<%BANNER%>

Characterization of In Vitro Epileptiform Activity and Propagation in the in utero Irradiated Rat Model

Permanent Link: http://ufdc.ufl.edu/UFE0041642/00001

Material Information

Title: Characterization of In Vitro Epileptiform Activity and Propagation in the in utero Irradiated Rat Model
Physical Description: 1 online resource (174 p.)
Language: english
Creator: GRIMES,JOHN NICHOLAS
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: CAUSALITY -- CORTEX -- CORTICAL -- DYSPLASIA -- EPILEPSY -- GRANGER -- ICTAL -- IN -- PROPAGATION -- SEIZURE
Biomedical Engineering -- Dissertations, Academic -- UF
Genre: Biomedical Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Epilepsy, one of the most common serious neurological conditions, is characterized by spontaneous recurrent seizures, the catastrophic synchronization of brain activity. While anti-epileptic drugs are used as the first-line treatment, they fail in approximately 20 ? 30% of patients. Further, this failure rate is dependent on the type of epilepsy and its cause. Malformations of the cortex have been highly correlated with epilepsy, with an estimated 75% of patients having a cortical malformation presenting with epilepsy at some point. Further, cortical malformations are typically less responsive to drug therapies. Although estimated to only exist in 14% of epilepsy patients, cortical malformations account for between 25 ? 40% of all medically intractable childhood epilepsies. While the normally laminated cortex has been studied for over 100 years, only more recently has attention been devoted to the malformed cortex. In order to study the role that cortical malformations play in epileptogenesis, animal models of cortical dysplasia, such as the in utero irradiated rat model, have been employed by researchers. Unfortunately, the bulk of this work has been focused on histological and intrinsic neuron property differences with little consideration given to the emergent network properties in dysplastic slices. In this work I utilized the in utero irradiated rat model to induce cortical dysplasia and compared electrophysiological differences between normal and dysplastic cortex using microelectrode arrays (MEAs). Several metrics were used in the analysis and were categorized as classic, spatial, and novel. Classic metrics were those used by previous researchers to compare epileptiform activity in dysplastic and normally laminated cortices and included ictal event lengths, ictal event distributions, and number of field potentials per ictal event. Spatial metrics were those used by previous researchers to analyze normally laminated cortical slices but have not been applied to dysplastic slices or used to compare the two. New metrics consisted of the application of Granger Causality methods to the data, which have been used previously in neuroscience, but not to compare functional differences between normal and dysplastic tissue. Results supported and expanded upon previous counter-intuitive and anecdotal evidence that low dose irradiated subjects are functionally more dissimilar from controls than high dose irradiated subjects. This was evidenced by key classic and spatial metrics, such as mean number of ictal events per recording time, mean event length, mean inter-event interval, bias in LFP extrema peaks during events, and wave speed propagation through cortical layers. Surprisingly, and of most potential clinical relevance, results from novel application of Granger Causality analysis to in vitro slice suggests that localized areas drive ictal activity when a dominant initiation site is absent and do not coincide with dominant initiation sites when present.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by JOHN NICHOLAS GRIMES.
Thesis: Thesis (Ph.D.)--University of Florida, 2011.
Local: Adviser: Demarse, Thomas.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0041642:00001

Permanent Link: http://ufdc.ufl.edu/UFE0041642/00001

Material Information

Title: Characterization of In Vitro Epileptiform Activity and Propagation in the in utero Irradiated Rat Model
Physical Description: 1 online resource (174 p.)
Language: english
Creator: GRIMES,JOHN NICHOLAS
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: CAUSALITY -- CORTEX -- CORTICAL -- DYSPLASIA -- EPILEPSY -- GRANGER -- ICTAL -- IN -- PROPAGATION -- SEIZURE
Biomedical Engineering -- Dissertations, Academic -- UF
Genre: Biomedical Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Epilepsy, one of the most common serious neurological conditions, is characterized by spontaneous recurrent seizures, the catastrophic synchronization of brain activity. While anti-epileptic drugs are used as the first-line treatment, they fail in approximately 20 ? 30% of patients. Further, this failure rate is dependent on the type of epilepsy and its cause. Malformations of the cortex have been highly correlated with epilepsy, with an estimated 75% of patients having a cortical malformation presenting with epilepsy at some point. Further, cortical malformations are typically less responsive to drug therapies. Although estimated to only exist in 14% of epilepsy patients, cortical malformations account for between 25 ? 40% of all medically intractable childhood epilepsies. While the normally laminated cortex has been studied for over 100 years, only more recently has attention been devoted to the malformed cortex. In order to study the role that cortical malformations play in epileptogenesis, animal models of cortical dysplasia, such as the in utero irradiated rat model, have been employed by researchers. Unfortunately, the bulk of this work has been focused on histological and intrinsic neuron property differences with little consideration given to the emergent network properties in dysplastic slices. In this work I utilized the in utero irradiated rat model to induce cortical dysplasia and compared electrophysiological differences between normal and dysplastic cortex using microelectrode arrays (MEAs). Several metrics were used in the analysis and were categorized as classic, spatial, and novel. Classic metrics were those used by previous researchers to compare epileptiform activity in dysplastic and normally laminated cortices and included ictal event lengths, ictal event distributions, and number of field potentials per ictal event. Spatial metrics were those used by previous researchers to analyze normally laminated cortical slices but have not been applied to dysplastic slices or used to compare the two. New metrics consisted of the application of Granger Causality methods to the data, which have been used previously in neuroscience, but not to compare functional differences between normal and dysplastic tissue. Results supported and expanded upon previous counter-intuitive and anecdotal evidence that low dose irradiated subjects are functionally more dissimilar from controls than high dose irradiated subjects. This was evidenced by key classic and spatial metrics, such as mean number of ictal events per recording time, mean event length, mean inter-event interval, bias in LFP extrema peaks during events, and wave speed propagation through cortical layers. Surprisingly, and of most potential clinical relevance, results from novel application of Granger Causality analysis to in vitro slice suggests that localized areas drive ictal activity when a dominant initiation site is absent and do not coincide with dominant initiation sites when present.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by JOHN NICHOLAS GRIMES.
Thesis: Thesis (Ph.D.)--University of Florida, 2011.
Local: Adviser: Demarse, Thomas.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0041642:00001


This item has the following downloads:


Full Text

PAGE 1

1 CHARACTERIZATION OF IN VITRO EPILEPTIFORM ACTIVITY AND PROPAGATION IN THE IN UTERO IRRADIATED RAT MODEL By JOHN NICHOLAS GRIMES A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 201 1

PAGE 2

2 201 1 John Nicholas Grimes

PAGE 3

3 To my parents, Tom and LeVinuia Grimes

PAGE 4

4 ACKNOWLEDGMENTS I would like to take a moment to give special thanks to the people and entities that have provided the support necessary for this work to come to fruition. I thank Dr. Tom DeMarse, my committee chair for his inspiration and advice I than k the other members on my committee Dr. Benjamin Keselowsky, Dr. Bill Ogle, and Dr. Eric Laywell for their feedback and guidance. I thank Dr. Huan Xin Chen and Dr. Stev en Roper for providing the tissue on which these experiments were performed. I thank Dr. Roman Garnett for his insights and help with various statistical tests. I give thanks to the J. Crayton Pruitt Family and the University of Florida Alumni Foundation and for their generosity which has enabled m e to pursue my graduate studies. I would also like to thank my close friends and family I give special thanks to Pam Anderson who has motivated and given support during this difficult time. Finally, I give my greatest thanks to my loving parents: my dad, Tom, and my mom LeViniua, who passed away unexpectedly before this work was completed.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................. 4 LIST OF TABLES ............................................................................................................ 8 LIST OF FIGURES ........................................................................................................ 10 LIST OF ABBREVIATI ONS ........................................................................................... 13 ABSTRACT ................................................................................................................... 15 CHAPTER 1 INTRODUCTION AND BACKGROUND ................................................................. 17 Introduction ............................................................................................................. 17 Causes of Epilepsy ........................................................................................... 18 Cortex Anatomy and Physiology ...................................................................... 18 Co rtical Dysplasia and Epilepsy ....................................................................... 20 Models of Epilepsy ........................................................................................... 21 Animal selection ......................................................................................... 22 Epilepsy induction protocols ....................................................................... 23 Animal models of cortical dysplasia ........................................................... 23 The in utero irradiated rat model of cortical dysplasia ................................ 25 in vitro Techniques ........................................................................................... 29 Whole cell patch recording ......................................................................... 29 Optical imaging with voltage and calcium sensitive dyes ........................... 30 Microelectrode arrays ................................................................................ 31 Previous Work and Literature Review ..................................................................... 32 Normally Laminated Slices ............................................................................... 33 Salient anatomical structures and physiological properties ........................ 33 Initiation of epileptiform events ................................................................... 36 Propagation of epileptiform activity ............................................................ 38 Event termination ....................................................................................... 43 In Utero Irradiated Induced Dysplastic Slices ................................................... 45 In vivo EEG ................................................................................................ 45 In vitro epileptiform waveforms .................................................................. 47 Neuron distributions ................................................................................... 48 Functional differences of neurons in radiation induced dysplastic cortex ... 49 Objectives and Central Hypothesis ......................................................................... 51

PAGE 6

6 2 EXPERIMENT AND METRICS ............................................................................... 54 Experimental Protocol ............................................................................................. 54 Animals and Irradiation Protocol ....................................................................... 54 Brain Slice Preparation ..................................................................................... 54 Recording ......................................................................................................... 56 Post Processing and Data Analysis .................................................................. 56 Definition of Metrics ................................................................................................ 57 Verification of Slice Stationarity ........................................................................ 58 Classical and NonSpatial Metrics .................................................................... 61 Ictal events per recording time ................................................................... 61 Average event length ................................................................................. 62 Distribution of event lengths ....................................................................... 62 Av erage burst integral ................................................................................ 62 Average inter event interval ....................................................................... 63 Distribution of inter event intervals ............................................................. 63 Recovery ratio ............................................................................................ 64 Number of local fi eld potentials per epileptic event .................................... 64 Spatial Metrics .................................................................................................. 64 Initiation site ............................................................................................... 65 Ictal wave propagation speeds and pattern ............................................... 65 Termination site ......................................................................................... 66 New Metrics ...................................................................................................... 66 Overview of Granger Causality .................................................................. 66 Pairwise Granger Causality ........................................................................ 67 Conditional Granger Causality ................................................................... 71 3 VERIFICATION OF SLICE STABILITY ................................................................... 78 Control Slice Stability .............................................................................................. 78 Low Dose Slice Stability .......................................................................................... 80 High Dose Slice Stability ......................................................................................... 81 Stability Conclusions ............................................................................................... 82 4 CLASSICA L AND SPATIAL RESULTS AND ANALYSIS ....................................... 98 Results of Classical Analysis Techniques ............................................................... 98 Ictal Events per Recording Time ...................................................................... 98 Average Event Length ...................................................................................... 99 Distribution of Event Lengths ............................................................................ 99 Average Inter Event Interval ........................................................................... 101 Distribution of Inter Event Intervals ................................................................ 101 Recovery Ratio ............................................................................................... 103 Number of Field Potentials per Epileptic Event .............................................. 104 Summary of Classical Analysis ...................................................................... 105 Results of Spatial Analysis Techniques ................................................................ 106

PAGE 7

7 Initiation Site ................................................................................................... 106 Propagation .................................................................................................... 107 Wave speed ............................................................................................. 1 08 Propagation pattern ................................................................................. 109 Termination Site ............................................................................................. 110 Summary of Spatial Metrics ............................................................................ 112 5 GR ANGER CAUSALITY RESULTS AND ANALYSIS .......................................... 131 Results From Pairwise Granger Causality Analysis .............................................. 131 Analysis Methods ........................................................................................... 131 PWGC Mean Values and Distributions ........................................................... 132 Mean PWGC Values by Cell Layer ................................................................ 133 PWGC Changes Over Time ........................................................................... 136 Initiation Site vs. PWGC Source Focus .......................................................... 137 Results From Conditional Granger Causality Analysis .......................................... 138 Analysis Methods ........................................................................................... 138 CGC Mean Values and Distributions .............................................................. 139 Mean CGC Values by Cell Layer .................................................................... 140 CGC Changes Over Time .............................................................................. 141 Initiation Site vs. CGC Source Focus ............................................................. 142 Summary of Granger Causality Analysis .............................................................. 143 6 CONCLUSIONS ................................................................................................... 154 Conclusions .......................................................................................................... 154 Future Work .......................................................................................................... 155 LIST OF REFERENCES ............................................................................................. 157 BIOGRAPHICAL SKETCH .......................................................................................... 174

PAGE 8

8 LIST OF TABLES Table page 4 1 Ictal duration mean, standard deviation, and 95% confidence interval ............. 113 4 2 Inter event interval mean, standard deviation, and 95% confidence interval .... 113 4 3 Recovery ratio mean and standard deviation ................................................... 113 4 4 Positive extrema rate mean and standard deviation ......................................... 113 4 5 Negative extrema rate mean and standard deviation ....................................... 113 4 6 Total extrema rate mean and standard deviation ............................................. 113 4 7 Results of two sample t tests on measured extrema rate means ..................... 114 4 8 Mean horizontal and vertical wave speeds through tissue ............................... 114 4 9 Results of two sample Kolmogorov Smirnov tests on wave propagation speeds between groups ................................................................................... 114 4 10 Mean horizontal propagation speed through cell layers ................................... 114 4 11 Results of intra group twosample Kolmogorov Smirnov tests on horizontal wave propagation speeds between cell layers ................................................. 114 4 12 Results of inter group twosample Kolmogorov Smirnov tests on horizontal wave propagation speeds between cell layers ................................................. 115 5 1 Mean pairwise Granger Causality values of each group .................................. 144 5 2 Results of two sample Kolmogorov Smirnov tests on pairwise Granger Causality distributions between groups ............................................................ 144 5 3 Mean PWGC in each cell layer ......................................................................... 144 5 4 Results of two sample Kolmogorov Smirnov tests on pairwise Granger Causality d istributions between cell layers ....................................................... 144 5 5 Mean conditional Granger Causality values of each group .............................. 145 5 6 Results of two sample Kolmogorov Smirnov tests on conditional Granger Causality values between groups ..................................................................... 145 5 7 Mean CGC in each cell layer ............................................................................ 145

PAGE 9

9 5 8 Results of two sam ple Kolmogorov Smirnov tests on conditional Granger Causality values between cell layers ................................................................ 145

PAGE 10

10 LIST OF FIGURES Figure page 1 1 Data collected in the DeMarse lab from acute rat cortical slices treated with 4 channel 3D MEAs ........................... 52 1 2 An acute rat hippocampal slice with CA1 and the dentate gyrus placed over the electrodes of an Ayanda Biosystems MEA60 200 3D. ................................. 53 2 1 Visual representation of the automated post processing algorithm. Here four seconds from one channel of raw data in which an ictal event occurred is plotted ................................................................................................................. 74 2 2 Concentration of solutes in the recording chamber vs. time. .............................. 75 2 3 Example of an erroneous direct causal connection generated by Pairwise Granger Causality on a network in which a mediated connection exists. ........... 76 2 4 Example of erroneously increasing connection weights generated by PWGC in which a series of mediated connections exists ............................................... 77 3 1 Ictal event number vs. ictal event start time for a normally laminated slice used to verify activity stability. ............................................................................ 84 3 2 Control slice demonstrating three linear phases. ................................................ 85 3 3 Ictal event duration vs. ictal event number. Data are from the same recording used to generate Figure 32, which appeared to have three distinct activity phases .................................................................................................... 86 3 4 Ictal event duration vs. ictal event start time, verifying epileptiform activity was stable over the course of the recording period. ........................................... 87 3 5 Ictal event duration vs. ictal event number for the same normally laminated slice data used in Figure 31. .............................................................................. 88 3 6 Ictal event duration vs. ictal event start time number for the same normally laminated slice data used in Figure 31 and Figure 35. ..................................... 89 3 7 Plot of ictal event number vs. ictal event start time for a control slice subject illustrating stair step pattern. .............................................................................. 90 3 8 Ictal event durations vs. ictal event number for a control subject. Data used are the same used to generate Figure 37. ........................................................ 91 3 9 Ictal event duration vs. ictal event start time for a control subject. Data are from the same subject used to generate Figure 37 and Figure 38 ................... 92

PAGE 11

11 3 10 Plot of ictal event number vs. ictal event start time for a low dose irradiated subject illustrating characteristic pronounced stair step pattern. ........................ 93 3 11 Ictal event durations vs. ictal event number for a low dose irradiated subject. Data used are the same used to generate Figure 310. ..................................... 94 3 12 Ictal event duration vs. ictal event start time for a low dose irradiated subject. Data are from the same subject used to generate Figure 310 and Figure 311. ...................................................................................................................... 95 3 13 Characteristic stability plots for the high dose subject exhibiting a simple linear trend. ......................................................................................................... 96 3 14 Characteristic stability plots for high dose subject exhibiting the stair step tr end. .................................................................................................................. 97 4 1 Histograms of event duration normalized by number of events. For readability, events over 10 s in length have been om itted. ............................... 116 4 2 Log normal fits of the event durations for each group. ..................................... 117 4 3 Histograms of inter event intervals normalized by number of event intervals. For readability, intervals over 50 s in length have been omitted ....................... 118 4 4 Exponential fit of the inter event intervals for each group. ................................ 119 4 5 Recovery scatter plots. The recovery scatter plot is a plot of inter event intervals vs. time. .............................................................................................. 120 4 6 Histograms of recovery ratio for subjects. Recovery ratios larger than 20 were omitted for visual readability. ................................................................... 121 4 7 Histograms of positive extrema rates in control, low dose, and high dose groups. ............................................................................................................. 122 4 8 Histograms of negative extrema rates in control, low dose, and high dose groups. ............................................................................................................. 123 4 9 Histograms of all extrema rates in control, low dose, and high dose groups. ... 124 4 10 Normal fits of the positive, negative, and total extrema rates for each group. .. 125 4 11 Analysis plots for a control subject displaying a strongly dominant initiation focus ................................................................................................................. 126 4 12 Analysis plots for a control subject without a dominant initiation focus ............. 127 4 13 Wave propagation analysis plots ...................................................................... 128

PAGE 12

12 4 14 Analysis plots for a control subject di splaying a dominant termination focus.. .. 129 4 15 Analysis plots for a control subject displaying diffuse termination locations. .... 130 5 1 Probability density estimates of pairwise Granger Causality values by group. 146 5 2 Probability density estimates of pairwise Granger source and sink values among groups by cell layer. .............................................................................. 147 5 3 Comparison between initiation focus and PWGC source. ................................ 148 5 4 PWGC driving region over time. ....................................................................... 149 5 5 Probability density estimates of conditional Granger Causali ty values by group. ............................................................................................................... 150 5 6 CGC Source and Sink values by layer. ............................................................ 151 5 7 Comparison between initiation focus and CGC source. ................................... 152 5 8 CGC driving region over time. .......................................................................... 153

PAGE 13

13 LIST OF ABBREVIATION S 4 AP 4 aminopyridine m micrometer ACSF Artificial Cerebrospinal Fluid AED Antiepileptic Drugs AR Autoregressive BMI bicuculline methiodide CCD charged coupled device CD Cortical Dysplasia CGC Conditional Granger Causality cGy centigray DNA Deoxyribonucleic acid E# Embryonic Day # EEG Electroencephalogram EPSC Excitatory Post Synaptic Current EPSP Excitatory Post Synaptic Potential GABA gamma a minobutyric acid GC Granger Causality IB Intrinsically Bursting neuron IBI Inter Burst Interval or Inter Event Interval IGER Iharas Genetically Epileptic Rat IPSC Inhibitory Post Synaptic Current IPSP Inhibitory Post Synaptic Potential LFP Local Field Potential MAM methylazoxymethanol

PAGE 14

14 MCD Malformations of Cortical Development MEA Micro or Multi electrode Array mM millimolar MRI Magnetic Resonance Imaging ms millisecond mV millivolt MVAR Multivariate Autoregressive NFP Negative Field Potential NFP /EE Negative Field Potentials per Epileptic Event P# Postnatal Day # PFP Positive Field Potential PFP/EE Positive Field Potentials per Epileptic Event PET Positron Emission Tomography PS Population Spikes PWGC Pairwise Grang er Causality RS Regularly Spiking neuron SVAR Single variate autoregressive tish telencephalic internal structural heterotopia

PAGE 15

15 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CHARACTERIZATION OF IN VITRO EPILEPTIFORM ACTIVITY AND PROPAGATION IN THE IN UTERO IRRADIATED RAT MODEL By John Nicholas Grimes May 2011 Chair: Thomas B. DeMarse Major: Biomedical Engineering E pilepsy, one of the most common serious neurological conditions, is characterized by spontaneous recurrent seizures, the catastrophic synchronization of brain activity. While anti epileptic drugs ar e used as the first line treatment, they fail in approxim ately 20 30% of patients. Further, this failure rate is dependent on the type of epilepsy and its cause. Malformations of the cortex have been hig hly correlated with epilepsy, with an estimated 75% of patients having a cortical malformation presenting with epilepsy at some point. Further, cortical malformations are typically less responsive to drug therapies. Although estimated to only exist in 14% of epilepsy patients, cortical malformations account for between 25 40% of all medically i ntractable childhood epilepsies. While the normally laminated cortex has been s tudied for over 100 years, only more recent ly has attention been devoted to the malformed cortex. In order to study the role that cortical malformations play in epileptogenesis animal models of cortical dysplasia, such as the in utero irradiated rat model, have been employed by researchers. Unfortunately, the bulk of this work has been focused on histological and

PAGE 16

16 intrinsic neuron property differences with little consideration given to the emergent network properties in dysplastic slices. In this work I utilized the in utero irradiated rat model to induce cortical dysplasia and compared electrophysiological differences between normal and dysplastic cortex using microelectrode ar Several metrics were used in the analysis and were categorized as classic, spatial, and novel. Classic metrics were those used by previous researchers to compare epileptiform activity in dysplastic and normally laminated cortices and include d ictal event lengths ictal event distributions, and number of field potentials per ictal event Spatial metrics were those used by previous researchers to analyze normally laminated cortical slices but have not been applied to dysplastic slices or used to compare the two. New metrics consisted of the application of Granger Causality methods to the data, which have been used previously in neuroscience, but not to compare functional differences between normal and dysplastic tissue. Results supported and expanded upon previous counter intuitive and anecdotal evidence that low dose irradiated subjects are functionally more dissimilar from controls tha n high dose irradiated subjects. This was evidenced by key classic and spatial metric s, such as mean number of ictal events per recording time, mean event length, mean inter event interval, bias in LFP extrema peaks during events, and wave speed propagation through cortical layers. Surprisingly, and of most potential clinical relevance, r esults from novel application of Granger Causality analysis to in vitro slice suggests that localized areas driv e ictal activity when a dominant initiation site is absent and do not coincide with dominant initiation sites when present

PAGE 17

17 CHAPTER 1 INTRODUCTION AND BACKGROUND Introduction Epilepsy is a chronic neurological condition characterized by spontaneous, recurrent seizures, the catastrophic synchronization of brain activity. One of the nearly 1% o f the population, including over 2.3 million Americans (Begley et al., 2000; Anti carbamazepine and phenytoin, serve as the first line treat ment of epilepsy (Chisholm and WHO these pharmacological agents are able to reduce both the frequency and severity of seizure events. Unfortunately, between 2030% of patients are unresponsive to pharmacological treatment. ( Leppik, 1992; Kwan and Brodie, 2000; Additionally, a significant portion of patients who do respond to AEDs suffer from intolerable side effects, such as cognitive problems, liver toxicities, severe rash, and low white blood cell or platelet counts (Greenwood, 2000; cases of intractable, pharmacologically unresponsive or intolerable epilepsy, additional treatments are employed. These treatments include highly restrictive ketogenic diets, electrical s timulation of the nervous system, and surgery to remove epileptogenic portions of the brain. The success rate of these treatments is heavily dependent on the type of epilepsy, its cause, and region of the brain in which the seizures originate. For instance, epilepsies originating in the neocortex, or outer layer of tissue surrounding the brain, are typically the least responsive to pharmacological agents and even surgical

PAGE 18

18 outcomes are far less favorable than other types of epilepsy. Post surgical seizurefree rates for patients with neocortical epilepsy range between 20 and 40%, compared to between 70 and 80% for epilepsies that have a clearly defined lesion causing the seizures (Aykut Bingol et al., 1998; Mosewich et al., 2000; Causes of Epilepsy Approximately 70% of epilepsy cases are classified as idiopathic or cryptogenic, are caused by a number of factors, both exogenous and endogenous. Exogenous cause s include brain trauma, hypoxia, exposure to certain toxins, and infection. Endogenous causes include congenital conditions like Downs or Angelmans syndrome, brain tumors, and genetic factors. Cortex Anatomy and Physiology In order to better understand previous experiments and interpret their findings on activity propagation, a brief overview of cortex anatomy and physiology is first required. For a more indepth description of the human cortex, see Nieuwenhuys et al. For the r at cortex, see Kol b and Tees ( The cerebral cortex is the outer most tissue surrounding the cerebrum and is classified into three regions based upon the number of layers formed by cytoarchitecture, or distribution of neuron cell bodies: the archicortex, the paleocort ex, and the neocortex. The archicortex includes the hippocampal formation and is comprised of three distinct cell layers. The paleocortex includes the olfaction, entorhinal, and periamygdaloid cortices and varies between three to five cell layers. The paleocortex and archicortex together are referred to as the allocortex. The third region, which accounts for approximately 90% of the cerebral cortex, is classified as

PAGE 19

19 neocortex or isocortex. The neocortex is evolutionarily the youngest portion of the mam malian brain and is the location where higher brain functions, such as spatial reasoning, language processing, and conscious thought, arise. Two main structural features are present throughout the neocortex (Lynch, 1997; Nieuwenhuys et al., The p rimary structural and most visible feature of the neocortex is its laminar Vergleichende Lokalisationslehre der Grosshirnrinde in ihren Prinzipien dargestellt auf Grund des Zel lenbaues six distinct horizontal cytoarchitectual layers of cells are present throughout the entire neocortex. These six cell layers are denoted by Roman numerals, with layer I denoting the outermost layer and layer VI denoting the basal layer. Although a number of neuron types are present, the pyramidal cell is by far the most common and is almost the sole output pathway for information out of the cerebral cortex, with other cell types serving as interneurons. Further, pyramidal neurons within the cort ex project to specific locations depending on which layer the cell body resides. Typically, cell bodies in layers II and III have axons that either terminate on cortical cells in the same hemisphere or run through the corpus callosum and synapse on neurons in the opposite hemisphere. Layer V neurons project to subcortical structures. Finally, layer VI neurons terminate on multiple areas, such as the thalamic nuclei or other cortical regions. The second structural feature of the neocortex is its columnar organization. Across the neocortex, neurons are grouped together into modular and repeating vertical columns that run perpendicular to the surface and span cell layers II VI (Mountcastle,

PAGE 20

20 aving cats and monkeys, first observed these cortical columns. He discovered that a small area of cortex received information from one location, such as a finger, while a nearby area of cortex received information from a different area, such as the palm. Further, it was shown that different columns that received information from the same location were responsible for processing different sensory submodalities, such as light touch on the skin or heavy pressure on deeper tissues. Cortical Dysplasia and Epil epsy Cortical dysplasia refers to abnormal deviations in the structure of the cortex First examined as a causative factor of epilepsy found. Both during and after embryological development, a number of factors such as genetic mutations (Granata et al., 1997; Fox et al., 1998; Gleeson et al., 1998; in utero injuries (Palmini et al., 1994; Roper et al., 1995; Roper et al., 1997; Roper, 1998; can lead to regular laminar and/or columnar structure of the cortex and have been classified extensively using different schemes, such as the timing of the insult leading to the malformation (Barkovich et al., 2001; Leventer et al., 2008; epending on the cause, malformations may focal, or localized to one small area of the cortex; multi focal; or generaliz ed throughout the entire cortex Although cortical dysplasia alone does not assure epilepsy and animal models of corti cal malformations, such as the r eeler mouse, that neither present nor display an

PAGE 21

21 increased propensity for seizures exist that a striking 75% of patients having a cortical malformation will have epileps y east 14% of all epilepsy cases (Meencke and Veith, 1992; in the presence of cortical malformations tend to respond poorly or not at al l to AEDs and other therapies. An estimated 25% to 40% of all intractable or medically resistant childhood epilepsies are due to a cortical malformation (Hardiman et al., 1988; Farrell et al., 1992; Guerrini et al., 2002; removal of the dysplastic tissue is the only predictive factor for seizurefree surgical outcomes for these epilepsies and differs from other forms of epilepsy which are highly correlated to demographics, risk factors, and natural his tories The need to fully resect affected tissue for the best surgical outcomes becomes exceedingly problematic in cases where the dysplasia has multiple focal points, is generalized, or involves tissue in necessary areas, such as the motor cortex. Models of Epilepsy The exact mechanisms by which cortical dysplasia contributes to or causes epilepsy and other neurological problems remain unknown. Advanced imaging n electrodes, and microwire arrays, are used to localize cortical malformations and seizure origin in patients. While beneficial, these techniques are limited in their utility to elucidate the underlying mechanisms and more invasive methods must be employed. D ue to patient safety and ethical concerns, animal models are frequently employed by researchers to study the underlying causes of human epilepsy and

PAGE 22

22 epileptogenesis the processes by which a normal brain develops epilepsy (McNamara A large number of animals, from drosophila to nonhuman primates, and epilepsy inducing protocols injection of kani c acid have been chosen to investigate the mechanisms underlying epilepsy and to ascertain the safety and effectiveness of AEDs and other treatments prior to use in human subjects. In the last decade, at least two books Models of Seizure and Epilepsy ( Animal Models of Epilepsy ( and two detailed reviews on animal models of epilepsy have been published. Sarkisian presented a general overview of various epilepsy types and animal models used to study them, while Wong focused specifically on animal models of focal cortical dysplasia and tuberous sclerosis complex. In the following sections, I give a brief overview of the animal selection and epilepsy induction protocol used in this study and refer more detai led discussion to these reviews. Anima l s election Although a wide range of animals are used to model human epilepsy, research has been focused on mammals due to the structural similarities found across all placental mammal brains ( Kolb and Tees being housed and trained, the rat has been the principle choice of all mammals used and is by far the best studied. This has produced a rich corpus of information about the rat nervous system including in depth atlases (Zilles, 1985; Paxinos and Watson, 1997; Swanson, 1998; Paxinos et al., 1999; Ashwell and Paxinos, 2008; Altman and Bayer, 1995; Foster, 1998; Kolb and Tees 1990; d publications either focusing specifically on the neurobiology of the rat or utilizing a rat model for neuroscience research.

PAGE 23

23 Epilepsy i nduction p rotocols In addition to animal selection, a valid epilepsy induction protocol must produce epileptogenesis w hose presentation and underlying mechanisms most closely match the specific class of human epilepsy being studied is necessary. Since Goddard first published the effect of low intensity electrical stimulation, or kindling, on the rat brain (Goddard, 1967 in vivo protocols that induce epilepsy in the rat and/or produce hyperexcitibility or decrease inhibition in vitro have been identified. However, there is a fundamental difference between seizure models and epilepsy models. While a seizure is defined as a catastrophic neurological event due to the synchronization of brain activity epilepsy is defined as a chronic condition marked by multiple recurrent seizures re event, such as the administration of the neurotoxin, seizures is not a viable model of epilepsy (Engel, 1992; Mody and Schwartzkroin, Animal m odels of c ortical d ysplasia Since cortical malformations are associated with epilepsies that are notoriously pharmacologically and surgically resistant, a number of animal model are aimed at inducing cortical dysplasia to examine the correlation with epilepsy. Many of these models have been described by previous researc hers and can be categorized based upon when the protocol initiates the malformation: genetic, in utero insul ts, and neonatal manipulations (Schwartzkroin and Walsh, 2000; Although the number of genetic based models that accurately mimic human cortical dysplasia is limited, a few popular ones do exist. These include the telencephalic internal structural heterotopia ( tish

PAGE 24

24 Tish ra ts display a characteristic double cortex, or band heterotopia, and a layer of gray matter below the normal gray matter in the frontal and parietal cortices Some of these rats exhibit spontaneous seizures. The p35 gene is key in the development of normal cortical structure. Loss of this gene causes the cortical structure to be inverted, which is similar to that of the reeler mouse caused by a different mechanism. Additionally, while the reeler mouse does not exhibit seizures or an increased propensity for seizures, one study has shown that a quarter of p35 knockout mice exhibit seizures and an additional 50% of them display aberrant s seizures sta rting at around 5 months of age Two main classes of in utero insults exist: exposure to chemicals and radiation. Intraperitoneal injection of various drugs, such as the DNA alkylating agents carmustine ( 1 3 bis chloroethyl nitrosourea or BCNU in vitro hyperexcitaility, and in some models spontaneous epilepsy (Baraban et al., 2000; Benardete and Kriegstein, 2002; Harrington et al., 2007; in utero irradiated rat model is discussed in more detail in the following section. Cortical dysplasia models based upon manipulations to the neonate have also been examined. Closed head injury was first reported to have caused focal cortical been studied. The more popular of which include cranial injection of ibotenate and freeze lesioning. Ibontenate is an analogue of the excitatory neurotransmi tter

PAGE 25

25 glutamate. When injected intracortically into the neonate rat it induces localized abnormal cortical lamination, heterotopic neurons, and deviant sulcus formation, thus approxim ating focal cortical dysplasia (Marret et al., 1995; Redecker et al., 1998a; Redecker et al., 2005; the surface of the neonates skull, thus killing cortical cells near the probe. Over time, a four layered microgyrus appears at the site o f insult (Jacobs et al., 19 96; Jacobs et al., 1999b; ow increased hyperexcitability (Redecker et al., 1998b; Jacobs et al., 1999a; but no spontaneous epileptic activity has been observed in living animals. (Redecker et al., 1998b; Holmes et al., 1999; The in utero i rradiated r at m odel of c ortical d ysplasia The original motivation behind irradiating rats during gestation was to study the process of normal cortical development by interrupting the process at various points in time (McGrath et al., 1956; dysplasia and the correlation between cortical malformations and intractable epilepsy, Roper began utilizing the mode I subsequently present an overview, it is to his 1998 review and Models of Seizures and Epilepsy that I refer the reader to for a detailed analysis of the model (Roper, 1998; Lin The protocol used for this model is among the most simple and easy to perform. Timed pregnant rats are exposed to radiation from an external source. Doses typically range from between 150 to 250 cGy in most laboratories, and no significant differences between linear accelerators or cobalt sources have been noted (Rope care and handling is otherwise no different from controls. Aside from reports of learning

PAGE 26

26 normally, and live long life spans (Lin and Roper, 200 In early studies, it was discovered that the development of the cortex was drastically altered by radiation and is highly dependant on the time of exposure (Hicks et histological effects that irradiation at different timings play. Exposure early in the produces a significantly thinner but otherwise regularly laminated cortex Exposure on E13 through E1 4 generates large clumps of subcortical grey matter in the white matter beneath a on E16, which causes loss of laminar and columnar structure and abnormal orientation of pyramidal cells. As development continues, the effects of irradiation diminish. Exposure on E17 and E18 are similar to but less severe than, E16. Finally, the effect of exposure on E19 and E20 are limited to the superficial layers, which are the last to In addition to timing, the degree of dysplasia has been correlated to the amount of radiation administered (Jensh and Brent, 1987; Fukui et al., 1991; Fushiki et al., 1996; More recently, Kellinghaus et al analyzed the histological and electrophysiological effects that different amounts of radiation caused. In this study, rats were irradiated with either 100, 145, or 175 cGy on E17. While increased dysplasia was correlated with higher radiation doses, only animals in the group receiving 145 cGy displayed spontaneous seizures.

PAGE 27

27 The basic mechanism by which in utero irradiation induces cortical dysplasia is through to consist of a twostep process First, cells ar e initially damaged or killed due to exposure to radiation. Cell death of mitotically active and immature cells in the cerebral cortex and hippocampus have been shown to be mitigated by cycloheximide, indicating that protein synthesis is a deterministic f actor in cell death (Ferrer et al., 1992; also indicated that immature migrating neurons were severely damaged by radiation after the initial insult, cortical development proceeds in an altered cellular environment, which disturbs the complex embryological cascade, ultimately leading to cortical dysplasia (Marnglia, which act as important guidance structures for migrating neurons, are also vulnerable to radiation (Rakic, 1972; irradiation on E17 disrupted glial fibers that did not recover during the perinatal period and that the dorsomedial cortex was the site of both the most severe glia disruption postnatal and cortical dysplasia as adults In order to verify its utility, a number of comparisons between human cases of cortical dysplasia and this model have been made. Several critical similarities between the two are present, but some minor differences do exist. Foremost, this model has been shown to produce acute tissue preparations that are both disinhibited and hyperexcitable in vitro (Roper et al., 1995; Roper et al., 1997; Zhu and Roper, 2000;

PAGE 28

28 Chen and Roper, as well as animals that seize spontaneously in vivo (Kondo et al., 2001; Therefore, this model properly meets the requirements of an epilepsy model. The technical fulfillment of this definition is crucial since other fr equently utilized models, such as freeze lesioning, produces hyperexcitable tissue in vitro but does not produce spontaneous epileptogenicity in vivo histological abnormalities similar to those seen in human patients. While attempts to provide a universal classification of human cortical dyspl asia have yet to be successful Michel et al. produced a list of nine pathological features present. This model consistently provides three of these histological features: loss of normal laminar structure, neurons present in differences between the animal model and hum an model have been noted. Although the in utero irradiated rat model does produce cortical dysplasia, it is diffuse and is therefore does not accurately model focal or multifocal dysplasia like other protocols, such as freeze lesioning (Lin and Roper, 200 morphology is produced, no giant neurons or balloon cells have been reported. Thus, this model more closely approximates mild to moderate, but not severe or Taylors type cortical dysplasia or cortical tubers (Mis chel et al., 1995; Roper, 1998; Lin and Roper, that irradiated animals also have abnormal cell dispersions in the CA1 and CA3 regions of the hippocampus, which has yet to be seen in humans (Roper et al., 1995; Kondo et

PAGE 29

29 in vitro Techniques Although much has been learned about epilepsy from behaving animals exhibiting chronic recurrent seizures, underlying mechanisms responsible require more in depth study. Due to the limitations on what information can be obtained from subjects in vivo tissue samples from animals have been studied in vitro To accomplish this, animals are deeply anesthetized, sacrificed by decapitation, and their brains quickly removed and placed in an oxygenated nutrient medium. The brains are then sliced into thin sections and then studied, often in the presence of various pharmacological treatments such as bicuculline methiodide, 4 AP, and picrotoxin. Three primary techniques have been uti lized for studying epileptiform activity from living brain slices. These techniques allow real time monitoring and/or stimulation of single neurons or neuron populations in vitro Whole c ell p atch r ecording The workhorse of modern neuron electrophysiology wholecell patch recording is a technique that evolved from voltage clamping to achieve intracellular recording from a single neuron. In this technique, a glass micropipette filled with an ionic solution similar to the intracellular matrix with an approxima tely 1 and is placed on the surface of the cell. Suction is then applied to remove a small section of the cell membrane, allowing access to the intracellular matrix. This allows recording of the total current through all ion cha nnels across the cell membrane (Dhillon and Jones, 2000; Kandel et al., 2000; While the patch clamp method provides the most information possible about activity of a cell, there method has some limitations The primar y drawback is the difficulty in recording from more than one cell at the same time. Due to space

PAGE 30

30 limitations, the technique does not scale, and it is not possible to obtain information from more than three or four neurons within a slice at the same time. Being limited to two or three electrodes leads to reliance on signal processing techniques, such as triangulation based on waveform delays and propagation velocity or multiple stimulations and recordings at different to ascertain information about the initiation location of epileptiform activity and propagation patterns. These techniques may be unreliable since signals in different cortical layers may propagate at different veloc ities. Optical i maging with v oltage and calcium sensitive d yes Due to the physical constraints of monitoring activity at more than a few neurons simultaneously with patch clamp methods, new methods to ascertain neural activity were sought. The application of voltage sensitive dyes to electrophysiology has allowed researchers to observe changes in membrane potentials and neuron activity both in vitro (Cohen and Lesher, 1986; Yuste and Katz, 1991; Tsau et al., 1999; Baker et al., 2005; Trevelyan and Yuste, 2 in vivo (Grinvald et al., 1984; Kleinfeld and Delaney, 1996; In this technique, a voltage or calcium sensitive dye is added to the ACSF. Changes in voltage or calcium ion concentration cause a conformational change in the dye, which in turn changes the absorbance spectrum of the dye. A light source with a known spectrum is continuously passed through the tissue. Absorbance changes are he While this technique allows a much greater amount of activity of the slice to be observed, three primary drawbacks exist. First, although activity over a large area of

PAGE 31

31 the tissue can be observed, stimulation cannot and must be performed by other means. Second, a great deal of experimental calibration and optimization is necessary to achieve a high signal to noise ratio. Finally, prolonged optical recording is limited due to dye bleachi ng and because the dyes may have an effect on the cells, such as modulating activity or damaging them. (Wu et al., 1993; Momose Sato et al., 1995; Tsau et al., 1999; Microelectrode a rrays the late 1970s using semiconductor fabrication techniques and was designed to circumvent the problems inherent with patchclamp methods of recording from or stimulating multiple neurons simultaneously while also allowing for a far greater number of sites t o be recorded from than previously possible (Thomas, 1972; Gross et al., 1977; device possessed only 30 electrodes, MEAs have benefited greatly from advances in electronics and fabrication techniques, increasing the number of electrodes and sampling rate. The current standard MEA system in use at most research facilities consists of a 60 or 64 channel system, although commercial MEA systems can be purchased capable of acquiring data from 256 electrodes simultaneously ( Ayanda B iosystems Fundamentally different than the intracellular recordings gained by the patchclamp method, the MEA records changes in extracellular potentials caused by changes in ion concentrations due to cellular activity. As illust rated in Figure 11, MEAs are

PAGE 32

32 field inhibitory post synaptic potentials (IPSP single unit action potentials, also known as spikes. Although chiefly used in neuroscience research, MEAs have been utilized on a broad array of cell and tissue types exhibiting electrical properties, cardiac ( Thomas, 1972; Israel et al., 1984; Connolly et al., 1990; retinal (Wong et al., 1993; Meister et al., 1994; Grumet et al., 2000; cells (Rothermel et al., 2006; cience applications, MEAs are used primarily to study dissociated neuron cultures, but have gained considerable traction by researchers using cultured organotypic tissue and acute slices of neural tissue (Novak and Wheeler, 1988; Egert et al., 1998; Oka et al., 1999; Egert et al., 2002; Heuschkel et al., 2002; In order to increase signal transduction between the tissue slice and electrodes, adaptations to the planar MEA have been made in which the electrodes are cones measuring on the base and these 3D MEAs are able to pierce the outer layer of cells in the slice, which are typically dead or damaged during the specimen preparation process and insulate the active cells from planar electrodes. Figure 11 shows electrophysiological recordings from an acute rat cortical slice collected with a 3D MEA, while Figure 12 shows an acute rat hippocampal slice on a 3D MEA. Due to these advances, MEAs are now being employed to examine slicewide activity of acute cortic al rat and hippocampal slices (Egert et al., 2002; Previous Work and Literature Review Here I present a brief overview of the electrophysiological activi ty that has been reported to occur in both normally laminated and in utero irradiated induced dysplastic

PAGE 33

33 acute rat cortical slices. Topics covered include stages of epileptiform events, propagation patterns and velocities, signal amplitudes, and waveform features. Normally Laminated Slices In order to understand the functional abnormalities caused by in utero irradiation induced cortical dysplasia, a review of electrophysiological activity in normal cortical slices is necessary. Although the role of the neocortex in pharmacologically resistant epilepsies has been increasingly apparent over the last two decades the functional connectivity of the hippocampus structure has garnered the bulk of investigation R ecently significant effor ts have increasingly been focused on characterizing epileptiform activity in the neocortex. Broadly, the results of these efforts can be placed into one of the following four categories: salient anatomical structures and physiologic properties, initiation of an epileptiform event, propagation of the activity to other neuron populations, and event termination. Salient a natomical s tructures and p hysiological p roperties The foundations on which all neocortical processing, oscillations, and epileptiform activity are built is the anatomical and physiological properties of the neocortical neurons. Used frequently for both in vitro and in vivo epilepsy experiments, the struc tural organization of the rat somatosensory cortex has been fairly well examined elucidated several cytoarchitectural features of the neocortex, which are thought to be spec ifically related to the formation and maintenance of epileptiform activity. First, two main types of pyramidal neurons have been described in the neocortex: al., 1982; McCormick et al., 1985; Chagnac Amitai et al., 1990; Larkman and Mason,

PAGE 34

34 1990; Amitai, 1994; Williams and Stuart, 1999; pyramidal neuron types differ in their distribution throughout the neocortical layers, electrophysiol ogical properties, and connectivity. Although both neuron types are found in layer V of the neocortex, RS neurons are also found in layers II and III (Gottlieb and bur more regular discharges (Chagnac significant morphological and functional connectivity differences. IB cells are typified by large somas, thick apical trunks, and heavily branched dendrites, while RS cells have smaller somas, thinner apical trunks, and less branched dendrites (Chagnac Amitai et al., 1990; while layer V RS cells horizontal afferents project more strongly into layer VI and more weakly to layers II/III inhibitory inputs from all cortical cellular layers, although the laminar distribution of inputs differs between the two types. Layer VI serves as the predominant source of excitatory input into IB cells whereas layer V is for RS cells (Schubert et al., 2 onto IB cells arrive from layers II/III, IV, and V but exhibit far less control on these cells than the inhibitory inputs arriving from layers II/III and V on RS cells (Schubert et al., Second, electrophysiological properti es intrinsic to different cortical layers, which in healthy tissue are thought to be responsible for cortical information processing, may play a significant role in the prevention and control of epileptiform activity. A strong

PAGE 35

35 dichotomy between the electr ophysiological action of superficial and deeper cortical layers has been documented (Barkai et al., 1995; van Brederode and Spain, 1995; Yang et al., 1998; Dhillon and Jones, 2000; Yang and Benardo, 2000; Yang and cortical layers appear to have a primarily inhibitory action whereas deeper cortical layers appear primarily excitatory (Barkai et al., 1995; Yang and Benardo, 1997; series of experiments on intact and lami nar strips of cortical slices that reaffirmed and isolated superficial layers are the only layers capable of glutamateindependent synchronized activity and that the synchronization was due to GABAergic activity. Second, excitatory activity in the superficial layers was weak, even though the tissue was bathed in the convulsant 4AP. Third, activity in the middle and deep layers is primarily excitatory, since it is mediated through glutamatergic transmission. And finally, that the superficial inhibitory network can exert considerable control over the deeper layers. These intrinsic laminar properties have been postulated to play a significant role in epileptiform activity. The excitatory / inhibitory stratification of the cortex has led some to postulate that the deep layers may be seizure sensitive while the superficial layers may be seizure resistant (Lopantsev and Avoli, 1998; suggeste d that ictal events in rat neocortical slices are initiated by excitatory neurons isolated in the deep layers and then spread to the superficial layers (Connors, 1984; Hoffman and Prince, 1995; For instance, Connors used 400 500 m thick slices of guine a pig sensorimotor neocortex and measured the

PAGE 36

36 amount of focally applied L glutamate or increased K+ co ncentration to each layer necessary to evoke a paroxysmal field potential in the slice. It was found that a drasticall y smaller quantity applied to Layer IV or upper Layer V was necessary to evoke a slic e wide response, as compared to the other layers. In another experiment, Connors found that focally applied bicuculline applied to the superficial and deep layers caused no effect or a marginal increase in amplitude from the recording site nearest the application. However, when applied to the middle layers, the same amount of bicuculline caused field potentials of prolonged and variable latency. Further, one study docume nted that ictal discharges from the deep cortical layers are attenuated in layer II and hypothesized that this layer actively filters and controls epileptiform activity Initiation of e pileptiform e vents Initiation is the first phase of an epileptiform event. Having particular clinical relevance to epilepsy, knowing the causes and locations of initiation are imperative to designing treatments and preventative measures. Earlier studies have pointed towards IB neurons located in l ayer IV or V as potential pacemakers for these events (Connors, 1984; Chagnac Amitai and Connors, 1989b; was unknown whether the initiation process was dependent on a small number of cells or was an emergent property due to the interactions of a large number of cells throughout tissue. Tsau et al sought to solve this question using optical imaging and two electrodes to monitor spontaneous epileptiform activity in acute cortical slices. They reasoned that if event s were initiated by a small number of cells, the initiation site would be confined to a localized area of tissue. While if events were initiated by the

PAGE 37

37 interactions of a large number of cells, the initiation site would be a large, diffuse area with no app about initiation of epileptiform activity were discovered. First, every observed event began in a confined spot measuring less than the area of one of the photodetector elements used (0.141 m2 tissue. Second, all spontaneous events were initiated from a few confined initiation foci further lending credence to the idea that initiation is started by a small localized grouping of pacemaker neurons. Finally, it was always found that one focus dominates over the other foci for a period of time before yielding to one of the other foci. Unfortunately, the locations of these foci were estimated from data obtained using only two elect rodes, estimated horizontal and vertical propagation speeds, and clustering upon latency times. Using other techniques to verify the existence and location of initiation foci is desirable. While Tsau reported that ictal events were always generated from t wo to three focal sites within a slice and these sites exhibited temporal dominance, other researchers have reported conflicting findings For instance, using organotypic slices on microelectrode arrays and either BIC or kainic acid, Jimbo and Robinson (J imbo and ictal events initiated in the superficial cortical layers and did not ar ise from a dominant focal point. Further, others have noted that the dominant layer of ictal generation depends on the type of treatment used. In most preparations, the deep layers serve as the ictal generation zone, but it has been observed that the upper layers are responsible when treated with kainic acid

PAGE 38

38 Propagation of e pileptiform a ctivity Once a successful initiati on event occurs, the activity propagates from the focus to surrounding tissue. Similarly as important as initiation in clinical relevance and a potential target for therapeutic advancements, propagation of epileptiform activity through the neocortex has r eceived considerable study. Previous researchers have described the process of propagation not as a series of sequential initiations of neurons to another (Chervin et al., 1988; Chagnac Amitai and Connors, 1989a; Golomb and Amitai, 1997; consistently have been focused upon three main features: the pattern and speed by which the activity spreads, the importance and necessity of certain cortical laminae to support propagation, and mechanistic dependencies on which propagation relies. Studies evaluating the propagation of neuronal activity in acute cortical slices have primarily observed evoked response to white matter electrical stimulation. Though distinctively different than endogenous activity, they provide key insight into the underlying functional architecture by which spontaneous activity also travels. For instance, Contreras and Llins patterns, calculated velocities, and showed propagation dependency on stimulation frequency. In their study, acute slices of guinea pig somatosensory cortex were electrically stimulated in the white matter and evoked responses were observed using voltage sensitive dyes and intracellular recordings. During single stimulation, imaging revealed that tissue in layer VI immediately above the stimulation electrode was activated. Activity from this activation site immediately propagated vertically towards the pial surface and, nearly simultaneously, horizontally through layers V and VI. Once the wave front reached

PAGE 39

39 layers II and III, activity propagated through those layers as well, although at a slightly slower velocity than in the infragranular layers. Although the majority of studies observing propagation of evoked activity through cortical tissue stimulate sites in the white matter, similar propagation patterns are generated when stimulating any layer of cortex. For instance, Tsau et al. micro electrode and progressively stimulated each cortical layer and observed the evoked with response with voltage sensitive dyes. In every case a response was evoked, activity propagated vertically from the site towards the pia and almost simultaneously horiz ontally through the laminae. Interestingly, propagation patterns display a strong dependency on stimulation frequency. Two exper iments by Contreras and Llins clearly demonstrated this frequency dependency by stimulating the white matter with different pulse trains, one at 10 Hz and one at 40 Hz. These frequencies were chosen due to their biological relevance. During slow wave sleep, EEG is dominated by low frequency oscillations less than 15 Hz (Steriade et al., 1993; Contreras et al., 1996; while high frequency oscillations in the range of 30 50 Hz dominate while awake (Steriade et al., 1993; were applied at each frequency and the optical response measured. When the 10 Hz pulse train was applied, there was no appreciable difference in activity propagation between the first and fifth stimulation sequence. However, when the 40 Hz pulse train was applied, the activation profile gradually shrank, and by the fifth stimulation the wave propagation was restricted to a small, columnar area directly above the stimulation electrode. In the second experiment, one or both of two sites 1.5 mm apart were

PAGE 40

40 stimulated with the pulse trains. When only one or the other site was stimulated, the response was as previously described. When both sites were stimulated simultaneously at 10 Hz, two waves spread from each site and coalesced in the middle of the slice. However, when both sites were stimulated at 40 Hz, acti vity propagation was limited to two separated columnar areas over both electrodes Intrinsic to propagation patterns, the velocity at which activity propagates has also been studied in neocortical slices. Estimates of propagation veloci ty vary widely by a ctivity type, pharmacological treatments and direction of propagation reported velocities from two types of activity, ensemble activity and a fast response to stimulation that precedes ensemble activity, and compared them with literat ure reports of a third type, interictal spikes in pharmacologically disinhibited slices. Using optical imaging to observe evoked activity propagation, they reported ensemble activity propagated horizontally through cortical laminae much slower than fast r esponse or literature reported interictal spikes at 1 20 mm/s, compared to 40 160 mm/s and 40 160 mm/s, respectively (Chervin et al., 1988; Wadman and Gutnick, 1993; Tanifuji et al., 1994; Golomb and Amitai, 1997; Tsau et al., 1998; Demir et al., 199 ensemble activity propagation velocities were drastically slower than interictal spikes in disinhibited slices, Wu et al. directly compared the velocity of these two activity types in the same slice. Each slice was first perfused with nor mal ACSF and the velocity of ensemble activity was measured with optical imaging. Then, a modified ACSF containing 10 20 M of bicuculline was perfused for 20 30 minutes and the velocity of interictal spikes was measured. It was found that the ensemble activity propagated over an order of magnitude slower than the disinhibited interictal spikes (11 6 mm/s

PAGE 41

41 vs. 125 24 mm/s when treated with bicuculline further corroborated earlier reports that suggested inhibition is essential for c ontrolling propagation velocity Direction of propagation also strongly affects velocity and signal amplitude. As previously discussed, activity propagating from an initiation sit e moves both vertically towards the white matter and pia and almost simultaneously horizontally through the cortical laminae. Contreras and Llinas showed velocity dependencies on both direction of propagation and cortical layer. Activity propagating vertically from layer VI to layer I was fastest and measured at 265 48 mm/s. Horizontal deep layer propagation was faster than superficial propagation and was measured at 217 53 mm/s and 181 44 mm/s, respectivel y. Similarly, activity amplitude also is dependent on cortical layer. While activity propagates more quickly through the deep cortical layers, larger signal amplitudes are observed in the superficial layers (Connors, 1984; Connors and Amitai, 1993; Sheph erd and Koch, 1998; Telfeian and Connors, 1999; In addition to playing pivotal roles in the initiation of epileptiform activity, propagation velocity, and signal amplitude, specific cortical layers are required for sustaining epileptifor m discharges. Three sets of experiments con ducted by Telfeian and Connors mildly disinhibited cortical slices. In these experiments, neocortical slices were bathed in ACSF cont picrotoxin and activity was evoked by electrical stimulation of the white matter.

PAGE 42

42 made, leaving bridges of intact tissue. Sites in the white matter were stimulated and two recording electrodes placed at opposite ends of the slice in layer II/III monitored if propagation across the slice was successful. In slices bathed in high concentrations of picrotoxin, epileptiform activity successfully propagated across the slice as long as an intact tissue bridge at least 350 m wide existed, independent of the layer the bridge was in. However, in slices bathed in low doses of picrotoxin, propagation across the slice w as only successful if the tissue bridge was in layers IV and V or V and VI. In the next set of experiments, propagation was compared between intact slices superficial, middl e, and deep cortical layers. Intact and sectioned slices were able to successful propagate activity in the presence of both high and low doses of picrotoxin. In the presence of low doses of picrotoxin, only intact and sectioned middle layers (layers IV a is necessary for propagation if considerable inhibition is still present. To further test this notion, intact slices were stimulated in the white matter and a recording electrode was placed in either layer I or layer VI to measure the horizontal distance of propagation. When the recording electrode was in layer I, horizontal cuts were progressively made between layers starting with the layer VI V interface and moving upwards. Whe n the recording electrode was in layer VI, horizontal cuts began at the layer I II/III interface and moved downward. In both cases cuts were not complete and an intact vertical strip with all layers always remained above the white matter stimulation site. Other than one exception, none of these horizontal cuts, which

PAGE 43

43 eliminated successive lower or upper layers, had any effect on lateral propagation distance unless part of layer Va ( the superficial strata or Vb (the deep er strata wa s involved. When cuts were near the layer VaVb border, propagation distance was severely affected. Additionally, propagation was never entirely successful unless part of layer Vb remained intact. In the final set of experiments, one recording electrode was placed in layer II/III above the stimulation site and another recording electrode was laterally at least 2 mm away. A series of focal GABA applications was made starting at the pia and moving downward in 200 m increments. One second after focal GABA application, the white mater was stimulated and propagation was monitored. Trials were spaced at 1minute intervals to allow for complete recovery. GABA application to layer V caused the greatest effect on propagation latency. Further, dosages of GABA applied to layer V that completely blocked propagation were unable to similarly block propagation when applied to other layers. Event t ermination The third and final phase of an epileptiform event is termination. Although logically and functionally different from the processes of initiation and propagation, termination has received considerably less attention. Studies examining termination have been almost completely focused on underlying cellular mechanisms, nearly ignoring the interactions among neuron gr oups and the emergent properties that may contribute extensively to the process. A number of intrinsic neuron properties and cellular mechanisms are evidenced to play key roles in the termination process For instance, Bikson et al. demonstrated th at epileptiform activity in hippocampal slices does not depend on

PAGE 44

44 continuous somatic neuron firing. This may be attributed to the fact that although the neurons enter a depolarization block arresting the soma, the axon terminals may still be active (Gutni ck and Prince, 1972; importance of synaptic interactions in the formation of secondary waves, which is demonstrated by effectively blocking them with NMDA receptor antagonists (Miles et al., 1984; Lee and Hablitz, 1991; Traub et al., 1993; since glial cells play a role in the clearance and maintenance of neurotransmitters in the extracellular space, they may have a significant function in all three stages of epileptiform events (Duffy and MacVicar, 1999; Kettenmann, 1999; Pinto et al. ( 2005 cellular and network levels. Here, the concept of termination was examined in two different ways. Sp atially, termination was approached as propagation failure (Bressloff, patterns following an initial wave front eventually return to rest (Miles et al., 1984; Traub et al., 19 wholecell patch techniques to record individual neurons in layers II/III and V, evoked responses to layer IV stimulation were observed. Network level results showed that, like initiation, termination occurred at discrete locations that remained constant from trial to trial. Moreover, in twothirds of trials, termination coincided with the end of the initial depolarization wave. However, in onethird of trials, complex sec ondary patterns arose. Compared to secondary waves in the disinhibited hippocampal slices (Miles et al., 1984; trial to trial. Cellular level results showed that term ination coincided with the patched

PAGE 45

45 neurons entering a depolarization block. Further, individual neurons in both layers II/III and V were active during secondary activity. In Utero Irradiated Induced Dysplastic Slices Although increasingly used to study hu man epilepsy, animal models of cortical dysplasia have yet to receive the amount or depth of study of normal l y laminated slices. Little work has been done in understanding the emergent properties of neuronal populations in dysplastic cortex. Of the few s tudies conducted, n early all utilized dysplasia models other than irradiation, particularly freeze lesioning Since the freeze lesioning model aims to replicate polymicrogyria, while the in utero irradiated model causes generalized cortical dysplasia, and since freeze lesioning fails to produce animals that are epileptic, applicability of results from these studies to the in utero irradiated model is questionable. Further, no literature currently exists examining differences in the three logical phases of epileptiform events in the in utero irradiated rat model of cortical dysplasia. The overwhelming majority of experiments conducted on the in utero irradiated model have been primarily focused on in vivo EEG features individual in vitro epileptiform waveforms differences in neuron distributions and functional differences of neurons in dysplastic cortex. In vivo EEG Roper et al. in utero irradiated animals in the first analysis of the model. In this study, EEG recordings from surgically placed skull screws were taken from control and E17 pups irradiated with 196 cGy of gamma radiation. Neither group displayed ictal activity under normal conditions, but one experimental animal exhibited interictal activity at 0.8 spikes per minute. Once the anesthetic agents acepromazine or xylazine were administered,

PAGE 46

46 experimental animals displayed ictal activity, while controls did not. Further, both control and experimental animals displayed interictal activity under these anesthetic agents, but the mean frequency of interictal discharges was higher in experimental animals than controls, 2.7 0.79 Hz and 1.1 0.42 Hz respectively. Later, Kondo et al. continuous EEG monitoring of the in utero irradiated rat model, an injury based cortical dysplasia model. The experimental group was composed of rats that were exposed to 145 cGy of radiation on E17 Recording electrodes were implanted into 7 control and 7 experimental rats at P45 and EEG signals were monitored for 71 297 hours (mean periods, interictal discharges were observed only in one of seven control rats, but all 7 of the irradiated rats. Further, while none of the control rats exhibited spontaneo us seizures measured with EEG, four of the irradiated rats did. Mos t recently, Kellinghaus et al. ed the effects of varying radiation doses and performed long term video and EEG recordings. Three experimental groups were formed by exposure to 100, 145, or 175 cGy of radiation on E17 and EEG electrodes were surgically implanted on P60. No interictal e pileptiform discharges were observed in the control rats, but occurred at a median frequency of 0.89 per hour in the 100 cGy dosed group, 0.83 per hour in the 145 cGy dosed group, and 0.29 per hour in the 175 cGy dosed group. Interestingly, spontaneous seizures were observed only in the 145 cGy dosed group, even though a positive correlation between radiation dose and the dysplasia severity was found.

PAGE 47

47 In vitro epileptiform w aveforms After the first study demonstrating that in utero irradiated rats displayed ictal EEG recordings and may serve as a m odel of epilepsy, Roper et al. examined whether the protocol generated a greater propensity for epileptiform activity than normally laminated slices in the presence of the GABAA ant agonist bicuculline of radiation on day E17. After weaning on day P21, 400 m thick coronal in vitro brain slices were prepared with a vibraslicer. Electrical stimulation was delivered at t he white matter interface and extracellular field potentials were recorded from the superficial layers of the somatosensory cortex. Slices were then disinhibited with 10 M BMI and the median number of negative field potentials per epileptic event (NFP/EE In the absence of BMI, both normally laminated and dysplastic slices only produced a single negative field potential in response to stimulation. However when slightly disinhibited with BMI, control and experimental slices displayed epileptiform activity both in response to stimulation and spontaneously. While the total number of epileptiform events was not significantly different between groups, dysplastic slices displayed a greater number of NFP/EE (4.1 1.03 vs. 1.6 0.4 for evoked ev ents and 5.1 1.2 vs. 1.5 Following the finding of in vitro hyperexcitability of in utero irradiated cortical slices by Roper et al. mechanisms responsible. These studies primarily have focused on examining the differences in neuron distributions and the intrinsic cellular properties of neurons in dysplastic cortex.

PAGE 48

48 Neuron distributions Since many theories of epileptogenesis are based on an imbalance in the normal d istribution of excitatory and inhibitory neurons and their connections, Roper et al. slices. The experimental group was created by in utero irradiation to 225 cGy of r immunohistochemical studies were performed on brain slices. Results were c onsistent with previous studies (McGrath et al., 1956; Cowen et al., 1970; Roper et al., 1995; a thinned neocortex lacking lamination, subcortical grey matter, heterotopic neurons, heterotopic neurons in the hippocampus, and absence of the corpus callosum. Aft er application of staining techniques, it was found that two types of cortical inhibitory 50% in irradiated animals, but that overall neuron density was unchanged. These data suggested that interneurons are pref erentially reduced by in utero irradiation. Deukmedjian et al. up study aimed at better understanding the progression by which the irradiation protocol preferentially reduced the neocortical GABAergic system. To chart the proces s of the reduction, experimental animals exposed to 225 cGy of radiation on day E17 and controls were examined using immunohistochemical techniques at two different time points: E21 and P6. In irradiated rats, the total number of neurons was about 50% of controls at both time points. During the time interval, the total number of neurons doubled in both groups. Surprisingly, during the do ubling, the distribution of neurons changed. While GABAeric neurons increased by an order of magnitude in control animals, no change was reported in

PAGE 49

49 irradiated animals. This led to a drastic shift in the distribution of inhibitory neurons in the experimental group, from 18% at E21, down to 9% at P6. Thus, it seems neocortical GABAergic neurons are less capable at recovering from radiationinduced injury. Functional differences of neurons in radiation induced dysplastic cortex More recent studies have been focused on identifying the functional differences of neurons in in utero radiation induced dysplastic cortex. Zhu and Roper performed the first of such studies. Spontaneous events and evoked responses were obtained from pyramidal neurons in neocortical slices from control and rats irradiated with 225 cGy at E17 using wholeseveral functional changes leading to reduced inhibition in the dysplastic cortex. First, both the amplitude and frequency of sIPSCs were reduced, 35% and 70% respectively in dysplastic cortical neurons. However the decay time constant and 10 90% rise times were unchanged between the groups. Second, the frequency of mIPSCs was reduced in dysplastic cortex by 86%. Third, monosynaptic evoked IPSCs in dysplastic cortex showed a 48% decrease in mean maximal amplitude compared to controls. Fourth, evoked EPSCs from dysplastic cortex showed both greater peak amplitude and averaged area than controls. Finally, spontaneous EPSC in dysplastic cortex showed a 42% increase in amplitude and 77% increase in frequency. Combined, these data show considerable inhibitory system impairment in the in utero radiation induced cortex. A follow of heterotopic cortex, gray matter residing deep within the subcortical white matter of irradiated slices, and corroborated the selective impairment of the inhibitory sys tem in

PAGE 50

50 dysplastic cortex. Wholecell voltage clamp recordings showed a similar reduction in the average frequency of sIPSCs in heterotopic cortex compared to normal pyramidal neurons in layer II/III (4.1 0.4 Hz vs. 7.4 However, sIPSC amplitudes and kinetic properties, such as 10 90% rise times, were not significantly different between groups. Similarly to sIPSCs, the frequency of mIPSCs was also reduced in dysplastic cortex (2.2 0.4 Hz vs. 3.3 a mplitude and kinetic properties were unchanged. Differences in short term plasticity in inhibitory and excitatory neurons were also examined. This was done by stimulating near the cells with a 5pulse train at 10 or 20 Hz and observing the response. The IPSC responses evoked from cells at 10 Hz behaved similarly in both groups and depressed over successive pulses in the train. However, the EPSC responses evoked at 20 Hz differed. In control cortex, successive EPSC responses were greater than the first r esponse, with the second pulse in the train causing the largest amplitude response, while in dysplastic cortex subsequent responses gradually decreased in amplitude. While this study effectively demonstrates impairment in the function of interneurons on pyramidal neurons in heterotopic cortex, it was unable to tell if it was due to a decreased number of interneurons or other changes in presynaptic release mechanisms. Subsequent studies have continued to illustrate the intrinsic differences between individu al neurons in in utero irradiate rats and controls These differences include a decreased response of interneurons to excitatory drive (Xiang et al., 2006; Zhou et al., and supporting evidence for arrested neuron

PAGE 51

51 Obj ectives and Central Hypothesis In order to further characterize the in utero irradiated rat model of cortical dysplasia for use in epilepsy research, a greater emphasis on network level interactions must be given. In this study, I utilize MEAs to analyze spontaneous epileptiform discharges in cortical slices from animals exposed in utero to one of two different radiation doses on E18 and compare results with normally laminated slices. By using MEAs to acquire electrophysiological data, both classic and novel statistical techniques, notably Grang er Causality, can be utilized. Through these techniques, I seek to achieve two main goals F irst, since previous work devoted to studying the spatial properties of ictal propagation in acute rat cortical slices has focus solely on normally laminated slices, I a i m to begin providing a comparable understanding about epileptiform activity in in utero radiation induced dysplastic slices. Second, utilizing new metrics, such as Granger Causality, I aim to gain deeper ins ight into the spatial regions driving ictal activity in both normally laminated and dysplastic cortical tissue. In particular, it is currently unknown whether the origination of ictal events, as defined by the location that has the first local field potent ial peak, is also the region responsible for driving the ictal event. The answer to this question potentially has high clinical relevance, since methods of surgical resection in cases of intractable epilepsy often focus removal of the seizure initiation s ite localized by sca lp or sub cranial EEG recordings. Due to the poor surgical outcomes in cases of intractable neocortical epilepsy unless the entire dysplastic region is removed, I hypothesize that the location of ictal initiation does not coincide spat ially with the location responsible for driving the ictal event.

PAGE 52

52 A B Figure 11 Data collected in the DeMarse lab from acute rat cortical slices treated with 4 recorded during an interictal period. Each window represents activity on one channel over the course of the previous second. The amplitude of each This is an example of a s pontaneous population spike recorded from multiple channels throughout the tissue with individual spikes progresses to an ictal event. Time and amplitude scales the same as A.

PAGE 53

53 Figure 12 An acute rat hippocampal slice with CA1 and the dentate gyrus placed over the electrodes of an Ayanda Biosystems MEA60 200 3D. The electrodes are pyramidshaped with a square base measur ing 40 m by 40 m and have a height of 50 to 70 m with a center to center spacing of 200 m.

PAGE 54

54 CHAPTER 2 EXPERIMENT AND METRICS Experimental Protocol Animal acquisition, irradiation, housing and brain slice preparation were performed by the Roper Laboratory ( Epilepsy Neurophysiology Lab, University of al., 1997; Zhu and Roper, 2000; and is briefly discussed in the following text. Animals and Irradiation Protocol Timed pregnant Sprague Dawley rats were acquired from Harlan ( Indianapolis IN, were split into one control group and two experimental groups. The two experimental groups were further classified as either low dose irradiated or high dose irradiated. On E17, pregnant rats from the experimental groups were placed in a well ventilated acrylic corresponding to low dose or high dose respectively, of external radiation from a linear accelerator source. Otherwise, all litters were housed identically. Offspring were born naturally and weaned on postnatal day 21 (P2 were kept on 12hour dark/light cycles and received food and water ad libitum. All procedures followed guidelines approved by the Institutional Animal Care and Use Committee at the University of Florida. Brain Slice Preparation Rats were anesthetized via isoflurane inhalation and sacrificed by decapitation. Their brains were quickly removed and placed in chilled (0-

PAGE 55

55 2 and 5% CO2 and containing the following solute concentrations in mM: 124 NaCl, 2.5 KCl, 1.25 NaH2P O4, 6.0 MgSO4, 1.0 CaCl2, 26.2 NaHCO3, 10.0 glucose. A low potassium concentration and high magnesium concentration were chosen to block activity during dissection and slicing. Brains were allowed to cool for approximately 90 seconds and then were removed from the dissection ACSF, affixed to a Vibratome (Leica VT1000s, Leica from 8.7 to 4.8 mm ant erior to the interaural line oxygenated with carbogen for at least 1 hour prior to recording. Following slice equilibration, slices were bisected along the medial longitudinal fissure and individually transferred to a 3D MEA (Ayanda Biosystems, Lausanne, an inverted microscope (Nikon beneath the cortical layers. Recording locations ranged medially from M2, the secondary motor cortex, to the barrel field of the primary s omatosensory cortex After tissue positioning a mesh harp (ALA Scientific Instruments, Farmingdale, placed on top of the slice to increase slice contact with the electrodes and prevent slice movement during recording. The MEA was then placed into the

PAGE 56

56 concentrations in mM: 124 NaCl, 5.0 KCl, 1.25 NaH2PO4, 0.5 MgSO4, 2.0 CaCl2, 26.2 NaHCO3, 10 glucose, and 0.14 4aminopyridine (4the MEA recording chamber at a rate of 23 ml/min via a microfluid chamber level controller (LeveLock LL+ and Ca2+ concentrations, dec reased Mg2+ concentration, and 4AP were administered in the recording ACSF to encourage spontaneous epileptiform activity, which is uncommon i n submerged slice preparations (D'Arcangelo et al., 2001; Recording A ground electrode w as connected to the system and submerged in the MEA recording chamber. Electrophysiological signals were amplified, band pass filtered from 1 Hz 10 KHz and digitized by a data acquisition card (Multichannel Systems MEA1060BC Reutlingen, Germany sampling rate o f 25 kHz with 12 bit resolution. Real time monitoring of voltages on all 60 channels was performed using internally developed software built upon the Open Source MEABench suite (Wagenaar et al., Recordings were started immediately upon the onset of the first ictal event for each slice and continued for at least 15 minutes. However, i f an ictal event was not observed within 20 minut es, the slice was discarded. In total, data from 11 slices from 7 control animals, 7 slices from 4 low dose animals, and 11 slices from 5 high dose animals were analyzed. Post Processing and Data Analysis After recording, the raw signals were post processed by software to identify and separate ictal events i nto a single file for each slice Information stored for each ictal event included: time of the event, the channel that triggered the event, the leng th of the

PAGE 57

57 event, and voltages throughout the event at the 60 electrodes To detect ictal events, a voltag e threshold of 2. 8 3. 5 times the standard deviation of the mean voltage measured in the absence of activity was used. If at any time any channel had a voltage exceeding the threshold for more than 150 ms declared on that channel and data from 100 ms prior to the time of peak voltage and 500 ms after was collected. During the 500 ms after the peak other LFP events could occur. If three or more contiguous LFP events were detected an ictal event was declared. Recording of the ictal continued until no LFPs had occur red for 500 ms. At this time the end of the ictal event was declared and the event was stored to disk Figure 2 1 gives a visual representation of the automated post processing algorithm. After automated post processing, data from every channel of each ictal event was plotted with custom software written in Objective C and verifi ed by visual inspection to be genuine. Events were required to have at least three successive super threshold peaks on at least one channel to distinguish them from LFPs. Since ictal events were typically measured simultaneously across the entire 8x8 electrode grid they were easy to distinguish. Events that were erroneously classified as ictal events (e.g. : electrical noise produced by t he per by the automated program due to recording artifacts or that did not have at least three successive super threshold peaks were removed from analysis. Subsequent data processing and analysis was performed using programs wri tten in MATLAB and ObjectiveC. Definition of Metrics The two questions that I aim to answer are What are the differences between in vitro cortical ictal, or seizure like, activity in control and irradiated animals? and, Does the initiation site of ict al events coincide spatially with the region driving ictal activity ?

PAGE 58

58 In order to answer these questions a set of metrics must be defined to measure and quantify the ictal activity recorded. I first present a simple test to ascertain if the slice activi ty is stable with respect to time. This simple test provides a solid foundation necessary for all subsequent qualitative and quantitative analysis performed. Next, I present the actual metrics, which are subdivided into three groups based upon their use by previous researchers and their complexity. First, I introduce and define a set of classic metrics used by previous researchers that do not incorporate spatial or topological information and do not vary over time. Second, I progress to more complex met rics, which capitalize on the rich spatial information about the slice provided by the MEAs Finally, I present a brief overview of Granger Causality and its application to quantifying causal interactions during propagation within and across the cortical layers during ictal events. Verification of Slice Stationarity In order for any of the metrics or statistical analysis performed to have meaningful results, the electrophysiological activity of the slice must be stable with respect to time. This is especially important for metrics that depend on temporal components. If electrophysiological activity is not sufficiently stable over the recording period, results from statistical analysis will become meaningless. Specifically, two important scenarios exist which could invalidate temporal ly dependent metrics. First, there is an amount of time the recording solution, which contains 4AP, an elevated KCl concentration, decreased Mg2+ concentration, and is heated to 35 C, must suf ficiently wash in the tissue and begin to have a physiological effect. This could cause an increase in activity during the washin period. Second, gradual cell /slice death over the course of the recording could result in a decrease in

PAGE 59

59 activity over the course of the recording. In order to ensure that neither of these phenomena plays a significant role, two provisions were made. The first provision was the experimental protocol, which was designed to mitigate the effects of wash in. Recording was only started after the first observed ictal event. This typically took between 5 and 15 minutes. I modeled the change in concentrations of solutes in the recording chamber by assuming the chamber initially held 1 ml of dissection ACSF, the average in and out flow rates were 2.5 ml/minute, and the fluids mixed perfectly. Next, I constructed a differential equation based on these assumptions to approximate the solute concentrations in the recording chamber over time as follows: The number of moles of solute in chamber at time t is equal to the initial number of moles plus the amount of solute flowed in at time t minus the amount of solute flowed out at time t. This gives Eq. 2 1, the general differential equation for recording chamber solute concentration as a function of time. C (t) = [solute]in flowout* C(t) (2 w to time, [solute]in is the number of moles of solute flowing into the chamber per minute, flowout is the number of moles of moles of solute in the chamber at time t. For 4 AP, the specific equation is C (t)4 AP = 0.35 10-6 2.5 C(t)4 AP with an initial condition of C ( 0 ) 0 Using the general solution method gives

PAGE 60

60 C (t)4 AP = e-2.5* t( 0.35 10 6* e2.5* tdt k) Solving the indefinite integral gives C (t)4 AP = e-2.5* t(0.14 10 6* e2.5* t k ) (2 Reducing E q 2 2 yields C (t)4 AP = 0.14 10 6 k e2.5* t Finally, solving for the constant k using the initial condition produces C (t)4 AP = 0.14 10 6* ( 1 e2.5* t) Similarly, Eq. 2 3 through Eq. 2 5 gives the amount of K+, Mg2+, and Ca2+ present in the chamber at a given time, respectively C (t)K+ = (2.5 e 2.5* t 5.0 ) 10 6 (2 C (t)M g2 + = ( 5.5 e 2.5* t 0.5 ) 10 6 (2 C (t)C a2+ = 0.00025 e 2.5* t 0.000005 (2 Using these equations to plot the concentration of each solute as a function of time gives the graphs shown in Figure 2 2 In the case of Mg2+, which has the longest delay, the chamber concentration reaches within 5% of the final concentration within 131 seconds and within 0.5% o f the final concentration within 186 seconds. This shows that complete washin should be completed well before the 5 15 minutes before the first ictal event is observed and recording is started. The second provision made to ensure that washin and necrosis had no appreciable effect was examining the rate at which ictal events occurred over time. It is hypothesized that ictal events should occur more or less linearly over time. The ictal event number versus event start time for each recording was plotted and a linear fit was

PAGE 61

61 performed to ascertain the level of linearity. An initial, rapid increase in activity could be attributed to washin, while a gradual decline throughout or rapid decline in activity later in longer recordings could be attributed to necrosis. Should such an inc rease or decrease in activity be seen, data from only the linear region should be included for further statistical analysis (e.g.: after any rapid initial activity increase and before any Classical and N onSpatial Metrics Ictal e vents per r ecording t ime The first, and most fundamental metric used is the number of ictal events per unit of recording time. For the purposes of this document, an ictal event or ictal is defined as at least three consecutive lo 3.2 standard deviations above the noise floor recorded on at least one of the sixty channels and includes 100 ms before the first LFP breaks the threshold and continues 500 ms after the activity levels return to the noise floor. Since recording lengths vary between slices, normalizing by recording length is necessary. This metric has widespread use and has been used by previous researchers to characterize epileptiform activity in different in vitro models (Gulys Kov cs et al., caused by cortical dysplasia (Kondo et al., 2001; In this study, an elevation in the number of ictal events per time is expected with dysplastic slices, due to a loss of laminar structure and regulatory inhibition networks, thus lowering the ictal threshold and allowing for more synchronous activity to occur

PAGE 62

62 Further, in line with r esults reported by Kellinghaus et al. ( dose irradiated subjects will have a greater increase than high dose irradiated subjects. Average e vent l ength Employed by previous researchers to quantify the effectiveness of various pharmacological and electrical stimulation protocols aimed at lessening epileptiform activity, the average event length is the second fundamental metric used to quantify electrophysiological activity differences between groups (D'A rcangelo et al., 2005; D'Antuono et al., 2002; Dysplastic slices are expected to have a longer average event length, due to loss of regular morphology and impeding inhibitory networks that would normally terminat e synchronous events. Distribution of event lengths E xpanding further on the average event length metric, I present histograms of event lengths, normalized by recording time. This provides a more descriptive metric than average event length, since information about the presence of clusters of event lengths would be hidden by a simple average. Decreased inhibition is expected to lead to distributions tending towards longer event lengths in irradiated subjects as compared to control subjects. Average burst integral The epilep hyperexcitabilty of dysplastic cortex due to fr eeze lesioning in the rat model (Kellinghaus This metric was employed using data collected from intracellular methods and is dependent on the amplitude of the signals collected. Since our protocol used extracellular recording techniques, namely 3D MEAs, to acquire electrophysiological signals, th is metric becomes less reliable.

PAGE 63

63 A mplitude variations arise due to a number of factors including differences in slice placement on the array, tissue contact with the 3 D electrodes t hickness of the insulating deadcell layer caused by the tissue slicing process, and distances between active neuron populations and the electrodes Due to the average burst integrals reliance on uniform signal amplitude collection and variability in this due to our experimental protocol I ch ose not to include incorporate t his metric into our analysis Average i nter e vent i nterval The inter in vitro experiment ers using dissociat ed cultures and computer models (Canepari et al., 1997; Chiappalone et al., 2003; Chiappalone et and it has been adapted by researchers using acute slices (Slutzky et al., 2001; in vivo experiments with EEG In this study, I define the IBI as the length of time between the end of one ictal event and the beginning of the next Due to cortical dysplasia leading to a loss of inhibition, normal oscillatory activity, which normally would terminate via inhibitory networks, are expected to more effectively recruit neuron ensembles, resulting in a lower IBI for irradiated subjects. Further, it is hypothesized that the mean IBI from the low dose group will be shorter than the mean IBI from the high dose group. Distribution of inter event intervals Also expanding on the average IBI metric, I present histograms of IBI lengths, normalized by recording time. This provides a more descriptive metric than average IBI length, since information about the presence of clusters IBI lengths would be hidden by a simple average. Decreased inhibition is expected to lead to distributions tending towards shorter IBIs in irradiated subjects as compared to control subjects. Further, it is

PAGE 64

64 anticipated that the IBI distribution from the low dose group will tend towards shorter lengths than the high dose group. Recovery ratio Tying event duration and IBIs together, I seek to quantify the recovery ratio after an ictal event. I define the recovery ratio as the inter event interval divided by the duration of the immediately preceding event. I expect three main trends in the recovery ratio. First, it is expected that longer events will be followed by a longer IBI, due to depletion of neurotransmitters. Second it is hypothesized that due to loss of lamination and inhibitory networks, irradiated subjects will in general have a decreased recovery ratio Finally, I expect a larger decrease in recovery ratios from low dose group than recovery ratios from the high dose group. Number of local field potentials per epileptic event Modifying the number of negative f and NFP per epileptic metric s previously employed by researchers such as Roper et al ( 1 I developed MATLAB software based on the public domain peakdet program ( Billauer www .billauer.co.il to down sample the signals from 25 kHz to 250 Hz and then count the number of positive and negative local field potentials appearing on each channel during each ictal event to assess if there are differences bet ween experimental and control groups Due to previous studies, I expect to find a greater number of field Spatial Metrics A great deal of research has been devoted to exploring the spatial properties of seizure like activity in acute rat cortical and hippocampal slices. However, this previous work has focused solely on normally laminated tissue. Here I describe the previously

PAGE 65

65 used and expanded spatial metrics I applied to normally laminated and dysplastic corti cal slices to compare seizure like events. Initiation s ite Previous studies using 1D electrode arrays and optical imaging have shown that the initiation for spontaneous epileptiform activity is a localized process usually constrained to a few localized g roups of neur ons in layer V of the neocortex with most preparations Further, literature reports conflicting views on the existence of dominant initiation foci. Some researchers indicate that there is no dominant initiation site typically two or three initiation foci exist and the dominant foci changes over time Here, I denote the initiation location as the first channel to reach its first peak in each ictal event. It is anticipated that control slices will have a few dominant initiation foci contained chiefly in the lower cortical layers, while the irradiated slices will have a greater number of initiation foci spread throughout cortical layers due to loss of lamination. Ictal w ave p ropagation s peed s and pattern Modifying metric s employed by previous researchers (Bakker et al., 2009; Chervin et al., 1988; Demir et al., 1999; Golomb and Amitai, 1997; I analyzed the propagation of ictal activity through the tissue. This was done quantitatively by calculat ing wave propagation velocities through the tissue slices both horizontally through layers and vertically across layers. Further, I qualitatively analy zed w ave propagation patterns originating from dominant initiation foci using color gradients generated from event first peak times Results from these metrics were then compared between groups.

PAGE 66

66 Termination s ite Previously examined in normally laminated linear array of electrodes, I define the termination site as the electrode with the last peak in an ictal event. In accordance with the findings of Pinto et al., I expect termination to be confined to a few foci in the upper layers in normally laminated slices. As with event initiation, d ue to loss of lamination in dysplastic slices, I expect less regular termination sites as compared to control subjects New Metrics Overview of Granger Causality While classic and spati a l metrics allow for characterization of ictal events, it is possible that these observations are the result of more complex and unobserved underlying processes. For instance, although a dominant initiation focus may be present in a slice, this may not be the location actually driving the seizure event. This has particular clinical relevance in cases of intractable epilepsy where surgical resection of brain tiss ue is employed. Therefore, it is desirable to utilize a technique, such as Granger Causality, that allows analysis into underlying processes. Although originally used for economic data, Granger Causality has gained significant traction in the neuroscience community over the last half decade. Specifically, it has been most attractive in instances where researchers wish to gain greater insight into causal interactions between neuron populations or brain region in behaving animal or human subjec ts and where experimental physical or ethical boundaries limit data collection to EEG or MRI (Goebel et al., 2003; Hesse et al., 2003; Bollimunta et al., 2008; Roebroeck et al., 2005;

PAGE 67

67 First, it provides a mathematical quantificatio n of causal relationships, providing an effective means to compare the strength of causal interactions among groups. Secondly it separates causal relationships from merely coincidental occurrences Third, using an improved technique, known as conditional Granger causality, it allows for the elimination of mediating and confounding influences. The fundamental idea behind Granger Causality can be tracked back to Wiener ca n be said to be causal to B if past information about A allows a more accurate prediction about the future behavior of B. Granger later formalized this idea in 1969 for linear regression models of stochastic processes. This is accomplishe d by evaluating whether or not the variance of the prediction error for B at the present time is reduced by including past measurements from A in a linear regression model. If it is reduced, then A is said to have a causal, direct, or driving influence on B. Granger Causality analysis has three main advantages over traditional crosscorrelation metrics. First, GC is less sensitive to changes in overall activity rate. Second, it provides a directional measure of the causal influence, whereas cross correlation does not. Finally, GC may require less data to successfully complete a meaningful analysis. Pairwise Granger Causality Here I present an overview of the mathematical underpinnings of Pairwise Granger Causality in both the time and frequency domains wh ich closely follows the Geweke (1982; treatment of the subject matter.

PAGE 68

68 simultaneously acquired stationary time series: X given by X1, X2, ..., Xn and Y given by Y1,Y2, ...,Yn these time series based on previous samples can be given by: X ( t ) ax( j ) X ( t j ) t j 1 p Y ( t ) ax( j ) Y ( t j ) t j 1 p where the variance of each error series, X ( t ) and Y ( t ) respectively, is a measure of the model, W ( t ) where X and Y are calculated on previous values of both time series gives W ( t ) X ( t ) aXX( j ) X ( t j ) aXY( j ) Y ( t j ) t j 1 pj 1 pY ( t ) aYX( j ) Y ( t j ) aYY( j ) Y ( t j ) t j 1 pj 1 p (2 Using Wieners idea, Granger formulated that if the X prediction is improved by using past information about the Y series, then Y can said to have a causal influence on the X time series. This influence can be quantified by FY X ln var(t) var(t) The variance of the predicti on error for the AR model of X is compared to the variance of the prediction error when Y is included in the model. Evaluating whether or not there is a causal influence in the opposite direction, from X to Y merely requires reversing the roles of the two time series.

PAGE 69

69 Since natural time series, such as economic data, contains oscillatory components, Geweke (1982; 1 later developed the mechanics to handle a spectral version of Granger Causality, thus allowing the quantification of causal influences at specific frequencies. This method compares the power of a time series generated by the stochastic processes intrinsic to that time series, or its intrinsic power to the total power of the same time series. Since the total power of the time series includes any causal influences from other time series in the frequency domain, there may be a difference in the intrinsic power and total power. Thus, when no causal i nfluence exists upon the time series, the total power is equal to the intrinsic power and the log ratio of the total to intrinsic power is equal to zero. Before being able to compare the total and intrinsic powers of a time series, a number of steps must be taken. In order to generate the frequency domain function W ( f ) a Fourier transform is performed on the MVAR W ( t ) defined in Eq. 26 yields, and yields A ( f ) W ( f ) E ( f ) (2 where A ( f ) is the spectral coefficient matrix and E ( f ) is the spectral noise vector resulting from the Fourier transform. Next, Eq. 27 is rearranged by setting the transfer function H ( f ) equal to A ( f ) 1 which pr oduces W ( f ) H ( f ) E ( f ) Next, the spectral matrix S ( f ) containing the auto spectra and cross spectra of W ( t ) can be calculated, as shown in Eq. 28. S ( f ) H ( f ) 3H*( f ) (2

PAGE 70

70 Where H*( f ) is the transpose and conjugate of H ( f ) and 3 is the covariance matrix of the noise terms from W ( t ) of a given frequency f for one time series into the sum of the intrinsic power and causal power from an influencing series. Each time a new spectral matrix is calculated in the same manner, except both sides of Eq. 27 are left multiplied by Gewekes norma lization transform P 1 0 cov(t,t) var(t) 1 The normalization transform isolates the intrinsic power of X by calculating the intrinsic power component of the total power separately from the causal power X component of the new transfer function, shown in Eq. (2 X time series. H~ xx( f ) Hxx( f ) cov(t,t) var(t) Hxy( f ) (2 The Granger Causality spectrum is then calculated by taking the logarithm of the ratio of the total power over the intrinsic power Fy x( f ) ln Sxx ( f )H~ xx( f ) var(t) Hxx ~( f ) Finally, Using the value of the causality spectrum, it is then possible to evaluate how much of the variance of a signal at f is due to the second signal exerting a causal influence.

PAGE 71

71 Conditional Granger Causality While useful at determining connections between timeseries, Pairwise Granger particularly in networks where mediated connections exist: falsepositive connections and over estimated connection strengths. These shortcomings are illustrated in Figures 2 3 and 24. Suppose I have the simple threeneuron network with connectivity as shown in F igure 23A. Neuron X is connected to Neuron Y and Neuron Y is connected to Neuron Z. No other connections exist. Performing a traditional PWGC Analysis on this network will likely describe the network shown in Figure 23B. The most obvious difference between the actual network and the one reconstructed through PWGC analysis is the erroneous direct connection from Neuron X to Neuron Z. This is due to the fact that Neuron X has a mediated connection to Neuron Z through Neuron Y. In other words, the firin g of Neuron X induces Neuron Y to fire, which then induces Neuron Z to fire. Secondly, in the PWGC reconstructed network, the connection strength between Neuron Y and Neuron Z is over estimated. This connection over estimation is the sum of two component s: the actual endogenous firing of Neuron Y causing Neuron Z to fire and the mediated connection from Neuron X to Z through Y in which Neuron X fires, evoking a response from Neuron Y causing Neuron Z to fire. False positive connection inferences and over estimated connection strengths are compounded in networks possessing a cascade of mediated connections, such as the network shown in Figure 24. As the length of the neuron chain increases, the estimated PWGC reconstructed connection weights towards the end of the chain approach 1.

PAGE 72

72 In order to tease out these mediated connections, a modification of PWGC and remove mediated connections erroneously interpreted as direct connections, can be performed. Algorithms have been developed to perform this analysis in both the time and frequency domains For brevity, I only present the time domain case and use the simple threeneuron network shown in Figure 2 3A for our discussion. In the timedomain, CGC compares the prediction of Z including Y in a bivariate MVAR model, V ( t ) shown in Eq. 210, with the prediction of Z including Y and X in the trivariate MVAR model, U ( t ) shown in Eq. 211. V ( t ) Y ( t ) aY Z( j ) Z ( t j )j 1 p aY Y(j ) Y ( t j )j 1 ptZ ( t ) aZ Z( j ) Z ( t j )j 1 p aZ Y( j ) Y ( t j )j 1 pt (2 U ( t ) X ( t ) Ax x( j ) X ( t j )j 1 p Ax y( j ) Y ( t j )j 1 p Ax z( j ) Z ( t j )j 1 ptY ( t ) Ay x( j ) X ( t j )j 1 p Ay y( j ) Y ( t j )j 1 p Ay z( j ) Z ( t j )j 1 ptZ ( t ) Az x( j ) X ( t j )j 1 p Az y( j ) Y ( t j )j 1 p Az z( j ) Z ( t j )j 1 pt (2 Since the influence of Neuron X on Neuron Z is entirely mediated through Neuron Y, the trivariate model prediction of Z is no more accurate than the bivariate model. This is determined by taking the log of the ratio of the variance of the prediction error of Z from V ( t ) to the prediction error of Z from U ( t ) and is shown in Eq. 212. FY Z X l n va r (t) va r (t) (2

PAGE 73

73 Thus yielding a value of zero in the case where the influence of Neuron X on Neuron Z is completely mediated through Neuron Y, but a value greater than zero when Neuron X has a direct, nonmediated influence on Neuron Z.

PAGE 74

74 Figure 2 1 Visual representation of the automated post processing algorithm. Here four seconds from one channel of raw data in which an ictal event occurred is plotted. The two black horizontal lines centered about 0 V are the thresholds. Together, the three shaded area s represent the packaged ictal event. The redyellow interface is the peak of the thresholdcrossing event. The red area denotes 100 ms of data prior to the event peak that is collected, yellow represents the main part of the event, and green is the last 500 ms of data collected after the voltage on the channel returns within threshold.

PAGE 75

75 A B C D Figure 2 2 Concentration of solutes in the recording chamber vs. time. Since the concentrations of some solutes differ between the dissection and recor ding ACSF, there is a delay at the beginning of each experiment when the recording fluid is washing in. Blue indicates the solute concentration within the recording chamber and the red indicates a concentration within 5% of the + Concentration vs. 2+ 2+ Concentration vs. Time

PAGE 76

76 A B Figure 23. Example of an erroneous direct causal connection generated by Pairwise Granger Causality on a network in The actual network possessing a mediated connection from Neuron X to possesses an erroneous direct connection from Neuron X to Neuron Z due to the aforementioned mediated connection. Note: differences in connection weights were omitted from B for clarity.

PAGE 77

77 A B Figure 24. Example of erroneously increasing connection weights generated by PWGC in which a series of mediated connections exists. chain network possessing a series of mediated connections from Neuron V weights are erroneously over emphasized progressively down the chain of mediated connections. Note: erroneous direct connections, as described in Figure 23, would also be inferred by PWGC but were omitted from B for clarity.

PAGE 78

78 CHAPTER 3 VERIFICATION OF S LICE STABILITY Although the experimental protocol was designed to limit the potential effect s of chemical washin and tissue necrosis, and was validated with mathematical models in Chapter 2, further validation of the protocol is required to ensure that our subsequent statistical analysis will be meaningful. I verify slice stability by plotting the ictal event number versus the start time of the event A linear trend is expected indicating that the activity is stable or stationary with respect to time. However, if washin of K+ or 4 AP and washout of Mg2+ is gradual, a super linear trend of ictal onset would be observed. Additionally, if tissue slowly dies over the course of a recording session, a sublinear trend would be observed. Control Slice Stability The stability of the control slice s was first ass essed In order to evaluate stability a serie s of three plots was generated from the electrophysiological data collected for each slice. Initially, the ictal event number was plotted versus the start time of the event using MATLAB. A simple linear fit was performed on the data and superimposed on the graph. One such plot, which is representative of the normally laminated control slices, is presented in Figure 31 As illustrated, the data show that the ictal events occur at a fairly regular, rhythmic rate. This i ndicat es that washin of the recording ACSF and simultaneous washout of the dissection ACSF have no appreciable effect on the rate of ictal events. While Figure 31 is repr esentative of the data for near ly all of the control slices, one appear s to have a change in ictal event generation rate over the course of the recording. Figure 32 presents this slice. T he plot appears to consist of three linear

PAGE 79

79 phases. The first phase occur s from 0 to 6 minutes. Beginning around 6 minutes into the recording, the slope of the linear trend appear s to increase drastically. After approximately 2 minutes of this increased linear activity, a near instantaneous drop occurs where the frequency of i ctal events returns to the earlier rate. This pattern warranted further investigation. The data from the recording w ere split into each of the three phases and analyzed separately. All three phases demonstrate linear trends. The slopes of the linear fi ts are 6.5, 25, and 7.6 for the first, second, and third phases, respectively. It is unlikely that this change of slope can be attributed to ACSF wash in/washout for a number of reasons. First, although there is a marked change in slope between the phas es the trend within each phase i s always linear. An e ffect from washin/washout and tissue necrosis would be expected to be an exponential rise early in the recording or asymptotic decrease late in the recording, respectively. Second, the increase in s lope occur s after almost 5.5 minutes, well after the effect of washin/washout would be expected from the models discussed in Chapter 2. Third, the second change in slope revert s back to a state very similar to the first phase of the recording and does so after around the 7 minute mark. Tissue necrosis is unlikely after only 7 minutes. Finally, no other control slices display appreciable nonlinear phases T hus it is unlikely that only one of 11 such slices is affected by these phenomena. In order to further verify that the changes in linear slope demonstrated in Figure 32 were not attributable to washin/washout and tissue necrosis, two further plots were constructed. In addition to the rate of ictal event generation, the duration of ictal events could be modified by washin/washout and tissue necrosis. If these processes have an appreciable effect, ictal events durations should become longer during the period while

PAGE 80

80 the dissection ACSF is replac ing the recording ACSF. Additionally, as neurons die, ic tal event s should shorten in duration because a monotonically decreasing number of neurons would be able to sustain activity. To address this issue, I first plotted ictal event length vs. the event number ; the results are shown in Fi gure 33 N o r eliable trend in ictal durations is evident from this plot. I also compared ictal event durations vs. ictal event start times ; the results are shown in Figure 34. Although the rate of event generation elevates around 5.5 minutes into the recording and returns to the earlier rate near the 7minute mark, there is no signif icant change in ictal durations. These plots were generated for all slices. Figure 3 5 and 36 present these plots for the representative data set initially plo tted in Figure 31. Finally, three slice s from three different control animals exhibited a n interesting stairstep trend. Figure 37 illu strates one such stability plot. Although not strictly linear, the stair step pattern is consistent over the recording period and does not appreciably change. Next, plots of event duration vs. event number and event duration vs. event start time for the se slices were inspected to see if there w as a correlation between the stair step pattern and event lengths These plots are shown in Figure 38 and Figure 39 respectively No apparent relationship between ictal event duration and the pattern is seen and neither washin/washout nor tissue necrosis a p pear s to affect ictal activity, indicat ing that the activity remai n s stable for these slices as well Low Dose Slice Stability Initial stability analysis for low dose slices began with examining ictal event number vs. ictal event start time. This analysis demonstrates a very pronounced stair step trend in five of the seven slices from four of the five animals Further, the remaining animal and two low dose slices demonstrate a less pronounced stair step pattern

PAGE 81

81 instead of the simple linear trend present in the majority of control slices. Figure 310 sho ws one example of the more pronounced stair step stability plots Again, although not simply linear, the stair step pattern is consistent and does not appreciably change, even over an extended recording time exceeding 40 minutes. Next, plots of ictal even t duration vs. ictal event number were generated for low dose subject s Figure 311 presents this plot for the data set used to construct Figure 3 10. Although the majority of events are short in duration (approximately 1 4 m uch longer event durations (approximately 25 40 every 20 30 events No increase in event duration early in the recording or decrease in event duration late in the recording wa s observed, indicating that activity remained stable. Finally ictal event duration vs. ictal event start time plots were generated for low dose subjects. Figure 312 presents this plot for the data set used to construct Figures 3 10 and 311. Again, no event duration increase early in the recording or event durat ion decrease in the recording was observed. This further corroborates that the epileptiform activity generated is stable. Surprisingly, the longer duration events occur at relatively periodic rates, with an approximate period of 6 minutes in this example. Further, these long ev ents are followed by clusters of short duration events which are in turn followed by periods of no activity. High Dose Slice Stability Lastly, stability for high dose irradiated subjects was assessed. O nly o ne s lice from th i s group exhibited a simple linear trend. The stair step pattern was observed i n six of the eleven slices and four of the seven animals The remaining four slices demonstrated a primary linear trend with a weak stair step influence. Figure 313

PAGE 82

82 presents al l three stability plots for the high dose subject demonstrating a linear trend and Figure 31 4 present s these plots for a slice demonstrating the stair step. Regardless of whether the pattern demonstrated was linear, stair step, or an amalgam of the two, the trend remained stable over the course of the recording period. This indicates that electrophysiological activity from high dose subjects, like control and low dose subjects, was stable over time. Stability Conclusions Overall, a nalysis of slice stabili ty revealed acceptable stationarity of slice electrophysiology during the recording periods. Normally laminated control tissue demonstrated typically constant rates of activity A few control slices displayed changes in ictal initiation rate but maintained linearity. Three of the eleven control slices from three of the seven control animals displayed a stair step pattern in the occurrence of events. Examining the duration of the ictal events over time further validated stability and corroborated initial results suggesting that neither washin/washout n or tissue necrosis played a significant role in epileptiform activity over the course of the recording periods. Five of seven slices from four of five low dose irradiated animals exhibited the stair step pattern on their ictal event number vs. event starting time plot. Further analysis showed that longer duration events occurred at relatively regular intervals, both in terms of event number and time. Additionally, these long duration events preceded clusters of subsequent short duration events, which were in turn followed by periods of no activity. These trends remained unchanged over recording periods exceeding 40 minutes thus indicating that the activity was stable.

PAGE 83

83 Finally, high dose subj ects displayed a simple linear trend in one slice and the stair step pattern in six of eleven slices from four of seven animals. The remaining high dose subject displayed a primarily linear trend with a weak stair step influence. Regardless the trends remained constant over the course of the recording This analysis suggests that neither wash in/washout nor tissue necrosis had an appreciable effect on the electrophysiological activity of any slice over any of the recording periods.

PAGE 84

84 Figure 31 I ct al event number vs. ictal event start time for a normally laminated slice used to verify activity stability. The blue plot is the ictal event number vs. the start time. The red line is a linear fit of the data. As expected, the data signify that the was h in / washout process did not have an appreciable effect on ictal activity.

PAGE 85

85 A B C D Figure 32. Control slice demonstrating three linear phases ictal event start time over the entire 13.3minute recording. T he slice activity has three dist inct linear phases and are color coded red, yellow, and green for phase 1, phase 2, and phase 3, respectively. phase from A from the beginning until approximately 6 minutes into the recordi ng. The slope of the linear fit for this phase Enlarged plot of the second phase from A, from approximately 6 to 8 minutes during the recording. The slope of the linear fit for this phase is 25. from the third phase of A. Here the activity returns to a state similar to the first phase with a slope of 7.6.

PAGE 86

86 Figure 33. Ictal event duratio n vs. ictal event number. Data are from the same recording used to generate Figure 32, which appeared to have three dis tinct activity phases. N o observable increase in ictal durations occur over the course of earlier events and no observable decreases occur over the course of later events, further suggesting ACSF washin / washout and tissue necrosis d id not have an appr eciable effect o n epileptiform activity over the course of the recording period.

PAGE 87

87 Figure 34. Ictal event duration vs. ictal event start time, verifying epileptiform activity wa s stable over the c ourse of the recording period. Although an increased rate of event occurrence wa s observed during the second phase of the recording period (Figure 32 panel A ( yellow and Figure 32 panel no trend in event duration exists.

PAGE 88

88 Figure 35 Ictal event duration vs. ictal event number for the same normally laminated slice data used in Figure 31. Again the data do not indicate a trend attributable to washin / washout.

PAGE 89

89 Figure 36 Ictal event duration vs. ictal event start time number for the same normally laminated slice data used in Figure 31 and Figure 35 This plot factors in th e start time of the ictal event rather than just the event number. Again, the data suggest that washin / washout has no appreciable effect on activity.

PAGE 90

90 Figure 37. Plot of ictal event number vs. ictal event start time for a control slice subject illustrating stair step pattern. Although not strictly a linear trend, the trend wa s maintained for a 2 0 minute period without significant deviation, suggesting stable activity during the recording period.

PAGE 91

91 Figure 38. Ictal event durations vs. ictal event number for a control subject. Data used are the same used to generate Figure 3 7. No definite trend wa s seen for ictal durations, indicating that washin/washout and tissue necrosis were not factors leading to the stair step pattern.

PAGE 92

92 Figure 39 Ictal event duration vs. ictal event start time for a control subject. Data are f rom the same subject used to generate Figure 37 and Figure 38. This clearly makes the periodic event clustering more apparent

PAGE 93

93 Figure 310. Plot of ictal event number vs. ictal event start time for a low dose irradiated subject illustrating characteristic pronounced stair step pattern. Although not simply linear, the trend wa s maintained for over 40 minutes without significant deviation, suggesting stable activity during the recording period.

PAGE 94

94 Figure 311. Ictal event durations vs. ictal event number for a low dose irradiated subject. Data used are the same used to generate Figure 310. As expected, an increasing trend early in the recording period and decreasing trend late in the recording period were absent, signifying stability over the recording period. Interestingly, long events appear to occur every 20 30 events.

PAGE 95

95 Figure 312. Ictal event duration vs. ictal event start time for a low dose irradiated subject. Data are from the same subject used to generate Figure 310 and Figure 311. This more clearly illustrates the periodic trend. Interestingly, the events with long durations occur at a relatively periodic rate and precede series of subsequent short duration events.

PAGE 96

96 A B C Figure 3 1 3 Characteristic stability plots for the high dose subject exhibiting a simple plot demonstrates a linear trend and was observed in most control slices The ictal event duration vs. ictal event number plot for this subject, display s no trends in event duration o ver time, suggest ing duration vs. ictal event start time plot also display s no trends over the recording period. Together, these data suggest that high dose slices, like control and low dose slices, display stable activity.

PAGE 97

97 A B C Figure 31 4 Characteristic stability plots for high dose subject exhibiting the stair step event start time plot, demonstrating the stair step pattern also seen in nearly all low dose slices and over half of ictal event duration vs. ictal event number plot for this subject display s no trend in event duration over time suggest ing stability. s trend changes in event duration over the recording period. However, as seen with low dose slices, long duration events appear to precede clusters of short duration events, which are then followed by periods of inactivity. Together, these data suggest that high dose slices, like control and low dose slices, display stable activity.

PAGE 98

98 CHAPTER 4 CLASSIC AL AND SPATIAL RESULTS AND AN ALYSIS Results of Classical Analysis Techniques In this section I present the results of the experiments and analyze the data using classical, non spatial metrics. These metrics have previously been employed by researchers to quantify the differences in s eizurelike activity in a number of experimental protocols, suc h as the comparing the efficacy of different AEDs or electrical stimulation protocols. Ictal Events per Recording Time The occurrence of ictal events were modeled as being generated by a Poisso n process a stochastic process that counts the average number of events occurring per a unit time interval, where the events are generated independent ly at a constant average rate. I assumed this P oisson process to be homogeneous, with each class having a fixe d but unknown rate parameter. T he maximum likelihood estimator for this parameter is simply the mean number of events per minute for each class, and t he standard error is the square root of the sample mean. Control subjects produced an average 3.2298 1.7972 events per minut e, low dose subjects produced an average 4.0957 2.0238 events per minute, and high dose subjects produced an average 2.6091 1.6153 events per minute. These results partially confirm the hypothesis that dysplastic tissue w ould have a higher mean rate of ictal events per time and low dose subjects would present a greater increase in ictal rate than the high dose subjects. Although low dose subjects presented an elevated mean rate of ictal events compared to control subjects high dose subjects produc ed a lower mean rate of ictal events. However, this decline in ictal

PAGE 99

99 rate generation in high dose subjects is not completely surprising, as behaving high dose subjects have not been shown to display epilepsy in vivo Average Event Length Because event lengths are always positive a distribution over event lengths with probability mass on negative values is inappropriate. Thus data were modeled using a log normal distribution. First, the lowest and highest 2.5% of samples were di scarded to remove outliers The n the event duration mean and standard deviation were calculated for each group, as well as the 95% confidence interval for the event length, according to Equations 41, 4 2, and 43 respectively These results are shown in Table 41 E [ X ] e 12 2 (4 std dev .[ X ] e 1 2 2* e2 1 (4 CI95%[ X ] e 1.96, e 1.96 (4 w here and are the mean and standard deviation of the natural logarithm of the event durations respectively T hese calculations show that t he low dose group had both a shorter mean event length and smaller standard deviation than either the control or high dose gr oups. Distribution of Event Lengths To further investigate differences in event durations between groups the distributions of event lengt hs were compared. Again, the lowest and highest 2.5% of samples from each group were discarded. Then normalized histograms were generated and are shown in Figure 41. Visually, three observations are apparent. First, the minimum event lengths generated by high dose subjects was longer than the minimum lengths generated by either the low dose or control subject s. Second, over

PAGE 100

100 12% of the low dose events were of minimum length. Finally, the high dose and control groups contain a larger portion of longer ev ents than the low dose group. Next, the event lengths were fit to a log normal distribution ; the result ing fits are shown in Figure 42 along with a rug plot for reference. Although the most likely event duration was similar for all three groups, the fitted probability density functions show that events from low dose subjects are much more likely to be shorter in duration that either control or high dose events. Further, events from high dose appear to likely be longer in duration than control events. Finally, two simple two sample t test s were performed The first was done to investigate whether the measured event durations have significantly different means across groups It was found that the mean ictal duration in low dose subjects was different than control and high dose subjects (pcontrol, low dose = 2.746 x 1052, plow dose, high dose = 6.997 x 1055control, high dose The second twosample t test was to investigate whether the mean of the m easured low dose events was less than the high dose or control events. The second twosample t test did not assume the classes had equal variances. The results indicate that there is significant evidence that mean ictal duration of the low dose group was lower than both the control and high dose group ( p Together, these results reject the hypothes es that events generated by dysplastic tissue would have a longer mean duration and have a distribution of event lengths tending towards longer event lengths compared to controls Surprisingly, low dose subjects the only group to display epilepsy in vivo ha ve the lowest mean event duration, standard deviation, and 95% confidence interval and have a distribution of

PAGE 101

101 event lengths that tend s towa rds shorter events compared to the other two groups Additionally, control and high doses groups have a similar mean event duration, standard deviation, and distribution of event lengths. Average Inter Event Interval The inter event interval s were modeled using an exponential distribution, which is commonly used to model the time between event occurrences that are generated independently with a constant rate. First, the lowest and highest 2.5% of samples were discarded. Then, the event duration mean and standard deviation for each group were calculated, as well as a 95% confidence interval for the mean, according to Equations 44, 4 5, and 46, respectively Results are shown in Table 42 E [ X ] 1 (4 4 std dev .[ X ] 1 (4 5 CI95%1 1 2 n2 n ; 0.025 2 1 2 n2 n ; 0.975 2 (4 6 where is the reciprocal of the sample mean and 2 k;x is the critical value of the chi squared distribution with k degrees of freedom at the level specified by x From these calculations, the low dose group had shorter mean inter e vent interval s, a smaller standard deviation, and smaller confidence intervals than the control or high dose groups. The control and high dose groups had similar means, standard deviations, and confidence intervals. Distribution of Inter Event Intervals To further investigate potential differences in inter event intervals between groups, the distributions of intervals were compared. Again, the lowest and highest 2.5% of

PAGE 102

102 samples from each group were discarded. Then normalized histograms were generated and are shown in Figure 43 As with the event duration histograms, three observations are apparent. First, the histograms appear to follow an exponential distribution for all groups Second, a higher percentage of inter event intervals from low dose subjects a re mu ch smaller than those from control or high dose subjects. F inally, with the exception of the shortest bin, the normalized histograms for inter event intervals from high dose and control subjects appear to be similar. Next, the inter event durations were fitted to an exponential distribution ; the result ing fits are shown in Figure 44 along with a rug plot for reference. T he fitted probability density functions show that inter event interval s from low dose subjects are much more likely to be shorter in duration tha n either control or high dose events. Further, inter event intervals from high dose and control subjects appear to be very similar The two sample Kolmogorov S mirnov confirmed that the inter interval distribution of the low dose subjects is significantly different than the control and high dose subjects (pcontrol, low dose = 3.552 x 1014, plow dose, high dose = 1.985 x 1011 and that the control and high dose subjects are not significantly different from one another at the 5% confidence level (pcontrol, high dose These results partially confirm the hypothes e s that the mean inter event intervals in dysplastic subjec ts tend to be shorter in irradiated subjects and that the inter event interval distribution of irradiated subjects tends towards shorter lengths Although the mean inter event intervals from the low dose group is significantly lower than the control and high dose groups, the mean inter event intervals of the contr o l and high dose groups are not significantly different from one another.

PAGE 103

103 Recovery Ratio After the lowest and highest 2.5% of samples from the three groups were discarded, scatter plot s of inter event intervals vs. event durations w ere generated and are shown in Figure 45. Because the points for the groups frequently overlapped, m ultiple scatter plots were necessary for readability Although some points from control and high dose subj ects are distal from the central group t he vast majority of points for all groups cluster near the origin. No linear trend is observed in the groups, which was verified by calculating the correlation between event duration and inter event interval for each group ( event duration, inter event interval = 0.0725, 0.2367, and 0.0088 for control, low dose, and high dose groups Next, the recovery ratios for each group and their mean and standard deviations were calculated and are shown in Table 43. The sample means of the three groups are within 7.23 % of each other, although the standard deviations of the groups vary much more To gain insight into the distribution of recovery ratios, histograms of the recovery ratios were generated and are shown in Figure 46. Recovery ratios in all groups tend to be short (less than 5 uniform recovery ratio distribution than the experimental groups. Additionally, the decay in recovery ratios f rom low dose subjects is the most rapid, with over 59% being less than 3 F inally, because the variables were not independent, no attempt was made to fit the se data to a distribution. Instead, two nonparametric two sample Kolmogorov Smirnov test s were again performed All groups are statistically different from one another at the 5% confidence level ( pcontrol, low dose = 3.8237 x 105, pcontrol, high dose =

PAGE 104

104 0.0018, plow dose, high dose = 0.00 23, tests were performed to ascertain whether the di stributions of irradiated group recovery ratios were less than the distribution from the control group. Only the distribution from the low dose group was found to be statistically likely to be smaller than the high dose group ( p These results m ostly reject the three hypotheses made about recovery ratios. First, no linear trend in dicating an increased recovery ratio with increasing event lengths is observed. Second, both low and high dose groups have longer mean recovery ratios than the control group. Finally, the mean recovery ratio of the low dose group is greater than the high dose group. Number of Field Potentials per Epileptic Event The rates of positive negative and total field potentials among each group w ere assessed. First, the number of extrema occurring on each channel for every event was calculated and divided by the event duration. Next, since poor tissue electrode contact could cause a large number of channels to register significant ly fewer or no extrema but not an elevation in the number of local field potentials, the extrema rate of the top six channels were averaged together for each event. This mean was used as the extrema rate for the event. The mean and standard deviation for the positive, negative, and total extrema rate were calculated and are presented in Tables 4 4 4 5, and 4 6, respectively Next, the distributions of the positive extrema rate, negative extrema rate, and total extrema rate between each group were compar ed. After discarding the upper and lower 2.5% of samples, normalized histograms were generated and are shown in Figure 4 7, 4 8, and 4 9 for control, low dose, and high dose groups, respecti vely. The

PAGE 105

105 extrema rates were then fitted to a normal distribution and the resulting fits are shown in Figure 410, along with a rug plot for reference. Finally, a series of twosample t tests were performed to investigate whether the measured number of positive, negative, and total extrema rates during ictal events had significantly different means across groups. These results are presented in Table 47 and show that both the mean positive and negative extrema rates are significantly different among group s. Similarly, t he mean total extrema rate of the control group is significantly different than the high dose group. However, there are not significant differenc e s between t he total extrema rates of the control and low dose groups or the high dose and low dose groups. The hypothesis that dysplastic tissue would exhibit a higher extrema rate was partially confirmed. Low dose subjects displayed an elev ated mean positive extrema rate compared to the control group, and the mean positive extrema rate in high d ose subjects was lower than in the control group. The low dose group also displayed the lowest mean negative extrema rate, and the control group had the highest. Finally the mean total extrema rate was found to be statistically different only between t he control and high dose groups. Thus, the low dose group produced a disproportionately higher positive extrema rate and disproportionately lower negative extrema rate, compared to the other groups. Summary of Classical Analysis The results from classical analysis provide quantitative results that greatly expand upon the previous counter intuitive anecdotal findings that high dose irradiated animals, which exhibit severe cortical dysplasia, are functionally more similar to control animals than low dose irr adiated animals which exhibit milder cortical dysplasia. Specifically,

PAGE 106

106 low dose irradiated animals differ notably from control and high dose animals in that they have a higher mean number of ictal events per recording time, lower mean event length, small er mean inter event interval, greater number of positive extrema per epileptic event, lower number of negative extrema per ictal event. These findings suggest that a Goldilocks Principle applies to the amount of cortical dysplasia necessary to generate epilepsy. In other words, there may exist an ideal degree of dysplasia necessary to generate epilepsy, while dysplasia that is too minor or too severe fails to do so. Further, the combinat ion of a higher mean number of ictal events and smaller inter event interval in low dose subjects may be act to sat urate the inhibitory networks allowing seizure events to occur in vivo Results of Spatial Analysis Techniques In this section I present fur ther results of the experiments and analyze the data using spatial metrics. Previously, these spatial metrics were used exclusively on normally laminated slices to study the phases of ictal events and propagation of epileptiform ac tivity. Initiation Site After calculating the initiation location for each event in all slices, two types of plots were generated for each recording to analyze the data. The first plot generated was a heat map, which shows the percentage of ictal events starting at each locati on on the MEA. If one or more dominant initiation foci were observed, the cell layer at which it existed was estimated using micrographs of the slice placed over the MEA The second type of plot used was a graph of initiation channel number vs. event number. D ominant initiation foci are only observed in three control subjects (two from one animal and the three low dose subject s ( all from different animals and no

PAGE 107

107 high dose subject s. Of these, all of the control subjects and one of the low dose subjects demonstrate a strongly dominant focus; the other foci are weakly dominant. No other subjects displayed localized initiation foci. Figure 411 displays both plot types and the associated micrograph for one of the control subjects displaying dominant initiation foci and Figure 4 12 displays both plot types and associated micrograph for a different control subject without dominant initiation foci. In all control subject s, the dominant initiation foc i are located in the deep (or infra granular layer s. Due to severe delaminat ion in irradiated groups estimation of the initiation focus location wa s based upon the layers that would be present in normally laminated subjects Initiation foci are located i n the approximated deep l ayers for one of the low dose subjects and in the upper, or supragranular, layers in two low dose subjects. A final property observed i n some slices displaying more than one dominant initiation foc us was temporal dominanc e Figure 4 11 part C illustrates this property. Here, two dominant initiation foci are present. Initially, electrode 40 serve s as the dominant initiation foc us However after approximately onesixth of the events occur electrode 42 becomes the dominant initiation focus and remains so throughout the experiment. T emporal dominance of initiation foci was observed in one control subject and one low dose subject. All other slices exhibiting dominant initiation foci merely had a higher number of initiations occurring at that location and did not display any temporal dominance. Propagation Two previously employed metrics for propagation were utilized: wave speed, a quantitative measure, and propagation pattern, a qualitative analysis of the flow of

PAGE 108

108 epileptiform activity throughout the tissue. Because complex wave patterns could arise from p henomena, such as reflected waves, I chose to only include the first 500 ms of each event in the analysis. Next, the time of the first peak on each channel for all events was calculated. These data were then used to calculate the wave speeds and generate a visual representation of the wave propagation pattern. Wave speed Using the first peak times, a number of wave speed metrics were utilized First, the horizontal and vertical wave speeds through the three groups were calculated. Wave speeds slower tha n 1 mm/s and faster than 275 mm/s were discarded from analysis. After discarding the upper and lower 2.5% of samples, the means for each group were calculated and two sample Kolmogorov Smirnov test s performed. Mean speeds are presented in Table 48 and results of the twosample Kolmogorov Smirnov tests between groups are presented in Table 49. In all groups, and in accordance with literature, the mean speed of vertical wave propagation was greater than horizontal pr opagation within the same group. These results were also found to be statistically significant by the twosample Kolmogorov Smirnov test ( p = 2.0475 x 1083, 1.1113 x 1028, and 8.3396 x 1024 for control, low dose, Again, the high dose group appeared to be more f unctionally similar to controls than the low dose group, with high dose subjects having the lowest mean wave speeds in both horizontal and vertical directions The low dose group presented the fastest mean wave speed times, and the mean horizontal speed w as faster than the mean vertical wave speed for either the control or high dose groups. Next, using mic rographs for registration, the mean horizontal wave speed through was

PAGE 109

109 ca lculated and twosample Kolmogorov Smirnov tests were performed. As with initiation sites, estimation of cell layer in delaminated subjects was based upon approximation to normally laminated slices. Mean horizontal wave speeds through each layer grouping are presented in Table 4 10 and results of the twosample Kolmogorov Smirnov tests are presented in Table 411 and Table 4 12. Interestingly and differing from the experiments performed by Contreras and Llins in guinea pig cortex the horizontal waves propagated fastest t hro ugh the supragranular layers. Additionally a key functional difference in the low dose group was that the propagation speed in the infragranular layer was slowest while in the control and high dose groups, wave propagation through the granular layer was slowest. Results from two sample Kolmogorov Smirnov tests showed strong statistical significance between mean propagation speeds through each layer classification within each group and corresponding layers across groups. Propagation pattern In slices in which dominant initiation foci were present, colored gradients were used to qualitatively assess the propagation pattern originating from the foci As previously mentioned, only thr ee control and three low dose irradiated slices demonstrated dominant initiation foci. To create the colored gradients, the first peak times for all the ictal events originating from the dominant foci were averaged together, while events originating at ot her locations wer e discarded for this analysis. Two main differences were observed between the control and low dose groups. First, wave propagation in control slices was typically uniform from the initiation site and the wave front eventually was observed at all locations. In low dose slices, the wave front did not move uniformly throughout the slice and some areas did not observe a

PAGE 110

110 peak until far later in time, even though the sites were adjacent to one another. Second, wave propagation originating from a dominant focus in low dose slices appeared to move much more quickly through the slice than in control slices. Figure 41 3 illustrate s these findings. Termination Site The procedure used to analyze termination locations was similar to the initiation lo cation analysis After calculating the termination location for each event in all slices, three types of plots were generated for each recording to analyze the data. The first two plots were the same type used in the initiation analysis: a heat map which shows the percentage of ictal events terminating at each location on the MEA and a graph of termination channel number vs. event number. If one or more dominant termination foci were observed, the cell layer at which it existed was estimated usin g micrographs of the slice placed over the MEA. The final plot type plotted both initiation and termination channel locations vs. event number. This was used to ascertain whether a dominant initiation focus was coupled to a dominant termination focus. In general, a dominant termination focus was present in subjects also having a dominant initiation focus. Dominant termination foci were observed in two of the three control subjects that displayed dominant initiation foci and in two of the three low dose subjects with dominant initiation foci. However, one control subject that did not display a dominant initiation focus displayed two dominant termination foci. No dominant termination foci were observed in any high dose subject Figure 4 14 displays termi nation analysis plots from the cont rol subject used in Figure 411, and Figure 415 displays termination analysis plots for a control slice displaying no dominant termination foci.

PAGE 111

111 Termination foci were found in all anatomical layers in control subjects i n the infragranular or granular layers in low dose subjects and were not observed in high dose subjects Again, estimation of cell layer in delaminated subjects was based upon approximat ion to normally laminated slices. One control subject had a single dominant termination focus located in the supragranular layers, one control subject had a single termination focus in the infragranular layers, and one control subject had one dominant termination focus in the granular layer and anoth e r in the infr agranular layers. A dominant termination focus was seen in the infragranular layers of one low dose slice and the granular layer of another low dose slice. Like initiation foci, termination foci also demonstrated temporal dominance. Only one control subj ect, shown in Figure 414 panel C and one low dose subject had a termination focus displaying strong temporal dominance. Two control slices and one low dose slice displayed weak termination temporal dominance. The other low dose slice with a dominant ter mination foc us merely had a higher percentage of events ending at that location and did not display temporal dominance. Finally, coupling of a termination focus to an initiation focus was examined. Figure 4 14 panel D illustrates this occurrence. Here the dominant initiation focus located at column 8 row 2, is strongly coupled to the dominant termination focus, located at column 1 row 2. Strong coupling was observed in only one control subject and one low dose subject. Weak coupling was observed only in one control dose slice. The hypothesis that normally laminated subjects would display highly focused initiation and termination sites while dysplastic slices would display more diffuse initiation and termination sites was partially rejected. Dominant initiation foci were

PAGE 112

112 observed in a minority of subjects while no high s displayed dominant initiation or termination locations. The majority of slices in all groups displayed diff use initiation and termination sites throughout the tissue. Summary of Spatial Metrics The three phases of epileptiform activity: initiation, propagation, and termination in both normally laminated and dysplastic cortex were analyzed. As with conflicting reports in literature about the existence of dominant initiation foci, some control and low dose subjects displayed such a phenomena, while the majority of subjects and all high dose irradiated subject s displayed diffuse initiation sites. In agreement with literature, w ave propagation speeds were found to be faster traveling vertically across cell layers than horizontal ly through them. Wave velocities and propagation pattern in dysplastic cortex was provided and compared to normall y laminated tissue, further showing functional similarities between high dose irradiated and control slices. Finally, dominant termination foci were observed in a small minority of control and low dose subjects, but not in high dose subjects.

PAGE 113

113 Table 41. Ictal duration mean, standard deviation, and 95% confidence interval Group Mean Standard deviation Control 2. 5623 1.1783 0.9866 5.4930 Low dose 1.7541 0.6459 0. 8183 3.3111 High dose 2.7853 1.5925 0. 8526 6.8573 Table 42 Inter event interval mean, standard deviation, and 95% confidence interval Group Mean Standard deviation Control 12.1508 12.1508 11.2180 13.2055 Low dose 7.8289 7.8289 7.2921 8.4277 High dose 11.7032 11.7032 10.8879 12.6141 Table 43 Recovery ratio mean and standard deviation Group Mean Standard deviation Control 4.6 889 5. 8627 Low dose 5.0280 9.1528 High dose 4.8501 7.5043 Table 44 Positive extrema rate mean and standard deviation Group Mean (positive Standard deviation 95% Confidence interval Control 4.8246 1.5838 4.6967 4.9296 Low dose 5.4449 1.7579 5.3008 5.5199 High dose 4.5027 4.5028 4.3742 4.5884 Table 45 Negative extrema rate mean and standard deviation Group Mean (negative Standard deviation 95% Confidence interval Control 4.6345 1.5287 4.5024 4.7218 Low dose 3.9797 1.1564 3.8969 4.0373 High dose 4.5028 2.2451 4.2656 4.5437 Table 46 Total extrema rate mean and standard deviation Group Standard deviation 95% Confidence interval Control 8.8641 2.7362 8.6364 9.0389 Low dose 8.6512 2.2476 8.4994 8.7769 High dose 8.4680 3.1588 8.2192 8.6286

PAGE 114

114 Table 47 Results of t wo sample t t ests on measured extrema rate means Positive e xtrema r ate Negative e xtrema r ate Total e xtrema r ate Control vs. Low d ose 5.3335 x 10 13 1.7433 x 10 18 0.0993 Control vs. High d ose 4.1827 x 10 5 0.0262 0.0052 Low d ose vs. High d ose 4.7256 x 10 32 3.2117 x 10 8 0.0876 Table 48 Mean horizontal and vertical wave speeds through tissue Group Mean horizontal speed ( mm Mean vertical speed ( mm Control 33.7410 43.0345 Low d ose 45.8539 52.9533 High dose 30.9584 34.2324 Table 49 Results of two sample Kolmogorov Smirnov tests on wave propagation speeds between groups Mean h orizontal s peed Mean v ertical s peed Control vs. Low d ose 8.8093 x 10 116 1.7101 x 10 66 Control vs. High d ose 3.3318 x 10 81 9.0150 x 10 131 Low Dose vs. High d ose 8.0999 x 10 143 2.4330 x 10 169 Table 410. Mean horizontal propagation speed through cell layers Mean supragranular layer Mean granular layer Mean infragranular layer Control 33.7783 29.9520 33.0372 Low D ose 48.2533 45.9235 40.0221 High D ose 30.8285 22.4406 23.3160 Table 411. Results of intra group two sample Kolmogorov Smirnov tests on horizontal wave propagation speeds between cell layers p value Control supragranular vs. Control granular 9.5269 x 10 40 Control supragranular vs. Control infragranular 4.2351 x 10 14 Control granular vs. Control infragranular 1.3191 x 10 12 Low dose supragranular vs. Low dose granular 2.7296 x 10 13 Low dose s upragranular vs. Low dose infragranular 8.2805 x 10 62 Low dose granular vs. Low dose infragranular 1.3214 x 10 9 High dose supragranular vs. High dose granular 5.9380 x 10 11 High dose supragranular vs. High dose infragranular 1.4839 x 10 31 High dose granular vs. High dose infragranular 1.3852 x 10 12

PAGE 115

115 Table 412. Results of inter group two sample Kolmogorov Smirnov tests on horizontal wave propagation speeds between cell layers Supragranular layer Granular layer I nfragranular layer Control vs. Low dose 2.3866 x 10 29 4.4604 x 10 38 2.3577 x 10 22 Control vs. High dose 6.3804 x 10 53 5.8648 x 10 35 2.0605 x 10 167 Low dose vs. High dose 2.2467 x 10 78 2.0162 x 10 75 2.0494 x 10 130

PAGE 116

116 A B C Figure 41. Histograms of event duration normalized by number of events. For readability, events over 10 s in length have been omitted. histogram of event event durations in low durations in high dose subjects.

PAGE 117

117 Figure 42 L og normal fit s of the event durations for each group. The ticks below the x axis correspond to events occurri ng with the respective duration. For readability, this figure omits events greater than 20 s The mean event duration from control and high dose subjects are significantly different from low dose subjects (p insignificant higher probability of having shorter events than high dose ow dose subjects wa s statistically lower than the mean from either high or control groups ( p

PAGE 118

118 A B C Figure 43. Histograms of inter event intervals normalized by number of event intervals Normalized histogram of inter event Normalized histogram of inter Normalized histogram of inter event intervals in high dose subjects.

PAGE 119

119 Figure 44. E xponential fit of the inter event intervals for each group. The ticks below the x axis correspond to event intervals occurring with the respective duration. For readability, this figure omits interval s greater than 45 s. The mean inter event interval from l ow dose subjects wa s statistically different from the means of either control subjects or high dose subjects (p bot the mean inter event interval from control subjects was not statistically different from high dose subjects (p = 0. 0603

PAGE 120

120 A B C D Figure 45. Recovery s catter plots The recovery scatter plot is a plot of inter event intervals vs. time. plotted, with the control group Zoomed in region with low dose group pl h igh dose plotted last.

PAGE 121

121 A B C Figure 46. Histograms of recovery ratio for subjects. Recovery ratios larger than 20 control subjects. Histogram of recovery ratios from high dose subjects.

PAGE 122

122 Figure 47. Histograms of positive extrema rates in control, low dose, and high dose groups.

PAGE 123

123 Figure 48. Histograms of negative extr ema rates in control, low dose, and high dose groups.

PAGE 124

124 Figure 49. Histograms of all extrema rates in control, low dose, and high dose groups.

PAGE 125

125 A B C Figure 410. Normal fits of the positive, negative, and total extrema rates for each extrema rate for each group.

PAGE 126

126 A B C Figure 411. Analysis plot s for a control subject displaying a stro ngly dominant initiation foc us row 8 column 2 Teflon harp to ensure good tissueelectrode contact. The dominant foc us app ear ed to arise from the infra event number illustrati ng that initially, a second focus was strongly dominant for the first sixth of the events b efore the primary focus became dominant.

PAGE 127

127 A B C Figure 4 1 2 of the slice, using a Teflon harp to ensure good tissuePlot of initiation channel vs. event number illustrating absence any dominant initiation foci, including temporally local dominant initiation foci.

PAGE 128

128 A B C D Figure 4Control slice h eat map showing the dominant initia formed from averag ing first peak times for events originating at row 8, column 2. Note that the wave propagates rather uniformly from the origination point and the wave front is observed at all l Low dose slice heat map Colored gradient formed from averaging first peak times for ev ents originating at row 1, colu mn 2. Note that the wave does not propagate uniformly throughout the slide Additionally, the wave appears to propagate m uch faster to the preferred areas as compared to the control slice.

PAGE 129

129 A B C D Figure 414. Analysis plots for a control subject displaying a dominant termination foc us termination events showing that the electrode located at row 1 column 2 was dominant harp to ensure good tissueelectrode contact. The dominant termination foc us appear s to arise from the sup ra termination channel vs. event number illustrating that the termination focus was dominant throughout the recording locations as a function of event number. From this plot, the dominant termination focus appears to be strongly coupled to the initiation focus.

PAGE 130

130 A B C Figure 415. Analysis plots for a control subject displaying diffuse termination locations. there appears to be no coupling between initiation and termination sites.

PAGE 131

131 CHAPTER 5 GRANGER CAUSALITY RESULTS AND ANALYSIS Results From Pairwise Granger Causality Analysis In this section I present the results of pairwise Granger Causality analysis in an attempt to uncover previously unobserved localized driving regions within slice during ictal events. While these analysis techniques, originating in the economics field, h ave recently gained prominence in the neuroscience community their application to acute slice dat a is novel. Both quantitative comparisons, such as differences in PWGC values and distributions between experimental groups and anatomical cell layers, and qualitative comparisons, such as the spatial locations of driving PWGC sites, are made. Analysis Method s Pairwise Granger causality was calculated using a custom program written in our laboratory using the C programming language. The recorded data, which were post processed with custom automated tools I wrote and visually verified as discussed in Chapter 2, were used. In order to reduce computation time, data were first downsampled from 25 kHz to 250 Hz P airwise G ranger causality then was calculated using a moving window of 200 ms with a 50% overlap. The output of pairwise Granger causality produces a single complex valued spectral matrix based upon the total number of trials, or in this case ictal events. Three steps were taken to simplify analysis. First, the magnitudes of the complex values were taken to provide real number values. Second, spectral information was collapsed through summation to provide the total PWGC values across all f r equencies. Finally, since events could have varying lengths, all events were at least one second in duration, and previous literature reports that the first part of an ictal event is the most

PAGE 132

132 first 1000 ms of time windows were used in the analysis. Subsequent calculations were then performed using custom programs I wrote in MATLAB. PWGC Mean Value s and Distributions T he mean PWGC source value for each group was calculated by first averaging over the f irst 1000 ms of time windows and all channels then truncating the top and bottom 2.5% of values. Results are shown in Table 51. PWGC sink mean values were not calculated separately since the total PWGC s ource and sink values are equal. Differing from t he trend of control and high dose subjects being functionally similar using classic and spatial metrics, I found that control and low dose subjects have similar PWGC values, differing by less than 0.5%, while high dose subjects w ere found to have an approx imately 9% decrease. Next probability density estimates of PWGC source and sink values by group were generated and are show n in Figure 51. T wo sample Kolmogorov Smirnov tests were performed on the t otal PWGC source and sink distributions. Results are presented in Table 52. S ignificant differences exist between both the PWGC source and sink distributions generated by control and high dose subjects and low dose and high dose subjects. However, the distributions of PWGC source values generated fro m the control and low dose subjects are far more similar in comparison ( source pcontrol, low dose and that the PWGC distribution of sink values among these two groups is not statistically significant ( sink pcontrol, low dose Although t hese results initially suggest a greater functional similarity between the control and low dose groups, a marked departure from the results of both classic and spatial metrics discussed previously, careful examination indicates that these findings

PAGE 133

133 are not unexpected. As previously discussed, a greater degree of cortical dysplasia is generated in high dose subjects than in low dose subjects. Therefore, the morphology of low dose subjects is more similar to the control subject s and more organized compared to the high dose subjects. S ince PWGC provides a measure of the causality between the neuron ensembles (more precisely the causality of the voltages recorded higher mean PWCG values can be interpreted as the occurrence of a more ordered system. This is because a system that is more organized, with units that deliberately caus e the activity of other units produces a higher mean PWGC value, while a more random system where no units exert influen c e over the activity of other units produces a low mean PWGC value. Thus, the mean PWGC values verify that the control subjects exhibit the most organized structure, the high dose subjects exhibit the most random and the low dose subjects are structurally far more similar to the controls. These findings are reinforced by the results of the twosample Kolmogorov Smirnov tests and visual inspection of the PWGC source and sink probability density estimates presente d Figure 5 1. Mean PWGC Values b y Cell Layer As set forth in Chapter 1, literature suggests that the infragranular layers are the site of initiation and driving force behind ictal events, while the supragranular layers are thought to mitigate epileptic act ivity. In order to examine this notion further, t he PWGC source and sink values of each of the gross cell layers across all groups were calculated and are presented in Table 53. Next, p robability density estimates based on these distributions were plott ed and twosample Kolmogorov Smirnov tests were

PAGE 134

134 performed on the distributions Results are presented in Figure 52 and Table 54 respectively While the magnitude of the PW GC values varies among groups, four trends are m aintained across all groups First, the supragranular layers have the lowest PWGC source values Second, the infragranular layers have the highest PWGC source values Third, the infragranular layers have the lowest PWGC sink values Finally, the granular layers have the highest PWG C sink values While these trends hold for all groups, results from the two sample Kolmogorov Smirnov tests s how that the PWGC sink distributions are not statistically significant among any cell layers in low dose subjects or between the supragranular and infragranular layers in high dose subjects. Additionally, the PWGC source distributions ar e not statistically significant. The first thr ee observed trends were expected from previous reports As discussed earlier, literature indicates that the infragranular layers are necessary for ictal event propagation Thus, it was expected that the infragranular layers would act as both the chief PWGC source and weakest PWGC sink S imilarly, since the supragranular layers are thought to mitigate ictal activity once it has begun and only the first 1000 ms of the ictal events were included in this analysis it was unsurprising that supragranular layers s erve d as the weakest PWGC source. Had a later or lo nger time segment of ictal activity been included in the analysis, it may have been shown that the supragranular layers were activated to a greater extent and acted as a stronger PWGC source While it was initially surprising that the granular layers served as the site of strongest PWGC sink values, closer analysis provides a reasonable rationale. I t was

PAGE 135

135 expected that the supragranular layers would be the site of greatest PWGC source values for three main reasons. First, as presented in the overview of PW GC, mediated connections generate an artificial increase in PWGC values Second, literature report s that ictal wave s propagat e from the infragranular layers to the supragranular layers through the granular layers Third, experimental results in this work discussed in Chapter 4 confirm the literature and demonstrate that ictal wave s propagate from the infragranular layers to the supragranular layers through a connection mediated by the granular layers. One possibility for the apparent discrepancy may be t hat while the infragranular layers exert an ictal prone effect to the granular layers, the supragranular layers simultaneously exert a control ling signal. Thus, the granular layers are effectively caught in the crossfire between ictal inducing and ictal controlling regions and their associated signals Finally, results from the two sample Kolomogorov Smirnov tests performed on the PWGC distributions by cell layer further corroborate anatomical and functional diffe rences between the three groups. In control slices, b oth source and sink distributions for all cell layers are significantly different suggesting that these three regions are anatomically and functionally distinct. Meanwhile, t he two irradiated groups show a lost of significance in some PW GC distributions indicating that some areas are affected, both anatomically and functionally, by the protocol. This is expected due to the loss of regular cortical morphology. Further leading credence to the notion of a Goldilocks principle applying t o the amount of cortical dysplasia necessary to cause differences leading to epilepsy low dose subjects, which exhibit less cortical dysplasia, had more distr ibutions become homogeneous than high dose subjects which exhibit

PAGE 136

136 greater cortical dysplasia. T his loss of significance in experimental groups is seen almost exclusively among PWGC sink distributions, which may arise from literature reports that the model preferentially induces changes in the inhibitory network. PWGC Changes O ver T ime Finally, since ictal events are dynamic processes, I examined changes in P WGC driving regions over time. In order to simplify analysis, I compared between PWGC source values between the supragranular and infragranular layer. This was done by plotting the ratio of the mean PWGC source values of the supragranular layers divided by the sum of the PWGC source values of i nfragranular layers and the PWGC source values of the supragranular layers for each time window Thus, a value greater than 0.5 means that t he supragranular layers were exerting gr eater control on activity than the infragranular layers and a value less than 0.5 means that the infragranular layers were dominant over the supragranular layers A combined plot was generated for all three groups and is shown in Figure 54. As shown from the graph, control and high dose subjects do not have a change in dominant driving region during the first 1000 ms of ictal activity, but differ in the layer of dominance. I nfragranular layers serve as the dominant driving region in control subjects, while supragranular layers serve as the dominant driving region in high dose subjects. Low dose subjects on the other hand beg in being driven by the infragranular layers and at some point between 400 and 600 ms transi tion to being driven by the supragranular layers. All three of these results were unanticipated. In the case of control subjects, and to a lesser extent the high dose subjects, it was expected that the inf ra granular layers, known to be necessary for ictal event propagation, to drive activity early in the event.

PAGE 137

137 A s the event progressed in time, it was expected that the supragranular layers, thought to mitigate ictal activity, would activate and contribute a greater role in driving the event. Since low dos e subjects are the only ones to exhibit epilepsy in vivo and only the first 1000 ms was included in the analysis, it was expected that the s upragranular layers would be the chief driving force throughout. A possible explanation for these aberrations may be due to the aforementioned errors in PWGC analysis. Initiation Site vs. PWGC Source Focus One of the prime questions I seek to answer through the use of Granger causality analysis is whether the observed initiation sites and do minant foci, if present, correspond to areas of maximum source values of Granger causality. To perform this analysis, total PWGC source was calculated at each electrode for the first 1000 ms of time windows and averaged together. Data were normalized and used to generate PWGC source value heat maps. These images were then compared to the plots used to ascertain if and where dominant initiation foci existed. In total, eight control slices, five low dose slices, and five high dose slices exhibited PWGC source foci. Of these, one control and one low dose slice had their dominant PWGC source in the upper layers, each of the three groups had one PWGC in the middle layers, and the rest of were in the deep layers. Further, all but one low dose slice displaying a single dominant initiation focus also displayed a localized PWGC source focus. However, in the two control slices displaying two dominant initiation foci, a more diffuse PWGC source pattern was seen. It is possible that, in these slices, two separ ate processes generated the distinct foci, thus confounding the PWGC source image analysis.

PAGE 138

138 In support of my central hypothesis that ictal driving regions do not coincide spatially with ictal initiation sites, two interesting phenomena were observed and ar e illustrated in Figure 53. First, the PWGC source foci do not correspond to the dominant initiation focus. Second, slices without dominant initiation foci often have concentrated PWGC source foci. These findings may hold significant importance in a cl inical setting when resorting to surgical resection of brain tissue to deal with intractable epilepsy. Specifically, if applicable in vivo localization and removal of Granger source areas, which may not correspond to seizure epicenters residing within dy splastic regions, may mitigate the loss of brain tissue while simultaneously improving surgical outcomes. Further, the finding that a localized ictal driving region can exist in tissue without a dominant initiation focus may be beneficial at targeting cor tical areas for surgical resection in patients who present with cortical epilepsy in the absence of an area of dysplastic tissue. Results From Conditional Granger Causality Analysis Finally, conditional Granger causality analysis was performed on the ictal event data. The main goal of incorporating CGC analysis was an attempt to remove mediated connections that could have confounded the aforementioned PW GC analysis. I n this section I apply the same techniques performed on PWCG data to CGC data, report the results, and note salient differences between Granger causality methods. Analysis Methods T he same post processed and manually verified data were used for CGC analysis. Processing methods and parameters similar to those described above with PWGC was perf ormed with some modifications As with PWGC, data were first downsampled from 25 kHz to 250 and then processed through the CGC algorithms using a custom program

PAGE 139

139 written in the C programming language. Again, a 200 ms window with 50% overlap parameter was used. However, CGC is a more complex analysis technique than PWGC and requires a far greater number of calculations to tease out mediated connections Due to its intensive computational requirements, spectral information was not computed using CGC. E ve n omitting spectral calculations the custom CGC program took over a week to finish processing the data on a 24 node computing cluster Again, only the first 1 000 ms of time windows were used in the analysis. Differing from PWGC, the output of the conditional Granger causality program produces only real values and additionally generates an accompanying matrix indicating the statistical pvalues of the corresponding CGC value obtained. For analysis, only CGC values with associated pvalues less than 0.001 were used. Subsequent calculations were performed using custom programs I wrote in MATLAB. CGC Mean Values and Distributions The mean CGC value for each group was calculated by averaging over the first 1000 ms of time windows and all channels then truncating the top and bottom 2.5% of values. Results are shown in Table 55 Next probability density estimates of CGC source and sink values by group were generated and are show in Figure 55 Finally, two sample Kolmogorov Smirnov tests were performe d on the total CGC source and sink distributions Results from the statistical tests are shown in Table 56 The mean CGC values show three marked differences when compared to the mean PWGC using the same data. First, the overall mean CGC values are dramatically smaller over four orders of magnitude, than the mean PWGC values. Second, instead of the control and high dose groups being similar to one another, the two dysplastic groups have much greater commonality with each other S pecifically the m ean CGC

PAGE 140

140 values of the low dose and high dose subjects are within approximately 30% of each other, while the control group mean CGC value is over 226% times larger. Third, the low dose group has the lowest mean CGC value while the high dose group had the l owest mean PWGC value. The departures from the earlier discussed PWGC values may be the result of the removal of CGC values having an associated p value of greater than 0.001, differences between CGC and PWGC processing (such as the removal of confounding mediating connections generating erroneous valu combination of the two Thus, CGC may be giving a more accurate result of mean causality in the groups. Mean CGC Values b y Cell Layer The mean CGC source and sink values of each gross cell layer of all groups were calculated and results are presented in Table 57. Probability density estimates based on the CGC value distributions of each cell layer were plotted and twosample Kolmogorov Sm irnov tests were performed on the distributions. Results are presented in Figure 5 6 and Table 58, respectively. The t rends observed in the results from PWGC were not preserved. Instead, the infragranular layer s exhibit both the lowest mean CGC source and sink values and with the exception of the low dose source group, the greatest mean CGC values are found in the supragranular layer The statistical tests indicate that the CGC source value distributions are not significantly different bet ween the infragranular and granular layers in low dose and high dose groups and the granular and infragranular layers in the high dose group. Similarly, the CGC sink value distributions are not significantly different between the supragranular and granular control group and granular and infragranular layers in the high dose group.

PAGE 141

141 Thus, CGC results differ greatly from the PWGC results and suggest that the supragranular layer is the chief driving region of activity during ictal th e early phase of ictal events, rather than the infragranular layer Although appearing to directly defy both literature and PWGC results indicating that the infragranular layers are necessary for ictal event propagation, there are at least two possible reasons the CGC results may be complimentary, rather than conflicting. F irst, it is possible that the infragranular layers trigger a region in the supragranular layers that then continue to drive the event. Second, the supragranular layers may be responding to ictal activity in an attempt to abate the event. Additionally the differences in observed dominant source values with CGC may be attributable to the fact that the process successfully removed the influence of the mediated connections that were artificial ly inflating the infragranular layer source values in PWGC analysis. CGC Changes O ver T ime Finally, the changes in dominant CGC driving layers over time were assessed. As with PWGC, a nalysis was simplified by comparing the ratio of the mean CGC source va lues of the supragranular layers divided by the sum of the CGC source values of infragranular layers and the CGC source values of the supragranular layers for each time window. A combined plot was generated for all three groups and is shown in Figure 58 Likely due to CGC removing the confounding effects of mediated connections, the CGC changes over time more closely follows the expected pattern previously discussed T he control and high dose subjects exhibit a functionally similar pattern, with the high dose subjects displaying an apparent lag. I n these groups, the supragranular layer appear s to be the driving region early during the ictal events, howev er as time

PAGE 142

142 progressed, processes shift and the infragranular layers serves as the dominant region befor e being once again usur ped by the infragranular layers. Meanwhile, the low dose subjects began ictal events with a slight bias towards the supragranular layers as the driving region. Dominance quickly shif t s to the infragranular layers and continues throug hout the analyzed time window T his functional difference in the low d ose group is also expected, since it suggests that the supragranular layers thought to suppress ictal activity, are unable to sufficiently activate and thus abate the event. Ini tiation Site vs. CGC Source Focus In order to provide deeper investigation into whether dominant driving regions within slice correspond to dominant initiation foci, heat maps were generated from total CGC source values in the first 1000 ms of time windows and compared to initiation plots. In total, four control subjects, two low dose subjects, and four high dose subjects demonstrated strong localized CGC driving foci. Again, as with PWGC, CGC show s strong driving foci in locations other than the dominant initiation focus and localized driving foci in slices that d o not have a dominant initiation focus. Figure 57 provides two such examples. Two expected differences are observed between the CGC analysis and the PWGC analysis. First, some slices do not d isplay dominant driving CGC foci, but do display dominant driving PWGC foci. Second, some slices absent of dominant PWGC foci display strong CGC driving foci. Additionally, an unexpected change from the PWGC also is noted. PWGC driving regions often appeared to be located towards the middle of the electrode array, while CGC driving regions were not restricted to the center region of the array. This discrepancy may arise due to two possible mechanisms. First, mediating connections may confound PWGC anal ysis. Since waves propagate from the lower layers to the upper layers through the middle layers,

PAGE 143

1 43 PWGC may erroneously over estimate the causal values of the center layers. Second, CGC may be able to successfully remove the over estimated value of the mediated connections towards the center of the array, while electrodes on the border of the array have less possible connections to remove. As with PWGC analysis, results from CGC analysis provide further evidence that the central hypothesis is correct. Summary of Granger Causality Analysis Pairwise and Conditional Granger Causality were used to analyze spontaneous epileptiform activity recorded from both normally laminated and dysplastic cortical acute rat cortical slices Some c onflicting results arose from the two techniques, but were likely due to known shortcomings with PWGC Both techniques provided evidence suggesting key functional differences, almost certainly arising from morphological changes between normally laminated and dysplastic slices. Most importantly, the findings from this anal y sis suggest that the central hypothesis was confirmed.

PAGE 144

144 Table 51 Mean pairwise Granger Causality values of each group Mean PWGC Value Control 344.4622 Low d ose 342.9090 High d ose 311.3465 Table 52 Results of two sample Kolmogorov Smirnov tests on pairwise Granger Causality distributions between groups PWGC Source Values PWGC Sink Values Control vs. Low dose 0.0335 0.0694 Control vs. High dose 2.4679 x 10 79 1.4588 x 10 64 Low Dose vs. High d ose 1.8455 x 10 81 1.1910 x 10 51 Table 53 Mean PWGC in each cell layer Mean supragranular PWGC Value Mean granular layer PWGC Value Mean infragranular PWGC Value Control s ource 295.9778 344.6329 370.4592 Control s ink 352.3007 372.7259 318.5575 Low d ose s ource 266.6269 326.1246 351.9793 Low dose s ink 328.2719 334.2669 317.7433 High dose s ource 274.0517 327.9370 405.4087 High dose s ink 354.9639 409.3771 326.6135 Table 54 Results of two sample Kolmogorov Smirnov tests on pairwise Granger Causality distributions between cell layers PWGC Source Values PWGC Sink Values Control supragranular vs. Control granular 0.0140 0.0300 Control supragranular vs. Control infragranular 4.1510 x 10 6 0.0224 Control granular vs. Control infragranular 0.0068 0.0018 Low dose supragranular vs. Low dose granular 0.0290 0.0875 Low dose supragranular vs. Low dose infragranular 2.3955 x 10 6 0.2163 Low dose granular vs. Low dose infragranular 0.2704 0.0825 High dose supragranular vs. High dose granular 0.0187 0.0364 High dose supragranular vs. High dose infragranular 1.6708 x 10 9 0.0754 High dose granular vs. High dose infragranular 0.0064 0.0045

PAGE 145

145 Table 55 Mean conditional Granger Causality values of each group Mean CGC Value Control 0.1764 Low d ose 0.0538 High d ose 0.0778 Table 56 Results of two sample Kolmogorov Smirnov tests on conditional Granger Causality values between groups CGC Source Values CGC Sink Values Control vs. Low d ose 2.6076 x 10 20 7.9976 x 10 23 Control vs. High d ose 7.6837 x 10 5 0.0023 Low Dose vs. High d ose 3.2594 x 10 33 3.0699 x 10 1 3 Table 57. M ean CGC in each cell layer Mean supragranular CG C Value Mean granular layer C GC Value Mean infragranular C GC Value Control s ource 0.1519 0. 1343 0. 1118 Control s ink 0.1475 0.1422 0.1190 Low dose s ource 0.0775 0.0815 0.0592 Low dose s ink 0.0993 0.0569 0.0418 High d ose s ource 0.1088 0.1080 0.0921 High dose s ink 0.1287 0.0992 0.0957 Table 58 Results of two sample Kolmogorov Smirnov tests on conditional Granger Causality values between cell layers C GC Source Values CG C Sink Values Control supragranular vs. Control granular 0. 0 160 0.0908 Control supragranular vs. Control infragranular 2. 0 7 6 5 x 10 4 0.0211 Control granular vs. Control infragranular 0.0331 0.0310 Low dose supragranular vs. Low dose granular 0.5618 0.0033 Low dose supragranular vs. Low dose infragranular 3.8213 x 10 4 4.0837 x 10 10 Low dose granular vs. Low dose infragranular 0.0198 0.0406 High dose supragranular vs. High dose granular 0.9427 0.0044 High dose supragranular vs. High dose infragranular 0.0075 0.0012 High dose granular vs. High dose infragranular 0.0533 0.5434

PAGE 146

146 A B Figure 5 1. P robability density estimates of pairwise Granger Causality values by group. Probability density estimate of PWGC source values. Probability densi ty estimate of PWGC sink values.

PAGE 147

147 A B C D E F Figure 52 Probability density estimates of pairwise Granger s ource and sink values among groups by cell layer Probability density estimates of PWGC source values in control subjects Probability density estimates of PWGC sink values in control subjects. Probability density estimates of PWGC source values in low dose subjects Probability density estimates of PWGC sink values in low dose subjects. Probability density estimates of PWGC source values in high dose subjects. Probability density estimates of PWGC sink values in high dose subjects.

PAGE 148

148 A B C D Figure 53. Comparison between initiation focus and PWGC sourceInitiation site heat map of a c ontrol slice exhibiting strongly dominant initiation focus in th e infragranular layer Normalized t otal PWGC source plot of slice in A demonstrating that the causal driving focus is separate from the dominant initiation site. Initiation site heat map of a control slice showing a diffuse initiation pattern. Normalized total PWGC source plot of slice in C demonstrating that even though a dominant initiation site is absent, a strongly localized PWGC source is present

PAGE 149

149 Figure 54 PWGC driving region over time. Values greater than 0.5 indicate that the supragranular layers are driving the ictal event while values less than 0.5 indicate that the infragranular layers are driving the event.

PAGE 150

150 A B Figure 55. Probability density estimates of conditional Granger Causality values by density estimate of C density estimate of C GC sink values.

PAGE 151

151 A B C D E F Figure 56 CGC Source and Sink values by layer. CGC CGC CGC CGC sink values in low dose subj CGC CGC sink values in high dose subjects.

PAGE 152

152 A B C D Figure 57 Comparison between initiation focus and C Initiation site heat map of a control slice exhibiting strongly dominant initiation focus in the C GC source plot of slice in A demonstrating that the CG driving focus is may be a diffuse process even if a dominant initiation site is present showing a diffuse initiation patternC GC source plot of slice in C demonstrating that strongly localized C GC source s may be present, even though a dominant initiation site is absent.

PAGE 153

153 Figure 58 C GC driving region over time. Values greater than 0.5 indicate that the supragranular layers are driving the ictal event while values less than 0.5 indicate that the infragranular layers are driving the event.

PAGE 154

154 CHAPTER 6 CONCLUSIONS Conclusions This work provides a detailed analysis and comparison of electrophysiological activity during seizurelike events in normally laminated and in utero irradiation induced dysplastic acute rat cortical slices. Novel application of spatial metrics were performed on dysplastic tissue and compared with normally laminated controls. Further, pairwise Granger causality and conditional Granger causality analy sis were applied to data collected using MEAs from both normally laminated and dysplastic cortical slices. In general the results indicate that the activity from low dose irradiated subjects, or mildly dysplasti c tissue is functionally more dissimilar fr om controls than from high dose irradiated subjects, or severely dysplastic tissue. This was observed particularly with classic metrics, such as the mean number of ictal events per recording time, mean event length, mean inter event interval, bias in LFP extrema peaks during events and spatial metrics, such as wave speed propagation through cortical layers These results are consistent with previous reports that in utero exposure to lower doses of radiation generates spontaneously epileptic animals, whil e higher doses of radiation, although generating more severe dysplasia does not. A number of novel contributions to the field were provided, including providing wave speed calculations and qualitative comparisons of wave propagation in dysplastic cortical tissue to that of normally laminated. M ost notable was the application of Granger causality analysis to acute in vitro slice recordings. Holding the most promise for potential clinical applications were the findings sugges ting that the central hypothesis : that localized areas may serve to drive ictal events in the absence of a well

PAGE 155

155 define d dominant ictal initiation focus and that even if a well defined dominant initiation focus exists its location does not necessarily coincide with the location of the driving regionis true Future Work The focus of this work was to further characterize differences in seizure like activity in normally laminated and dysplastic cortical slices caused by in utero irradiation, which has been demonstrated to generate epilept ic animals in vivo These techniques could be employed to compare o ther current models of epilepsy, including those that have also been observed to generate epileptic animals, such as in utero application of MAM and kindling and those that have not, such as n eonatal freeze lesioning model, to one another. Additionally, thi s body of work has produced a rich source of data that can later be mined. A number of other data analysis methods could eas ily be applied to this data set. Recommendations include using principle component analysis coupled with clustering algorithms to characterize and compare individual ictal events between one another and experimental groups, spectral analysis of the ictal events using power spectral density or spectrograms, and finally PWGC and CGC spectral analysis to see if causality measures differ at known brain oscillations frequencies (e.g.: the alpha, beta, Expanding on the theme of using MEAs to examine hyperexcitable cortical tissue, additional experiments could be performed on neural tissue. Recommendations include compar ing electrophysiological activity in control, in utero irradiated, and secondhit animals, or animals which have suffered t wo distinct insults and has been demonstrated

PAGE 156

156 to have a higher success rate at generating spontaneously epileptic animals than singular insults. Finally, future work should focus on if the central hypothesis holds for in vivo animals and whether or not the distances separating ictal initiation site and driving region are clinically significant This should be approached on three fronts. First, more detailed and computationally intensive Granger Causality analysis of the present data set should be performe d. This includes the aforementioned spectral analysis, as well as p Second, in vivo studies on animal models should be performed. Finally, analysis of electrophysiological data from human patients should be analyzed with these techniques.

PAGE 157

157 LIST OF REFERENCES cortical formation: MR imaging features. AJNR Am J Neuroradiol 30:411. term synaptic plasticity causes suppression of epileptiform activity in rat hippocampal slices. Brain Res 998:5664. Differential radiosensitivity of stationary and migratory primitive cells in the brains of infant rats. Exp Neurol 22:52 74. CRC Press, Inc.. Amano S, Ihara N, Uemura S, Yokoyama M, Ikeda M, Serikawa T, Sasahara M, spontaneous limbic like seizures. Am J Pathol 149:329336. neocortical neurons. Neuroscience 63:151161. Diego, CA: Academic Press. Aykut Bingol C, Bronen R in occipital lobe epilepsy: implications for pathophysiology. Ann Neurol 44:6069. oscillation in neocortical slices : coupled local oscillators. J Neurophysiol 96:25282538. Imaging brain activity with voltageand calcium sensitive dyes. Cell Mol Neurobiol 25:245282. Bakker R, Schubert D cortical microcircuits based on microelectrode array data from slices of rat barrel cortex. Neural Netw 22:11591168. ion of heterotopic cell clusters in the hippocampus of rats exposed to methylazoxymethanol in utero. Epilepsy Res 39:87102.

PAGE 158

158 inhibition during convulsive activity induced by 4aminopyridine in neocortical slices. J Neurophysiol 73:14621467. 248. Burlington, MA: Elsevier Academic Press. Barkovich AJ, K classification scheme for malformations of cortical development. Neuropediatrics 27:5963. system for malformations of cortical development: update 2001. Neurology 57:21682178. patterns that are stable for many hours in cortical slice cultures. J Neurosci 24:52165229. Begley CE, Famulari M, Annegers JF, Lairson DR, Reynolds TF, Coan S, Dubinsky S, United States: an estimate from popul ation based clinical and survey data. Epilepsia 41:342351. GABA in an animal model of dysplastic cortex. Epilepsia 43:970982. Bikson M, Hahn PJ, Fox JE, Jefferys JG (2 maintenance of electrographic seizures. J Neurophysiol 90:24022408. alpha oscillations in awakebehaving macaques. J Neurosci 28: 9976 9988. initiation and spread of epileptiform discharges in three in vitro models. Brain Res Bull 69:161167. failure in an inhomogeneous neural network. PHYSICA D 155:83100. Vergleichende Lokalisationslehre der Grohirnrinde in ihren Prinzipien dargestellt auf Grund des Zellenbaues Leipzig: JA Barth.

PAGE 159

159 Canepari M, Bove M, Maeda E, Cappello M, Ka neuronal dynamics in cultured cortical networks and transitions between different patterns of activity. Biol Cybern 77:153162. Press. neuronal specific activator of Cdk5, display cortical lamination defects, seizures, and adult lethality. Neuron 18:2942. Chagnac of synchronized activity in neocortex and its control by GABA mediated inhibition. J Neurophysiol 61:747758. Chagnac by intrinsically bursting neurons in neocortex. J Neurophysiol 62:11491162. Chagnac layer 5 pyramidal neurons of rat neocortex have different morphological features. J Comp Neurol 296:598613. inhibitory synaptic activity in experimental heterotopic gray matter. J Neurophysiol 89:150158. term plasticity in pyramidal cells of dysplastic cortex. Epilepsia 48:141148. Chervi propagation of epileptiform discharges across neocortex. J Neurophysiol 60:16951713. networks show s pontaneously correlated activity patterns during in vitro development. Brain Res 1093:4153. Networks of neurons coupled to microelectrode arrays: a neuronal sensory system for phar macological applications. Biosens Bioelectron 18:627 634. Chisholm D, WHO effectiveness of first line antiepileptic drug treatments in the developing world: a populationlevel analysis. Epilepsia 46:751759. ical monitoring of membrane potential: methods of multisite optical measurement. Soc Gen Physiol Ser 40:7199.

PAGE 160

160 microelectrode array for monitoring electrogenic cells in culture. Biosens Bioelectron 5:223234. 310:685687. the Neocortex. In: Epilepsy (Schwartzkroin 423. Camridge, UK: Cambridge University Press. neocortical neurons in vitro. J Neurophysiol 48:13021320. Control of spatiotemporal coherence of a thalamic oscillation by corticothalamic feedback. Science 274:771774. spindle oscillations in cortex and thalamus. J Neurosci 17:1179 1196. sensitive dye imaging of neocortical spatiotemporal dynamics to afferent activation frequency. J Neurosci 21:94039413. Cowe closed head focal injury. J Neuropathol Exp Neurol 29:2142. D'Antuono M, Benini R, Biagini G, D'Arcangelo G, Barbarosie M, Tancredi V, Avoli M actions leading to hyperexcitability in a model of temporal lobe epilepsy. J Neurophysiol 87:634639. frequency stimulation reduces epileptiform synchronization in limbic neuronal network s. Neurobiol Dis 19:119128. seizures in the rat limbic system. Neurobiol Dis 8:9931005. two discrete sites generate epileptiform discharges in slices of piriform cortex. J Neurosci 19:12941306. potentials and unit discharges in cat cerebral cortex during natural wake and sleep states. J Neurosci 19:45954608.

PAGE 161

161 developing neocortex has a reduced capacity to recover from in utero injury in experimental cortical dysplasia. J Neuropathol Exp Neurol 63:12651273. the rat entorhinal cortex in vitro. Neuroscience 99:413422. Neuroscience. In: Handbook of Time Series Analysis: Recent Theoretical pp437460. Berlin: Wiley VCH Verlag GMBh & Co. Induced Status Epilepticus: a Chronic Model of Acquired Epilepsy. In: Models of Seizure and 432. Burlington, MA: Elsevier Academic Press. tability by astrocytes. Adv Neurol 79:573581. dimensional monitoring of spiking networks in acute brain slices. Exp Brain Res 142:268274. Egert U, Schlosshauer B, Fennrich S, Nisch W, Fejtl M, Knott T, Muller T, Hamm erle H term culture of the rat hippocampus on substrateintegrated multielectrode arrays. Brain Res Brain Res Protoc 2:229242. human epi leptic phenomena. Epilepsy Res Suppl 8:920. temporal patternforming systems. REP PROG PHYS 61:353 430. Farrell MA, DeRosa MJ, Curran JG, Secor DL, Cornford ME, Comair YG, Peacock WJ, Shields WD, Vinters HV (1 childhood epilepsy. Acta Neuropathol 83:246259. Ferrer I, Pineda M, Tallada M, Oliver B, Russi A, Oller L, Noboa R, Zjar MJ, Alc ntara circuit neurons in epilepsia partialis continua associated with focal cortical dysplasia. Acta Neuropathol 83:647652. masses induced by prenatal irradiation in the rat. Virchows Arch A Pathol Anat Histopathol 422:16.

PAGE 162

162 ray induced cell death in the developing hippocampal complex involves neurons and requires protein synthesis. J Neuropathol Exp Neurol 52:370378. Epileptic seizures and epilepsy: definitions proposed by the International League Epilepsia 46:470472. by distinct glutamate receptor subtypes and neuronal populations. J Neurophysiol 75:951957. al Rat Brain. New York, NY: Oxford University Press. Dobyns WB, Hirsch BA, Radtke RA, Berkovic SF, Huttenlocher PR, Walsh CA ral cortical neurons in human periventricular heterotopia. Neuron 21:13151325. French JA, Kanner AM, Bautista J, AbouKhalil B, Browne T, Harden CL, Theodore WH, Bazil C, Stern J, Schachter SC, Bergen D, Hirtz D, Montouris GD, Nespeca M, Gidal B, Marks WJ Turk WR, Fischer JH, Bourgeois B, Wilner A, Faught RE, Sachdeo RC, Beydoun A, Glauser TA, American Academy of Neurology Therapeutics and Technology Assessment Subcommittee, American Academy of Neurology Quality Standards Subcommittee, American Epilepsy S ociety Quality Standards Subcommittee, American Epilepsy Society Therapeutics and new antiepileptic drugs, I: Treatment of new onset epilepsy: report of the TTA and QSS Subcommittees of the American Academy of Neurology and the American Epilepsy Society. Epilepsia 45:401 409. disturbance of rat cerebral cortex following prenatal low dose gammairradiation: a qu antitative study. Exp Neurol 112:292298. doses of ionizing radiation decelerates neuronal migration in the developing rat brain. Int J Radiat Biol 70:5360. ntiepileptic drug treatment: outcomes and adherence. Pharmacotherapy 20:191S 199S. Dependence and Feedback between Multiple Time Series. Journal of the American Statistical Association 77:304313.

PAGE 163

163 ures of Conditional Linear Dependence and Feedback between TimeSeries. Journal of the American Statistical Association 79:907915. Gleeson JG, Allen KM, Fox JW, Lamperti ED, Berkovic S, Scheffer I, Cooper EC, Dobyns WB, Minnerath SR, Ross ME, Walsh CA (19specific gene mutated in human X linked lissencephaly and double cortex syndrome, encodes a putative signaling protein. Cell 92:6372. intensity. Nature 214:10201021. interactions in timeresolved fMRI data using vector autoregressive modeling and Granger causality mapping. Magn Reson Imaging 21:12511261. Golomb D, computational and experimental study. J Neurophysiol 78:11991211. neurons in rat barrel cortex. Exp Brain Res 115:4760. Familial schizencephaly associated with EMX2 mutation. Neurology 48:14031406. tric models and cross spectral methods. Econometrics 37:424438. S52. time optical imaging of natur ally evoked electrical activity in intact frog brain. Nature 308:848850. array multielectrode system designed for long term monitoring of extracellular single unit neuronal activity in vitro. Neurosci Lett 6:101105. electrode stimulation and recording in the isolated retina. J Neurosci Methods 101:3142. of the cerebral cortex. In: Epileptic Syndromes in Infancy, Childhood and 479. London, UK: John Libbeyerd.

PAGE 164

164 Gulys Kovcs A, Dczi J, Tarnawa I, Dtri L, Banczerowski Comparison of spontaneous and evoked epileptif orm activity in three in vitro epilepsy models. Brain Res 945:174180. spikes from a cortical epileptogenic focus. Science 176:424426. de Guzman P, Inaba Y, Baldelli E, de hyperexcitability within the deep layers of the pilocarpinetreated rat entorhinal cortex. J Physiol 586:18671883. Microdysgenesis in resected temporal neocortex: incidence and clinical significance in focal epilepsy. Neurology 38:10411047. transporter expression and spontaneous seizures in rats exposed to methylazoxymethanol in utero. Epilepsia 48:158168. Demos Medical Pub. auditory cortic Neurophysiol 83:26262638. variant EEG Granger causality for inspecting directed interdependencies of neural assemblies. J Neurosci Methods 124:2744. dimensional multi electrode array for multi site stimulation and recording in acute brain slices. J Neurosci Methods 114:135148. e development of the mammalian nervous system. I. Malformations of the brain, especially the cerebral cortex, induced in rats by radiation. II. Some mechanisms of the malformations of the cortex. J Comp Neurol 113:435469. Hirtz D, Thurman DJ, Gwinn Hardy How common are the "common" neurologic disorders? Neurology 68:326337. by 4 aminopyridine. Brain Res Dev Brain Res 85:6470. Holmes GL, Sarkisian M, BenAri Y, Liu Z, Chevassus Au Consequences of cortical dysplasia during development in rats. Epilepsia 40:537544.

PAGE 165

165 and record fro m cardiac cells in culture. Am J Physiol 247:H669H674. maldevelopment. Cereb Cortex 6:514523. polymicrogyria. J Neurophysiol 81:159173. cortical malformations. Epilepsy Res 36:165188. level prenatal X irradiation on postnatal development in the Wistar rat. Proc Soc Exp Biol Med 184:256263. cortical slice cultures recorded by planar electrode arrays. Bioelectrochemistry 51:107115. Kandel McGraw Hill, Health Professions Division. the rat visual cortex. I. Correlation among cell morp hology, intrinsic electrophysiological properties, and axon targets. J Comp Neurol 339:459474. novel mutation of the doublecortin gene in Japanese patients with X linked lis sencephaly and subcortical band heterotopia. Hum Genet 104:341344. histopathologic abnormalities and in vivo epileptogenicity in the in utero radiation model of rats is dose dependent. Epilepsia 45:583591. Kellinghaus C, Mddel G, Shigeto H, Ying Z, Jacobsson B, Gonzalez Martinez J, epileptogenicity in a rat model of cortical dysplasia. Epileptic Dis ord 9:1119. 571. upper layers of somatosensory cortex revealed with voltagesensitive dyes. J Comp Neurol 375:89108. Electroencephalographic characterization of an adult rat model of radiationinduced cortical dysplasia. Epilepsia 42:12211227.

PAGE 166

166 Krsek P, Maton B, Jayakar P, Dean P, Korman B Rey G, Dunoyer C, PachecoJacome Incomplete resection of focal cortical dysplasia is the main predictor of poor postsurgical outcome. Neurology 72:217223. dentification of refractory epilepsy. N Engl J Med 342:314319. pyramidal neurons in slices of rat visual cortex. I. Establishment of cell classes. J Neurosci 10:14071414. Lee KS, Schottler F, Collins JL, Lanzino G, Couture D, Rao A, Hiramatsu K, Goto Y, Hong SC, Caner H, Yamamoto H, Chen ZF, Bertram E, Berr S, Omary R, human neocortical heterotopia associated with seizures. J Neurosci 17:62366242. receptors in the disinhibited rat neocortex. J Neurophysiol 65:8795. Lehohla M, Russell V, Kellaway L, Govender A (200 evaluate glutamate receptor function in rat barrel cortex slices. Metab Brain Dis 15:305314. 11. Amsterdam: Elsevier. epilepsy. Dialogues Clin Neurosci 10:4762. Clinical and imaging features of cortical malformatio ns in childhood. Neurology 53:715722. pp315321. Burlington, MA: Elsevier Academic Press. Lomb Epilepsia 41:245253. 796. Lynch J pp455470. Singapore: Churchhill Livingstone Inc.

PAGE 167

167 Marn neuronal morphology in irradiationinduced cortical dysplasia: a Golgi Cox study. J Neuropathol Exp Neurol 62:11291143. brain development: an excitotoxic mouse model of microgyria and posthypoxic like lesions. J Neuropathol Exp Neurol 54:358 370. electrophysiology of pyramidal and sparsely spiny stellate neurons of the neocortex. J Neurophysiol 54:782806. ion of the adult brain (albino 46; discussion, 4467. pp351363. Burlington, MA: Elsevier Academic Press. underlying epileptogenesis. Sci STKE 2006:re12. 9: 319; discussion 3940. induced spontaneous epileptiform activity in combined rat entorhinal cortex hippocampal slices. Neuroreport 8:18571861. Meister M, Pine J, neuronal signals from the retina: acquisition and analysis. J Neurosci Methods 51:95106. of hippocampal mossy fibers in rats exposed to X irradiati on in utero. Brain Res Dev Brain Res 112:275280. contribution of local synaptic interactions. Neuroscience 12:11791189. ortical dysplasia associated with pediatric epilepsy. Review of neuropathologic features and proposal for a grading system. J Neuropathol Exp Neurol 54:137153. n, MA: lsevier Academic Press.

PAGE 168

168 404. Philadelphia, PA: Lippincott Raven. MomoseSato Y, Sato K, Sakai T, Hirota A, optimal voltagesensitive dyes for optical monitoring of embryonic neural activity. J Membr Biol 144:167176. Mosewich RK, So EL, O'Brien TJ, Cascino GD, Sharbrough FW, Marsh WR, Meyer FB, Jack CR, O'Brien PC (20 epilepsy surgery. Epilepsia 41:843849. 50. Cambridge MA: MIT Press. Md del G, Jacobson B, Ying Z, Janigro D, Bingaman W, Gonzlez Martnez J, contributes to epileptogenesis in human cortical dysplasia. Brain Res 1046:1023. Najm IM, Tilelli CQ, Oghlakian cortical dysplasia: a critical review of human tissue studies and animal models. Epilepsia 48 Suppl 2:2132. array recordings of cardiac action potentials as a high throughput method to evaluate pesticide toxicity. Toxicol In Vitro 20:375381. System, pp491679. New York: Springer. Noebels JL, Prince DA (1 axon terminal bursting. J Neurophysiol 41:12671281. using a 32 element microelectrode array. J Neurosci Methods 23:149159. multielectrode array for extracellular recording: application to hippocampal acute slice. J Neurosci Methods 93:6167. vents and genetic factors in epileptic patients with neuronal migration disorders. Epilepsia 35:965973. Rat Forebrain. San Diego, CA: Academic Press.

PAGE 169

169 Academic Press. responses in layer 2/3 of rat barr el cortex measured in vivo by voltagesensitive dye imaging combined with wholecell voltage recordings and neuron reconstructions. J Neurosci 23:12981309. microcircuit ele ctrodes. J Neurosci Methods 2:1931. termination of epileptiform activity in rodent neocortex in vitro involve distinct mechanisms. J Neurosci 25:81318140. of cell migration to the superficial layers of fetal monkey neocortex. J Comp Neurol 145:6183. 176. Optical imaging of epileptiform activity in experimentally induced cortical malformations. Exp Neurol 192:288298. term evolution of excitotoxic cortical dysgenesis induced in the devel oping rat brain. Brain Res Dev Brain Res 109:109113. Excitability changes and glucose metabolism in experimentally induced focal cortical dysplasias. Cereb Cortex 8:623634. Roeb using Granger causality and fMRI. Neuroimage 25:230242. in the brain using fMRI: Model sel ection, causality and deconvolution. Neuroimage. dysgenesis: a review. Epilepsy Res 32:6374. disruption of radial glia in rats. Dev Neurosci 19:521528. and calbindin D28immunoreactive neurons in experimental cortical dysplasia. Epilepsy Res 37:6371.

PAGE 170

170 Roper SN, Gilmore RL, Ho migration produce an increased propensity for electrographic seizures in rats. Epilepsy Res 21:205219. slices co ntaining experimentally induced cortical dysplasia demonstrate hyperexcitability. Epilepsy Res 26:443449. time measurement of PMA induced cellular alterations by microelectrode ar ray based impedance spectroscopy. Biotechniques 41:445450. Epileptic Disorders. Epilepsy Behav 2:201216. natal period. Am J Dis Child 141:969980. glutamatergic and GABAergic neurotransmission. Neurochem Res 28:347352. Schubert D, Staiger JF, Cho N, Ktter R, Zilles K, Luhmann HJ (200 specific intracolumnar and transcolumnar functional connectivity of layer V pyramidal cells in rat barrel cortex. J Neurosci 21:35803592. Dev Disabil Res Rev 6:268280. countries: where do we go from here? Bull World Health Organ 79:344351. of Neocortical Epilepsies. In: Neocortical Epilepsies (Williamson PD, Siegel AM, 139. Philadelphia, PA: Lippincott Williams & Wilkins. Organ 36. New York, NY: Oxford University Press. layer 5 pyramidal neurons. Science 251:432435. rmations of cortical development causing 1091.

PAGE 171

171 hippocampus using chaos control and adaptive techniques. IEEE Trans Biomed Eng 50:55957 0. vitro hippocampal models of epilepsy. Ann Biomed Eng 29:607618. sleeping and aroused brain. Science 262:679 685. stimulation of the isolated chicken retina. Vision Res 40:17851795. visual cortical slices revealed by optical imaging. Science 266:10571059. cortex in epilepsy. J Neurol Neurosurg Psychiatry 34:369387. specific pathways for the horizontal propagation of epileptiform discharges in neocortex. Epilepsia 39:700708. n patterns mediated by NMDA and nonNMDA receptors in rat neocortex. Epilepsia 40:14991506. cultured cells. Exp. Cell Res. 74:6166. Traub RD, Miles R, Jefferys JG (1 picrotoxininduced synchronized after discharges in the guineapig hippocampal slice. J Physiol 461:525547. 161. New York, N Y: Humana Press. neocortical slice. J Neurophysiol 80:978982. neocortical layers: an optical imaging study. J Neurophysiol 82:19651973. layer II III and layer V neurons of the cat neocortex. J Neurophysiol 74:11491166.

PAGE 172

172 y of recurrent multiple brain states in nonconvulsive status epilepticus. Clin Neurophysiol 118:27982804. uniform propagation of epileptiform discharge in brain slices of rat neocortex. Neuroscience 52:255262. Wagenaar D cultures with closedloop multi electrode stimulation. J Neurosci 25:680688. 29. Wenzel HJ, Robbins C and functional organization of the hippocampus in a p35 mutant model of cortical dysplasia associated with spontaneous seizures. J Neurosci 21:983998. Modern Mathematics for the engineer Hill. firing in rat neocortical pyramidal neurons. J Physiol 521 Pt 2:467482. Wolf P, Rothermel A, Beck real time monitoring of vital MCF 7 mamma carcinoma cells by impedance spectroscopy. Biosens Bioelectron 24:253259. ntial antiepileptogenic therapy: From tuberous sclerosis to common acquired epilepsies. Epilepsia. during development of the mammalian retina. Neuron 11:923938. Woolsey T field composed of discrete cytoarchitectonic units. Brain Res 17:205242. Wu JY, Falk CX, Cohen L, Tsa potential activity in invertebrate ganglia. Jpn J Physiol 43 Suppl 1:S21S29. population activity in rat sensory cortical slices. J Neurophysiol 86:24612474. interneurons in an animal model of cortical dysplasia. J Neurophysiol 96:569578.

PAGE 173

173 auma from two sources of disinhibition. J Neurophysiol 78:28042810. vitro model in neocortical slices. Epilepsia 41:15071513. properties of 4 aminopyridineinduced synchronous network activities in rat neocortex. Neuroscience 111:303313. reorganization following kainic acidinduced seizures during dev elopment. Brain Res Dev Brain Res 107:169177. excitatory and inhibitory neurotransmitters. Neuron 6:333344. excitatory synaptic activity is altered in fast spiking interneurons in experimental cortical dysplasia. J Neurophysiol 102:25142525. Neurosci 20:89258931. Zilles K ( Verlag.

PAGE 174

174 BIOGRAPHICAL SKETCH Born in Pensacola, F lorida John Nicholas Grimes moved to Crestview, F lorida when he was adopted at 5 months of age into a home by his loving family : parents, Tom and LeViniua; brother Ricky ; and sister Vicki Grimes graduated from Crestview High School in 1999 and mo ved to Gainesville, F lorida to attend the University of Florida where he graduated cum laude with a Bachelor of Science in Computer E ngineering in 2004. After a stint working for the United States Department of Defense, he enrolled in graduate school to pursue a Ph.D. in B iomedical E ngineering also at the University of Florida. Grimes received his Master of Engineering in 2009 and his Ph.D. in 2011