject had any region significant at a threshold of Z > 4.9 and extent greater than 106 voxels. Two of the five patients had regions that were significantly different from the control group at this threshold (table 11.2 and figure 11.2). KS showed significantly greater decreases in rCBF in occipital cortex bilaterally (figure 11.6). KT had a significantly greater increase in rCBF in the inferior (hand) region of the right precentral gyrus (figure 11.7).

There was no significant difference between the task performance of the patient g^up and the normal control group. Furthermore, using task performance as a covariate, no differences between the control group and any patient were found to be due to differential task performance.

The results are summarised in table 11.3 and illustrated in figures 11.2. 11.6 and 188 n .7 . 189

1 1.4. DISCUSSION

We have shown that eight out of ten patients with MCD showed significant changes in rCBF in affected brain regions with appropriate tasks. These findings included changes in rCBF seen in regions of focal cortical dysgenesis, in the cortex lining schizencephalic clefts, in subependymal grey matter heterotopia and in the cortex overlying band and subependymal heterotopia. Two subjects showed no significant change in rCBF in an area affected by MCD; in both, focal cortical dysgenesis failed to activate. In addition, there was a significant alteration in the overall activation pattern in five patients, compared with a normal control group, using a very high statistical threshold that was shown to exclude outliers from the normal control group. In three of these patients this alteration in overall pattern of activation included regions of cortex which appeared entirely unaffected by MCD on MRI. Two further subjects with generalised MCD (one with subependymal heterotopia and one with anterior band heterotopia and posterior subependymal heterotopia) showed extensive reorganisation of the activation pattern in cortex overlying the heterotopia that appeared structurally normal on MRI. The rigorous methodological approach has identified regions of brain in these patients which were significantly involved in the task and were remote from brain regions showing similar changes in rCBF in the normal control group.

Generally, the brain regions recruited during the performance of the visual and motor tasks by the patients were similar to those recruited by the normal group. Furthermore, measures of task performance indicated that the patients were performing the tasks in a comparable manner to the controls and differential task performance did not explain the differences seen between the patients and normal controls. Additionally, the alterations in the activation patterns seen were related anatomically to the site of dysgenesis in most of the patients: for example, in subject TP the dysgenetic occipital cortex did not activate normally and the major activations in the posterior hemisphere were "displaced" anteriorly into the boundary area between structurally normal and affected cortex. KT had dysgenesis encroaching into the inferior part of the left primary motor area which would normally be strongly activated by the motor learning task but, in contrast, activation was seen in the inferior right primary motor cortex. AF, EM, DL and KS showed activation of the subependymal heterotopia underlying the motor or visual cortical areas. These findings argue strongly that the alterations in activation pattern seen in our patients were directly related to the presence of MCD.

It is well-recognised that the region of brain involved in the performance of the same task as determined with PET in different individuals may vary in terms of position and magnitude of blood flow change [503]. Thus, significant differences might be 190 observed between any one normal individual and the mean and variance of a separate normal group. The study of Watson et al. showed that although the position of the maximally activated voxel in the vicinity of occipital area V5 (the visual motion area) varied by up to 27mm, this peak voxel was always found within a particular sulcus of the occipital region. The detection of any difference between a patient and a normal control group in our study may be no more than an illustration of this normal idiosyncrasy. Indeed, it was shown that the majority of our normal control group did have significant differences from the remainder of the control group. However, the method enabled the range of this normal variability to be determined and to show that the differences between five of the patients and the normal subjects were of a significantly greater order of magnitude. This distinction between normal variability and unequivocal abnormality is crucial in any attempt to reveal cortical reorganisation. These methods, available within the SPM software package, could be easily applied in other areas of research in which cortical reorganisation or plasticity is hypothesised.

The activation paradigms used were known to activate a distributed network and involve more than one aspect of cognition. An alternative might have been to use apparently simple tasks intended only to activate the dysgenetic area, employing only one simple cognitive or motor function. However, the aim was simply to determine whether any activation could be observed in dysgenetic regions and whether reorganisation of cortical functions would be seen in patients with MCD. The approach was to use statistically robust, established activation paradigms, rather than to design new, potentially simpler paradigms because the nature of the activation pattern was not relevant to the question being addressed. In addition, SPM was chosen rather than a region of interest based approach because it was not possible to predict a priori which regions of brain would be involved in the task performance in the subjects with MCD. A volumetric transformation to a standard template was used as in the previous study of MCD patients using ^^C- flumazenil and SPM described in Chapter 5 [475, 504]. Although the MCD patients had anatomical abnormalities, excellent volumetric normalisation was seen.

The ability of areas of MCD to participate in cognitive processing has not been demonstrated previously. Other studies, however, indicated that this might be the case. Evoked potentials may, in some cases, be recorded overlying regions of MCD [422]. Electrical stimulation of brain regions affected by schizencephaly has also been shown to give rise to motor responses [124]. Using ^®F-deoxyglucose PET, heterotopic nodules and bands were shown to have a resting metabolic activity that was similar to normal cortex [380, 384]. Furthermore, in a single case studied twice with PET, once at rest and once with a task, increased task-related metabolic activity was observed in the nodule [387]. The finding in this chapter of activation in heterotopic nodules used a 191 more robust method and amply confirms this prior suggestion.

Three patients showed a failure of activation of the region affected by MCD: 2 showed no significant activation of the region of focal cortical dysgenesis; 1 showed no activation of a heterotopic nodule. However, these patients have been tested with only one task; it is possible that other paradigms might activate these regions. The limitations imposed by radiation exposure with PET prevent the evaluation of each patient with more than one paradigm; functional MRI has a potential advantage in this respect. Such studies may be relevant in the presurgical evaluation of patients with MCD if the major functions relevant to the area in question could be tested.

The detection of a different pattern of rCBF changes in the patients with MCD compared to the normal controls raises some interesting neurobiological questions. Firstly, in the presence of cortical malformation, does the observation of a change in blood flow imply a change in neuronal activity, or could the blood flow responses be occurring independently of neuronal activity? The blood flow changes seen were of a similar relative order of magnitude in the patients as in the controls (ie. the Z-scores peaks were of similar height) and the patients and normal controls performed the tasks at a similar level of competence. It would be most unlikely that the patients performed the same task to the same level as the normal controls, producing regions of significant blood flow change of the same order of magnitude as in the normals, but these blood flow changes were unrelated to the neuronal task. Secondly, what are the effects of a chronic seizure disorder, anticonvulsant drugs and interictal spikes on the patterns of blood flow responses to a task? There are n(?p%y{^data specifically addressing these issues; the current study indicates that in some cases there was no significant difference between the patient and the normal control group, implying that these factors do not necessarily result in rCBF differences between a patients with MCD and normal controls.

In addition to differences in local gross anatomy or histological appearance, it is possible that there may be changes in connectivity within the brains of subjects with MCD. As a follow-up to the current study we are examining the “effective” connectivity [505] of the brains of the subjects with MCD in comparison with the normal controls by using an “eigenimage” analysis, looking for a generalised eigenimage solution that shows the pattern of effective connectivity maximally expressed in the normals but minimally expressed in the patient with MCD and vice versa.

In conclusion, it has been demonstrated that regions of MCD may participate in normal cognitive activity and that the presence of even localised MCD can be associated with an abnormal organisation of cortical function. This atypical organisation of cognitive 192 function may reflect abnormal connectivity between areas of MCD participating in the task, may be an adaptive response of the brain in order to compensate for the MCD, or may be a reflection of the more widespread nature of MCD than is evident with MRI. In some patients with MCD a lack of activation is seen in the affected area, suggesting that this region may be relatively non-functional. The use of functional imaging studies with PET and functional MRI to reveal such findings may in the future become an important preoperative evaluation in such patients in whom surgery is being considered. 193

Patient Age Duration of Hand- MRI findings Seizures Seizure Time since last EEG findings AED Gender epilepsy edness frequency seizure (mg/day) (years)

Occipital malformation of cortical development

DL Male 24 5 L Bilateral frontoparietal band SPS 3/month 2 weeks Some slow; no CBZ 800 heterotopia; bilateral CPS focal or VPA 500 occipital subependymal epileptiform heterotopia features

EM Female 25 10 R Single R parieto-occipital CPS 6/year 2 months No definite CBZ 1200 heterotopic nodule abnormality

SN Female 26 7 R Single R parieto-occipital SPS Rare 6 months No definite CBZ 700 heterotopic nodule abnormality

TP Female 29 27 R R temporo-parieto- CPS CPS 1/day 8 days R posterior VPA 2500 occipital and L occipital 2° gen. SPS 1/week hemisphere CBZ 600 cortical dysgenesis and interictal LTG 100 heterotopias epileptiform activity

MW Male 18 13 R Right temporo-ocdpital SPS CPS 3/year 2 weeks R posterior CBZ 800 dysgenetic cortex CPS 2* gen, 2/year hemisphere 2" gen. interictal epileptiform activity

Left frontal malformation of cortical development

AF Female 47 21 R Bilateral subependymal CPS CPS 3/year 2 months Bilateral temporal PHT 75 heterotopia 2° gen. 2* gen. 1/month interictal tonic epileptiform activity

EG Female 24 11 R Bilateral band heterotopia SPS CPS 2/week 3 days No definite GVG 3000 CPS abnormality GBP 1800

RH Female 34 33 R Bilateral cortical L arm focal 1 /week 4 days No definite CBZ 800 dysgenesis and motor abnormality schizencephaly L and R 2“ gen. frontal

KS Female 30 6 R Bilateral subependymal R arm focal 3/week 4 days No definite VPA 600 heterotopia sensory abnormality PHT 300 2" gen.

KT Female 19 12 R Extensive L hemisphere CPS 2/month 7 days Widespread L CBZ 800 cortical dysgenesis slow GVG 3000

Table 77.7. Clinical details of patients with MCD. L=left, R=right, SPS=simple partial seizures, CPS=complex partial seizures, 2°gen.= secondary generalised seizures, CBZ=carbamazepine, VPA=sodium valproate, LTG= lamotrigine, PHT-phenytoin, GVG^vigabatrin, GBP=gabapentin, AED=antiepileptic drugs. 194

Subject Relative rCBF Region size Peak Z-score Location No. significant change compared (2x2x4mm regions with controls voxels)

Occipital cortex activation

DL 3 Increase 178 7.88 L middle frontal gyrus

Decrease 105 6.68 R superior temporal gyrus

Decrease 100 7.97 L occipital cortex

EM 4 Increase 615 9.3 L occipital cortex overlying nodule

Decrease 219 7.61 R precentral gyrus

Decrease 151 7.54 R superior temporal gyrus

Decrease 168 6.92 Mid-cingulate gyrus

TP 1 Increase 169 7.5 L posteroinferior temporal

Left frontal cortex activation

KS 2 Decrease 584 6.58 R occipital

Decrease 515 6.19 L occipital

KT 1 Increase 244 6.2 R precentral gyrus

Table 11.2. Atypical cortical organisation in the patients with MCD. The brain regions found significant in the patients, using an interaction effect to compare each patient with the appropriate control group, as described in the text. The number of such regions, relative change in rCBF, extent in number of voxels, peak Z-score height and anatomical position are described. 195

Patient rCBF change in rCBF change in rCBF change in Atypical organisation region of MCD region of MCD cortex overlying in apparently p<0.01 p<0.001 heterotopia p<0.001 normal cortex

Occipital cortex activation

DL No No Yes Yes

EM Yes No No Yes

SN No No Yes Yes

TP Yes Yes Overlying cortex Yes affected by MCD

MW No No Patient has no No heterotopia

Left frontal cortex activation

AF Yes No Yes No

EG Yes No Yes No

RH Yes Yes Patient has no No heterotopia

KS Yes Yes Yes Yes

KT No No Patient has no Yes heterotopia

Table 7 7.3. Summary of the cerebral activation findings in 10 patients with MCD. 196

Subject Anatomical Increases in Decreases in Greater tendency to Greater tendency to abnormality rCBF rCBF increase rCBF decrease rCBF on MRI Z > 3 Z > 3 compared with normals compared with normals Z > 6 Z > 6

DL

#' i EM M

No significant No significant SN voxels voxels

No significant TP voxels

goÿttol corord

No significant No significant MW voxels voxels

Normal control group

Figure 11.1. Summary of RVIP results. The distribution of MCD and the brain regions showing significant Increases and decreases In rCBF with the RVIP task and the brain regions found significant under the group by condition Interaction decrlbed In the text. The results are displayed as maximum Intensity projections Ie. as though looking at the relevant aresa In a "glass brain"; upper left: lateral projection, upper right: coronal projection, lower left: axial projection. In all projections right of the brain Is right of the Image. 197

Subject Anatomical Increases in Decreases in Greater tendency to Greater tendency to abnormality rCBF rCBF increase rCBF decrease rCBF on MRI Z > 3 Z > 3 compared with normalscompared with normals Z > 4.9 Z> 4.9

No significant No significant No significant AF voxels voxels voxels

EG No significant No significant No significant voxels voxels voxels

No significant No significant RH voxels voxels

No significant KS voxels

No significant KT voxels

Normal control group

Figure 11.2. Summary of MSL results. The distribution of MCD and the brain regions showing significant increases and decreases in rCBF with the MSL task and the brain regions found significant under the group by condition interaction decribed in the text. The results are displayed as maximum intensity projections as in Figure 1. 198

i

« M

ncrease p<0.001 ■ Increase p<0.0001 ■ Decrease p<0.001 ■ Decrease p<0.0001

Relatively greater increase Z>6 Relatively greater decrease Z>6

Figure 11.3. Subject D L Bilateral subependymal heterotopia Is present In the posterior hemispheres, band heterotopia Is present anteriorly. Selected transverse sections from the 3D dataset are Illustrated, superimposed on the coregistered and volumetrlcally normalised T1- welghted MRI. Rows (a) and (c) show the regions with significantly Increased or decreased rCBF with the task versus rest at the thresholds shown. Rows (b) and (d), which are the same transverse sections as In the rows Immediately above, show the result of the comparison of the Individual patient’s activation pattern with that of the control group, at the thresholds shown. Significant Increase In rCBF Is seen In the left frontal lobe overlying band heterotopia and In a region of subependymal heterotopia In the left temporal lobe; significant decrease In rCBF Is seen In the right temporal lobe. Comparison with the normal control group shows relatively greater Increase In rCBF In the left frontal cortex and relatively greater decrease In rCBF In the left temporal lobe. 199

i » »

Increase p<0.001 ■ Increase p<0.0001 ■ Decrease p<0.001 ■ Decrease p<0.0001

Relatively greater increase Z> Relatively greater decrease Z>6

Figure 11 A. Subject EM (a and b) and subject SN (c and d). Subject EM has bilateral subependymal nodules in the occipital lobes. Subject SN has a single large heterotopic nodule in the right parietal lobe. Selected transverse sections from the 3D datasets are illustrated, superimposed on the coregistered and volumetrically normalised T1-weighted MRI. Subject EM: Row (a) shows the regions with significant increased or decreased rCBF with the task versus rest at the thresholds shown; row (b), which is the same transverse sections as in the row immediately above, shows the result of the comparison of the individual patient’s activation pattern with that of the control group, at the thresholds shown. Significant increase in rCBF is seen in the left occipital cortex overlying the heterotopia and in the heterotopia itself. Comparison with the normal control group shows relatively greater increase in rCBF in the left occipital cortex overlying the heterotopia and in the heterotopia itself and relatively greater decrease in rCBF in the right superior temporal and precentral gyri. Subject SN: Rows (c) and (d) show the regions with significant increased or decreased rCBF with the task versus rest at the thresholds shown. This pattern was indistinguishable from the normal controls. 200

K

Increase p<0.GDI ■ Increase p<0.0001 ■ Decrease p<0.001 ■ Decrease p<0.0001

Relatively greater increase Z>6 Relatively greater decrease Z>6

Figure 11.5. Subject TP (a and b) and subject MW (c and d). Subject TP has subependymal heterotopic nodules in the temproal and ocipital lobes with abnormal cortex at the occipital pole bilaterally. Subject MW has a single region of abnormal cortex in the right temporal lobe. Selected transverse sections from the 3D datasets are illustrated, superimposed on the coregistered and volumetrically normalised T1-w eighted MRI. Subject TP: Row (a) shows the regions with significant increased or decreased rCBF with the task versus rest at the thresholds shown; row (b), which is the same transverse sections as in the row immediately above, shows the result of the comparison of the individual patient's activation pattern with that of the control group, at the thresholds shown. Significant increase in rCBF is seen in bilateral posterior temporal/occipital cortex. Comparison with the normal control group shows relatively greater increase in rCBF in the left posterioinferior temporal cortex. Subject MW: Rows (c) and (d) show the regions with significant increased or decreased rCBF with the task versus rest at the thresholds shown. No activation is shown in the abnormal region or in the adjacent occifital cortex; however, this pattern was indistinguishable from the normal controls. 201

«

ncrease p<0.001 ■ Increase p<0.0001 ■ Decrease p<0.001 ■ Decrease p<0.0001

Relatively greater increase Z>4.9 Relatively greater decrease Z>4.9

Figure 11.6. Subject RH (a and b) and subject KS (c and d). Subject RH has two schizencephalic clefts, one at the left frontal pole and the other in the right centrl region. Subject KS has bilateral generalised subependymal nodular heterotopia. Selected transverse sections from the 3D datasets are illustrated, superimposed on the coregistered and volumetrically normalised T1-weighted MRI. Subject RH: Rows (a) and (b) show the regions with significant increased or decreased rCBF with the task versus rest at the thresholds shown. Activation was seen in the cortex associated with both schizencephalic clefts; however, this pattern was indistinguishable from the normal controls. Subject KS: Row (c) shows the regions with significant increased or decreased rCBF with the task versus rest at the thresholds shown; row (d), which is the same transverse sections as in the row immediately above, shows the result of the comparison of the individual patient’s activation pattern with that of the control group, at the thresholds shown. Significant increase in rCBF is seen in the heterotopia on the left. Comparison with the normal control group shows relatively greater decrease in rCBF in bilateral occipital cortex. 202

ncrease p<0.001 ■ Increase p<0.0001 ■ Decrease p<0.001 ■ Decrease p<0.0001

Relatively greater increase Z>4.9 Relatively greater decrease Z>4.9

Figure 11.7. Subject KT. Extensive abnormal cortex is present in much of the left hemisphere. Selected transverse sections from the 3D dataset are illustrated, superimposed on the coregistered and volumetrically normalised T1 -weighted MRI. Row (a) shows the regions with significant increased or decreased rCBF with the task versus rest at the thresholds shown. Row (b), which is the same transverse sections as in the row immediately above, shows the result of the comparison of the individual patient’s activation pattern with that of the control group, at the thresholds shown. Significant increase in rCBF is seen only in the right precentral area. Comparison with the normal control group shows relatively greater increase in rCBF in this same area. 203

Chapter 1 2.

Discussion.

12.1. SUMMARY OF THE MAIN FINDINGS

The main findings of this thesis were:

- Malformations of cortical development (MOD) are associated with abnormalities of regional ^’C-flumazenil volume of distribution (FMZVD). In particular, multiple regions of both increased and decreased FMZVD may be seen in patients with MCD, both within lesions visible on MRI and extending beyond. The finding of a region of increased FMZVD may be indicative of the presence of MCD.

- Acquired cerebral lesions, such as head injury and stroke, are associated with regions of reduced FMZVD, usually confined to the MRI lesion. FMZVD abnormalities are not seen in remote brain areas.

- Dysembryoplastic neuroepithelial tumours (DNET) are associated with a single well- circumscribed region of reduced FMZVD confined to the area of the MRI lesion. Remote changes in FMZVD may exceptionally be seen, perhaps associated with occult MCD.

- In patients with neocortical localisation-related epilepsy and normal MRI, abnormalities of regional FMZVD are seen in the majority of cases. These abnormalities may be single or multiple and consist of either increases or decreases in FMZVD or both. Concordance of these FMZVD abnormalities with EEC data is strongest in patients with unilateral frontal lobe epilepsy. The finding of decreased FMZVD in patients with normal MRI may be evidence of prior cerebral injury; the finding of increased FMZVD suggests the presence of occult MCD.

- MRI data may be analysed at the voxel level with statistical parametric mapping (SPM) to identify regions of anatomical abnormality, in a manner analagous to the analysis of PET data. Findings so obtained are in accord with those using regional volumetric techniques in patients with MCD. MRI data may also be analysed in conjunction with PET data at a voxel level with SPM to allow structural abnormalities to 204 be taken into account when analysing function. Using such an approach, the abnormalities of FMZVD in patients with MCD were shown to be over and above any abnormalities of anatomy. In some cases, simultaneous analysis of structure and function allowed abnormalities of FMZVD to be detected which had not been detected when functional data were examined alone.

- In patients with localisation-related epilepsy and normal MRI, a cortical-subcortical motor circuit was identified in which abnormalities of FMZVD were observed. In particular, increased FMZVD was seen in the ipsilateral striatum and pallidum, anterior cingulate and ipsilateral premotor areas, with contralteral areas also but less significantly affected. In the basal ganglia, the greater the increase in FMZVD the fewer the seizures experienced by the patient; in the ipsilateral frontal cortex of patients with unilateral frontal lobe seizures, greater increases in FMZVD reflected greater seizure frequency.

- Regions of MCD may participate in normal cortical functions as indicated by significant changes in rCBF in these regions during behavioural tasks. Reorganisation of the cortical pattems of activation can be seen in some of these patients.

12.2. NEUROBIOLOGICAL IMPLICATIONS

The main theme of the investigations detailed in this thesis is to determine the functional integrity and consequences of malformations of cortical development. The main theme of the findings is the identification of widespread functional change as a consequence of apparently circumscribed and localised malformations. It is plausible to suggest that a developmental lesion would have an extensive impact: not only might a localised region contain neurons with abnormal orientation, structure and local intracortical connections, but distant regions connected directly or even indirectly through polysynaptic pathways might be functionally abnormal as a consequence of misdirected axons, abnormal dendrites or abnormal synaptogenesis, for which there is much suggestive evidence [40, 51-54, 70]. In the patients with MCD, abnormalities of local benzodiazepine receptor binding were seen; in many patients there were also remote regions of abnormal cBZR. This has been interpreted as evidence of the presence of distant, in some cases occult, malformation, but these regions could have abnormal cBZR as a result of functional changes in otherwise normal areas of cortex, as a consequence of remote malformation. The rCBF studies in these patients showed, in some cases, a widespread abnormality of the activation pattern, often involving normal-appearing cortex. This again could reflect functional changes in otherwise normal areas of cortex, as a consequence of remote malformation. The widespread nature of the functional 205 consequences of MCD is in accord with the recognised poor response to surgery in these patients and their widespread EEC abnormalities [94, 100]. Furthermore, the possibility of reorganisation of cortical functions has been suggested by other lines of evidence [92, 421]. Activation of a region of brain affected by MCD was observed in many patients in the rCBF study described in this thesis. Although the involvemnt of a “lesion” in normal brain function is perhaps surprising, electrophysiological studies also show normal responses from regions affected by MCD [46 , 124, 422].

In contrast, acquired lesions (eg. stroke, trauma) were associated with single regions of reduced cBZR confined to the locality of the MRI lesion. As they occur after the period of synaptogenesis in embryonic life, acquired lesions might not be expected to result in extensive neuroreceptor changes in remote areas. The local reduction in cBZR may be due to a loss of all cell types, or a selective loss of GABAergic synapses or a specific cell type. Surgical or autopsy histology are required to clarify this.

The findings in DNET were similar to those in acquired lesions - single, well- circumscribed regions of reduced cBZR. These lesions are possibly part of the spectrum of MCD - and the finding in one case of a remote region of increased cBZR might tend to confirm this. However, the finding of isolated lesions in the majority of cases suggests that these lesions behave as acquired rather than congenital abnormalities and is in accord with the good response to surgery in such patients.

The aetiology of epilepsy in patients with focal seizure disorders but normal MRI is frequently obscure. Having identified typical cBZR binding abnormalities in the patients with MCD, acquired lesions and DNET, in combination with the known cBZR findings in hippocampal sclerosis from other studies, it is possible to attempt to interpret the cBZR changes seen in the patients with normal MRI. It is suggested that such patients with single areas of reduced cBZR are most likely to have acquired brain lesions while patients with regions of increased cBZR may have occult MCD.

A study of a group of patients with normal gross anatomy and seizure foci all in neocortex provided the opportunity to look for abnormalities of the GABA^-cBZR system in the basal ganglia. The identification of a cortical-subcortical motor system showing marked cBZR binding abnormalites suggests a model of the control of cortical excitability, and hence seizure frequency, via an increase in GABA^-cBZR in the striatum. Such a model appears plausible given current knowledge of basal ganglia- motor cortical interactions and the influence of the basal ganglia in animal models of epilepsy [180, 236-241]. This model is testable and implies that control of seizures might be improved by increasing inhibitory basal ganglia output; this would be a novel 206

approach,

12.3. FUTURE STUDIES

Most hypotheses concerning the mechanism of seizures postulate a relative imbalance between inhibitory and excitatory neurotransmitter systems. It would be of great interest to undertake a parallel set of studies examining glutamatergic and GABA receptor binding. Unfortunately, PET tracers for excitatory amino acid receptors do not yet exist.

Of those PET tracers in current use, a number might provide interesting and complimentary data. In particular, diprenorphine (a non-selective opioid receptor ligand) is an obvious candidate. Opioids are co-secreted at some GABA^ synapses, hence diprenorphine binding may be a further way of investigating the GABAergic system indirectly. Furthermore, abnormalities of carfentanil binding to ^-receptors have been noted in TLE; investigation of this system in MCD might be of interest.

The finding of abnormalities of cBZR in the basal ganglia in partial seizure disorders could be developed via studies of other neurotransmitter systems with PET, particularly the opioid system and the dopaminergic system. Of great interest would be the performance of multiple tracer experiments in the same individual, allowing a cross- correldbn between changes in different systems to be made and also to allow the evaluation of the impact of drugs and changes in seizure frequency. The evolution of a newer generation of PET scanners with increased sensitivity would potentially allow an increased number of studies to be undertaken in one individual by using lower radioactivity doses.

The hypothesis explaining the changes in cBZR in the motor system in localisation- related epilepsy is testable. It would be possible to examine changes in receptor density in these regions in animal models; in addition it would be possible to undertake lesioning or stimulation experiments in such animals specifically to test hypotheses regarding the influence of specific struc tres on seizures. In man, fMRI could be employed to look for signal changes occurring with interictal spikes, again addressing issues surrounding cortical-subcortical interactions.

The activation studies in MCD reveal changes in overall activation pattern; these studies do not necessarily imply changes in connectivity. Newer SPM techniques in development, specifically eigenimaging, allow the elucidation of effectively connected networks. Such analyses could be applied to data already obtained. 207

It would be valuable to have confirmation of the nature of cBZR binding changes observed with PET via in vitro studies of surgical resection specimens. Such specimens are not often obtained due to the infrequent nature of such relatively unsuccessful surgery. The planning of such surgery might be influenced by fMRI studies of a range of cognitive and motor paradigms aiming to characterise the functional capacity of regions of MCD. 208 References

^. Li LM, Fish DR, Sisodiya SM, et al. High resolution magnetic resonance imaging in adults with partial or secondary generalised epilepsy attending a tertiary referral unit. J Neurol Neurosurg Psychiatry 1995;59:384-7.

2. Brodtkorb E, Nilsen G, Smevikm 0, Rinck PA. Epilepsy and anomalies of neuronal migration: MRI and clinical aspects. Acta Neurol Scand 1992;86:24-32.

3. Ohtahara S, Ohtsuka Y, Yamatogi Y, et al. Prenatal etiologies of West syndrome. Epilepsia 1993;34:716-22.

4. Jellinger K. Neuropathological aspects of infantile spasms. Brain Dev 1987;9:349-57.

5. Raymond AA, Fish DR, Boyd SG, et al. Cortical dysgenesis: serial EEG findings in children and adults. Electroencephalogr Clin Neurophysiol 1995;94:389-97.

6. Raymond AA, Fish DR, Stevens JM, et al. Association of hippocampal sclerosis with cortical dysgenesis in patients with epilepsy. Neurology 1994;44:1841-5.

7. Sidman RL, Rakic P. Neuronal migration, with special reference to developing human brain: a review. Brain Res 1973;61:1-35.

8. Muller F, O'Rahilly R. The first appearance of the major divisions of the human brain at stage 9. Anat Embryol 1983; 168:419-32.

9. Caviness V. Normal development of the cerebral cortex. In: Evrard D, Minkowski A, eds. Developmental Neurobiology. New York: Raven, 1989:1-10.

10. Okada N, Oppenheim RW. Cell death of motoneurons in the chick embryo spinal cord. J Neurosci 1984;4:1639-52.

11. Chenn A, McConnell SK. Cleavage orientation and asymmetric inheritance of Notch 1 immunoreactivity in mammalian neurogenesis. Cell 1995;82:631-41.

12. Zhong W, Feder J, Jiang M, Jan L, Jan Y. Asymmetric localisation of a mammalian 209

Numb homolog during mouse cortical neurogenesis. Neuron 1996;17:43-53.

13. Austin C, Cepko C. Cellular migration patterns in the mouse cerebral cortex. Development 1990;110:713-32.

14. Angevine JB, Sidman RL Autoradiographic study of cell migration during histogenesis of cerebral cortex in the mouse. Nature 1961;192:766-8.

15. Walsh C, Cepko CL. Clonal dispersion in proliferative layers of developing cerebral cortex. Nature 1993;362:623-5.

16. O'Rourke NA, Sullivan DP, Kaznowski CE, Jacobs AA, McConnell SK. Tangential migration of neurons in the developing cerebral cortex. Development 1995;121:2165-76.

17. Caviness VS. Patterns of cell and fiber distribution in the neocortex of the reeler mutant mouse. J Comp Neurol 1976;170:435-48.

18. Rakic P, Sidman RL. Weaver mutant mouse cerebellum: defective neuronal migration secondary to abnormality of Bergmann glia. Proc Natl Acad Sci 1973;70:240-4.

19. Hatten ME. Riding the glial monorail: a common mechanism for glial-guided neuronal migration in different regions of the developing mammalian brain. Trends Neurosci 1990;13:179-84.

20. McConnell SK. Fates of visual cortical neurons in the ferret after isochronic and heterochronic transplantation. J Neurosci 1988;8:945-74.

21. Hatten ME. Molecular mechanisms of glial-guided neuronal migration. Discuss Neurosci 1991;7:74-81.

22. Anton ES, Cameron RS, Rakic P. Role of neuron-glial junctional domain proteins in the maintenance and termination of neuronal migration across the embryonic cerebral wall. J Neurosci 1996;16:2283-93.

23. Choi BH. Role of the basement membrane in neurogenesis and repair of injury in the central nervous system. Microsc Res Tech 1994;28:193-203. 210

24. del Rio JA, Martinez A, Fonseca M, Auladell C, Soriano E. Glutamate-like immunoreactivity and fate of Cajal-Retzius cells in the murine cortex as identified with calretinin antibody. Cereb Cortex 1995;5:13-21.

25. D'Arcangelo G, Miao GG, Chen SC, et al. A protein related to extracellular matrix proteins deleted in the mouse mutant reeler. Nature 1995;374:719-23.

26. Ogawa M, Miyata T, Nakajima K, et al. The reeler gene-associated antigen on Cajal-Retzius neurons is a crucial molecule for laminar organization of cortical neurons. Neuron 1995;14:899-912.

27. Ohshima T, Ward J, Huh C-G, et al. Targeted disruption of the cyclin-dependent kinase 5 gene results in abnormal corticogenesis, neuronal pathology and perinatal death. Proc Nat Acad Sci 1996;93:11173-8.

28. Kuzniecky Rl. Magnetic resonance imaging in developmental disorders of the cerebral cortex. Epilepsia 1994;35 SuppI 6:S44-56.

29. Sarnat MB. Developmental disorders of the nervous system. In: Bradley WG, Daroff RB, Fenichel RB, Marsden CD, eds. Neurology in clinical prectice. Boston: Butterworth-Heinemann, 1991:1239-68.

30. Choi BH, Kim RC, Lapham LW. Do radial glia give rise to both astroglia and oligodendroglial ceiss? Dev Brain Res 1983;8:119-30.

31. Gressens P, Richelme C, Kadhim HJ, Gadisseux JF, Evrard P. The germinative zone produces the most cortical astrocytes after neuronal migration in the developing mammalian brain. Biol Neonate 1992;61:4-24.

32. Takashima S, Becker LE. Developmental changes of glial fibrillary acidic protein in cerebral white matter. Arch Neurol 1983;40:14-8.

33. Geelen JA, Dormans JA, Verhoef A. The early effects of methylmercury on the developing rat brain. Acta Neuropathol Berl 1990;80:432-8.

34. Barkovich AJ, Lindan CE. Congenital cytomegalovirus infection of the brain: imaging analysis and embryologie considerations. AJNR Am J Neuroradiol 1994;15:703-15. 211

35. Ferrer I, Alcantara S, Catala I, Zujar MJ. Experimentally induced laminar necrosis, status verrucosus, focal cortical dysplasia reminiscent of microgyria, and porencephaly in the rat. Exp Brain Res 1993;94:261-9.

36. Majkowski J, Lee MH, Kozlowski PB, Haddad R. EEG and seizure threshold in normal and lissencephalic ferrets. Brain Res 1984;307:29-38.

37. Germano IM, Sperber EF, Friedman LK, Zuckin RS, Moshe SL. Méthylation of nucleic acid during neuroblastic division induces experimental neuronal migration disorders. Proc Soc Neurosci 1995;21:1516.

38. Germano IM, Sperber EF, Friedman LK, Zuckin RS, Moshe SL. Neuronal migration disorders and epileptogenesis. Epilepsia 1994;35:2.

39. Sing SC. Redirected perforant and commissural connections of eutopic and ectopic neurons in the hippocampus of methylazoxymethanol acetate treated rats. Acta Neuropathol Berl 1978;44:197-202.

40. Sing SC. Deformed dendrites and reduced spine numvers on ectopic neurones in the hippocampus of rats exposed to methylazoxy methanol acetate. Acta Neuropathol Berl 1 9 8 0;49:193-8.

41. Germano IM, Friedman LK, Sperber EF, et al. Expression of GABA and glutamate receptors in a rat model of epilepsy. Proc Soc Neurosci 1993;17:1466.

42. Roper SN, Gilmore RL, Houser CR. Experimentally induced disorders of neuronal migration produce an increased propensity for electrographic seizures in rats. Epilepsy Res 1995;21:205-19.

43. Schmahl W, Knoedlseder M, Favor J, Davidson D. Defects of neuronal migration and the pathogenesis of cortical malformations are associated with Small eye (Sey) in the mouse, a point mutation at the Pax-6-locus. Acta Neuropathol Berl 1993;86:126-35.

44. Fushiki S, Matsushita K, Schull WJ. Decelerated migration of neocortical neurones in explant culture after exposure to radiation. Neuroreport 1993;5:353-6.

45. Reiner 0, Albrecht U, Gordon M, et al. Lisse nee phaly gene (LIS1) expression in the CNS suggests a role in neuronal migration. J Neurosci 1995;15:3730-8. 212

46. Silva LR, Gutnick MJ, Connors BW. Laminar distribution of neuronal membrane properties in neocortex of normal and reeler mouse. J Neurophysiol 1991;66:2034-40.

47. Becker LE, Armstrong DL, Chan F, Wood MM. Dendritic development in human occipital cortical neurons. Brain Res 1984;31 5:117-24.

48. Purpura DP. Dendritic differentiation in human cerebral cortex and aberrant developmental patterns. Adv Neurol 1975;12:139-48.

49. Sheen VL, Macklis JD. Targeted neocortical cell death in adult mice guides migration and differentiation of transplanted embryonic neurons. J Neurosci 1995;15:8378-92.

50. Koester SE, O'Leary DD. Connectional distinction between callosal and subcortically projecting cortical neurons is determined prior to axon extension. Dev Biol 1993;160:1-14.

51. Goldman-Rakic PS. Morphological consequences of prenatal injury to the primate brain. Prog Brain Res 1980;53:3-19.

52. Jones EG, Valentino KL, Fleshman JW. Adjustment of connectivity in rat neocortex after prenatal destruction of precursor cells of layers ll-IV. Brain Res 1982;254:425-31.

53. Fabregues I, Ferrer I, Cusi MV, et al. Fine structure based on the golgi method of the abnormal cortex and heterotopic nodules in pachygyria. Brain Dev 1984;6:317-22.

54. Sisodiya SM, Free SL, Stevens JM, Fish DR, Shorvon SD. Widespread cerebral structural changes in patients with cortical dysgenesis and epilepsy. Brain 1995;118:1039-50.

55. Dambska M, Wisniewski K, Sher J. Lissencephaly: two distinct clinico- pathological types. Brain Dev 1984;5:302-10.

56. Barkovich AJ. Magnetic resonance imaging of lissencephaly, polymicrogyria, schizencephaly, hemimegalencephaly and band heterotopia. In: Guerrini R, 213

Andermann F, Canapicchi R, Roger J, Zifkin BJ, Pfanner P, eds. Dysplasias of cerebral cortex and epilepsy. Philadelphia: Lippincott-Raven, 1996:115-29.

57. Reiner 0, Carrozzo R, Shen Y, et al. Isolation of a Miller-Dieker lissencephaly gene containing G protein beta-subunit-like repeats. Nature 1993;364:717-21.

58. Ledbetter SA, Kuwano A, Dobyns WB, Ledbetter DH. Microdeletions of chromosome 17p13 as a cause of isolated lissencephaly. Am J Hum Genet 1992;50:182-9.

59. Hattori M, Adachi H, Tsujimoto M. Miller-Deiker lissencephaly gene encodes a subunit of brain platelet-activating factor acetylhydrolase. Nature 1994;370:216-8.

60. De Rosa MJ, Secor DL, Barsom M, Fisher RS, Vinters HV. Neuropathologic findings in surgically treated hemimegalencephaly: immunohistochemical, morphometric, and ultrastructural study. Acta Neuropathol Berl 1992;84:250-60.

61. Dobyns W, Stratton R, Greenberg F. Syndromes with lissencephaly. I: Miller- Deiker and Norman-Roberts syndromes and isolated lissencephaly. Am J Med Genet 1984;18:509-26.

62. Sebire G, Goutieres F, Tardieu M, Landrieu P, Aicardi J. Extensive macrogyri or no visible gyri: distinct clinical, electroencephalographic, and genetic features according to different imaging patterns. Neurology 1995;45:1105-11.

63. de Rijk van Andel JF, Arts WF, de Weerd AW. EEG and evoked potentials in a series of 21 patients with lissencephaly type I. Neuropediatrics 1992;23:4-9.

64. Barkovich A, Koch I , Carrol C. The spectrum of lissencephaly: report of ten cases analysed by magnetic resonance imaging. Ann Neurol 1991;30:139-46.

65. Squier MV. Development of the cortical dysplasia of type II lissencephaly. Neuropathol AppI Neurobiol 1993;19:209-13.

66. Aicardi J. The Agyria-pachygyria complex: a spectrum of cortical malformations. Brain Dev 1991;13:1-8.

67. Barkovich AJ, Kjos BO. Gray matter heterotopias: MR characteristics and correlation with developmental and neurologic manifestations. Radiology 214

1992;182:493-9.

68. Raymond AA, Fish DR, Stevens JM, et al. Subependymal heterotopia: a distinct neuronal migration disorder associated with epilepsy. J Neurol Neurosurg Psychiat 1994;57:1195-202.

69. Huttenlocher PR, Taravath S, Mojtahedi S. Periventricular heterotopia and epilepsy. Neurology 1994;44:51-5.

70. Eksioglu YZ, Scheffer IE, Cardenas P, et al. Periventricular heterotopia: an X- linked dominant epilepsy locus causing aberrant cerebral cortical development. Neuron 1996;16:77-87.

71. Palmini A, Andermann F, Aicardi J, et al. Diffuse cortical dysplasia, or the 'double cortex' syndrome: the clinical and epileptic spectrum in 10 patients. Neurology 1991;41:1656-62.

72. Barkovich AJ, Guerinni R, Battaglia G, et al. Band heterotopia: correlation of outcome with magnetic resonance imaging parameters. Ann Neurol 1994;36:609- 17.

73. Pinard J-M, Motte J, Chiron C, et al. Subcortical laminar heterotopia and lissencephaly in two families: a single X-linked dominant gene. J Neurol Neurosurg Psychiat 1994;57:914-20.

74. Dobyns WB, Truwit CL. Lissencephaly and other malformations of cortical development: 1995 update. Neuropediatrics 1995;26:132-47.

75. Billette de Villemeur T, Chiron C, Robain 0. Unlayered polymicrogyria and agenesis of the corpus callosum: a relevant association? Acta Neuropathol Berl 1992;83:265-70.

76. McBride MC, Kemper TLM. Pathogenesis of four-layered ,icrogyric cortex in man. Acta Neurpathol Berl 1992;57:93-8.

77. Ferrie CD, Jackson GD, Giannakodimos S, Panayiotopoulos CP. Posterior agyria- pachygyria with polymicrogyria: evidence for an inherited neuronal migration disorder. Neurology 1995;45:150-3. 215

78. Richman DP, Stewart RM, Caviness VS. Cerebral microgyria in a 27 week fetus: an architectonic and topographic analysis. J Neuropathol Exp Neurol 1974;33:374-84.

79. Rosen GD, Press DM, Sherman GF, Galaburda AM. The development of induced cerebrocortical microgyria in the rat. J Neuropathol Exp Neurol 1992;51:601- 11.

80. Kuzniecky R, Andermann F, Tampieri D, et al. Bilateral central macrogyria: epilepsy, pseudobulbar palsy, and mental retardation-a recognizable neuronal migration disorder. Ann Neurol 1989;25:547-54.

81. Graff Radford NR, Bosch EP, Stears JC, Tranel D. Developmental Foix-Chavany- Marie syndrome in identical twins. Ann Neurol 1986;20:632-5.

82. Guerrini R, Dravet C, Raybaud C, et al. Neurological findings and seizure outcome in children with bilateral opercular macrogyric-like changes detected by MRI. Dev Med Child Neurol 1992;34:694-705.

83. Guerrini R, Dravet C, Raybaud C, et al. Epilepsy and focal gyral anomalies detected by MRI: electroclinico-morphological correlations and follow-up. Dev Med Child Neurol 1992;34:706-18.

84. Barkovich AJ, Kjos BO. Schizencephaly: correlation of clinical findings with MR characteristics. AJNR Am J Neuroradiol 1992;13:85-94.

85. Bird C, Gilles F. Type I schizencephaly: CT and neuropathological findings. AJNR 1987;8:451-4.

86. Landrieu P, Lacroix C. Schizencephaly, consequence of a developmental vasculopathy? A clinicopathological report. Clin Neuropathol 1994;13:192-6.

87. Brunelli S, Faiella A, Capra V, et al. Germline mutations in the homeobox gene EMX2 in patients with severe schizencephaly. Nature Genet 1996;12:94-6.

88. Taylor DC, Falconer MA, Bruton CJ, Corsellis JA. Focal dysplasia of the cerebral cortex in epilepsy. J Neurol Neurosurg Psychiatry 1971;34:369-87.

89. Leuzzi V, Ricciotti V, Pelliccia A. Hemiplegic dystonia associated with regional 216

cortical dysplasia (pachygyria). Mov Disord 1993;8:242-4.

90. Schifter T, Hoffman JM, Hatten HP, Jr., et al. Neuroimaging in infantile autism. J Child Neurol 1994;9:155-61.

91. Lehericy S, Dormont D, Semah F, et al. Developmental abnormalities of the medial temporal lobe in patients with temporal lobe epilepsy. AJNR Am J Neuroradiol 1995;16:617-26.

92. Barkovich AJ, Kjos 80. Nonlissencephalic cortical dysplasias: correlation of imaging findings with clinical deficits [see comments]. AJNR Am J Neuroradiol 1992;13:95-103.

93. Hirabayashi S, Binnie CD, Janota I, Polkey CE. Surgical treatment of epilepsy due to cortical dysplasia: clinical and EEG findings. J Neurol Neurosurg Psychiatry 1993;56:765-70.

94. Palmini A, Andermann F, Olivier A, et al. Focal neuronal migration disorders and intractable partial epilepsy: a study of 30 patients. Ann Neurol 1991;30:741-9.

95. Ferrer I, Pineda M, Tallada M, et al. Abnormal local-circuit neurons in epilepsia partialis continua associated with focal cortical dysplasia. Acta Neuropathol Berl 1992;83:647-52.

96. Fusco L, Bertini E, Vigevano F. Epilepsia partialis continua and neuronal migration anomalies. Brain Dev 1992;14:323-8.

97. Desbiens R, Berkovic SF, Dubeau F, et al. Life-threatening focal status epilepticus due to occult cortical dysplasia. Arch Neurol 1993;50:695-700.

98. Barkovich AJ, Chuang SH, Norman D. MR of neuronal migration anomalies. AJNR 1987;8:1009-117.

99. Vital A, Marchai C, Loiseau H, et al. Glial and neuronoglial malformative lesions associated with medically intractable epilepsy. Acta Neuropathol Berl 1994;87:196-201.

100. Palmini A, Gambardella A, Andermann F, et al. Operative strategies for patients with cortical dysplastic lesions and intractable epilepsy. Epilepsia 1994;35 217

SuppI 6:S57-71.

101. Prayson RA, Estes ML. Cortical dysplasia: a histopathologic study of 52 cases of partial lobectomy in patients with epilepsy. Hum Pathol 1995;26:493-500.

102. Guerrini R, Dravet C, Raybaud C, et al. Epilepsy and focal gyral anomalies detected by MRI: electroclinico-morphological correlations and follow-up. Dev Med Child Neurol 1992;34:706-18.

103. Kuzniecky R, Murro A, King D, et al. Magnetic resonance imaging in childhood intractable partial epilepsies: pathologic correlations. Neurology 1993;43:681- 7.

104. Quirk JA, Kendall B, Kingsley DP, Boyd SG, Pitt MC. EEG features of cortical dysplasia in children. Neuropediatrics 1993;24:193-9.

105. Gambardella A, Palmini A, Andermann F, et al. Usefulness of focal rhythmic discharges on scalp EEG of patients with focal cortical dysplasia and intractable epilepsy. Electroencephalogr Clin Neurophysiol 1996;98:243-9.

106. Becker LE. Synaptic dysgenesis. Can J Neurol Sci 1991;18:170-80.

107. Vinters HV, Fisher RS, Cornford ME, et al. Morphological substrates of infantile spasms: studies based on surgically resected cerebral tissue. Childs Nerv Syst 1992;8:8-17.

108. Duong T, De Rosa MJ, Poukens V, Vinters HV, Fisher RS. Neuronal cytoskeletal abnormalities in human cerebral cortical dysplasia. Acta Neuropathol Berl 1994;87:493-503.

109. Mattia D, Olivier A, Avoli M. Seizure-like discharges recorded in human dysplastic neocortex maintained in vitro. Neurology 1995;45:1391-5.

110. Trottier S, Evrard B, Biraben A, Chauve I P. Altered patterns of catecholaminergic fibers in focal cortical dysplasia in two patients with partial seizures. Epilepsy Res 1994;19:161-79.

111. Trottier S, Berger B, Chauvel P, Dedek Y, Gay M. Alterations of the cortical noradrenergic system in chronic cobalt epileptogenic foci in the rat. Neuroscience 218

1981;6:1069-80.

112. Trottier S, LIndvall 0, Chauvel P, Bjorkiund A. Facilitation of focal cobalt-induced epilepsy after lesions of the noradrenergic locus coeruleus system. Brain Res 1988;454:308-14.

113. Raymond AA, Halpin SF, Alsanjari N, et al. Dysembryoplastic neuroepithelial tumor. Features in 16 patients. Brain 1994;117:461-75.

114. Prayson RA, Estes ML. Dysembryoplastic neuroepithelial tumor. Am J Clin Pathol 1992;97:398-401.

115. Leung SY, Gwi E, Ng HK, Fung CF, Yam KY. Dysembryoplastic neuroepithelial tumor. A tumor with small neuronal cells resembling oligodendroglioma. Am J Surg Pathol 1994;18:604-14.

116. Taratuto AL, Pomata H, Sevlever G, Gallo G, Monges J. Dysembryoplastic neuroepithelial tumor: morphological, immunocytochemical, and deoxyribonucleic acid analyses in a pediatric series. Neurosurgery 1995;36:474-81.

117. Prayson RA, Estes ML, Morris HH. Coexistence of neoplasia and cortical dysplasia in patients presenting with seizures. Epilepsia 1993;34:609-15.

118. Vigevano F, Bertini E, Boldrini R, et al. Hemimegalencephaly and intractable epilepsy:benefits of hemispherectomy. Epilepsia 1989;30:833-43.

119. Cieuta C, Guerrini R, Ferrari AR, Dulac 0. Antiepileptic drug treatment and intractability of epilepsy related to cortical dysplasia. In: Guerrini R, Andermann F, Canapicchi R, Roger J, Zifkin BJ, Pfanner P, eds. Dysplasias of cerebral cortex and epilepsy. Philadelphia: Lippincott-Raven, 1996:337-50.

120. Marchai G, Andermann F, Tampieri D, et al. Generalized cortical dysplasia manifested by diffusely thick cerebral cortex. Arch Neurol 1989;46:430-4 .

121. Villemure JG, Rasmussen T. Functional hemispherectomy in children. Neuropediatrics 1993;24:53-5.

122. Di Rocco C. Surgical treatment for hemimegalencephaly. In: Guerrini R, Andermann F, Canapicchi R, Roger J, Zifkin BJ, Pfanner P, eds. Dysplasias of 219

cerebral cortex and epilepsy. Philadelphia: Lippincott-Raven, 1996:295-304.

123. Dubeau F, Tampieri D, Andermann E, et al. Periventricular and subcortical nodular hetrotopia: comparison of clinical findings and results of surgical treatment. In: Guerrini R, Andermann F, Canapicchi R, Roger J, Zifkin BJ, Pfanner P, eds. Dysplasias of cerebral cortex and epilepsy. Philadelphia: Lippincott-Raven, 1996:395-406.

124. Leblanc R, Tampieri D, Robitaille Y, Feindel W, Andermann F. Surgical treatment of intractable epilepsy associated with schizencephaly. Neurosurgery 1991;29:421-9.

125. Palmini A, Andermann F, Olivier A, Tampieri D, Robitaille Y. Focal neuronal migration disorders and intractable partial epilepsy: results of surgical treatment. Ann Neurol 1991;30:750-7.

126. Kuzniecky R, Garcia JFI, Faught E, Morawetz RB. Cortical dysplasia in temporal lobe epilepsy: magnetic resonance imaging correlations. Ann Neurol 1991;29:293-8.

127. Sawhney IM, Robertson U, Polkey CE, Binnie CD, Elwes RD. Multiple subpial transection: a review of 21 cases. J Neurol Neurosurg Psychiatry 1995;58:344- 9.

128. Barr ML, Kiernan JA. The human nervous system: an anatomical viewpoint. Philadelphia: Lippincott, 1988.

129. Braak H. Architectonics of the human telencephalic cortex. Berlin: Springer- Verlag, 1980.

130. Hubei TH, Weisel TN. Functional architecture of macaque monkey visual cortex. Proc R Soc Lond 1977;198:1-59.

131. Mody I, De Koninck Y, Otis TS, Soltesz I. Bridging the cleft at GABA synapses in the brain. Trends Neurosci 1994;17:517-25.

132. Krnjevic K. Chemical nature of synaptic transmission in vertebrates. Physiol Rev 1974;54:418-540. 220

133. Hill DR, Bowery NG. 3H-baclofen and 3H-GABA bind to bicucufline-insensitive GABAB sites in rat brain. Nature 1981;290:149-52.

134. Sivilotti L, Nistri A. GABA receptor mechanisms in the central nervous system. Prog Neurobiol 1991;36:35-92.

135. Cutting GR, Lu L, O'Hara BF, et al. Cloning of the y-aminobutyric acid (GABA) pi cDNA: a GABA receptor subunit highly expressed in the retina. Proc Natl Acad Sci USA 1991;88:2673-7.

136. Bormann J, Feigenspan A. GABA-C receptors. Trends Neurosci 1995; 18:515-9.

137. Schofield PR, Darlison MG, Fujita N, et al. Sequence and functional expression of the GABAA receptor shows a ligand-gated receptor super-family. Nature 1987;328:221-7.

138. DeLorey TM, Olsen RW. y-aniinobutyric acid-A receptor strucure and function. J Biol Chem 1992;267:16747-50.

139. Schumacher M, McEwen BS. Steroid and barbiturate modulation of the GABAa receptor. Possible mechanisms. Mol Neurobiol 1989;3:275-304.

140. Schoch P, Richards JG, Haring P, et al. Co-localization of GABA receptors and benzodiazepine receptors in the brain shown by monoclonal antibodies. Nature 1985;314:168-71.

141. Vicini S, Mienville JM, Costa E. Actions of benzodiazepine and p-carboline derivatives on y-arninobutyric acid-activated Cl-channels recorded from membrane patches of neonatal rat cortical neurons in culture. J Pharmacol Exp Ther 1987;243:1195-201.

142. Fisher TE, Johnson DD, Tuchek JM, Crawford RD. Evidence for the pharmacological relevance of benzodiazepine receptors to anticonvulsant activity. Can J Physiol Pharmacol 1985;63:1477-9.

143. Verma A, Snyder SH. Peripheral type benzodiazepine receptors. Annu Rev Pharmacol Toxicol 1989;29:307-22.

144. Sieghart W. Heterogeneity of GABAA-benzodiazepine receptors. Biochem Soc Trans 221

1991;19:129-32.

146. Pritchett DB, Luddens H, Seeburg PH. Type I and Type II GABAA-benzodiazepine receptors produced in transfected cells. Science 1989;245:1389-92.

147. Shivers BD, Killisch I, Sprengel R, et al. Two novel GABAA receptor subunits exist in distinct neuronal subpopulations. Neuron 1989;3:327-37.

148. Rabow LE, Russek SJ, Farb DH. From ion currents to genomic analysis: recent advances in GABAA receptor research. Synapse 1995;21:189-274.

149. Verdoorn TA, Draguhn A, Ymer S, Seeburg PH, Sakmann B. Functional properties of recombinant rat GABAA receptors depend upon subunit composition. Neuron 1990;4:919-28.

150. Burt DR, Kamatchi GL. GABAA receptor subtypes: from pharmacology to molecular biology. Faseb J 1991;5:2916-23.

151. Luddens H, Wisden W. Function and pharmacology of multiple GABAA receptor subunits. Trends Pharmacol Sci 1991 ;12:49-51.

152. Doble A, Martin IL. Multiple benzodiazepine receptors - no reason for anxiety. Trends Pharmacol Sci 1992;13:76-81.

153. Mohler H. GABAergic synaptic transmission. Regulation by drugs. Arzneimittelforschung 1992;42:211-4.

154. Mohler H, Okada R. Benzodiazepine receptor: demonstration in the central nervous system. Science 1977;198:849-51.

155. Braestrup C, Albrechtsen R, Squires RF. High densities of benzodiazepine receptors in human cortical areas. Nature 1977;269:702-4.

156. Richards JG, Schoch P, Haring P, Takacs B, Mohler H. Resolving GABAA/benzodiazepine receptors: cellular and subcellular localization in the CNS with monoclonal antibodies. J Neurosci 1987;7:1866-86.

157. Meldrum BS, Chapman AG. Benzodiazepine receptors and their relationship to the treatment of epilepsy. Epilepsia 1986;27 SuppI 1:S3-13. 222

158. Young WS, Kuhar MJ. Autoradiographic localisation of benzodiazepine receptors in the brains of humans and animals. Nature 1979;280:393-5.

159. Zezula J, Cortes R, Probst A, Palacios JM. Benzodiazepine receptor sites in the human brain: autoradiographic mapping. Neuroscience 1988;25:771-95.

160. Suhara T, Inoue 0, Kobayashi K, et al. No age-related changes in human benzodiazepine receptor binding measured by PET with [11C]Ro 15-4513. Neurosci Lett 1993;159:207-10.

161. Morrison JH, Hof PR. The organisation of the cerebral cortex: from molecules to circuits. Discussions Neurosci 1992;9:29-41.

162. Ince P, Stout N, Shaw P, et al. Parvalbumin and calbindin D-28k in the human motor system and in motor neuron disease. Neuropathol App Neurobiol 1993;19:291-9.

163. Celio MR. Calbindin D-28k and parvalbumin in the rat nervous system. Neuroscience 1990;3:375-475.

164. Dichter MA. Emerging insights into mechanisms of epilepsy: implications for new antiepileptic drug development. Epilepsia 1994;35 SuppI 4:S51-7.

165. Ayala GF, Dichter M, Gumnit RJ, Matsumoto H, Spencer WA. Genesis of epileptic interictal spikes: new knowledge of cortical feedback systems suggests a neurophysiological explanation of brief paroxysms. Brain Res 1973;52:1.

166. Dichter MA, Ayala GF. Cellular mechanisms of epilepsy: a status report. Science 1987;237:157-64.

167. Prince DA, Wilder BJ. Control mechanisms in cortical epileptogenic foci. Arch Neurol 1967;16:194-202.

168. Yamamoto C, Matsumoto K, Takagi M. Potentiation of excitatory postsynaptic potentials during and after repetitive stimulation in thin hippocampal sections. Exp Brain Res 1 980;38:469-77.

169. Miles R, Wong RKS. Latent synaptic pathways revealed after tetanic stimulation in 223

the hippocampus. Nature 1987;329:724.

170. Lombardo AJ, Kuzniecky R, Powers RE, Brown GB. Altered brain sodium channel transcript levels in human epilepsy. Brain Res Mol Brain Res 1996;35:84-90.

171. Knowles W, Schwartzkroin P. Local circuit synaptic interactions in hippocampal brain slices. J Neurosci 1981;1:318-22.

172. Malenka RC, Kocsis JD, Ransom BR, Waxman SG. Modulation of parallel fiber excitability by postsynaptically mediated changes in extracellular potassium. Science 1981;214:339-41.

173. Faber OS, Korn H. Synergism at central synapses due to lateral diffusion of transmitter. Proc Natl Acad Sci 1988;85:8708.

174. MacVicar B, Dudek FE. Electrotonic coupling between granule cells of rat dentate gyrus: physiological and anatomical evidence. J Neurophysiol 1982;47:579-92.

1 76. Prince DA, Deisz RA, Thompson SM, Chagnac-Amitai Y. Functional alterations in GABAergic inhibition during activity. Epilepsy Res 1992;Suppl. 8:31-8.

177. Piguet P. GABAA- and GABAB-mediated inhibition in the rat dentate gyrus in vitro. Epilepsy Res 1993;16:111-22.

178. Mott DD, Bragdon AC, Lewis DV, Wilson WA. Baclofen has a proepileptic effect in the rat dentate gyrus. J Pharmacol Exp Ther 1989;249:721-5.

179. Schneiderman JH, Sterling CA, Luo R. Hippocampal plasticity following epileptiform bursting produced by GABAA antagonists. Neuroscience 1994;59:259-73.

180. McNamara JO. Kindling: an animal model of complex partial epilepsy. Ann Neurol 1984;16 Suppl:S72-6.

181. Loscher W, Jackel R, Czuczwar SJ. Is amygdala kindling in rats a model for drug- resistant partial epilepsy? Exp Neurol 1986;93:211-26.

182. Morimoto K, Sanei T, Sato K. Comparative study of the anticonvulsant effect of gamma-aminobutyric acid agonists in the feline kindling model of epilepsy. 224

Epilepsia 1993;34:1123-9.

183. Sato K, Morimoto K, Okamoto M, et al. An analysis of anticonvulsant actions of GABA agonists (progabide and baclofen) in the kindling model of epilepsy. Epilepsy Res 1990;5:117-24.

184. Shin C, Rigsbee LC, McNamara JO. Anti-seizure and anti-epileptogenic effect of gamma-vinyl gamma-aminobutyric acid in amygdaloid kindling. Brain Res 1986;398:370-4.

185. Shin C, Pedersen HB, McNamara JO. gamma-Aminobutyric acid and benzodiazepine receptors in the kindling model of epilepsy: a quantitative radiohistochemical study. J Neurosci 1985;5:2696-701.

186. Kamphuis W, De Rijk TC, Lopes da Silva FH. Expression of GABAA receptor subunit mRNAs in hippocampal pyramidal and granular neurons in the kindling model of epileptogenesis: an in situ hybridization study. Brain Res Mol Brain Res 1995;31:33-47.

187. Ben Attia M, N'Gouemo P, Belaidi M, Rondouin G, Chicheportiche R. Kindling and electrode effects on the benzodiazepine receptors density of olfactory bulb and hippocampus after olfactory bulb kindling. Neurosci Lett 1992;143:74-8.

188. Otis TS, De Koninck Y, Mody I. Lasting potentiation of inhibition is associated with an increased number of y -aminobutyric acid type A receptors activated during miniature inhibitory postsynaptic currents. Proc Natl Acad Sci USA 1994;91:7698-702.

189. Sato K, Morimoto K, Okamoto M. Anticonvulsant action of a non-competitive antagonist of NMDA receptors (MK-801) in the kndling model of epilepsy. Brain Res 1988;463:12-20.

190. Mody I, Heinemann I. NMDA receptors of dentate gyrus granule cells participate in synaptic transmission following kindling. Nature 1987;326:701-4.

191. Martin D, McNamara JO, Nadler JV. Kindling enhances sensitivity of CA3 hippocampal pyramidal cells to NMDA. J Neurosci 1992;12:1928-35.

192. Rainnie DG, Asprodini EK, Shinnickgallagher P. Kindling-induced long-lasting 225

changes in synaptic transmission in the basolateral amygdala. J Neurophysiol 1992;67:443-54.

193. Kamphuis W, Gorter JA, da Silva FL. A long-lasting decrease in the inhibitory effect of GABA on glutamate responses of hippocampal pyramidal neurons induced by kindling epileptogenesis. Neuroscience 1991;41:425-31.

194. Minamato Y, Itano T, Tokuda M, et al. In vivo microdialysis of amino acid neurotransmitters in the hippocampus of amygdaloid kindled rats. Brain Res 1992;573:345-8.

195. Prince HK, Conn PJ, Blackstone CD, Huganir RL, Levey Al. Down-regulation of AM PA receptor subunit GluR2 in amygdaloid kindling. J Neurochem 1995;64:462-5.

196. Ashwood TJ, Lancaster B, Wheal HV. Intracellular electrophysiology of CA1 pyramidal neurones in slices of the kainic acid lesioned hippocampus of the rat. Exp Brain Res 1986;62:189-98.

197. Franck JE, Kunkel DD, Baskin DG, Schwartzkroin PA. Inhibition in kainate- lesioned hyperexcitable hippocampi: physiologic, autoradiographic, and immunocytochemical observations. J Neurosci 1988;8:1991-2002.

198. Wheal FiV. Function of synapses in the CA1 region of the hippocampus: their contribution to the generation or control of epileptiform activity. Comp Biochem Physiol A 1989;93:211-20.

199. Friedman LK, Pellegrini Giampietro DE, Sperber EF, et al. Kainate-induced status epilepticus alters glutamate and GABAA receptor gene expression in adult rat hippocampus: an in situ hybridization study. J Neurosci 1994;14:2697-707.

200. Tauck DL, Nadler JV. Evidence of functional mossy fiber sprouting in hippocampal formation of kainic acid treated rats. J Neurotox 1985;5:1016-22.

201. Sutula T, Xiao-Xian H, Cavazos J, Scott G. Synaptic reorganisation in the hippocampus induced by abnormal functional activity. Science 1988:1147-50.

202. Ben-Ari Y, Represa A. Brief seizure episodes induce long-term potentiation and mossy fibre sprouting in the hippocampus. Trends Neurosci 1990;13:321. 226

203. Sutula TP. Reactive changes in epilepsy; cell death and axon sprouting induced by kindling. Epilepsy Res 1991;10:62-70.

204. Dragunov M, Robertson HA, Robertson GS. Amygdala kindling and c-fos protein(s). Neurosci Lett 1988;99:351.

205. Watanabe Y, Johnson RS, Butler LS, et al. Null mutation of c-fos impairs structural and functional plasticities in the kindling model of epilepsy. J Neurosci 1996;16:3827-36.

206. McNamara RK, Routtenberg A. NMDA receptor blockade prevents kainate induction of protein FI/GAP-43 mRNA in hippocampal granule cells and subsequent mossy fiber sprouting in the rat. Brain Res Mol Brain Res 1995;33:22-8.

207. Rashid K, Van der Zee CE, Ross GM, et al. A nerve growth factor peptide retards seizure development and inhibits neuronal sprouting in a rat model of epilepsy. Proc Natl Acad Sci U S A 1995;92:9495-9.

208. Morrison RS, Wenzel HJ, Kinoshita Y, et al. Loss of the p53 tumor suppressor gene protects neurons from kainate-induced cell death. J Neurosci 1996;16:1337-45.

209. Lothman EW, Bertram EH, Kapur J, Stringer JL. Recurrent spontaneous hippocampal seizures in the rat as a chronic sequela to limbic status epilepticus. Epilepsy Res 1990;6:110-8.

210. Bekenstein J, Rempe D, Lothman E. Decreased heterosynaptic and homosynaptic paired pulse inhibition in the rat hippocampus as a chronic sequela to limbic status epilepticus. Brain Res 1993;601:111 -20.

211. Sloviter RS. Permanently altered hippocampal structure, excitability, and inhibition after experimental status epilepticus in the rat: the "dormant basket cell" hypothesis and its possible relevance to temporal lobe epilepsy. Hippocampus 1991;1:41-66.

212. Bekenstein JW, Lothman EW. Dormancy of inhibitory interneurons in a model of temporal lobe epilepsy. Science 1993;259:97-100. 227

213. Sloviter RS. Decreased hippocampal inhibition and a selective loss of interneurons in experimental epilepsy. Science 1987;235:73-6.

214. Chagnac Amitai Y, Connors BW. Horizontal spread of synchronized activity in neocortex and its control by GABA-mediated inhibition. J Neurophysiol 1989;61:747-58.

215. Chervin RD, Pierce PA, Connors BW. Periodicity and directionality in the propagation of epileptiform discharges across neocortex. J Neurophysiol 1988;60:1695-713.

216. Chatt AB, Ebersole JS. Convulsant-induced neocortical epilepsy: neuronal networks and anticonvulsant actiona in vivo. In: Faingold CL, Fromm GH, eds. Drugs for control of epilepsy: actions on neuronal networks involved in seizure disorders. Boca Raton: CRC Press, 1992:143-69.

217. Ebersole JS, Chatt AB. Differences between strychnine and penicillin epileptogenesis suggest a laminar organization of neocortical inhibition. Brain Res 1985;340:390-6.

218. Witte OW. Afterpotentials of penicillin-induced epileptiform neuronal discharges in the motor cortex of the rat in vivo. Epilepsy Res 1994;18:43-55.

219. Brailowsky S, Kunimoto M, Menini C, et al. The GABA-withdrawal syndrome: a new model of focal epileptogenesis. Brain Res 1988;442:175-9.

220. Silva Barrat C, Araneda S, Menini C, Champagnat J, Naquet R. Burst generation in neocortical neurons after GABA withdrawal in the rat. J Neurophysiol 1992;67:715-27.

221. Craig CR, Colasanti BK. Cobalt-induced focal seizures: neuronal networks and actions of antiepleptic drugs. In: Faingold CL, Fromm GH, eds. Drugs for control of epilepsy: actions on neuronal networks involved in seizure disorders. Boca Raton: CRC Press, 1992:126-41.

222. Engel J. Secondary epileptogenesis in rats. Electroenceph Clin Neurophysiol 1968;25:494.

223. Craig CR, Colasanti BK. GABA receptors, lipids, and gangliosides in cobalt epileptic 228

focus. Adv Neurol 1986;44:379-91.

224. Esclapez M, Trottier S. Changes in GABA-immunoreactive cell density during motor focal epilepsy induced by cobalt in the rat. Exp Brain Res 1989;76:369- 85.

225. Pitkanen A, Saano V, Hyvonen K, Airaksinen MM, Riekkinen PJ. Decreased GABA, benzodiazepine, and picrotoxinin receptor binding in brains of rats after cobalt- induced epilepsy. Epilepsia 1987;28:11-6.

226. Kawakami Y, Koyama I. Changes in the strength of recurrent inhibition in cobalt- induced epilepsy. Epilepsia 1992;33:428-34.

227. Salin P, Tseng GF, Hoffman S, Parada I, Prince DA. Axonal sprouting in layer V pyramidal neurons of chronically injured cerebral cortex. J Neurosci 1995;15:8234-45.

228. Bakay RA, Harris AB. Neurotransmitter, receptor and biochemical changes in monkey cortical epileptic foci. Brain Res 1981;206:387-404.

229. Ribak CE, Harris AB, Vaughn JE, Roberts E. Inhibitory, GABAergic nerve terminals decrease at sites of focal epilepsy. Science 1979;205:211-4.

230. Ribak CE, Bradburne RM, Harris AB. A preferential loss of GABAergic, inhibitory synapses in epileptic foci: A quantitative ultrastructural analysis of monkey neocortex. J Neurosci 1982;2:1725-35.

231. Ribak CE, Hunt CA, Bakay RA, Oertel WH. A decrease in the number of GABAergic somata is associated with the preferential loss of GABAergic terminals at epileptic foci. Brain Res 1986;363:78-90.

232. Ribak CE, Joubran C, Kesslak JP, Bakay RA. A selective decrease in the number of GABAergic somata occurs in pre-seizing monkeys with alumina gel granuloma. Epilepsy Res 1989;4:126-38.

233. Houser CR, Harris AB, Vaughn JE. Time course of the reduction of GABA terminals in a model of focal epilepsy: a glutamic acid decarboxylase immunocytochemical study. Brain Res 1986;383:129-45. 229

234. Ribak CE. Axon terminals of GABAergic chandelier cells are lost at epileptic foci. Brain Res 1985;326:251-60.

235. Kunimoto M, Tanaka T, Fukuda H, Yonemasu Y, Naquet R. GABA withdrawal syndrome: a model of the epilepsia partialis continua. Jpn J Psychiatry Neurol 1990;44:438-9.

236. Morimoto K, Goddard GV. The substantia nigra is an important site for the containment of seizure generalization in the kindling model of epilepsy. Epilepsia 1987;28:1-10.

237. Turski L, Cavalheiro EA, Schwarz M, et al. Susceptibility to seizures produced by pilocarpine in rats after microinjection of isoniazid or gamma-vinyl-GABA into the substantia nigra. Brain Res 1986;370:294-309.

238. Loscher W, Schwark WS. Evidence for impaired GABAergic activity in the substantia nigra of amygdaloid-kindled rats. Brain Res 1985;339:146-56.

239. McNamara JO, Galloway MT, Rigsbee LC, Shin C. Evidence implicating substantia nigra in regulation of kindled seizure threshold. J Neurosci 1984;4:2410-7.

240. Hatelid JN, Lothman EW, Zorumski ZF. Electrophysiologic and metabolic mapping of hippocampal seizures. Neurosci Abstr 1982;8:1018.

241. Fine A, Meldrum BS, Patel S. Modulation of experimentally induced epilepsy by intracerebral grafts of fetal GABAergic neurons. Neuropsychologia 1990;28:627- 34.

242. Ono K, Wada JA. Facilitation of premotor cortical seizure development by intranigral muscimol. Brain Res 1987;405:183-6.

243. Wahnschaffe U, Loscher W. Selective bilateral destruction of substantia nigra has no effect on kindled seizures induced from stimulation of amygdala or piriform cortex in rats. Neurosci Lett 1990;11 3:205-10.

244. Sabatino M, La Grutta V, Ferraro G, La Grutta G. Relations between basal ganglia and hippocampus: action of substantia nigra and pallidum. Rev Electroencephalogr Neurophysiol Clin 1986;16:179-90. 230

245. Garant DS, Gale K. Substantia-nigra-mediated anticonvulsant actions: role of nigral output pathways. Exp Neurol 1987;97:143-59.

246. Gale K, Pazos A, Maggio R, Japikse K, Pritchard P. Blockade of GABA receptors in superior colliculus protects against focally evoked limbic motor seizures. Brain Res 1993;603:279-83.

247. Sabatino M, Ferraro G, La Grutta V. Relay stations and neurotransmitters between the pallidal region and the hippocampus. Electroencephalogr Clin Neurophysiol 1991;78:302-10.

248. Sabatino M, Ferraro G, Caravaglios G, et al. Accumbens-caudate-septal circuit as a system for hippocampal regulation: involvement of a GABAergic neurotransmission. Neurophysiol Clin 1992;22:3-16.

249. Flaherty AW, Graybiel AM. Anatomy of the basal ganglia. In: Marsden CD, Fahn S, eds. Movement Disorders 3. Oxford: Butterworth-Heinemann, 1994:1-27.

250. Detre JA, Alsop DC, Aguirre GK, Sperling MR. Coupling of cortical and thalamic ictal activity in human partial epilepsy: demonstration by functional magnetic resonance imaging. Epilepsia 1996;37:657-61.

251. Isokawa Akesson M, Wilson CL, Babb TL. Inhibition in synchronously firing human hippocampal neurons. Epilepsy Res 1989;3:236-47.

252. McDonald JW, Garofalo EA, Flood T, et al. Altered excitatory and inhibitory amino acid receptor binding in hippocampus of patients with temporal lobe epilepsy. Ann Neurol 1991;29:529-41.

253. Geddes JW, Cahan LD, Cooper SM, al. e. Altered distribution of excitatory amino acid receptors in temporal lobe epilepsy. Exp Neurol 1990;108:214-20.

254. Johnson EW, de Lanerolle NC, Kim JH, et al. "Central" and "peripheral" benzodiazepine receptors: opposite changes in human epileptogenic tissue. Neurology 1992;42:811-5.

255. de Lanerolle NC, Kim JFI, Robbins RJ, Spencer DD. Hippocampal interneuron loss and plasticity in human temporal lobe epilepsy. Brain Res 1989;495:387-95. 231

256. Mathern GW, Pretorius JK, Babb T L Quantified patterns of mossy fiber sprouting and neuron densities in hippocampal and lesional seizures. J Neurosurg 1995;82:211-9.

257. Isokawa M, Levesque MF. Increased NMDA responses and dendritic degeneration in human epileptic hippocampal neurons in slice. Neurosci Lett 1991;132:212-6.

258. During MJ, Spencer DD. Extracellular hippocampal glutamate and spontaneous seizure in the conscious human brain [see comments]. Lancet 1993;341:1607- 10.

259. Ronneengstrom E, Hillered L, Flink R, et al. Intracerebral microdialysis of extracellular amino acids in human epileptic focus. J Cereb Blood Flow Metab 1992;12:873-6.

260. Sherwin A, Quesney F, Gauthier S, et al. Enzyme changes in actively spiking areas of human epileptic cerebral cortex. Neurology 1984;34:927-33.

261. Sherwin A, Robitaille Y, Quesney F, et al. Excitatory amino acids are elevated in human epileptic cerebral cortex. Neurology 1988;38:920-3.

262. Lloyd KG, Bossi L, Morselli PL, et al. Alterations of GABA-mediated synaptic transmission in human epilepsy. Adv Neurol 1986;44:1033-44.

263. Sherwin A, Matthew E, Blain M, Guevremont D. Benzodiazepine receptor binding is not altered in human epileptogenic cortical foci. Neurology 1986;36:1380-2.

264. Sherwin AL, van Gelder NM. Amino acid and catecholamine markers of metabolic abnormalities in human focal epilepsy. Adv Neurol 1986;44:1011-32.

265. Haglund MM, Berger MS, Kunkel DD, et al. Changes in gamma-aminobutyric acid and somatostatin in epileptic cortex associated with low-grade gliomas. J Neurosurg 1992;77:209-16.

266. Ferrer I, Oliver B, Russi A, Casas R, Rivera R. Parvalbumin and calbindin-D28k immunocytochemistry in human neocortical epileptic foci. J Neurol Sci 1994;123:18-25.

267. Avoli M, Olivier A. Bursting in human epileptogenic neocortex is depressed by an 232

N-methyl-D-aspartate antagonist. Neurosci Lett 1987;76:249-54.

268. Avoli M, Hwa GGC, Mattia D, Olivier A, Villemure J-G. Microphysiology of human neocortex in vitro. In: Guerrini R, Andermann F, Canapicchi R, Roger J, Zifkin BJ, Pfanner P, eds. Dysplasias of cerebral cortex and epilepsy. Philadelphia: Lippincott-Raven, 1996:35-42.

269. Engel J. Functional explorations of the human epileptic brain and their therapeutic implications. Electroencephalogr Clin Neurophysiol 1990;76:296- 316.

270. Berkovic SF, McIntosh A, Howell RA, et al. Familial temporal lobe epilepsy: a common disorder identified in twins. Ann Neurol 1996;40:227-35.

271. Ottman R, Risch N, Hauser WA, et al. Localization of a gene for partial epilepsy to chromosome lOq [see comments]. Nat Genet 1995;10:56-60.

272. Phillips HA, Scheffer IE, Berkovic SF, et al. Localization of a gene for autosomal dominant noctumal frontal lobe epilepsy to chromosome 20q 13.2. Nat Genet 1995;10:117-8.

273. Steinlein OK, Mulley JC, Propping P, et al. A missense mutation in the neuronal nicotinic acetylcholine receptor alpha 4 subunit is associated with autosomal dominant nocturnal frontal lobe epilepsy. Nat Genet 1995; 11:201-3.

274. Bonte FJ, Devous MD, Sr., Stokely EM, Homan RW. Single-photon tomographic determination of regional cerebral blood flow in epilepsy. AJNR Am J Neuroradiol 1983;4:544-6.

275. Lee Bl, Markand ON, Siddiqui AR, et al. Single photon emission computed tomography (SPECT) brain imaging using N,N,N'-trimethyl-N'-(2 hydroxy-3- methyl-5-123l-iodobenzyl)-1,3-propanediamine 2 HCI (HIPDM): intractable complex partial seizures. Neurology 1986;36:1471-7.

276. Lee Bl, Markand ON, Wellman HN, et al. HIPDM single photon emission computed tomography brain imaging in partial onset secondarily generalized tonic-clonic seizures. Epilepsia 1987;28:305-11.

277. Lee Bl, Markand ON, Wellman HN, et al. HIPDM-SPECT in patients with medically 233

intractable complex partial seizures. Ictal study. Arch Neurol 1988;45:397- 402.

278. Stefan H, Kuhnen C, Biersack HJ, Reichmann K. Initial experience with 99m Tc- hexamethyl-propylene amine oxime (HM-PAO) single photon emission computed tomography (SPECT) in patients with focal epilepsy. Epilepsy Res 1987;1:134- 8 .

279. Stefan H, Pawlik G, Bocher Schwarz HG, et al. Functional and morphological abnormalities in temporal lobe epilepsy: a comparison of interictal and ictal EEG, CT, MRI, SPECT and PET. J Neurol 1987;234:377-84.

280. Denays R, Rubinstein M, Ham H, Piepsz A, Noel P. Single photon emission computed tomography in seizure disorders. Arch Dis Child 1988;63:1184-8.

281. Podreka I, Lang W, Suess E, et al. Hexa-methyl-propylene-amine-oxime (HMPAO) single photon emission computed tomography (SPECT) in epilepsy. Brain Topogr 1988;1:55-60.

282. Homan RW, Paulman RG, Devous MD, Sr., et al. Cognitive function and regional cerebral blood flow in partial seizures. Arch Neurol 1989;46:964-70.

283. LaManna MM, Sussman NM, Harner RN, et al. Initial experience with SPECT imaging of the brain using 1-123 p-iodoamphetamine in focal epilepsy. Clin NucI Med 1989;14:428-30.

284. Gelfand MJ, Stowens DW. 1-123 iofetamine single photon emission tomography in school-age children with difficult-to-control seizures. Clin NucI Med 1989;14:675-80.

285. Andersen AR, Waldemar G, Dam M, et al. SPECT in the presurgical evaluation of patients with temporal lobe epilepsy-a preliminary report. Acta Neurochir SuppI Wien 1990;50:80-3.

286. Vies JS, Demandt E, Ceulemans B, de Roo M, Casaer PJ. Single photon emission computed tomography (SPECT) in seizure disorders in childhood. Brain Dev 1990;12:385-9.

287. Cordes M, Christe W, Henkes H, et al. Focal epilepsies: HM-PAO SPECT compared 234

with CT, MR, and EEG. J Comput Assist Tomogr 1990; 14:402-9.

288. Duncan R, Patterson J, Hadley DM, et al. Tc99m HM-PAO single photon emission computed tomography in temporal lobe epilepsy. Acta Neurol Scand 1990;81:287-93.

289. Duncan R, Patterson J, Hadley DM, et al. CT, MR and SPECT imaging in temporal lobe epilepsy [see comments]. J Neurol Neurosurg Psychiatry 1990;53:11-5.

290. Dietrich ME, Bergen D, Smith MC, Fariello R, Ali A. Correlation of abnormalities of interictal n-isopropyl-p-iodoamphetamine single-photon emission tomography with focus of seizure onset in complex partial seizure disorders. Epilepsia 1991;32:187-94.

291. Shen W, Lee 81, Park HM, et al. HIPDM-SPECT brain imaging in the presurgical evaluation of patients with intractable seizures. J NucI Med 1990;31:1280-4.

292. Bauer J, Stefan H, Feistel H, et al. [Ictal and interictal 99mTc-HMPAO-SPECT studies in temporal lobe epilepsy with unilateral EEG focus]. Nervenarzt 1991;62:745-9.

293. Markand ON, Shen W, Park HM, et al. Single photon imaging computed tomography (SPECT) for localization of epileptogenic focus in patients with intractable complex partial seizures. Epilepsy Res SuppI 1992;5:121-6.

294. Newton MR, Austin MC, Chan JG, et al. Ictal SPECT using technetium-99m- HMPAO: methods for rapid preparation and optimal deployment of tracer during spontaneous seizures. J NucI Med 1993;34:666-70.

295. Duncan R, Patterson J, Roberts R, Hadley DM, Bone I. Ictal/postictal SPECT in the pre-surgical localisation of complex partial seizures. J Neurol Neurosurg Psychiatry 1993;56:141-8.

296. Harvey AS, Bowe JM, Hopkins IJ, et al. Ictal 99mTc-HMPAO single photon emission computed tomography in children with temporal lobe epilepsy. Epilepsia 1993;34:869-77.

297. Markand ON, Salanova V, Worth RM, Park HM, Wellman HH. Ictal brain imaging in presurgical evaluation of patients with medically intractable complex partial 235

seizures. Acta Neurol Scand SuppI 1994;152:137-44.

298. Cross JH, Jackson GD, Neville BG, et al. Early detection of abnormalities in partial epilepsy using magnetic resonance. Arch Dis Child 1993;69:104-9.

299. Spencer SS. The relative contributions of MRI, SPECT, and PET imaging in epilepsy. Epilepsia 1994;35 SuppI 6:S72-89.

300. Rowe CC, Berkovic SF, Austin MC, McKay WJ, Bladin PP. Patterns of postictal cerebral blood flow in temporal lobe epilepsy: qualitative and quantitative analysis. Neurology 1991;41:1096-103.

301. Newton MR, Berkovic SF, Austin MC, et al. Postictal switch in blood flow distribution and temporal lobe seizures. J Neurol Neurosurg Psychiatry 1992;55:891-4.

302. Park CM, Kim SM, Streletz LJ, Zhang J, Intenzo C. Reverse crossed cerebellar diaschisis in partial complex seizures related to herpes simplex encephalitis. Clin NucI Med 1992;17:732-5.

303. Katz A, Bose A, Lind SJ, Spencer SS. SPECT in patients with epilepsia partialis continua [see comments]. Neurology 1990;40:1848-50.

304. Newton MR, Berkovic SF, Austin MC, et al. Dystonia, clinical lateralization, and regional blood flow changes in temporal lobe seizures. Neurology 1992;42:371- 7.

305. Kawamura M, Murase K, Kimura H, et al. Single photon emission computed tomography (SPECT) using N-isopropyl-p-(123l) iodoamphetamine (IMP) in the evaluation of patients with epileptic seizures. Eur J NucI Med 1990; 16:285- 92.

306. Rowe CC, Berkovic SF, Austin MC, et al. Visual and quantitative analysis of interictal SPECT with technetium-99m-HMPAO in temporal lobe epilepsy. J NucI Med 1991;32:1688-94.

307. Jibiki I, Kubota T, Fujimoto K, et al. Regional relationships between focal hypofixation images in 1231-IMP single photon emission computed tomography and epileptic EEG foci in interictal periods in patients with partial epilepsy. Eur 236

Neurol 1991;31:360-5.

308. Duncan S, Gillen G, Adams FG, Duncan R, Brodie MJ. Interictal HM-PAO SPECT: a routine investigation in patients with medically intractable complex partial seizures? Epilepsy Res 1992;13:83-7.

309. Pantano P, Matteucci C, di Piero V, et al. Quantitative assessment of cerebral blood flow in partial epilepsy using Xe-133 inhalation and SPECT. Clin NucI Med 1991;16:898-903.

310. Jabbari B, Van Nostrand D, Gunderson CM, et al. EEG and neuroimaging localization in partial epilepsy. Electroencephalogr Clin Neurophysiol 1991;79:108-13.

311. Adams C, Hwang PA, Gilday DL, et al. Comparison of SPECT, EEG, CT, MRI, and pathology in partial epilepsy. Pediatr Neurol 1992;8:97-103.

312. Ryvlin P, Garcia Larrea L, Philippon B, et al. High signal intensity on T2- weighted MRI correlates with hypoperfusion in temporal lobe epilepsy. Epilepsia 1992;33:28-35.

313. Ryvlin P, Philippon B, Cinotti L, et al. Functional neuroimaging strategy in temporal lobe epilepsy: a comparative study of 18FDG-PET and 99mTc-HMPAO- SPECT. Ann Neurol 1992;31:650-6.

314. Jack CR, Jr., Mullan BP, Sharbrough FW, et al. Intractable nonlesional epilepsy of temporal lobe origin: lateralization by interictal SPECT versus MRI. Neurology 1994;44:829-36.

315. Grunwald F, Durwen HF, Bockisch A, et al. Technetium-99m-HMPAO brain SPECT in medically intractable temporal lobe epilepsy: a postoperative evaluation. J NucI Med 1991;32:388-94.

316. Hara M, Takahashi M, Kojima A, et al. Single photon emission computed tomography in children with idiopathic seizures. Radiat Med 1991;9:185-9.

317. Verhoeff NP, Weinstein HC, Aldenkamp AP, et al. Focus localization in patients with partial epilepsy with 99Tcm-HMPAO SPECT under continuous surface EEG monitoring. NucI Med Commun 1992;13:127-36. 237

318. Launes J, livanainen M, Salmi T, et al. Interictal brain 99Tc-HM-PAO SPECT hypoperfusion in patients with unstable partial epilepsy and normal CT. Acta Neurol Scand 1992;86:558-62.

319. Janicek MJ, O'Tuama LA, Treves ST. Quantitative evaluation of 99mTc- hexamethylpropylenamineoxime brain SPECT in childhood-onset epilepsy. J NucI Biol Med 1992;36:319-23.

320. Franceschi M, Messa C, Ferini Strambi L, et al. SPET imaging of cerebral perfusion in patients with non-refractory temporal lobe epilepsy. Acta Neurol Scand 1993;87:268-74.

321. Mitsuyoshi I, Tamaki K, Okuno T, et al. Regional cerebral blood flow in diagnosis of childhood onset partial epilepsy. Brain Dev 1993;15:97-102.

322. Heiskala H, Launes J, Pihko H, Nikkinen P, Santavuori P. Brain perfusion SPECT in children with frequent fits. Brain Dev 1993;15:214-8.

323. Newton MR, Berkovic SF, Austin MC, et al. SPECT in the localisation of extratemporal and temporal seizure foci. J Neurol Neurosurg Psychiatry 1995;59:26-30.

324. Marks DA, Katz A, Floffer P, Spencer SS. Localization of extratemporal epileptic foci during ictal single photon emission computed tomography. Ann Neurol 1992;31:250-5.

325. Harvey AS, Hopkins IJ, Bowe JM, et al. Frontal lobe epilepsy: clinical seizure characteristics and localization with ictal 99mTc-HMPAO SPECT. Neurology 1993;43:1966-80.

326. Ho SS, Berkovic SF, Newton MR, et al. Parietal lobe epilepsy: clinical features and seizure localization by ictal SPECT. Neurology. 1994;44:2277-84.

327. Zubal IG, Spencer SS, Imam K, et al. Difference images calculated from ictal and interictal technetium-99m-HMPAO SPECT scans of epilepsy. J NucI Med 1995;36:684-9.

328. Henkes H, Hosten N, Cordes M, Neumann K, Hansen ML. Increased rCBF in gray matter heterotopias detected by SPECT using 99mTc hexamethyl-propylenamine 238

oxime. Neuroradiology 1991;33:310-2.

329. Ryding E, Rosen I, Elmqvist D, ingvar DH. SPECT measurements with 99mTc-HM- PAO in focal epilepsy. J Cereb Blood Flow Metab 1988;8:395-1 GO.

330. Chiron C, Raynaud C, Jambaque I, et al. A serial study of regional cerebral blood flow before and after hemispherectomy in a child. Epilepsy Res 1991;8:232-40.

331. Matsuda H, Onuma T, Yagishita A. Brain SPECT imaging for laminar heterotopia. J NucI Med 1995;36:238-40.

332. Otsubo H, Hwang PA, Jay V, et al. Focal cortical dysplasia in children with localization-related epilepsy: EEG, MRI, and SPECT findings. Pediatr Neurol 1993;9:101-7.

333. Kuzniecky R, Mountz JM, Wheatley G, Morawetz R. Ictal single-photon emission computed tomography demonstrates localized epileptogenesis in cortical dysplasia. Ann Neurol 1993;34:627-31.

334. Noll K, Deisenhammer E, Dauth J, Carmann H, Schubiger PA. Imaging benzodiazepine receptors in the human brain by single photon emission computed tomography (SPECT). Int J Rad AppI Instrum B 1989;16:759-63.

335. Schubiger PA, Hasler PH, Beer Wohlfahrt H, et al. Evaluation of a multicentre study with lomazenil-a benzodiazepine receptor ligand. NucI Med Commun 1991;12:569-82.

336. van Huffelen AC, van Isselt JW, van Veelen CW, et al. Identification of the side of epileptic focus with 1231-lomazenil SPECT. A comparison with 18FDG-PET and ictal EEG findings in patients with medically intractable complex partial seizures. Acta Neurochir SuppI Wien 1990;50:95-9.

337. Bartenstein P, Ludolph A, Schober 0, et al. Benzodiazepine receptors and cerebral blood flow in partial epilepsy. Eur J NucI Med 1991 ;18:111-8.

338. Cordes M, Henkes H, FerstI F, et al. Evaluation of focal epilepsy: a SPECT scanning comparison of 123-l-iomazenil versus HM-PAO. AJNR Am J Neuroradiol 1992;13:249-53. 239

339. Duncan S, Gillen GJ, Brodie MJ. Lack of effect of concomitant clobazam on interictal 1231-iomazenil SPECT. Epilepsy Res 1993;15:61-6.

340. Bartenstein P, Lehmenkuhler C, Sciuk J, Schuierer G. [Cortical dysplasia as an epileptogenic focus: reduced binding of 1231-iomezanil with barely perceptible 99mTc-HMPAO SPECT]. Nuklearmedizin 1992;31:142-4.

341. Jibiki I, Kurokawa K, Matsuda H, et al. Comparison of benzodiazepine receptor and regional cerebral blood flow imaging of epileptiform foci in hippocampal kindled rabbits: a preliminary report. Pharmacol Biochem Behav 1994;48:553-6.

342. al Tikriti MS, Dey HM, Zoghbi SS, et al. Dual-isotope autoradiographic measurement of regional blood flow and benzodiazepine receptor availability following unilateral middle cerebral artery occlusion. Eur J NucI Med 1994;21:196-202.

343. Abi Dargham A, Laruelle M, Seibyl J, et al. SPECT measurement of benzodiazepine receptors in human brain with iodine-123-iomazenil: kinetic and equilibrium paradigms. J NucI Med 1994;35:228-38.

344. Laruelle M, Abi Dargham A, al Tikriti MS, et al. SPECT quantification of [123l]iomazenil binding to benzodiazepine receptors in nonhuman primates: II. Equilibrium analysis of constant infusion experiments and correlation with in vitro parameters. J Cereb Blood Flow Metab 1994;14:453-65.

345. Laruelle M, Baldwin RM, Rattner Z, et al. SPECT quantification of [123l]iomazenil binding to benzodiazepine receptors in nonhuman primates: I. Kinetic modeling of single bolus experiments. J Cereb Blood Flow Metab 1994;14:439-52.

346. Venz S, Cordes M, Schmitz B, et al. [1231-iomazenil- and 99mTc-HMPAO in the diagnosis of focal epilepsies: a comparison of untreated and treated patients]. Nuklearmedizin 1994;33:1-7.

347. Sjoholm H, Rosen I, Elmqvist D. Role of 1-123-iomazenil SPECT imaging in drug resistant epilepsy with complex partial seizures. Acta Neurol Scand 1995;92:41- 8 .

348. Cepeda C, Menini C, Naquet R, et al. Positron emission tomography in a case of 240

experimental focal epilepsy in the baboon. Electroencephalogr Clin Neurophysiol 1982;54:87-90.

349. Franceschi M, Lucignani G, Del Sole A, et al. Increased interictal cerebral glucose metabolism in a cortical-subcortical network in drug naive patients with cryptogenic temporal lobe epilepsy. J Neurol Neurosurg Psychiatry 1995;59:427-31.

350. Kuhl DE, Engel J, Phelps ME, Selin C. Epileptic patterns of local cerebral metabolism and perfusion in humans determined by emission computed tomography of 18FDG and 13NH3. Ann Neurol 1980;8:348-60.

351. Engel J, Kuhl DE, Phelps ME, Crandall PH. Comparative localization of epileptic foci in partial epilepsy by PET and EEG. Ann Neurol 1982; 12:529-37.

352. Engel J, Brown WG, Kuhl DE, et al. Pathological findings underlying focal temporal hypometabolism in partial epilepsy. Ann Neurol 1982;12:518-28.

353. Engel J, Kuhl DE, Phelps ME, Mazziotta JC. Interictal cerebral glucose metabolism in partial epilepsy and its relation to EEG changes. Ann Neurol 1982;12:510-7.

354. Theodore WH, Newmark ME, Sato S, et al. 18F-fluorodeoxyglucose positron emission tomography in refractory complex partial seizures. Ann Neurol 1983;14:429-37.

355. Abou Khalil BW, Siegel GJ, Sackellares JC, et al. Positron emission tomography studies of cerebral glucose metabolism in chronic partial epilepsy. Ann Neurol 1987;22:480-6.

356. Franck G, Sadzot B, Salmon E, et al. Regional cerebral blood flow and metabolic rates in human focal epilepsy and status epilepticus. Advances in Neurology 1986;44:935-48.

357. Ryvlin P, Cinotti L, Froment JC, et al. Metabolic patterns associated with non­ specific magnetic resonance imaging abnormalities in temporal lobe epilepsy. Brain 1991 ;1 14:2363-83.

358. Sadzot B, Debets RM, Maquet P, et al. Regional brain glucose metabolism in patients with complex partial seizures investigated by intracranial EEG. Epilepsy 241

Res 1992;12:121-9.

359. Olson DM, Chugani HT, Shewmon DA, Phelps ME, Peacock WJ. Electrocorticographic confirmation of focal positron emission tomographic abnormalities in children with intractable epilepsy. Epilepsia 1990;31:731-9.

360. Henry TR, Mazziotta JC, Engel J, Jr., et al. Quantifying interictal metabolic activity in human temporal lobe epilepsy. J Cereb Blood Flow Metab 1990;10:748-57.

361. Sperling MR, Gur RC, Alavi A, et al. Subcortical metabolic alterations in partial epilepsy. Epilepsia 1990;31:145-55.

362. Swartz BE, Halgren E, Delgado Escueta AV, et al. Neuroimaging in patients with seizures of probable frontal lobe origin. Epilepsia 1989;30:547-58.

363. Radtke RA, Hanson MW, Hoffman JM, et al. Temporal lobe hypometabolism on PET: predictor of seizure control after temporal lobectomy. Neurology 1993;43:1088-92.

364. Chee MW, Morris HHd, Antar MA, et al. Presurgical evaluation of temporal lobe epilepsy using interictal temporal spikes and positron emission tomography. Arch Neurol 1993;50:45-8.

365. Debets RM, van Veelen CW, Maquet P, et al. Quantitative analysis of 18FDG-PET in the presurgical evaluation of patients suffering from refractory partial epilepsy. Comparison with CT, MRI, and combined subdural and depth. EEG. Acta Neurochir SuppI Wien 1990;50:88-94.

366. Engel J, Jr., Henry TR, Risinger MW, et al. Presurgical evaluation for partial epilepsy: relative contributions of chronic depth-electrode recordings versus FDG-PET and scalp-sphenoidal ictal EEG. Neurology 1990;40:1670-7.

367. Henry TR, Sutherling WW, Engel J, Jr., et al. Interictal cerebral metabolism in partial epilepsies of neocortical origin. Epilepsy Res 1991;10:174-82.

368. Gaillard WD, Bhatia S, Bookheimer SY, et al. FDG-PET and volumetric MRI in the evaluation of patients with partial epilepsy. Neurology 1995;45:123-6. 242

369. Henry TR, Babb TL, Engel J, et al. Hippocampal neuronal loss and regional hypometabolism in temporal lobe epilepsy. Ann Neurol 1994;36:925-7.

370. Semah F, Baulac M, Hasboun D, et al. Is interictal temporal hypometabolism related to mesial temporal sclerosis? A positron emission tomography/magnetic resonance imaging confrontation. Epilepsia 1995;36:447-56.

371. Theodore WH, Fishbein D, Dietz M, Baldwin P, Deitz M. Complex partial seizures: cerebellar metabolism. Epilepsia 1987;28:319-23.

372. Pawlik G, Fink GR, Kessler J, Heiss WD. PET and functional testing in temporal lobe epilepsy. Acta Neurol Scand 1994;152:150-6.

373. Chugani HT, Mazziotta JC, Engel J, Jr., Phelps ME. The Lennox-Gastaut syndrome: metabolic subtypes determined by 2-deoxy-2[18F]fluoro-D-glucose positron emission tomography. Ann Neurol 1987;21:4-13.

374. Chugani HT, Shewmon DA, Peacock WJ, et al. Surgical treatment of intractable neonatal-onset seizures: the role of positron emission tomography. Neurology 1988;38:1178-88.

375. Chugani HT, Shields WD, Shewmon DA, et al. Infantile spasms: I. PET identifies focal cortical dysgenesis in cryptogenic cases for surgical treatment [see comments]. Ann Neurol 1990;27:406-13.

376. Hajek M, Antonini A, Leenders KL, Wieser HG. Mesiobasal versus lateral temporal lobe epilepsy: metabolic differences in the temporal lobe shown by interictal 18F-FDG positron emission tomography. Neurology 1993;43:79-86.

377. Holmes MD, Kelly K, Theodore WH. Complex partial seizures. Correlation of clinical and metabolic features. Arch Neurol 1988;45:1191-3.

378. Rausch R, Henry TR, Ary CM, Engel J, Jr., Mazziotta J. Asymmetric interictal glucose hypometabolism and cognitive performance in epileptic patients. Arch Neurol 1994;51:139-44.

379. Hajek M, Wieser HG, Khan N, et al. Preoperative and postoperative glucose consumption in mesiobasal and lateral temporal lobe epilepsy. Neurology 1994;44:2125-32. 243

380. Miura K, Watanabe K, Maeda N, et al. Magnetic resonance imaging and positron emission tomography of band heterotopia. Brain Dev 1993;15:288-90.

381. De Voider AG, Gadisseux JF, Michel CJ, et al. Brain glucose utilization in band heterotopia: synaptic activity of "double cortex". Pediatr Neurol 1994; 11:290-4.

382. Bairamian D, Di Chiro G, Theodore WH, et al. MR imaging and positron emission tomography of cortical heterotopia. J Comput Assist Tomogr 1985;9:1137-9.

383. Falconer J, Wada JA, Martin W, Li D. PET, CT, and MRI imaging of neuronal migration anomalies in epileptic patients. Can J Neurol Sci 1990;17:35-9.

384. Lee N, Radtke RA, Gray L, et al. Neuronal migration disorders: positron emission tomography correlations. Ann Neurol 1994;35:290-7.

385. Rintahaka PJ, Chugani HT, Messa C, Phelps ME. Hemimegalencephaly: evaluation with positron emission tomography. Pediatr Neurol 1993;9:21-8.

386. Theodore WH, Holmes MD, Dorwart RH, et al. Complex partial seizures: cerebral structure and cerebral function. Epilepsia 1986;27:576-82.

387. Calabrese P, Fink GR, Markowitsch HJ, et al. Left hemisheric neuronal heterotopia: a PET, MRI, EEG and neuropsychological investigation of a university student. Neurology 1994;44:302-5.

388. Chugani HT, Rintahaka PJ, Shewmon DA. Ictal patterns of cerebral glucose utilization in children with epilepsy. Epilepsia 1994;35:813-22.

389. Hajek M, Antonini A, Leenders KL, Wieser HG. Epilepsia partialis continua studied by PET. Epilepsy Res 1991;9:44-8.

390. Leiderman DB, Albert P, Balish M, Bromfield E, Theodore WH. The dynamics of metabolic change following seizures as measured by positron emission tomography with fluoro-deoxyglucose FI 8. Arch Neurol 1994;51:932-6.

391. Bajorek JG, Lee RL, Lomax P. Neuropeptides: anticonvulsant and convulsant mechanisms in epileptic model systems and and in humans. Adv Neurol 1986;44:489-500. 244

392. Frost JJ, Mayberg HS, Fisher RS, et al. Mu-opiate receptors measured by positron emission tomography are increased in temporal lobe epilepsy. Ann Neurol 1988;23:231-7.

393. Mayberg HS, Sadzot B, Meltzer CC, et al. Quantification of mu and non-mu opiate receptors in temporal lobe epilepsy using positron emission tomography. Ann Neurol 1991;30:3-11.

394. Bartenstein PA, Prevett MC, Duncan JS, Hajek M, Wieser HG. Quantification of opiate receptors in two patients with mesiobasal temporal lobe epilepsy, before and after selective amygdalohippocampectomy, using positron emission tomography. Epilepsy Re 1994; 18:119-25.

395. Theodore WH, Carson RE, Andreasen P, et al. PET imaging of opiate receptor binding in human epilepsy using [1 8F]cyclofoxy. Epilepsy Res 1992;13:129-39.

396. Kumlein E, Bergstrom M, Lilja A, et al. Positron emission tomography with [1 ICJdeuterium deprenyl in temporal lobe epilepsy. Epilepsia 1995;36:712-21.

397. linuma K, Yokoyama H, Otsuki T, et al. Histamine HI receptors in complex partial seizures. Lancet 1993;341:238.

398. Mossaddeghi M, Burke TF, Moerschbaeker JM. Effects of acute administration of alcohol on in vivo binding of [3H]-Ro15-1788 to mouse brain. J Pharmacol Exper Ther 1992;263:639-45.

399. Farde L, Pauli S, Litton JE, et al. PET-determination of benzodiazepine receptor binding in studies on alcoholism. Exs 1994;71:143-53.

400. Pauli S, Liljequist S, Farde L, et al. PET analysis of alcohol interaction with the brain disposition of [1 ICjflumazenil. Psychopharmacology Berl 1992;107:180- 5.

401. Litton JE, Neiman J, Pauli S, et al. PET analysis of [1 ICjflumazenil binding to benzodiazepine receptors in chronic alcohol-dependent men and healthy controls. Psychiatry Res 1993;50:1-13.

402. Chavoix C, Samson Y, Pappata S, et al. Positron emission tomography study of 245

brain benzodiazepine receptors in Friedreich's ataxia. Can J Neurol Sci 1990;17:404-9.

403. Holthoff VA, Koeppe RA, Frey KA, et ai. Positron emission tomography measures of benzodiazepine receptors in Huntington's disease. Ann Neurol 1993;34:76-81.

404. Lloyd C, Abrahams S, Brooks D, Leigh N. Cl 1-flumazenil PET studies in motor neurone disease. Neurology 1995;45:508P.

405. Chen R, Rothstein JD, Frey KA, Tevaarwerk GJ, Hahn AF. Idiopathic Recurring Stupor: A New Syndrome? Neurology 1995;45:65P.

406. Savic I, Persson A, Roland P, et al. In-vivo demonstration of reduced benzodiazepine receptor binding in human epileptic foci. Lancet 1988;2:863-6.

407. Savic I, Ingvar M, Stone Blander S. Comparison of [1 1C]flumazenil and [18FJFDG as PET markers of epileptic foci. J Neurol Neurosurg Psychiatry 1993;56:615- 21.

408. Sadzot B, Debets RM, Delfiore G, et al. [1 1C]Flumazenil positron emission tomography: an in vivo marker of neuronal loss in temporal lobe epilepsy. Epilepsia 1994;35:29.

409. Henry TR, Frey KA, Sackellares JC, et al. In vivo cerebral metabolism and central benzodiazepine-receptor binding in temporal lobe epilepsy. Neurology 1993;43:1998-2006.

410. Koepp MJ, Richardson MP, Brooks DJ, et al. Cerebral benzodiazepine receptors in hippocampal sclerosis: an objective in vivo analysis. Brain 1996;119:1677-87.

411. Koepp MJ, Labbé C, Richardson MP, et al. Regional hippocampal [1 ICjflumazenil PET in temporal lobe epilepsy with unilateral and bilateral hippocampal sclerosis. Brain 1 99 7;120:186 5-7 6.

412. Burdette DE, Sakurai SY, Henry TR, et al. Temporal lobe central benzodiazepine binding in unilateral mesial temporal lobe epilepsy. Neurology 1995;45:934-41.

413. Savic I, Thorell JO, Roland P. [1 ICJFIumazenil positron emission tomography visualizes frontal epileptogenic regions. Epilepsia 1995;36:1225-32. 246

414. Savic I, Svanborg E, Thorell JO, Cortical benzodiazepine receptor changes are related to frequency of partial seizures: a positron emission tomography study. Epilepsia 1996;37:236-44.

415. Liu Z, Vergnes M, Depaulis A, Marescaux C. Involvement of intrathalamic GABA-B transmission in the control of absence seizures in the rat. Neuroscience 1992;48:87-93.

416. Prevett MC, Lammertsma AA, Brooks DJ, et al. Benzodiazepine-GABAA receptors in idiopathic generalised epilepsy measured with [1 ICjflumazenil and positron emission tomography. Epilepsia 1995;36:113-21.

417. Savic I, Widen L, Thorell JO, et al. Cortical benzodiazepine receptor binding in patients with generalized and partial epilepsy. Epilepsia 1990;31:724-30.

418. Savic I, Pauli S, Thorell JO, Biomqvist G. In vivo demonstration of altered benzodiazepine receptor density in patients with generalised epilepsy. J Neurol Neurosurg Psychiatry 1994;57:797-804.

419. Liu Z, Vergnes M, Depaulis A, Marescaux C. Evidence for a critical role of GABAergic transmission within the thalamus in the genesis and control of absence seizures in the rat. Brain Res 1991;545:1-7.

420. Prevett MC, Lammertsma AA, Brooks DJ, et al. Benzodiazepine-GABAA receptor binding during absence seizures. Epilepsia 1995;36:592-9.

421. Duchowny M, Jayakar P, Harvey AS, et al. Language cortex representation: effects of developmental versus acquired pathology. Ann Neurol 1996;40:31-8.

422. Raymond AA, Jones SJ, Fish DR, Stevens JM. Somatosensory evoked potentials in adults with cortical dysgenesis and epilepsy. Electroenceph Clin Neurophysiol 1994;91:141P.

423. Pike VW, Halldin C, Crouzel C, et al. Radioligands for PET studies of central benzodiazepine receptors and PK (peripheral benzodiazepine) binding sites- current status. NucI Med Biol 1993;20:503-25.

424. Townsend DW, Defrise M, Geissbuhler A, et al. Normalisation and reconstruction 247

of PET data acquired by a multi-ring camera with septa retracted. Med Prog Techno! 1991;17:223-8.

425. Bailey DL. 3D acquisition and reconstruction in positron emission tomography. Ann Nuc Med 1992;6:121-30.

426. Bailey DL, Meikle SR. A convolution-subtraction scatter correction method for 3D PET. Phys Med Biol 1994;39:411-24.

427. Grootoonk S, Spinks TJ, Bloomfield PM, Sashin D, Jones T. The practical implementation and accuracy of dual window sactter corection in a neuroPET scanner with the septa retracted. IEEE MEd Imaging Conference Records 1993;5:942-4.

428. Malizia AL, Richardson MP. Benzodiazepine Receptors and positron emission tomography: ten years of experience. A new beginning? J Psychopharmacol 1995;9:335-68.

429. Comar D, Maziere M, Godot JM, et al. Visualisation of 11C flunitrazepam displacement in the brain of the live baboon. Nature 1979;280:329-31.

430. Frost JJ, Wagner HN, Dannals RF, et al. Imaging benzodiazepine receptors in man with l i e suriclone by Positron Emission Tomography. Eur J Pharmacol 1986;122:381-3.

431. Bottlaender M, Brouillet E, Varastet M, et al. In vivo high intrinsic affinity of triazolam: a positron emission tomography study in non human prmates. J Neurochem 1994;62:1102-11.

432. Halldin C, Farde L, Litton JE, Hall H, Sedvall G. [11C]Ro 15-4513, a ligand for visualization of benzodiazepine receptor binding. Preparation, autoradiography and positron emission tomography. Psychopharmacology Berl 1992;108:16-22.

433. Hall H, Litton JE, Halldin C, Kopp J, Sedvall G. Studies on the binding of [3H] flumazenil and [3H] sarmazenil in post mortem human brain. Human Psychopharmacol 1992;7:367-77.

434. Mintun M, Raichle M, Kilbourn M, Wooten F, Welch M. A quantitative method for the in vivo assessment of drug binding sites with positron emission tomography. 248

Ann Neurol 1983;15:217-27.

435. Koeppe RA, Holthoff VA, Frey KA, Kilbourn MR, Kuhl DE. Compartmental analysis of [C-11] flumazenil kinetics for the estimation of ligand transport rate and receptor distribution using positron emission tomography. J Cereb Blood Flow Metab 1991 ;11:735-44.

436. Cunningham VJ, Jones T. Spectral analysis of dynamic PET studies. J Cereb Blood Flow Metab 1993;13:15-23.

437. Price JC, Mayberg HS, Dannals RF, et al. Measurement of benzodiazepine receptro number and affinity in humans using tracer kinetic modeling, Positron Emission Tomography and [11C] flumazenil. J Cereb Blood Flow Metab 1993;13:656-67.

438. Lassen NA, Bartenstein PA, Lammertsma AA, et al. Benzodiazepine receptor quantitation in vivo in man using [1 1C]-flumazenil and PET: Application of the steady state principle. J Cereb Blood Flow Metab 1995; 15:152-65.

439. Persson A, Ehrin E, Eriksson L, et al. Imaging of [1 1C]-labelled Ro 1 5-1788 binding to benzodiazepine receptors in the human brain by positron emission tomography. J Psychiatr Res 1985;19:609-22.

440. Samson Y, Hantraye P, Baron JC, et al. Kinetics and displacement of [C-11] Ro 15-1788. A benzodiazepine antagonist studied in human brain in vivo by Positron Emission Tomography. Eur J Pharmacol 1985;110:247-51.

441. Persson A, Pauli S, Halldin C, et al. Saturation analysis of specific 11C Ro 15- 1788 binding to the human neocortex using positron emission tomography. Human Psychopharm 1989;4:21-31.

442. Kawai R, Carson RE, Dunn B, et al. Regional brain measurement of BMax and Kd with the opiate antagonist cyclofoxy: equilibrium studies in the conscious rat. J Cereb Blood Flow Metab 1991 ;11:529-44.

443. Delforge J, Pappata S, Mileet P, et al. Quantification of benzodiazepine receptors in human brain using PET, [C-11] flumazenil, and a single experimental protocol. J Cereb Blood Flow Metab 1995;15:284-300.

444. Litton JE, Hall H, Pauli S. Saturation analysis in PET—analysis of errors due to 249

imperfect reference regions. J Cereb Blood Flow Metab 1994;14:358-61.

445. Pappata S, Samson Y, Chavoix C, et al. Regional specific binding of [11C]R0 15 1788 to central type benzodiazepine receptors in human brain: quantitative evaluation by PET. J Cereb Blood Flow Metab 1988;8:304-13.

446. lyo M, Itoh T, Yamasaki T, et al. Quantitative in vivo analysis of benzodiazepine binding sites in the human brain using positron emission tomography. Neuropharmacology 1991 ;30:207-15.

447. Abadie P, Baron JC, Bisserbe JC, et al. Central benzodiazepine receptors in human brain: estimation of regional Bmax and KD values with positron emission tomography. Bur J Pharmacol 1992;213:107-15.

448. Blomqvist G, Pauli S, Farde L, et al. Maps of receptor binding parameters in the human brain-a kinetic analysis of PET measurements. Eur J NucI Med 1990;16:257-65.

449. Delforge J, Syrota A, Bottlaender M, et al. Modeling analysis of [1 1C]flumazenil kinetics studied by PET: application to a critical study of the equilibrium approaches. J Cereb Blood Flow Metab 1993; 13:454-68.

450. Muller Gartner HW, Links JM, Prince JL, et al. Measurement of radiotracer concentration in brain gray matter using positron emission tomography: MRI- based correction for partial volume effects. J Cereb Blood Flow Metab 1992;12:571-83.

451. Rousset OG, Ma Y, Léger GC, Gjedde AH, Evans AC. Correction for partial volume effects in PET using MRI-based 3D simulations of individual human brain metabolism. In: Uemura K, ed. Quantification of brain function, tracer kinetics and image analysis in brain PET. Elsevier Science Publishers B. V., 1993:113- 25.

452. Rousset OG, Ma Y, Marenco S, Wong DF, Evans AC. In vivo correction for partial volume effects in PET: accuracy and precision. Neuroimage 1995;2:S33.

453. Frost JJ, Meltzer CC, Zubieta JK, et al. MR-based correction of brain PET measurements for heterogeneous gray matter radioactivity distribution. Neuroimage 1995;2:S32. 250

454. Labbe C, Froment JC, Kennedy A, Ashburner J, Cinotti L. Positron emission tomography metabolic data corrected for cortical atrophy using magnetic resonance imaging. Alzheimer Dis Assoc Dis 1996;10:141-70.

455. Evans AC, Marrett S, Torrescorzo J, Ku S, Collins L. MRI-PET correlation in three dimensions using a volume-of-interest (VOI) atlas. J Cereb Blood Flow Metab 1991;1 l:A69-78.

456. Greitz T, Bohm C, Hoke S, Eriksson L. A computerised brain atlas: construction, anatomical content, and some applications. J Comput Assist Tomogr 1991 ;15:26- 38.

457. Friston KJ, Holmes AP, Worsley KJ, et al. Statistical parametric maps in functional imaging: A general linear approach. Human Brain Mapping 1995;2:189-210.

458. Friston KJ, Ashburner J, Poline JB, et al. Spatial registration and normalization of images. Human Brain Mapping 1995;2:165-89.

459. Friston KJ, Worsley KJ, Frackowiak RSJ, Mazziotta JC, Evans AC. Assessing the significance of focal activations using their spatial extent. Human Brain Mapping 1994;1:214-20.

460. Spinks TJ, Jones T, Bailey DL, et al. Physical performance of a positron tomograph for brain imaging with retractable septa. Phys Med Biol 1992;37:1637-55.

461. Frackowiak RSJ. Functional mapping of verbal memory and language. Trends Neurosci 1994;17:109-15.

462. Raichle ME, Martin WRW, Herscovitch P, Mintun MA, Markham J. Brain blood flow measured with intravenous H2150. II. Implemetation and validation. J NucI Med 1983;24:790-8.

463. Lammertsma AA, Cunningham VJ, Deiber MP. Combination of integral and dynamic methods for generating reproducible functional CBF images. J Cereb Blood Flow Metab 1990;10:675-86. 251

464. Fox PT, Mintun MA, Raichle ME, Herscovitch P. A non-invasive approach to human quantitative brain mapping with H2150 and positron emission tomography. J NucI Med 1984;30:141-9.

465. Fox PT, Mintun MA. Non-invasive functional brain mapping by change- distribution analysis of averaged PET images of H2150 tissue activity. J NucI Med 1989;30:141-9.

466. Maziere M, Hantraye P, Prenant C, Sastre J, Comar D. Synthesis of an ethyl 8- f I uor o- 5 , 6- di hydr o- 5- [ 1 1 C] methyl - 6- oxo- 4 H- i mi dazo[ 1 ,5 - a][1,4]benzodiazepine-3-carboxylate (RO 15.1788-11C): A specific radioligand for the in vivo study of central benzodiazepine receptors by positron emission tomography. Int J AppI Radiat Isot 1984;35:973-6.

467. Ranicar ASO, Williams CW, Schnorr L, et al. The on-line monitoring of continuously withdrawn arterial blood during PET studies using a single BGO/photomultiplier assembly and non-stick tubing. Medical Progress Through Technology 1991;17:259-64.

468. Luthra SK, Osman S, Turton DR, et al. An automated system based on solid phase extraction and HPLC for the routine determination in plasma of unchanged [1 1C]- L-deprenyl,; [1 1C]diprenorphine; [1 1C]raclopride; and [1 ICfScherring 23390. J Label Compds and Radiopharm 1993;32:518-20.

469. Lammertsma AA, Bench CJ, Price GW, al. e. Measurement of cerebral monoamine oxidase B activity using L-[11C]deprenyl and dynamic positron emission tomography. J Cereb Blood Flow Metab 1991 ;11:545-56.

470. Lammertsma AA, Lassen NA, Prevett MC, et al. Quantification of benzodiazepine receptors in vivo using [1 1C]flumazenil: application of the steady state principle. In: Uemura K, ed. Quantification of brain function. Tracer kinetic and image analysis in brain PET. Elsevier Science Publishers, 1993:

471. Robb RA, Hanson DP. ANALYZE: A software system for biomedical image analysis. First Conference on Visualisation in Biomedical Computing. Atlanta, 1990:507- 18 .

472. Cunningham VJ, Ashburner J, Byrne H, Jones T. Use of spectral analysis to obtain parametric images from dynamic PET studies. In: Uemura K, ed. Quantification of 252

brain function. Tracer kinetic and image analysis in brain PET. Elsevier Science Publishers, 1993:101-11.

473. Coull JT, Frith CD, Frackowiak RSJ, Grasby PM. A fronto-parietal network for rapid visual information processing: a PET study of sustained attention and working memory. Neuropsychologia 1996;34:1085-95.

474. Woods RP, Mazziotta JC, Cherry SR. MRI-PET registration with automated algorithm. J Comput Assist Tomogr 1993;17:536-46.

475. Richardson MP, Koepp MJ, Fish DR, Brooks DJ, Duncan JS. Benzodiazepine receptors in focal epilepsy with cortical dysgenesis: an 11C-flumazenil PET study. Ann Neurol 1996;40:188-98.

476. Friston KJ, Frith CD, Liddle PF, et al. The relationship between global and local changes in PET scans. J Cereb Blood Flow Metab 1990;10:458-66.

477. Koepp MJ, Richardson MP, Brooks DJ, Fish DR, Duncan JS. Benzodianzepine Receptor Density in Mesial Temporal Lobe Epilepsy: Comparison of Volumetric MRI and 3D Flumazenil PET. Neurology 1995;45:A403.

478. Henry TR, Mazziotta JC, Engel J. Interictal metabolic anatomy of mesial temporal lobe epilepsy. Arch Neurol 1993;50:582-9.

479. Sloper JJ, Johnson P, Powell TPS. Selective degeneration of interneurones in the motor cortex of infant monkeys following controlled hypoxia: a possible cause of epilepsy. Brain Res 1980;198:204-9.

480. Loscher W, Schwark WS. Evaluation of different GABA receptor agonists in the kindled amygdala seizure model in rats. Exp Neurol 1985;89:454-60.

481. Cascino GD. Commentary: how has neuroimaging improved patient care? Epilepsia 1994;35:S103-7.

482. Spencer SS. MRI and epilepsy surgery. Neurology 1995;45:1248-50.

483. Zentner J, Hufnagel A, Wolf HK, et al. Surgical treatment of temporal lobe epilepsy: clinical, radiological and histopathological findings in 178 patients. J Neurol Neurosurg Psychiat 1995;58:666-73. 253

484. Williamson PD, Thadani VM, Darcey TM, et al. Occipital lobe epilepsy: clinical characteristics, seizure spread patterns, and results of surgery. Ann Neurol 1992;31:3-13.

485. Cascino GD, Hulihan JF, Sharbrough FW, Kelly PJ. Parietal lobe lesional epilepsy: electroclinical correlation and operative outcome. Epilepsia 1993;34:522-7.

486. Montes JL, Rosenblatt B, Farmer JP, et al. Lesionectomy of MRI detected lesions in children with epilepsy. Pediatr Neurosurg 1995;22:167-73.

487. Duncan JS. Imaging and epilepsy. Brain 1997;120:376-89.

488. Cascino GD, Jack CR, Parisi JE, et al. MRI in the presurgical evaluation of patients with frontal lobe epilepsy and children with temporal lobe epilepsy: pathologic correlation and prognostic importance. Epilepsy Res 1992;11:51-9.

489. Lorenzo NY, Parisi JE, Cascino GD, et al. Intractable frontal lobe epilepsy: pathological and MRI features. Epilepsy Res 1995;20:171-8.

490. Richardson MP, Koepp MJ, Brooks DJ, Duncan JS. [1 1C]-flumazenil PET reveals central benzodiazepine receptor binding abnormalities in partial epilepsy with normal MRI. Neurology 1996;46:A 122-3.

491. Kumlien E, Hilton Brown P, Spannare B, Gillberg PG. In vitro quantitative autoradiography of [3H]-L-deprenyl and [3H]-PK 11195 binding sites in human epileptic hippocampus. Epilepsia 1992;33:610-7.

492. Prevett MC, Cunnungham VJ, Brooks DJ, Fish DR, Duncan JS. Opiate receptors in idiopathic generalised epilepsy measured with [1 1C]-diprenorphine and PET. Epilepsy Res 1994;19:71-7.

493. Brooks DJ. Functional imaging in relation to Parkinsonian syndromes. J Neurol Sci 1993;115:1-17.

494. Turjanski N, Weeks R, Dolan R, Harding AE, Brooks DJ. Striatal DI and D2 receptor binding in patients with Huntington's disease and other choreas. Brain 1995;118:689-96. 254

495. Weeks RA, Piccini P, Harding AE, Brooks DJ. Striatal DI and D2 dopamine receptor loss in asymptomatic carriers of Huntington's Disease. Ann Neurol 1996;40:49-54.

496. Wright 1C, McGuire PK, Poline JB, et al. A voxel-based method for the statistical analysis of gray and white matter density applied to schizophrenia. Neuroimage 1995;2:244-52.

497. Poline J-B, Worsley KJ, Holmes AP, Frackowiak RSJ, Friston KJ. Estimating smoothness in statistical parametric maps: variability of p-values. J Comp Ass Tomog 1995;19:788-96.

498. Jahanshahi M, Jenkins IH, Brown RG, et al. Self-initiated versus externally triggered movements. I. An investigation using measurement of regional cerebral blood flow with PET and movement-related potentials in normal and Parkinson's disease subjects. Brain 1995;118:913-33.

499. Gale K. Mechanisms of seizure control mediated by gamma-aminobutyric acid: role of the substantia nigra. Fed Proc 1985;44:2414-24.

500. Palmini A, Andermann F, Olivier A, et al. Neuronal migration disorders: a contribution of modern neuroimaging to the etiologic diagnosis of epilepsy. Can J Neurol Sci 1991 ;18:580-7.

501. Raymond AA, Fish DR, Sisodiya SM, et al. Abnormalities of gyration, heterotopias, tuberous sclerosis, focal cortical dysplasia, microdysgenesis, dysembryoplastic neuroepithelial tumour and dysgenesis of the archicortex in epilepsy. Clinical, EEG and neuroimaging features in 100 adult patients. Brain 1995;118:629-60.

502. Jenkins IH, Brooks DJ, Nixon PD, Frackowiak RSJ, Passingham RE. Motor sequence learning: a study with positron emission tomography. J Neurosci 1994;14:3775-90.

503. Watson JDG, Myers R, Frackowiak RSJ, et al. Area V5 of the human brain: evidence from a combined study using positron emission tomography and magnetic resonance imaging. Cerebral Cortex 1993;3:79-94.

504. Richardson MP, Friston KJ, Sisodiya SM, et al. Cortical grey matter and benzodiazepine receptors in malformations of cortical development: a voxel-based 255

comparison of structural and functional imaging data. Brain 1997; 120:1961- 1973.

505. Friston KJ, Frith CD, Liddle PF, Frackowiak RSJ. Functional connectivity: The principal component analysis of large (PET) data sets. J Cereb Blood Flow Metab

1993;13:5-14.

506. Somogyi P, Tamas G, Lujan R, Buhl EH. Salient features of synaptic organisation in the cerebral cortex. Brain Res Rev 1998;26:113-135.

507. McBain CJ, Freund TF, Mody I. Glutamatergic synapses onto hippocampal interneurons: precision timing without plasticity. Trends Neurosci 1999;22:228-235.

508. Hendry SHC, Jones EG. Reduction in number of immunostained GABAergic neurones in deprived-eye dominance columns of monkey area 17. Nature 1986;320:750- 753.

509. Jones EG, Hendry SHC, De Felipe J, Benson DL. GABA neurons and their role in activity-dependent plasticity of adult primate visual cortex. In: Peters A, Rockland KS, eds. Cerebral Cortex 10, 61-140 (Plenum, New York, 1994).

510. Jacobs KM, Donoghue JP. Reshaping the cortical motor map by unmasking latent intracortical connections. Science 1991 ;251944-7.

511. Xerri C, Coq JO, Merzenich MM, Jenkins WM. Experience-induced plasticity of cutaneous maps in the primary somatosensory cortex of adult monkeys and rats. J Physiol (Paris) 19 9 6 ;9 0 :2 7 7 -2 8 7 . 256 Publications

The following are based on material included in this thesis:

Peer-reviewed journal papers:

Richardson MP, Koepp MJ, Brooks DJ, Fish DR, Duncan JS. Benzodiazepine receptors in focal epilepsy with cortical dysgenesis: an ’ ^C-flumazenil PET study. Ann Neurol 1996;40:188-98.

Richardson MP, Friston KJ, Sisodiya SM, Koepp MJ, Ashburner J, Free SL, Brooks DJ, Duncan JS. Cortical grey matter and benzodiazepine receptors in malformations of cortical development: a voxel-based comparison of structural and functional imaging data. Brain 1997: 120; 1961-1973.

Richardson MP, Koepp MJ, Brooks DJ, Coull JT, Grasby P, Fish DR, Duncan JS. Cerebral activation in malformations of cortical development. Brain 1998; in press.

Review:

Malizia AL, Richardson MP. Benzodiazepine Receptors and positron emission tomography: ten years of experience. A new beginning? J Psychopharmacol 1995;9:335-368.

Invited book chapter:

Richardson MP, Koepp MJ, Duncan JS. In vivo neuroreceptor assays in cortical dysplasia using positron emission tomography. In: Dysplasias of cerebral cortex and epilepsy. Guerrini R, Anderman F, et al. Eds. Lippincott-Raven, New York, 1996:181- 190.

Prize-winning abstract:

Richardson MP, Koepp MJ, Brooks DJ, Duncan JS. (C -11 )flumazenil PET reveals central benzodiazepine receptor binding abnormalities in partial epilepsy with normal MRI. Neurology 1996;46:1027.