Citation
Biobehavioral sources of variance in presurgical neuropsychological performance among patients with temporal lobe epilepsy

Material Information

Title:
Biobehavioral sources of variance in presurgical neuropsychological performance among patients with temporal lobe epilepsy
Creator:
Moser, David J., 1967-
Publication Date:
Language:
English
Physical Description:
vi, 131 leaves : ; 29 cm.

Subjects

Subjects / Keywords:
Electrodes ( jstor )
Electroencephalography ( jstor )
Epilepsy ( jstor )
Hemispheres ( jstor )
Hippocampus ( jstor )
Memory ( jstor )
Neuropsychology ( jstor )
Seizures ( jstor )
Temporal lobe ( jstor )
Wechsler scales ( jstor )
Decision Making ( mesh )
Department of Clinical and Health Psychology thesis Ph.D ( mesh )
Dissertations, Academic -- College of Health Professions -- Department of Clinical and Health Psychology -- UF ( mesh )
Electroencephalography ( mesh )
Epilepsy, Temporal Lobe -- diagnosis ( mesh )
Epilepsy, Temporal Lobe -- surgery ( mesh )
Magnetic Resonance Imaging ( mesh )
Neuropsychological Tests ( mesh )
Predictive Value of Tests ( mesh )
Reproducibility of Results ( mesh )
Research ( mesh )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph.D.)--University of Florida, 1997.
Bibliography:
Bibliography: leaves 124-130.
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by David J. Moser.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
029433297 ( ALEPH )
50995085 ( OCLC )

Downloads

This item has the following downloads:


Full Text














BIOBEHAVIORAL SOURCES OF VARIANCE IN PRESURGICAL
NEUROPSYCHOLOGICAL PERFORMANCE AMONG PATIENTS WITH TEMPORAL LOBE EPILEPSY













By

DAVID J. MOSER


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


1997














ACKNOWLEDGEMENTS


Completion of this study would not have been possible were it not for the important

contributions of many individuals who deserve my recognition and gratitude. First and foremost, I would like to extend my sincere thanks to my dissertation chairperson, Dr. Russell Bauer, for the enormous amount of knowledge, time, effort, patience and humor that he brought to this project. I would also like to thank Dr. Eileen Fennell and Dr. Duane Dede for their insightful comments, support and valuable input during all phases of the study. The assistance and guidance of Dr. Robin Gilmore is greatly appreciated, as she was very giving of her knowledge and expertise regarding epilepsy and also allowed me access to Epilepsy Monitoring Unit resources. I would also like to extend my sincere gratitude to Dr. James Algina, who went to great lengths to provide valuable and timely statistical guidance.

Dr. Christiana Leonard also deserves recognition and thanks for her assistance and

generosity in allowing me access to the Image Processing Laboratory. I would also like to thank Dr. Tara Spevack for her valuable and thoughtful input throughout the study. Dr. Steven Roper, Mr. Timothy Lucas, Ms. Arlene Frank, Ms. Candace Lariz, Dr. Rita Jakus, Ms. Donna Lilly, Ms. Christie Snively, the graduate students in Dr. Russell Bauer's Neuropsychology Laboratory, and the Epilepsy Monitoring Unit technicians also played important roles in the formation and completion of the study.

Finally I would like to thank my wife, Becky, and our daughter, Amelia for their constant and unwavering love, support and good humor.















TABLE OF CONTENTS

page

A CKN OW LED G EM EN TS ............................................................................................. ii

A B ST R A C T ........................................................................................................ .............. v

CHAPTERS

1 OVERVIEW OF RELEVANT LITERATURE ....................................................... 1

Epilepsy: Description, Incidence, and Classification.................................... 1
Neuropsychological Assessment of Epilepsy Patients......................................... 4
Hippocampal Pathology and Neuropsychological Performance.......................... 25
Quantitative MRI Volumetric Studies in Epilepsy.............................................. 29
Hippocampal Volumes, Pathology, and Neuropsychological Performance........ 34 EEG Localization of Seizure Foci ...................................................................... 37
Neuropsychological Testing, MRI, and EEG..................................................... 50
Epilepsy Surgery at the University of Florida........................... ............... 52
Purpose of the Present Study ............................................................. ................. 54

2 METHODS........................... ......... ....................... 56

S ubjects................................................................................................................ 5 6
Measures........................................ . .... ...... ................ 56
Missing Neuropsychological Data ..................................................... ............... 58
M R I V ariables............................................................ ....... ............ ............... 59
EEG Variables ............................................. 59
Data Collection Procedures.............................. .................... 62
Experimental Hypotheses ................................................................................. ... 68

3 R E S U L T S ................................................................................................................. 78

Demographic and Illness-Related Variables .................................... ................. 78
Neuropsychological Variables............................... ................ 78
Hippocampal Volumetrics........................ .... .............. .... 84
EEG Variables........................................... .... 84

4 D ISC U SSIO N .................................. . ................................................ .................. 100

Discussion of Experimental and Exploratory Hypotheses................................ 100
General Discussion.............. ............................................... .............. ..... 113










Directions for Future Research........................... .............. ..... 117

APPENDIX......................... . ............. ............ .... 120

R EFE R EN C E S................................................................................................................. 124

BIOGRAPHICAL SKETCH ............................................................................................ 131














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


BIOBEHAVIORAL SOURCES OF VARIANCE IN PRESURGICAL
NEUROPSYCHOLOGICAL PERFORMANCE AMONG PATIENTS WITH TEMPORAL LOBE EPILEPSY

By

David J. Moser

December, 1997

Chairman: Russell M. Bauer, Ph.D.
Major Department: Clinical and Health Psychology

Most epilepsy centers use electroencephalographic (EEG), structural-anatomic (MRI),

and neuropsychological (NP) data to identify potentially resectable regions of epileptogenic tissue in epilepsy surgery candidates. In ideal cases, these data are convergent and identify a discrete region to be resected. Sometimes, however, NP data are less specific than EEG and MRI data in lateralizing and localizing seizure onset. In the current study, discriminant function analyses (DFA) were used to evaluate the statistical efficacy ofEEG, MRI, and NP in predicting side of seizure focus and eventual surgery in a group of complex partial seizure patients undergoing unilateral anterior temporal lobectomy (ATL). Subjects were then divided into two groups depending upon whether their NP data were convergent or divergent with EEG and MRI regarding prediction of eventual surgical hemisphere. These groups were then compared regarding demographic and illness-related variables.

Subjects were 61 surgical candidates with complex partial seizures who eventually underwent either right (n = 26) or left (n = 35) ATL. For each subject, scalp EEG onset was
v








coded and transformed to produce a single value reflecting degree of seizure onset laterality (SLI). Likewise, quantitative MRI data were used to produce a value reflecting amount of volumetric asymmetry between the two hippocampi (DHF). Neuropsychological performance was characterized by calculating average z-scores in language, verbal memory, nonverbal memory, motor and visuoconstructive domains. Then, SLI, DHF, and the five NP domain scores were entered as predictors of side of surgery and of side of epileptogenic focus using DFA.

Results indicated that SLI was a slightly better predictor of side of surgery and side of seizure focus than DHF, and that both were significantly superior to NP domain scores in this regard. Correct prediction rates generally improved when these SLI, DHF, and NP domain scores were used in combination with one another rather than in isolation. Group comparisons made between those subjects for whom NP data were convergent with EEG and MRI and those for whom these data were divergent yielded no significant differences on demographic or illnessrelated variables. Implications for making treatment decisions in epilepsy programs, and directions for future research, were discussed.














CHAPTER 1
OVERVIEW OF RELEVANT LITERATURE Epilepsy: Description, Incidence, and Classification


Epilepsy is a neurological disorder characterized by recurrent seizures (Locharernkul, Primrose, Pilcher, Ojemann, & Ojemann, 1992). A seizure occurs when a group of neurons becomes "irritable" and fires in repeated bursts (Engel, 1989), and may lead to clinically observable changes in behavior if sufficient numbers of neurons are involved. Epilepsy is the most common chronic neurological disorder in the United States, affecting more than 2 million adults and children (McIntosh, 1992).

Although widely used, the term "epilepsy" is not a diagnosis in itself, but is a general term, which subsumes many different syndromes (Kuzniecky & Jackson, 1995). In 1964, the Commission on Classification and Terminology (CCT) of the International League against Epilepsy (ILAE) sought to categorize these various seizure types in such a way that professionals worldwide could treat, research, and discuss these events using consistent criteria and terminology. Prior to that time, many separate classification systems had been developed for different purposes, a situation which gave rise to varying and confusing terminology (Wyllie & Luders, 1993). The efforts of the CCT were successful, and the classification system that was developed remained in use until 1981, when it was revised into the system which is currently in use. A brief description of the current system follows.

Currently, all seizures are initially placed in one of two large categories. A seizure is defined as "generalized" if its onset involves activation of a large number of neurons in both hemispheres. "Partial" or "focal" seizures, on the other hand, are those that begin with activation
1








2



of neurons in a limited part of one hemisphere. A second major classification concerns the presence or absence of alterations of consciousness during the seizure. Those accompanied by no such alterations are termed "simple", while those that produce changes in consciousness are defined as "complex." For the purposes of the present study, the following will concern only partial seizures.

According to the 1981 classification system, partial seizures are further subdivided into three categories. As can be inferred from the above, "simple partial seizures" are those that begin focally and involve no alteration of consciousness, while "complex partial seizures" (CPS) begin focally and do involve such changes either immediately or shortly after onset. Finally, there are seizures which are described as "secondarily generalized", which begin as simple or complex partial events and propagate until large areas of both hemispheres are involved. Simple and complex partial seizures are then described more specifically according to the presence or absence of concomitant events such as motor involvement, somatosensory experiences, autonomic signs, and cognitive and emotional disturbances (Commission on Classification and Terminology of the International League Against Epilepsy, 1981). Approximately 40 percent of all epileptic individuals suffer from CPS (Gastaut et al., 1975), and approximately 60 percent of these patients have seizure disorders which are refractory to medication (Rodin, 1968). Thus, finding an alternative treatment for patients with CPS is essential.

An increasingly available and effective treatment for patients whose epilepsy cannot be controlled with medication is surgical resection of an identified "epileptogenic focus", which has been defined as that brain location from which a patient's habitual seizures arise, and removal of which will theoretically result in complete cessation of seizures (Risinger, 1991). It has recently been estimated that approximately 1,000 of these resective seizure surgeries are performed each












year (Pilcher, Locharernkul, Primrose, Ojemann, & Ojemann, 1992), and that 80 to 90 percent of these patients benefit significantly from this intervention.

In order to provide healthcare professionals with a useful and consistent indicator of surgical outcome, Engel (1987) developed a classification system whereby each postsurgical patient receives a score of I through IV. Criteria for each classification are as follows: Class I (Seizure-free): A) Completely seizure-free since surgery, excluding any seizures which may have occurred during the first few postoperative days; B) Auras only since surgery; C) Some seizures after surgery, but seizure-free for at least two years; or D) Atypical generalized convulsions with antiepileptic drug withdrawal only. Class II (Rare Seizures or "almost seizure-free"): A) Initially seizure-free following surgery but has rare seizures now; B) Rare seizures since surgery; C) More than rare seizures after surgery, but rare seizures for at least two years; or D) Nocturnal seizures only, which cause no disability. Class III (Worthwhile Improvement): A) Worthwhile seizure reduction; or B) Prolonged seizure-free intervals amounting to greater than half the follow-up period, but not less than two years. Class IV (No Worthwhile Improvement): Significant seizure reduction; B) No appreciable change; or C) Seizures worse. As can be inferred from these criteria, a given patient's Engel Classification may change across the course of time following surgery.

Resective surgery has been shown to be particularly successful when the seizure focus and subsequent target of resection is in the temporal lobe (Chelune, 1981; Engel, Van Ness, Rasmussen, & Ojemann, 1993). Although the aforementioned rates of surgical success are impressive, it is important to note that the effectiveness of surgery relies heavily on the selection of appropriate surgery candidates, and must be judged in the context of several factors in addition to a given patient's degree of potential postoperative seizure relief.












When evaluating an individual for seizure surgery, the potential for seizure control and confidence with which a given seizure focus is identified must be estimated and compared with the potential for the serious cognitive and psychosocial impairments that may result from temporal lobe resection. This type of "risk-benefit" assessment allows both the physician and patient to make well-informed decisions regarding the likely effects of surgery (Ivnik, Sharbrough, & Laws, 1988). Because such a determination is complex and multifactorial, data in several domains is usually collected to assist in making surgery decisions. Those to be discussed in this study include neuropsychological evaluation, magnetic resonance imaging (MRI), and electroencephalographic (EEG) investigation.


Neuropsychological Assessment of Epilepsy Patients


Neuropsychological testing plays a unique role in the assessment of the epileptic patient, as it provides information which is complementary to physiological and anatomical measures such as EEG and MRI. Although EEG and MRI provide valuable information about the state of the brain, only neuropsychological testing can reveal the brain's actual behavioral capacities. In fact, properly administered and interpreted neuropsychological testing may reveal brain dysfunction before structural abnormalities are evident using other measures (Jones-Gotman, 1991).

One goal of neuropsychological assessment in the study of epilepsy is to identify

impairments that may suggest that seizure onset is lateralized to a particular hemisphere, or even more specifically, localized within that hemisphere. Such information, when combined with data obtained through MRI, EEG, and clinical interview, may reveal an epileptogenic zone (Pilcher et al., 1992). In addition to identification of this area of seizure onset, neuropsychological data also plays an important role in determining how likely it is that removal of brain tissue will result in a












significant degree of seizure control (Dodrill, 1986). Factors which have been associated with greater postoperative seizure control include fewer neuropsychological impairments demonstrated on preoperative testing (Wannamaker & Matthews, 1976), and resection of tissue in the hemisphere or temporal lobe in which impairment is implicated by neuropsychological testing (Bengzon, Rasmussen, Gloor, Dussault, & Stevens, 1968). Finally, even though an epileptogenic zone may be identified, and prognosis for seizure control may be positive, it is essential to examine the neuropsychological data with regard to the potential cognitive and psychosocial impairments which may result from tissue resection (Rausch, 1987).

In order for neuropsychological data to be properly interpreted, the results of each measure employed must be considered within the context of other test scores as well as psychosocial information, as it is patterns within the data that are most informational, rather than a patient's score on one particular test (Rausch, 1987). Thus, it is important that neuropsychological batteries assess a wide range of brain functions. Although batteries vary widely across epilepsy centers, most thorough batteries include measures of general intellectual functioning, attention, mental control, language, visual perception, and auditory and visual memory (Rausch, 1987).

Due to the highly epileptogenic nature of the temporal lobes and the fact that 80 percent of all cortical excisions are made from this region (Jones-Gotman, 1987), it is particularly important that temporal lobe and hippocampal function be assessed using both sensitive and specific measures. The role of the temporal lobe and hippocampus in language and memory, and the resultant risk of amnesia that is posed by resection in these areas have been well documented (Bauer, Tobias, & Valenstein, 1994). Therefore, such assessment typically includes measures of verbal and visuospatial memory functioning and language (Jones-Gotman, 1991). Typically, verbal memory impairment has been associated with language-dominant hemisphere












involvement, while visuospatial memory impairment has been associated with non-dominant hemisphere involvement (Milner, 1967). This pattern of differing performance associated with the language-dominant and non-dominant hemispheres is referred to as material-specific memory impairment. There have been many investigations of temporal lobe functioning in epilepsy, including studies of both non-surgical and surgical epilepsy populations. Memory Functioning in Non-surgical and Surgical Epileptic Populations


In 1969, Fedio and Mirsky compared the performance of children with left temporal lobe seizure foci, right temporal lobe seizure foci, and a group with bilateral EEG abnormalities and no specific seizure foci on a battery of intelligence, memory and attention measures. Results showed that children with left temporal lobe epilepsy (TLE) demonstrated lower Verbal than Performance IQ scores on the Wechsler Intelligence Scale for Children, while those with right TLE showed the opposite pattern of performance. Although these groups did not differ in verbal or visual memory span, supraspan tests of verbal memory revealed deficits in the left TLE group, characterized by difficulty learning a word list and recalling the list after a delay. Supraspan tests of nonverbal memory revealed impairment in the right TLE group, and this group also showed significant impairment while attempting to reproduce the Rey-Osterrieth Complex Figure after a short delay. The third patient group showed no significant verbal or visual learning impairment, although these subjects did tend to have difficulty on measures of sustained attention, a deficit which was not evident in either the left or right TLE groups (Fedio & Mirsky, 1969).

In 1979, Ladavas and co-investigators sought to determine whether seizure foci in the left temporal lobe, right temporal lobe, left frontal lobe, or right frontal lobe would have different effects on visual and verbal tests of short and long term memory (Ladavas, Umilta, & Provinciali,












1979). The authors found that patients with right and left temporal lobe seizure foci did not show material-specific memory impairments on measures of memory span, such as Digit Span from the Wechsler-Bellvue Form I and the Corsi Block Tapping Test. These groups did, however, show material-specific effects on measures designed to assess long term memory including verbal tests such as a task which required subjects to learn supraspan strings of digits and another involving recall of a word list after a ten minute delay, and purportedly nonverbal tests such as learning supraspan Corsi block tapping sequences and reproducing the Rey-Osterrieth Complex Figure ten minutes after copying it. It was concluded that temporal lobe abnormalities appear to cause deficits in long term memory, while leaving short term memory tasks such as immediate recall of presented stimuli relatively unaffected. Patients with frontal lobe seizure foci were impaired in a material-specific manner on tests of recency judgment, but were unimpaired on all other measures. These findings are consistent with Baddeley and Warrington's (1970) data concerning the performance of temporal lobe amnesics on tests of short and long term memory and with the findings of Fedio and Mirsky (1969) mentioned earlier. Another interesting finding from Ladavas' study was that the degree of the material-specific long term memory impairments exhibited by the patients appeared to be directly related to length of illness (Ladavas et al., 1979).

Delaney (Delaney, Rosen, Mattson, & Novelly, 1980) surmised that some of the variation in memory performance demonstrated across studies may be due to differences in diagnostic criteria, seizure severity and duration, psychological measures, age of seizure onset, duration of epilepsy, seizure frequency, and types and levels of medication. These investigators carefully matched four groups of epilepsy patients on all of the above factors, and examined their performance on verbal and visual memory tasks relative to that of normal controls. Results revealed that, consistent with most previous studies, patients with a left temporal lobe seizure focus demonstrated marked impairment on verbal memory tasks, while those with a right












temporal lobe focus had greater difficulty with visual memory tasks. The performance of patients with either left or right frontal lobe seizure foci did not differ from the normal controls, or from one another.

A closer look at the data revealed that left and right temporal lobe patients did not differ significantly in their ability to immediately recall the stories and figures from the WMS, but after a delay, left temporal lobe patients had forgotten a significantly greater amount of the verbal material, while their counterparts had lost a significantly greater amount of the visual material. The authors concluded that the results of studies demonstrating the material-specific memory deficits among left and right temporal lobe epilepsy patients require delayed recall measures in order to reveal reliable differences (Delaney et al., 1980).

Mungas and co-investigators examined the performance of patients with left hemisphere TLE, right TLE, and normal controls on a verbal learning task (Mungas, Ehlers, Walton, & McCutchen, 1985) which was modeled after the Rey Auditory Verbal Learning Test. Consistent with the previous study, the groups did not differ significantly in their ability to immediately recall verbal material. However, subjects with left TLE were impaired in their ability to recall the list after a delay. This group's performance was most impaired when the subjects were asked to recall the list after given phonemic/graphemic cues, mildly impaired on delayed free recall, and relatively intact when provided with semantic cues. The right TLE patients and normal controls did not differ on this measure. These data are consistent with those found in earlier investigations (Delaney et al., 1980), in that performance on immediate recall measures did not discriminate the left and right TLE patients from one another, or from controls, and that the left hemisphere group did show impairment in verbal memory across a delay. This impairment further emphasizes the role of the left medial temporal lobe in verbal memory. With regard to the left TLE patients' impaired ability to utilize phonemic/graphemic cues, the authors suggest that temporal cortex












may perform relatively non-semantic language processing, and that the hippocampus' role in verbal memory may be purely learning and retrieval, and predominantly nonlinguistic (Mungas et al., 1985).

In another study involving a list-learning task, Hermann, Wyler, Richey, and Rea (1987) compared the performance of 15 patients with left TLE, fifteen with right TLE and fifteen healthy controls on the California Verbal Learning Test (CVLT). No significant differences existed between the groups with regard to demographic variables, and the two patient groups did not differ in terms of duration of illness, age of seizure onset, or level of intellectual functioning. Furthermore, 23 of the patients underwent intracarotid sodium amytal testing to determine which hemisphere was dominant for speech. In all of those cases, left hemisphere dominance was determined. Patients with a family history of left-handedness were excluded from the study.

Results showed that left TLE patients demonstrated significantly impaired verbal learning across the five CVLT trials and recalled significantly fewer total words on immediate free recall than did the right TLE and control groups, which were indistinguishable from one another on these variables. Furthermore, patients with left TLE also showed a less sophisticated strategy for organizing to-be-remembered words, in that they did not cluster the word list semantically as much as the other groups. No significant group effects were found on variables concerning recognition of list words, delayed recall, or the effects of an interfering word list.

The ability of an immediate recall measure to discriminate right and left TLE patients and the inability of a delayed recall task to do so is in contrast to earlier findings described above (Delaney et al., 1980; Mungas et al., 1985). This inconsistency was attributed to careful patient selection with regard to language dominance, the fact that patients in this study were being evaluated for surgery and therefore may have had more severe seizure disorders, and the fact that EEG assessment of actual seizure onset was used to determine seizure lateralization and












localization (Hermann et al., 1987). Absence of impairment of recognition memory is consistent with earlier findings (Delaney et al., 1980)

Despite that a preponderance of studies have demonstrated that patients with a right

hemisphere seizure focus show more severe visual than verbal memory deficits, and that patients with left hemisphere foci tend to show the opposite, it is important to note that not all studies have shown this. In 1973, Glowinski compared the performance of patients with left and right seizure foci on three subtests of the WMS. Glowinski's study revealed that patients with TLE demonstrated a significantly more severe memory deficit than did epileptic patients who did not have temporal lobe foci. Further analysis revealed that, although the expected pattern of performance was borne out among patients with right and left temporal lobe foci, the deficits exhibited by the two groups were not significantly different.

Taken together, the aforementioned studies have generally upheld the concept of

material-specific memory deficits with regard to memory performance in patients with left and right temporal lobe seizure foci. Most notably, material-specificity was shown most clearly on tests which included measures of long term memory, that is, memory after a delay. Tests of immediate recall yielded less consistent results.

In their discussion of the differing memory performance among temporal lobe epileptics, Hermann et al. (1987) implied that patients being considered for surgery and those actually qualifying for this treatment, may have seizure disorders which differ in severity and along other dimensions from those of nonsurgical populations. Subsequently, surgical populations may perform differently on neuropsychological tests, and this may reveal additional information about memory performance in TLE.

In 1987, Rausch and Babb examined the performance of 26 patients who had undergone left anterior temporal lobectomy (ATL) and 45 who had undergone right ATL on the Logical












Memory, Verbal Paired Associates, and Visual Reproduction subtests of the Wechsler Memory Scale. Results showed that, as a group, patients who underwent left ATL performed significantly more poorly on immediate and delayed recall of semantically unrelated word pairs and on a test of prose recall. In contrast, patients who underwent right ATL, performed significantly worse on immediate and delayed recall of geometric figures. It is important to note the existence of immediate memory impairment in surgical patients, as this effect was not shown in the aforementioned studies involving nonsurgical populations.

Using the same subtests of the Wechsler Memory Scale, Jones-Gotman et al. (1989) compared the verbal memory performance of epilepsy patients with left, right, and bilateral temporal lobe seizure foci. Results showed that patients with a left seizure focus and those with bilateral abnormalities with a predominance of left hemisphere involvement demonstrated significantly worse performance prior to resective surgery than their right-sided counterparts, demonstrated further impairment 14 days post-surgery, and ultimately returned to baseline at long-term follow-up. Patients with a right seizure focus and those with predominantly right-sided bilateral abnormalities demonstrated superior performance relative to their left-sided counterparts prior to surgery, and demonstrated increasing improvement during their two post-surgery testing sessions. Thus, all four groups of patients were distinguishable from one another using the verbal subtests of the WMS, with unilateral right hemisphere patients showing the least impairment, followed by those with predominantly right hemisphere bilateral abnormalities, those with unilateral left hemisphere seizure foci, and finally, those with predominantly left hemisphere bilateral abnormalities. With regard to performance on the Visual Reproduction subtest, results showed that the group with unilateral left hemisphere seizure foci performed significantly better than all other groups both pre- and post-operatively, and that the remaining groups could not be distinguished from one another using this measure. Thus, material-specific memory impairment












was shown with regard to both pre- and post-operative verbal and visual memory (Jones-Gotman et al., 1989). More detailed studies of neuropsychological performance among surgical populations will be discussed later.


LaWnguage Functioning in Epilepsy


In an effort to explain the sources of variation in the findings concerning verbal memory in TLE populations, Mayeux and co-investigators (Mayeux, Brandt, Rosen, & Benson, 1980) hypothesized that this ambiguity may be partially due to the varying degrees of language function and dysfunction experienced among these patients. The authors reasoned that few studies performed language assessments, as compared to the relatively large emphasis placed on verbal memory deficits. They also noted signs of anomia, circumlocution, and paraphasic errors among their patients, whose most common complaint was one of verbal memory.

The experimental population consisted of 14 patients with well-documented left TLE,

seven with right TLE, and eight patients whose seizures were non-focal and tended to secondarily generalize. The groups, which were matched for age, duration of illness, and seizure frequency, were administered a thorough battery including the WAIS-R, the WMS with delayed recall of prose passages, auditory and visual trigrams test, the Benton Visual Retention Test, copy and delayed recall of the Rey-Osterrieth Complex Figure, the Boston Naming Test and a measure of verbal fluency.

Data analyses revealed that the only significant differences among group performance

was on the Boston Naming Test, on which the left TLE group performed significantly worse than the other groups which did not differ from one another. This difference remained significant even after the effects of individual variations in vocabulary were statistically factored out. The authors concluded that subtle language impairment, particularly a naming impairment, could be the result












of epileptogenic foci in the left inferior temporal lobe. This "anomia" could be perceived by both the patient and physician to be a memory impairment (Mayeux et al., 1980). It should be noted that these data do not strongly indicate that naming deficits account for all apparent memory impairment in temporal lobe epilepsy. Furthermore, including tests of confrontational naming and verbal fluency mark an improvement on earlier studies, however, these do not constitute a thorough language battery.

In a similar study, Stanulius and Valentine (1988) examined the pre- and postoperative performance of patients undergoing left or right ATL on tasks of verbal fluency, confrontational naming (Boston Naming Test), and verbal learning (Rey Auditory Verbal Learning Test). Although the groups did not differ on the verbal fluency task, left ATL patients performed significantly poorer than right ATL patients, both pre- and postoperatively, on the confrontational naming task. Furthermore, left ATL patients showed a mild decline in postoperative performance, whereas the right ATL group improved significantly relative to their preoperative scores. The authors indicated that, among the left ATL patients, performance on the Boston Naming Test and RAVLT correlated at a value of .51, which they interpreted to mean that language and memory functions have been confounded in studies which have claimed to reveal material-specific memory deficits after ATL.

The "anomia" observed in the above studies has also been documented earlier in the literature (Heilman, Wilder, & Malzone, 1972). These investigators studied 10 patients who underwent dominant ATL, and found that four of them demonstrated signs of anomic aphasia on a test of confrontational naming of real objects. This deficit was persistent beyond six months postoperatively, and was not related to episodes of seizure, impairment of verbal fluency or comprehension.












Despite the existence of data suggesting that tests of verbal memory are confounded by language dysfunction, Hermann and Wyler (1988) provided evidence to the contrary when they compared pre- and post-operative performance of left and right ATL patients on the Multilingual Aphasia Exam. This is a comprehensive language battery which includes tests of Visual Naming, Sentence Repetition, Controlled Oral Word Association, Oral Spelling, Token Test (comprehension), Aural Comprehension of Words and Phrases, and Reading Comprehension of Words and Phrases. Patients were tested pre-operatively, and six months post-surgery, with alternate forms administered for four of the above subtests. Right ATL resections involved the removal of, on average, 5.1 cm of ATL, while left resections were slightly smaller and averaged

4.7 cm. The left hemisphere resections were tailored according to intraoperative electrocorticography and speech-mapping procedures.

Results showed that the two groups did not differ with regard to seizure outcome. Nine of 15 left ATL patients, and 10 of 14 right ATL patients became seizure-free following surgery, while all remaining subjects experienced a decrease in seizure frequency of at least 75 percent. The two groups did not differ significantly on any of the individual MAE subtests administered preoperatively, however, the left ATL group did perform more poorly on average. Postoperative performance also revealed no significant differences between the groups, although the left ATL group showed a nonsignificant trend toward demonstrating a greater pre-postoperative decline. Despite this trend, the left ATL group showed significant pre-postoperative improvement on the Token Test and Controlled Oral Word Association. These data strongly suggest that left and right ATL resection can be result in excellent seizure control without a significant loss in language ability, and that material-specific memory deficits observed in TLE are not likely the result of language dysfunction. The improvement shown by the left ATL group was attributed to a reduction in "neural noise", resulting in more effective ipsilateral brain












function. A similar hypothesis, with regard to contralateral functioning, has previously been proposed (Novelly, Augustine, & Mattson, 1984).

Chelune, Awad, and Luders (1989) generated consistent data, using relatively thorough memory and language battery to test 19 right ATL and 23 left ATL patients. Subjects were administered the RAVLT, Wechsler Memory Scale-Revised (WMS-R), Aphasia Screening Exam, Boston Naming Test and measures of Verbal IQ and verbal fluency. Testing was conducted prior to surgery, as well as six months postoperatively.

As a group, patients undergoing right ATL performed better than left ATL patients on Verbal IQ, Boston Naming Test, WMS-R verbal subtests, and the RAVLT. With regard to prepostoperative performance changes, the only changes reaching significance involved the WMS-R verbal memory index and delayed recall of prose from the WMS-R. In these cases, right ATL patients showed significant improvement postoperatively, while left ATL patients did not. It was also noted that a similar, but nonsignificant trend was observed with regard to performance on the BNT. The authors concluded that the general lack of change in performance on language tests suggests that the material-specific deficits observed before and after left ATL appear to be independent, not confounded, with language function. Obviously this is in marked contrast to earlier findings (Mayeux et al., 1980; Stanulius & Valentine, 1988).

The degree to which language impairment is associated with TLE and ATL remains

unclear. It appears, however, that those studies which utilized more thorough language batteries have shown that language dysfunction may be associated with temporal lobe involvement, but does not likely significantly confound the important memory findings yielded in other studies (Herman & Wyler, 1988; Chelune, Awad & Luders, 1989). The mixed findings of the above studies indicate the need for more thorough language assessment in epilepsy research studies.













Executive Functions in Epilepsy


Due to the fact that most seizure disorders and resective surgeries involve the temporal lobes, much of the research examining neuropsychological performance among epilepsy populations have focused most specifically on tests of memory and language. Recent data, however, suggests that patients with seizures of temporal lobe origin often demonstrate impairment on tests of executive functions which have traditionally been thought to be reflective of frontal lobe functioning. The precise nature of this impairment is unclear. At least two theories have been proposed, and these will be discussed below.

A measure of executive function which has been shown to be sensitive to temporal lobe seizure involvement is the Wisconsin Card Sorting Test (WCST). The WCST is a commonly used neuropsychological test which is thought to assess subject ability to create and test strategies in a problem-solving context, and to engage in abstract thinking (Grant 1948; Grant & Berg, 1948). The literature has traditionally suggested that patients with frontal lobe pathology, particularly in the dorsolateral prefrontal cortex, perform poorly on this test relative to normal controls and patients with pathology in other brain regions (Milner, 1963; Heaton, 1981). More recently, however, the specificity of the WCST performance with regard to detection of frontal lobe pathology has come into question.

In 1991, Anderson and co-investigators examined WCST performance in 91 patients with "stable focal brain lesions" detected on CT or MRI. The authors found no significant differences in WCST performance among patients with frontal as compared to non-frontal lobe lesions. Furthermore, some patients with extensive frontal lobe pathology performed within normal limits on this measure, while some with non-frontal lobe involvement failed outright. Overall, the authors were unable to identify specific frontal lobe regions which appeared to be specifically












involved in WCST performance, and were able to correctly categorize patients as frontal or nonfrontal in only sixty-two percent of their cases (Anderson, Damasio, Jones, & Tranel, 1991).

Previously, Hermann, Wyler, and Richey (1988) examined WCST performance among 35 subjects with complex partial seizures, 16 of who had language-dominant hemisphere temporal lobe onset, and the remainder having had non-dominant temporal lobe onset. A control group of five patients with primary generalized epilepsy and one with a parietal lobe seizure focus was also included in the study. Results showed that patients with non-dominant temporal lobe seizures made significantly more perseverative errors than did the other two groups, and that the patients with dominant temporal lobe seizures made more of these errors than did the controls. The groups did not differ on number of categories achieved. The authors attributed this, not to a temporal lobe role in WCST performance, but to an interference in frontal lobe functioning caused by "neural noise" generated by temporal lobe dysfunction. In support of this theory, the authors found that 17 patients showed improved WCST performance after undergoing ATL, which theoretically removed the temporal lobe source of "neural noise" which had been interfering with frontal lobe functioning.

To further examine the relative effects of frontal versus temporal lobe involvement on WCST, Corcoran and Upton (1993) administered a modified version of the task to 47 patients with unilateral seizure foci. The Stroop Test and a test of verbal fluency were also administered, as traditional tests of frontal lobe functioning. Sixteen of the 47 patients had documented hippocampal sclerosis as detected by MRI, with the cases equally divided between right and left pathology. Thirteen patients had evidence of unilateral temporal lobe seizure foci without apparent hippocampal sclerosis. Seven of these patients had left sided seizure foci, and six had right-sided foci. The remaining 18 patients had unilateral frontal lobe seizure foci, with 10 of these demonstrating left-sided onset and eight having right-sided onset.












The authors found that patients with hippocampal sclerosis took longer to complete the WCST and made more perseverative errors than did members of the other groups. They also achieved fewer categories than the temporal lobe patients without sclerosis, but not the frontal lobe group. Patients with frontal lobe seizure foci made more errors on the Stroop Test, and similar to those with non-sclerotic temporal lobe foci, performed more poorly on the test of verbal fluency than did patients with hippocampal sclerosis (Corcoran & Upton, 1993).

Looking at the performance of the patients with hippocampal sclerosis, those with right temporal lobe pathology made significantly more categorical errors and took longer to complete the task than did those with left temporal lobe pathology. These lateralization findings are consistent with those of Hermann et al. (1988). The authors interpreted their results within the context of Gray's (1982) model of the hippocampus as a "comparator" which compares incoming sensory information to internally stored information about previous responses. Damage to the hippocampus would theoretically lead to dysfunction in this aspect of "working memory", and lead to perseverative responses and failure to maintain set on the WCST (Corcoran & Upton, 1993).

Lateralization findings contradictory to those above were provided by Strauss et al., who examined WCST performance in 77 patients with complex partial seizures of temporal lobe origin, 35 of whom had left-sided seizure onset, and 42 of whom had right-sided onset (Strauss, Hunter, & Wada, 1993). The authors found that patients with left-sided onset, and an age of onset of less than one year, showed increased perseverative errors and an inability to shift set. Patients with right-sided onset showed more perseverative tendencies than would normals, but this was not as marked as in the left-sided group, and did not appear to be affected by age of seizure onset. Additionally, no relationships were shown between lateralization of language or general intellectual functioning and WCST performance for either group (Strauss et al., 1993).












Shigaki et al. (1995) found that left hippocampal volume was significantly correlated with WCST performance, as patients with decreased volumes achieved fewer categories, and made more perseverative responses and perseverative errors on the WCST, than did patients with decreased right hippocampal volumes. These results are more consistent with those of Strauss et al. (1993), than with the previous studies described above (Hermann et al., 1988; Upton & Corcoran). As can be seen, the studies conducted in this area have produced mixed results, leaving WCST performance among patients with temporal lobe seizures an issue which will require further research to clarify.


Neuropsychological Test Data as a Predictor of Surgical Outcome


Rausch (1987) has pointed out that, although many studies have used surgical resection of the temporal lobe and hippocampus as an independent variable in an effort to reveal the effects of temporal lobe resection on neuropsychological functioning, relatively few have explored how to use test performance to predict surgical outcome and select the most appropriate candidates for surgery.

In 1968, Bengzon and co-investigators examined data from 547 patients who underwent temporal lobe excision between 1953 and 1960 (Bengzon et al., 1968). The investigators selected all patients (N_= 50) who were either seizure-free following surgery or who had occasional seizures for 12 to 24 months post-operatively, followed by a cessation of seizures. Placed in a second group were those patients who experienced little or no change in seizure activity following surgery (N = 54). Patients who showed mild improvement were excluded from the study in order to accentuate the differences between the above two groups. Once the groups were formed, they were compared on seven domains of assessment including a) demographics; b) general seizure history; c) seizure pattern analysis and typology; d) skull X-rays,












pneumoencephalograms and angiograms, e) "operative and post-operative features"; f) neuropsychological exams; and g) pre- and postoperative EEG's. Findings in the neuropsychological domain were as follows.

The neuropsychological battery used was that recommended by Milner (1954). In the study described here, Bengzon examined IQ, normal vs. erratic performance patterns, presence vs. absence of performance deterioration, and speech lateralization and localization. Data analyses showed that, as mentioned above, the majority of patients in the "surgical success" group demonstrated neuropsychological exam performance consistent with the existence of a temporal lobe seizure focus. This was not true of the "surgical failure" group, most of whom showed evidence of extratemporal involvement, with or without concomitant temporal lobe abnormalities. With regard to lateralization, Bengzon et al. found that 40 percent of the "surgical success group" showed neuropsychological evidence of right hemisphere lateralization, 33 percent showed left lateralization, while the remaining patients were unlateralized, bilateral, or normal. Of the "surgical failure" group, 62 percent were either lateralized to the left hemisphere or bilateral, only 20 percent lateralized to the right hemisphere, and the remaining patients were unlateralized or normal. No other neuropsychological factors were significant (Bengzon et al., 1968).

The implications of this early study are important and deserve review. Characteristics of the surgical success group included neuropsychological lateralization to the hemisphere upon which surgery was performed, with greatest success occurring in those patients whose seizures were localized to the temporal lobe. Those with evidence of more diffuse involvement did not fare nearly as well postoperatively. The authors attribute the fact that more patients in the "surgery success" group had right hemisphere foci and resection to the fact that this likely limited the effects of disease and surgery on speech and language (Bengzon et al., 1968).












Wannamaker and Matthews (1976) examined the prognostic value of neuropsychological testing in a population of 14 patients who underwent seizure surgery between 1960 and 1973. Patients were divided into three groups as follows: five who were seizure free postoperatively, six who showed a clear reduction in seizure activity, and three who showed no benefit of surgery. The neuropsychological battery used in the assessment of these patients was not described in detail, but included the Wechsler Adult Intelligence Scale, tests developed by Halstead, and additional tests aimed at the assessment of motor and sensory abilities. An overall pre- and postoperative impairment index was generated for each patient based on performance on this battery. Results showed that the patients who were most impaired on preoperative neuropsychological measures were less likely to experience a postoperative decrease in seizure activity, and were also at greatest risk for exacerbated neuropsychological impairment (Wannamaker & Matthews, 1976). The authors also noted that patients undergoing right hemisphere surgery were more likely to show a postoperative decrease in seizure activity and less likely to show a decline in neuropsychological test performance following surgery. Although this is consistent with the results of Bengzon's (Bengzon et al., 1968) study, the authors not only point out that neuropsychological tests may be more sensitive to left hemisphere abnormalities, but also the fact that the patients undergoing right hemisphere surgery typically had lower preoperative seizure frequencies than their left hemisphere counterparts. Unfortunately, laterality and localization of seizures were not discussed in greater detail.

Studying a sample of 142 patients who underwent anterior temporal lobectomy, Ivnik,

Sharbrough, and Laws (1988) sought to determine effective ways of estimating potential surgical risks and benefits in order to counsel patients regarding surgery. Patients included in the study were considered medically intractable, had ictal EEG documentation of anterior temporal lobe












seizure focus, and clinical history consistent with TLE. Seizure foci were further confirmed intraoperatively through the use of electrocorticography with both surface and depth electrodes.

All patients were administered neuropsychological batteries pre- and postoperatively,

which consisted of the age-appropriate Wechsler Intelligence Scale, the Wechsler Memory Scale, and the Auditory Verbal Learning Test. Data analysis revealed that right lobectomy patients earned higher Verbal IQ and Verbal Capacity factor scores prior to surgery than did the left lobectomy group, and that this difference increased after surgery. With regard to Performance IQ and Perceptual Organization factor scores, both groups improved postoperatively, with the right lobectomy group showing the greater improvement. The authors carefully pointed out that the vast majority of patients in the study performed within normal limits on the measure of general intellectual functioning both pre- and postoperatively (Ivnik et al., 1988).

With regard to memory measures, the left lobectomy group performed significantly

worse than the right group preoperatively on tests of verbal memory. Following surgery, the left temporal lobectomy group showed a significant deterioration in performance on immediate verbal recall, but did not show a relative decrement in delayed verbal recall. Interestingly, surgery did not appear to impair visual memory in either surgical group, a finding which the authors attribute to the possibility that a delayed visual recall measure may have been necessary to reveal impairment. Regarding learning, the two surgical groups did not differ significantly preoperatively on trials one through five of the AVLT or on a measure of "total amount learned", however the right hemisphere group appeared to be less vulnerable to the effects of interference, and performed slightly better on delayed measures of free recall and recognition. Following surgery, the right lobectomy group showed improvement on every AVLT variable, while the left group showed a decline on all AVLT variables, resulting in significant group differences for each value (Ivnik et al., 1988).












The authors concluded that right and left anterior temporal lobectomy clearly had

differential postoperative effects, particularly involving learning and memory. It was noted that the degree of postoperative decline in learning and memory abilities was usually related to preoperative functioning, in that those subjects who go into surgery with these abilities intact are most likely to experience a significant postoperative decline. Patients with only mild preoperative impairments are more likely to experience a less severe postoperative decline (Ivnik et al., 1988). These findings are in contrast to those of Wannamaker and Matthews (1976).

In another, more thorough examination of pre- and postoperative neuropsychological functioning, Bauer and co-investigators studied 21 patients undergoing left temporal lobe resection and twelve undergoing right temporal lobe resection for the treatment of intractable epilepsy (Bauer et al., 1995). The neuropsychological battery, which was administered three months prior to surgery and approximately five months after surgery, consisted of the Wechsler Adult Intelligence Scale-Revised, the Logical Memory and Visual Reproduction subtests of the Wechsler Memory Scale-Revised, the California Verbal Learning Test, the Rey-Osterrieth Complex Figure, tests of confrontational naming and verbal fluency, the Benton Test of Facial Recognition, the Judgment of Line Orientation Test, and tests of executive functions including the Wisconsin Card Sorting Test.

Data analyses revealed that in general, this population scored below the mean on the

majority of tests, reflecting preoperative general neuropsychological impairment. A slight trend toward material-specific memory neuropsychological impairment, with left hemisphere patients showing impairment on verbal tasks and right hemisphere patients showing impairment on nonverbal tasks, was noted. However, the authors were careful to point out that this did not reach statistical significance (Bauer et al., 1995), and that this is consistent with previous findings (Delaney et al., 1980; Glowinski, 1973).












With regard to post-operative performance, patients undergoing left temporal lobe

excision performed significantly worse than those undergoing right hemisphere surgery on tests of verbal memory and language. The opposite effect was shown on tests of nonverbal memory and frontal functioning, however the significance of these data were not as robust as those concerning verbal measures. The authors characterized these results as a "widening" of differences that existed preoperatively between patients with left and right seizure foci, consistent with the findings detailed above (Ivnik, Sharbrough, & Laws, 1988). Also consistent with previous studies (Bengzon et al., 1968; Wannamaker & Matthews, 1976), patients undergoing right temporal lobe resection were more likely to show no change or improve on neuropsychological measures including tests of memory, language and frontal functions. This is partly attributed to the important role that the right hemisphere is believed to play in general attentional processes. It was also noted that resection of the left temporal lobe was associated with the greatest postoperative decline in test performance, although these patients did improve on some measures of nonverbal memory such as the Rey-Osterrieth Complex Figure. With regard to general neuropsychological functioning, those patients undergoing left temporal lobe resection who showed the best preoperative performance tended to show the greatest postoperative decrements, while patients undergoing right temporal lobe resection who had the lowest preoperative scores tended to show the greatest postoperative improvement (Bauer et al., 1995).

Although somewhat mixed, the results of the above studies suggest that

neuropsychological testing provides useful information regarding surgical prognosis. This information includes not only the likelihood that effective seizure control will be attained, but also is indicative of the potential for cognitive decline among different patient groups. Additional












inquiry in this area will be important for maximizing the utility of information obtained through noninvasive pre-operative assessment.


Hippocampal Pathology and Neuropsychological Performance


In a pioneering investigation of the neuropathological correlates of epilepsy, Margerison and Corsellis (1966) studied 55 epilepsy patients using clinical observation, EEG, and subsequent post-mortem histological study of the brain. Subsequent to clinical and EEG assessment, 26 of these subjects received a firm diagnosis of TLE. Of these 26, 22 were found to have hippocampal sclerosis. Eight of 13 subjects who were firmly diagnosed as not having TLE did not show any evidence of hippocampal sclerosis. Among the cases with pathology, the most commonly observed hippocampal abnormalities consisted of neuronal loss and fibrous gliosis in the Ammon's horn, also affecting field H1 (Sommer sector), H3 (end folium), and the dentate gymrus. Margerison and Corsellis surmised that the hippocampus has a "selective vulnerability" to the metabolic changes which occur during temporal lobe seizures, and that these changes lead to the development of sclerosis. In turn, sclerosis may maintain, or even increase, the likelihood of seizures occurring, thereby creating a "vicious circle." This important study gave rise to subsequent investigations of hippocampal pathology in epilepsy, its association with neuropsychological functioning and EEG, and how to assess it most accurately.

Due to the highly epileptogenic nature of the hippocampus (Lothman, 1991), and the association between epilepsy and hippocampal sclerosis (Margerison & Corsellis, 1966), it is essential that any neuropsychological assessment of the epileptic patient include specific measures of hippocampal function. Although the aforementioned retrospective and prognostic studies demonstrated the differential neuropsychological effects of left, right, and bilateral temporal lobe abnormalities and resection, measures of temporal lobe functioning vary












considerably in their sensitivity to hippocampal pathology. In addition to the differences among neuropsychological batteries used to assess these patients, Jones-Gotman (1987) has pointed out that the differences observed across studies examining the effects of temporal lobe resection may well be the result the varying amounts of tissue that is resected in different treatment centers.

At the Montreal Neurological Hospital, she has observed three general patterns of

performance on memory tasks used in the assessment of patients who have undergone minor or extensive hippocampal lobe resection. There are tasks which do not appear to be sensitive to the degree of hippocampal resection, such as recall of prose passages and paired associates from the WMS (Milner, 1967) and recall of the Rey-Osterrieth Complex Figure (Jones-Gotman, 1986). Other tasks, such as repetition of supraspan digit sequences (Milner, 1971) and recall of a complex figure which has been presented to the subject in a gradual, piecemeal manner (JonesGotman, 1986) appear to reflect significant impairment after small resections and significantly greater impairment after large resections. Finally, there are tasks such as subject-ordered pointing to abstract words and designs (Petrides and Milner, 1982), that reveal impairment only in patients that have had a large amount of the hippocampus removed (Jones-Gotman, 1987). Following is a closer look at some of the important studies in this area.

In an effort to reveal the important role of the hippocampus in visuospatial learning and memory, Jones-Gotman (1986), examined the abilities of normal controls and patients who had undergone temporal lobe resection to learn to copy a series of 13 abstract designs. The designs were presented to each subject as many times as necessary until he or she was able to correctly reproduce 12 of them from memory. Results showed that patients who underwent left temporal lobectomy did not differ significantly from controls in their performance, except for the fact that surgical patients required more trials to reach criterion than did the controls. Likewise, the performance of patients who had right temporal lobe resection with less than 1.5 cm of the












hippocampus removed did not differ from that of the control subjects. Right temporal lobectomy patients who had more than 1.5 cm ofhippocampal tissue removed showed significant impairment across the first five learning trials, and thereafter did not differ significantly from controls. All subjects were required to recall as many of the designs as possible after a 24 hour interval. Again, only subjects with large right hippocampal resections were significantly impaired, with the remaining groups performing similarly to one another. Jones-Gotman (1986) concluded that patients with extensive right hippocampal resections demonstrated a significant visuospatial learning impairment, in that they were not as able other subjects to benefit from repeated exposure to the stimuli.

With regard to the fact that patients who had undergone left temporal lobe resection had difficulty reaching criterion across the learning trials, Jones-Gotman (1986) surmised that this may be due to their having difficulty verbally encoding the figures, whereas their right hemisphere counterparts likely had difficulty encoding them in a spatial manner. The authors also explored primacy and recency effects among their subjects and found that patients with large right hippocampal lesions recalled almost no items from the primary part of the list, significantly fewer items from the middle parts of the list than other groups, and performed similarly to the other groups on items from the last part of the list. All groups showed a significant recency effect.

In an analogous study, Frisk and Milner (1990) studied patients with left and right

temporal resections and normal controls and their respective abilities to learn a short story. As in the previous study, the to-be-learned material was presented repeatedly until each subject was able to answer correctly a series of questions regarding the story's content. Results showed that patients with left temporal lobe resections evidenced slower rates of learning than the other groups, with the most significant impairment shown by patients whose left temporal lobe












resection had included part of the body of the hippocampus. The same pattern of performance emerged among the groups when subjects were asked to repeat the story and answer questions about the story after a twenty minute delay. The investigators concluded that excision of left temporal lobe tissue impaired learning and retention of a verbal passage, and that the severity of this impairment was strongly related to extent of hippocampal damage. It is surmised that these deficits arise from an abnormally high rate of forgetting among the patients with extensive hippocampal resections (Frisk & Milner, 1990).

In an attempt to reveal more specifically the specialization of memory roles within

different temporal lobe areas, Rausch and Babb (1987) examined resected tissue from 10 patients who had undergone left ATL. Pre-operatively, all of these patients had been administered the Wechsler Adult Intelligence Scale, and subtests of the Wechsler Memory Scale which required them to recall word pairs, prose, and geometric figures. Two independent examiners calculated cell densities for resected regions of the temporal lobe including the upper and lower dentate fascia, CA1-4, Prosubiculum, Subiculum, Presubiculum, and the hippocampal, fusiform, inferior temporal, and middle temporal gyri.

Correlational analyses demonstrated that lower cell counts in the hippocampus,

particularly CAl, were significantly associated with simple recall of word pairs, whereas low counts in the hippocampal gyrus were associated with the more complex task of immediate and delayed prose recall. An interesting, yet nonsignificant association was also demonstrated between low cell counts in the prosubiculum and recall of word pairs, while low cell counts in the subiculum appeared to associate inversely with prose recall. As expected given the fact that these patients underwent left ATL, none of the regional cell counts were significantly correlated with performance on the nonverbal memory task (Rausch & Babb, 1987).












The authors concluded that anatomically simpler areas, such as CAl and the prosubiculum appear to perform relatively simpler verbal memory tasks, whereas the hippocampal gyrus, with its widespread connections to association cortex, may support more complex verbal memory processes. Thus, evidence is provided that the temporal lobe regions are at least somewhat functionally divided (Rausch & Babb, 1987).

The aforementioned studies strongly suggest that the temporal lobe, and hippocampus in particular, may be at least somewhat functionally divided. More detailed studies will be needed in order to define more specifically the roles played by different temporal lobe and hippocampal regions. From a clinical standpoint, what is needed are ways to detect hippocampal sclerosis and to assess hippocampal functioning more effectively prior to surgery, rather than through the examination of resected tissue. In addition to ongoing progress in the neuropsychological testing of hippocampal function, great strides have recently been made in the use of MRI as it relates to epilepsy.


Quantitative MRI Volumetric Studies in Epilepsy


Magnetic Resonance Imaging refers to a variety of techniques used to image tissue. Unlike computed tomography (CT) and X-ray, which rely on ionizing radiation to produce an image, MRI is noninvasive and relies upon the inherent magnetic and electrical properties of the nuclei of tissue particles. It is of no known danger to the subject, and has very few contraindications, which include pacemakers, the presence of internal metal objects such as aneurysm clips, and severe claustrophobia (Andreasen, 1988). The basic principle of MRI relies upon the fact that certain nuclei, when placed in a magnetic field, will absorb energy within specific radio frequencies, and subsequently remit this energy as they return to their original state (Kuzniecky & Jackson, 1995).












In the case of brain MRI, the patient is placed within a magnetic field which causes hydrogen ions to line up with one another according to their magnetic properties. Once this alignment has occurred, the brain is exposed to a radio frequency pulse. This pulse rotates the ions from their position, and when the pulse is discontinued, the ions remit energy as they return to their alignment. The MR image is produced by a radio frequency receiver, which records the amount of energy or "signal" remitted from each small piece of brain tissue, each of which is referred to as a "voxel." These voxels are then arranged in image form in pixels, each of which will have a shade of grey corresponding to the amount of signal that was remitted from a given brain area. Different types of tissue contain varying degrees of hydrogen atoms, and therefore remit varying amounts of signal. Thus, these areas will be distinct from one another on the MR image (Andreasen, 1988). As MRI involves the assignation of quantitative values to the signal remitted by many small brain areas, these values can be easily stored in the scanner and reformatted to produce fine images in coronal, sagittal, or horizontal planes (Kuzniecky & Jackson, 1995).

Many MRI studies of epilepsy patients have involved quantitative measurement of the temporal lobes, hippocampi, or both, in an attempt to determine the existence of atrophy in one hemisphere relative to the other, or relative to previously collected norms. The simultaneous improvement of MRI resolution and development of new methods of quantitative measurement of brain structures has led to greatly advanced MRI detection of such atrophy since 1990. Abnormalities that were once evident only upon examination of resected tissue are now detectable in the pre-operative patient (Spencer, 1994). Quantitative MRI studies are considered preferable to visual assessment of MRI, as the former are reproducible, can correct for improper head rotation in the scanner, and are slightly more sensitive to atrophy (Bronen et al., 1994). As will be discussed below, quantitative MRI-based measures of hippocampal atrophy have been












shown to correlate with hippocampal sclerosis, lateralization of seizures by EEG, degree of hippocampal neuronal loss, performance on verbal memory measures, and post-operative seizure outcome.

In 1991, Cascino et al. investigated the correlation between MRI-based hippocampal volumes, and histopathology observed in the temporal lobe tissue that was removed from 24 patients with intractable partial seizures. All of these patients' seizures were lateralizeable using noninvasive methods such as scalp EEG. The investigators quantified neuronal cell loss in the subiculum, prosubiculum, and sectors Cl -C4 of the hippocampus. Of the 24 patients, 15 met criteria for mesial temporal lobe sclerosis (greater than 50 percent neuronal loss) upon histological examination of the resected tissue. Fourteen of those 15 had significant hippocampal atrophy on the same side as seizure origin, based on MRI. The authors added that MRI-based volume measurements were 93 percent sensitive and 100 percent specific in determination of seizure lateralization. The severity of pathology appeared to correlate with the hippocampal volumes, although the volumes were not able to predict pathology in cases of moderate to severe neuronal loss.

In order to establish normative values for MRI-based anterior temporal lobe and

hippocampal volumes in nonepileptic subjects, Jack et al. (1989) obtained these measurements from 52 healthy subjects between the ages of 20 and 40. After normalizing these measurements with regard to total intracranial volume, the authors found no significant effects of age or sex on volume of either the right or left anterior temporal lobes. A small, but significant difference in this measurement was found in right-handers, with the right anterior temporal lobe being slightly larger than the left. Age, sex, and handedness did not have a significant effect on right or left hippocampal volumes, however for the group of subjects it was found that the right hippocampus












was, on average, significantly larger than the left (2.8 versus 2.5 cm3, respectively.) Reasons underlying this asymmetry remain unclear.

In a retrospective study of 41 right-handed patients who underwent surgery for medically intractable complex partial seizures, Jack et al. (1990) compared the sensitivity and specificity of five measures in their abilities to correctly lateralize seizure onset. Seizure onset, of course, had been previously lateralized through the use video-recorded surface EEG monitoring, clinical observation, and intraoperative electrocorticography. MRI was administered preoperatively, but was not used to determine seizure lateralization. The following techniques were compared and were found to lateralize seizure onset in order of decreasing usefulness: MRI hippocampal volume measurements, visual grading of MRI to determine hippocampal atrophy, anterior temporal lobe volume measurements, visual grading of MRI to determine anterior temporal lobe atrophy, and unilateral signal intensity of MRI images with long repetition time. Quantitative MRI-based measurements of hippocampal volume correctly lateralized seizure onset in 76 percent of the cases, with no false lateralizations. Decisions regarding the significance of hippocampal atrophy were based on the norms collected previously (Jack et al., 1989). The authors considered atrophy to be significant if the difference between the two hippocampi were at least two standard deviations greater than is seen in normal individuals. A more liberal criterion of 0 cm3 difference between hippocampi would have correctly lateralized 90 percent of cases, with no false lateralizations, however the authors point out that such a liberal criteria could lead to important errors on a case by case basis, as it does not allow for measurement error. It was concluded that quantitative MRI evidence of hippocampal atrophy, combined with EEG data may render invasive recording unnecessary, and that disagreement between these two measures strongly indicates the need for invasive monitoring. It was also noted that visual grading of hippocampal atrophy was particularly difficult in cases of right hippocampal disease, as the right












hippocampus is typically larger than the left, and any atrophy tends to make the two appear more similar in size.

In a more recent investigation of the sensitivity and specificity of quantitative MRI volumetrics in the study of epilepsy, Spencer et al. (1993) studied 56 patients with intractable epilepsy. Twenty-nine had intracranial ictal EEG's suggestive of unilateral medial temporal lobe seizure onset. Of these, 21 had quantitatively measured hippocampal atrophy in the same hemisphere, and the remainder did not. Of 21 patients whose intracranial EEG's did not suggest temporal lobe onset, seven had unilateral hippocampal atrophy and 14 did not. Thus the authors concluded that their hippocampal volume measurements were 75 percent sensitive, and 64 percent specific in the detection of unilateral medial temporal lobe seizure onset. In this study, right and left hippocampal volume differences were expressed as a proportion of summed right and left hippocampal volumes, a strategy which the authors admit is not sensitive to bilateral atrophy. The authors also pointed out that only patients whose seizures were not adequately lateralized and localized with scalp EEG and other noninvasive measures were including in this study, suggesting that their measures of sensitivity and specificity may be low, as they were obtained through the study of their more challenging cases.

Spencer (1994) noted that the varying MRI protocols and methods of temporal lobe and hippocampal measurement across epilepsy research centers makes it difficult to assess the sensitivity and specificity of qualitative MRI assessment with regard to its correlation with EEG localization of seizure foci. Upon reviewing the literature, Spencer determined that of 809 patients who underwent EEG localization, 337 showed temporal lobe abnormalities on MRI. Of these, qualitative MRI assessment of temporal lobe abnormalities was 55 percent sensitive and 78 percent specific in its agreement with EEG localization of seizure foci. Across studies that












compared quantitative MRI measurements of hippocampal volume with EEG localization, MRI was shown to be 71 percent sensitive and specific in its correspondence with EEG.

Spencer (1994) further noted that studies comparing abnormalities detected through

qualitative MRI assessment to histological validation of pathology in resected tissue showed MRI to be 69 percent sensitive and 68 percent specific in the detection of temporal lobe abnormalities. Quantitative MRI measurements of hippocampal volume revealed 65 percent sensitivity and 80 percent specificity. Spencer cautions that her combination of the above studies did not take into account method of EEG localization, quality of MRI resolution or interpretation, or bias in patient selection.


Hippocampal Volumes, Pathology, and Neuropsychological Performance


In 1990 Sass administered the verbal Selective Reminding Test pre-operatively to 35

patients diagnosed with intractable temporal lobe epilepsy, and compared scores to neuronal cell densities found in areas CAl, CA2, CA3, the hilar area, and the granule cell layer of the dentate gyrus (Sass et al., 1990). Histological examination of the resected tissue revealed significant neuronal loss in all examined areas, relative to autopsy controls. Cell densities among patients with left seizure foci did not differ from those with right foci, or between those whose EEG lateralization and localization required invasive procedures compared to those whose did not. This latter finding, along with reportedly excellent post-operative seizure control, suggests that noninvasive methods were just as effective as invasive ones in terms of identifying the damaged hippocampus. Both left and right temporal lobe epilepsy patients performed significantly below the norm for healthy individuals on the Verbal Selective Reminding Test, however patients with a left seizure focus performed significantly worse than those with a right hemisphere focus. Significant correlations were found between neuronal densities in the CA3 and hilar areas, and












long-term memory retrieval scores on the Verbal Selective Reminding Task, but only for patients with a left hemisphere seizure focus.

In a study of 25 patients with medically intractable TLE and 14 right-handed control subjects, Lencz et al. (1990) investigated the relationships between MRI-based measurements, tissue pathology, and neuropsychological functioning. Results showed that, in control subjects, right and left hippocampal size was not significantly different, although the right temporal lobe was, on average, larger than the left. Among the patients, however, the hippocampus and temporal lobe ipsilateral to the seizure focus was smaller than in the contralateral hemisphere. Histological examination of resected hippocampal tissue revealed that neuronal densities were significantly and positively correlated with MRI-based hippocampal measurements for areas CAl, CA3, the Hilar area, and the granule cell layer of the dentate gyrus. Comparisons of temporal and hippocampal measurements and pre-operative performance on memory tests revealed significant relationships between left temporal lobe size and performance on the verbal Selective Reminding Task, particularly when left/right temporal lobe size ratios among patients with left hemisphere seizure foci were correlated with test scores. A significant correlation was also found when comparing hippocampal measurements of patients with left hemisphere seizure foci with percent retention scores from the Wechsler Memory Scale Logical Memory subtest. Right hemisphere hippocampal and temporal lobe measurements did not correlate significantly with performance on any of the verbal or nonverbal memory tests, indicating the need for more sensitive and specific neuropsychological tests of right hippocampal function.

Trenerry et al. (1993) examined the correlation between preoperative hippocampal volumes and pre- and post-operative performance on memory tests among patients who had temporal lobe resection, including the hippocampus and amygdala, in the treatment of intractable seizures. Hippocampal volumes were expressed in proportion to total intracranial volume. The












authors found that more severe pathology in the left hippocampus is associated with improved verbal and nonverbal memory performance subsequent to resection, and that resection of a relatively spared left hippocampus is associated with poorer memory performance in both domains. With regard to resection of the right hippocampus, this was associated only with a decline in visual learning, but not retention, when the resected hippocampus was relatively healthy. The authors concluded that right hippocampal volume, and the difference in right and left volumes, are more useful in predicting postoperative memory performance than is left hippocampal volume alone. Their results also suggested that a healthy left hippocampus may be responsible for aspects of both verbal and nonverbal learning, but that the right hippocampus may be more able to subsume the verbal memory responsibilities of a non-healthy left hippocampus, than a healthy left hippocampus can subsume the visual memory responsibilities of a non-healthy right hippocampus. On the whole, the investigators reported that the correlations between preoperative hippocampal volumes and pre-operative performance on memory tests were very small, suggesting that the tests they used (Wechsler Memory Scale-Revised, Rey Auditory Verbal Learning Test, and the Visual Spatial Learning Test) were not very sensitive to the memory functions specific to the hippocampal formation. Pre-operative hippocampal volumes were, as expected, smaller ipsilateral to the side of surgery.

In 1995, Shigaki et al. examined the relationships between hippocampal volumes and neuropsychological performance among a well-defined population of 27 epilepsy patients who were candidates for temporal lobe resection. Hippocampal volumes were expressed relative to total intracranial volume so as to adjust for individual differences in overall brain volume. Results showed significant relationships between left hippocampal volume and measures of verbal memory, such as the CVLT: Percent retained at long delay, delayed recall intrusions; WMS-R: Delayed recall performance, and percent of immediately recalled material retained after












a delay. Also, as described above, left hippocampal volume was shown to be significantly correlated with performance on the WCST, as patients with decreased volumes achieved fewer categories and made more perseverative responses and errors than did patients with decreased right hippocampal volumes.

The authors identified the importance of analyzing hippocampal volumes, not with regard to relative differences in right and left values, but as ratios which take total intracranial volume into account. It was also noted that several neuropsychological measures were highly correlated with WAIS-R Verbal IQ or Performance IQ, and statistical methods were utilized to partial out the effects of intellectual functioning in such cases. Several correlations which had appeared to be significant became nonsignificant after controlling for intellectual functioning (Shigaki et al., 1995).

These studies strongly suggest that MRI-based quantitative hippocampal volumes are an easily obtained, reliable, and useful diagnostic tool in the evaluation of hippocampal sclerosis. Further research will be necessary to determine precisely which parameters of hippocampal volume and asymmetry are the most sensitive and specific predictors of seizure lateralization and localization. MRI resolution is increasing dramatically with improved technology. This is expected to lead to more accurate and reliable measures of hippocampal volume.


EEG Localization of Seizure Foci


Although neuropsychological data and MRI studies reveal valuable information

regarding possible seizure foci, the electroencephalogram (EEG) has long been the traditional foundation for localization (Risinger, 1991). EEG allows for both non-invasive and invasive detailed analysis of electrical neural activity. Initially, EEG consisted of recordings made from electrodes attached to the scalp and was termed "surface EEG." Much research and development












led to the advent of invasive EEG, which is carried out by placing recording grids and strips directly on the brain surface, or stereotactically inserting depth electrodes into the brain.

While both surface and invasive EEG have unique advantages and limitations, it is widely agreed upon that invasive EEG should be avoided whenever possible. Any invasive procedure, including that required to place subdural strips, grids, and depth electrodes, carries with it a risk of serious infection, discomfort, and added financial burden to the patient (Barry, Sussmann, O'Connor, & Harner, 1992). Even when intracranial recording is a necessity, it is important to glean as much information as possible from surface EEG, as it can provide information which cannot be obtained from invasive procedures (Risinger, 1991).

Information obtained through the use of surface electrodes has several limitations which are important to understand. First, it has been shown that electrical signals are attenuated or weakened as they travel the distance from discharging neurons to the scalp electrode. In addition, brain tissue, the meninges, skull, and scalp tissue serve to impede the electrical signal. The angle produced by the surface electrode's orientation to the active neurons may also affect its ability to accurately measure neuronal activity. All of these factors serve to reduce the amplitude of the actual neuronal discharge as measured by surface electrodes, and also causes these electrodes to be sensitive mainly to lower frequency discharges, as these are less likely to be attenuated by distance and impeding materials (Risinger, 1991).

Other confounding factors which affect the utility of surface EEG include the existence of artifacts both from within the patient and in the external environment. Electrical activity associated with eye movements, the heartbeat, and musculature of the head and neck is often recorded by surface electrodes, as is activity produced by nearby electrical equipment. The development of improved electrodes has reduced the amount of such activity that is recorded, and additional progress in EEG technology has enabled physicians to filter out some degree of the












artifact which is detected. Despite this, surface electrodes continue to record a sufficient amount of artifact to render some EEG recordings unreadable (Ray, 1990).

Intracranial recording obviously reduces the distance between neurons and recording electrodes, places less impeding material between the two, and is less sensitive to artifact from muscle activity, however, this technique also has its limitations. Because intracranial electrodes are placed in such close proximity to discharging neurons, it is necessary to reduce their sensitivity to obtain an accurate measurement of neural activity. This reduction in "gain" on the EEG machine may cause weaker electrical signals from neurons that are distant from the recording electrode to be undetected. Furthermore, unlike surface electrodes, only a limited number of intracranial electrodes may be used, as their placement requires the creation of a burr hole or skull flap (Risinger, 1991).

The most commonly used array of surface electrodes is called the International 10-20

System, developed by Jasper (1958). This system is so named as electrodes are placed at 10 and 20 percent deviations from the following four cranial landmarks: a) the bridge of the nose or "nasion"; b) the bump on the back of the skull or "inion"; and c & d) the left and right preauricular areas, the small depressions above the cheekbones and anterior to the ears. Electrodes are labeled according to their locations and lateralization in the following manner: "F" = frontal, "P" = parietal, "C" = central, "T"= temporal, and "O" = occipital. Sphenoidal electrodes, implanted in the cheek, are labeled "SP". Electrode lateralization is also coded, with odd numbers referring to the left hemisphere , even numbers corresponding to the right hemisphere, and "Z" indicating a midline placement. For example, electrode T3 is placed on the left temporal lobe, while "FZ" is placed on the midline of the frontal lobe (Ray, 1990).

EEG recordings are actually the result of comparisons made between electrodes which are linked together and placed separately on the scalp. What emerges as one electrode's












recording is that electrical signal which is not common to both placements. An electrical signal which is identical at two sites is referred to as "isoelectric." Deciding which comparisons to make can be difficult, as different strategies have their advantages and limitations. One common technique is to link a number of electrodes together across the scalp and average the activity which they record. That averaged electrical activity is then used as the reference to which subsequent recordings are compared (Ray, 1990).

Using the International 10-20 System, (Jasper, 1958), Risinger (1991) has reported that the most reliable localizing information with regard to surface EEG is obtained through recording of actual seizure activity. EEG's of temporal lobe seizure onset typically begin with a desynchronization or flattening of background neural activity, with approximately half of all seizures subsequently revealing a pattern of rhythmic neuronal activity in the four to eight Hertz or theta frequency range. An electrode indicating an area of maximum amplitude in the theta frequency range within thirty seconds of seizure onset is highly suggestive of seizure focus. This information is thought to be most reliable when changes are noted on EEG prior to the patient's demonstrating any physical of "clinical" signs of seizure activity. When clinical signs precede EEG change, it is suspected that the seizure has spread from its place of origin, and that the electrode showing maximum amplitude may provide false localizing information. Other seizures, even those originating from a temporal lobe focus, may produce more ambiguous EEG patterns which provide little lateralizing or localizing information (Risinger, 1991).

In order to predict the lateralization and location of seizure onset most reliably, Risinger (1991) has suggested several criteria that must be met. As stated above, the optimal situation is one in which EEG change precedes clinical seizure semiology. Also, lateralization and localization suggested by ictal and interictal abnormalities should be consistent, or at least not contradictory to one another. Finally, surface EEG should suggest lateralization and localization












corroborative with that suggested by neuropsychological data and MRI. In instances in which these criteria are not met, or an extratemporal focus is suggested, invasive procedures are recommended.


Ictal and Interictal Surface EEG and Seizure Localization


Given the importance of maximizing the information provided by surface EEG, many investigators have sought to estimate the sensitivity and specificity of this focus-lateralizing strategy, as well as to improve the existing methods of doing so. In order to obtain a reference point against which to compare surface EEG's ability to localize seizure foci, some researchers have used invasive recording as the "gold standard" with which to compare surface EEG (e.g. Lieb, Walsh, Babb, Walter, & Crandall, 1976; Spencer et al., 1985; Risinger, 1989), while others have conducted retrospective studies of patients who underwent successful ATL and determined whether their seizure foci were correctly localized by surface EEG (e.g. Barry et al., 1992).

A major comparison between surface and depth EEG was conducted by Lieb (Lieb et al., 1976), who studied 34 patients with medically intractable seizures who underwent simultaneous surface and depth recordings as potential candidates for resective surgery. Surface electrodes were placed according to the International 10-20 system (Jasper, 1958), while depth electrodes were placed in the anterior, mid, and posterior portions of the hippocampal gyrus, in the anterior, mid, and posterior regions of the pes hippocampus, and occasionally in the amygdala and thalamus. Only seizures which originated unilaterally in the depths of one temporal lobe were considered. Bilateral surface and depth electrodes were recording during all events included in the study. It is important to note that the number of potential recording sites was greater than the number of available channels, and the areas of greatest electrical activity were sampled more heavily than others. Therefore, surface montages were incomplete during most events.












Lieb found that, of 161 total seizures originating unilaterally in the depth of the temporal lobe, 68.2 percent did not register on surface electrodes, and that in cases in which surface electrodes did pick up activity, it was bilateral and synchronous in 16.9 percent of events. In only 14.8 percent of the studied seizures was the event lateralized by ipsilateral surface electrodes, with no incidents of incorrect lateralization. With regard to seizures with varying concomitants, 86 percent of seizures accompanied by clinical behavior change produced alterations in surface electrodes, while 18 percent of those accompanied only by auras produced were registered on surface EEG. Only 10 percent of subclinical seizures produced changes on surface EEG (Lieb et al., 1976).

Lieb suggested that depth electrodes are significantly more useful than surface electrodes in localizing the onset of temporal lobe seizures and in estimating the time of seizure onset. The authors surmised that this is at least partially due to the fact that higher electrical frequencies are filtered out by the biological materials between neurons and surface electrodes, and that mostly incomplete surface montages were used in the study. It was also noted that surface electrodes were able to be used for seizure lateralization in some cases when several events were considered together (Lieb et al., 1976).

In a similar investigation using non-simultaneous depth and surface EEG, Spencer and co-investigators (Spencer et al., 1985) compared the surface and depth electrode recordings of 54 patients with complex partial seizures of temporal and frontal lobe origin. Three experienced electroencephalographers independently attempted to lateralize and localize seizure onset using at least three ictal surface EEG's from each patient. A team of four physicians, working together, attempted to do the same using at least three ictal depth recordings.

Of the 54 patients, 35 had seizure foci that were able to be localized using depth recordings. Twenty-seven of these patients showed temporal lobe onset, seven had frontal












seizure foci, and one patient had a parietal focus. With regard to the reliability of surface EEG, the independent raters showed 46 to 59 percent agreement with one another with regard to seizure lateralization and 28 to 35 percent concordance with one another concerning localization to a specific lobe. These values were corrected for chance. Although most disagreements on lateralization were the result of one or two raters being unable to lateralize a given seizure, raters lateralized seizures to opposite hemispheres two to four percent of the time. With regard to the concordance of surface EEG with depth recordings, levels of agreement on lateralization ranged from 46 to 49 percent, while that for localization to specific lobe ranged from 21 to 38 percent (Spencer et al., 1985).

Despite these rather low values, when the investigators analyzed data concerning only patients whose seizures had been depth localized to the temporal lobe, reliability estimates increased to 66 to 75 percent, while accuracy values increased to 57 to 60 percent. It is important to note, however, that surface EEG resulted in incorrect lateralization in 3 to 17 percent of these cases. Patients with frontal seizures were most difficult to assess (Spencer et al., 1985).

Important implications of this study include the authors' assertion that interpretation of surface EEG is very difficult even for experienced electroencephalographers. The findings that individuals rating surface EEG agreed more with one another than with depth localization, and that at least one third of diagnosed temporal lobe epilepsy patients were unlateralizeable using surface EEG were particularly troubling. It should also be noted, however, that seizure onset could not be determined by depth recording in all cases, and that the population studied in this investigation was inherently difficult to lateralize and localize due to the fact that they all showed sufficient surface EEG ambiguity as to require depth recordings. The authors surmised that development of a set of specific and consistent interpretation criteria may increase the accuracy and reliability of surface EEG (Spencer et al., 1985).












In an effort to create and evaluate such criteria, Risinger and co-investigators studied 110 patients with medically intractable complex partial seizures. All of these patients were suspected to have temporal lobe onset on the basis of seizure history and semiology, and all underwent surface and depth EEG. Risinger's hypothesis was that, at ictal onset, scalp electrodes would agree with subsequent depth recordings with regard to seizure lateralization and localization (Risinger, Engel, Van Ness, Henry, & Crandall, 1989).

Surface recording was done using the International 10-20 System (Jasper, 1958), with additional bilateral sphenoidal electrodes. Based on clinical experience, the authors developed the following set of EEG lateralization and localization criteria: a) A seizure showing a unilateral temporal or sphenoidal rhythmic discharge of five Hz or greater within 30 seconds of EEG ictal onset was defined as being "focal"; b) Ictal recordings that lacked a clear temporal or sphenoidal rhythm of five Hz or greater were defined as "nonlocalizing", even when some other lateralized or focal EEG change was apparent. The authors also sought to evaluate variations in timing of seizure onset, and developed the following criteria: a) Focal EEG recordings which showed a rhythmic pattern of five Hz or greater as the first EEG change (other than suppression of background activity), or those which began as slower temporal/sphenoidal frequencies but reached five Hz or greater within 30 seconds of EEG onset were collectively referred to as having "Initial Focal Onset"; b) A focal pattern of five Hz or greater which appeared after some other type of lateralized or diffuse electrical discharge was defined as having "Delayed Focal Onset." Patients whose ictal recordings showed evidence of both focal and nonlocalizing features were considered to have focal seizures, and those whose recordings showed features of both initial and delayed onset were considered to have initial onset. Overall, an initial or delayed focal ictal rhythmic discharge maximal at one temporal or sphenoidal electrode was used to localize onset and predict depth localization (Risinger et al., 1989).












Two independent raters analyzed 706 ictal recordings using the above system. Ictal recordings made with depth electrodes were analyzed more generally and were defined as lateralized and localized to one or the other temporal lobe, extratemporal regions, or were determined to be nonlocalizing. Of the 110 patients, 57 were determined to have focal seizure onsets using surface EEG. Of these 47 (82 percent) were correctly lateralized and localized when compared to their depth recordings. Patients with uniformly focal seizures were correctly localized 94 percent of the time, while those whose EEG's reflected a combination of focal and nonfocal seizures were correctly localized 67 percent of the time. Two patients were incorrectly lateralized, three were nonlocalizeable, and five showed evidence of extratemporal onset. Raters had "major disagreements" on five to 10 percent of selected recordings, however there were no disagreements regarding lateralization in instances in which localization was agreed upon. Of the remaining 53 patients, 34 were localized to one temporal lobe with depth recordings, and the remaining 19 patients were unable to be lateralized or localized (Risinger et al., 1989).

The authors concluded that distinguishing between initial and delayed seizure onset did not increase accuracy of seizure localization. However, they were able to show that experienced electroencephalographers could in fact predict seizure lateralization and localization reliably and accurately using a consistent set of interpretation criteria, particularly in patients whose seizures consistently met criteria for having focal onset. It was also noted that false localization occurred more commonly than false lateralization, and that complex partial seizures originating in extratemporal regions may show EEG characteristics which make them appear to have temporal lobe onset (Risinger et al., 1989).

The use of depth electrode localization as a "gold standard" against which to measure the accuracy and reliability of surface EEG was criticized by Walczak (Walczak, Radtke, & Lewis, 1992), who pointed out that depth recording is typically limited to a two week period, during












which a given patient may not experience all of his or her seizure types. Furthermore, it has been estimated that up to half of patients who undergo cortical resection following depth recordings have poor surgical outcome (Walczak et al., 1992). Instead of depth recording, Walczak considered freedom from seizures following resection of a given area to be the best indicator that the resected tissue was the seizure focus.

An early study which used both surface and depth electrodes for seizure localization was conducted by Lieb, Engel, Gevins, and Crandall (1981), who studied 52 patients who underwent surface and depth EEG prior to having ATL for medically intractable seizures. Surface recordings were made using the International 10-20 System (Jasper, 1958), while depth electrodes were placed bilaterally in the medical temporal region including the anterior, mid, and posterior hippocampal gyrus, anterior, mid, and posterior pes hippocampus, the amygdala, and occasionally in the uncus, thalamus, and supplementary motor cortex. As with Lieb's earlier study described above (Lieb et al., 1976), recording montages favored depth sites more heavily than surface sites.

Surgical outcome was categorized into four groups, ranging from total seizure relief to no worthwhile benefit from surgery. Two raters assessed between two and five interictal recordings and between one and six from each patient. A total of 455 seizures were assessed. With regard to surgery outcome, 73 percent of the patients fell into one of the three "improved" groups, meaning that they were either seizure free, had less than one seizure per year, or less than one seizure per month (Lieb et al., 1981).

Results showed that several interictal variables were associated with poor surgical outcome, including the presence of bilaterally synchronous surface spikes, sharp waves, and diffuse background slowing. Those associated with surgical success included the presence of multiple independent deep spike patterns. With regard to ictal recording, a high proportion of bilaterally synchronous onsets (surface, deep, or both) and the existence of variable onset












locations were associated with poor outcome. A high proportion of seizure onsets (surface, deep, or both) from the hemisphere later chosen for resective surgery, and a high proportion of deep focal onsets were associated with surgical success (Lieb et al., 1981).

Although this study did not directly compare the utility of surface and depth electrodes in the localization of seizure foci, it was shown that these two methods yield unique information with regard to surgical outcome. Using surface and depth recording together yielded significantly more accurate prognoses than using either one in isolation (Lieb et al., 1981). Subsequent studies evaluated the accuracy and reliability of these two techniques in greater detail.

Also employing the strategy of using surgical outcome as an indicator of the accuracy of preoperative surface EEG localization, Walczak and co-investigators attempted to determine the accuracy and reliability of surface EEG with regard to its ability to lateralize seizure foci and to determine whether some EEG features are more lateralizing than others (Walczak et al., 1992). Thirty-five patients were included in the study, all of whom had undergone ATL for the treatment of medically intractable complex partial seizures, and all of whom had been seizure free for a period of at least two years after surgery and prior to the study. Three independent raters evaluated 137 total seizures with regard to activity at onset, the presence or absence of rhythmic theta/alpha activity, topography of theta/alpha activity, and post-ictal findings. In order for rhythmic theta/alpha activity to be considered present, this activity must have occurred for at least ten seconds and have begun within forty seconds of EEG seizure onset.

Overall, surface EGG was determined to be 76 to 83 percent accurate with regard to lateralization of temporal lobe seizures, and only 47 to 65 percent accurate with regard to extratemporal lateralization. When seizures that were uninterpretable due to artifact were excluded, these ranges of accuracy increased to 92 to 99 percent, and 73 to 100 percent, respectively. With regard to reliability, two raters agreed on lateralization of temporal lobe












seizures 79 to 81 percent of the time, and 47 to 65 percent of the time on lateralization of extratemporal seizures. Seizure foci were incorrectly lateralized in 0 to 6 percent of temporal cases, and not at all in extratemporal cases (Walczak et al., 1992).

Analyses of specific EEG features revealed that activity at onset was indicative of onset 33 to 59 percent of the time in temporal lobe seizures, however, these values increased to 92 to 97 percent when uninterpretable or generalized events were excluded. The presence of rhythmic theta/alpha activity led to lateralization accuracy of 64 to 76 percent, which increased to 97 to 99 percent when uninterpretable or generalized events were excluded. Post-ictal features were slightly less accurate in predicting lateralization (Walczak et al., 1992).

Important implications of this study include the finding that surface EEG is able to provide lateralizing information with moderately good accuracy and reliability with regard to temporal lobe seizures, particularly when uninterpretable events are ignored. The presence of rhythmic alpha/theta activity was the best predictor of seizure lateralization when all events were considered, however onset activity and post-ictal features were also useful when assessing only interpretable events. Lateralization errors were found to be rare, particularly when more than one of the above features were used in lateralization decisions. It was strongly shown that rhythmic theta/alpha activity is highly correlated with temporal lobe localization, and that extratemporal seizures are quite difficult to assess with surface EEG. Similar to Risinger et al. (1989), Walczak was able to improve upon the accuracy/reliability findings of Spencer (Spencer et al., 1985) by employing a consistent set of EEG interpretation criteria (Walczak et al., 1992).

Studying 48 patients who underwent ATL for the treatment of medically intractable seizures, Barry examined pre-operative ictal and interictal scalp EEG's in an attempt to reveal electroencephalographic indicators of surgery outcome. Each patient had undergone at least two 30 minute periods of interictal scalp recording, using the International 10-20 System of electrode












placement (Jasper, 1958). Patients also underwent extended EEG monitoring and a tapering of medication levels, in an attempt to obtain recordings of ictal onset. Forty-five of the patients had seizures during this recording, and nineteen of them ultimately received depth electrode placement, due to bilateral or indeterminate seizure activity (Barry et al., 1992).

Two independent raters examined the interictal recordings, and using the site of

maximum voltage as a location indicator, coded interictal activity as temporal or extratemporal. Patients with 95 percent or more of their interictal spikes occurring in one hemisphere were defined as "unilateral", while the others were defined as "bilateral." Ictal onset was coded in a similar manner, with 80 percent being the cut-off between unilateral and bilateral/indeterminate seizure onset. In making both interictal and ictal localization decisions, electrodes NPl&2, F7&8, and T3-6 were defined as "temporal" while all others were defined as "extratemporal" (Barry et al., 1992).

Results showed that, of the 48 patients, 73 percent experienced fewer than three seizures per year following surgery, while 65 percent became seizure free. Twenty-seven percent continued to have frequent seizures. All 12 patients who had shown unilateral temporal lobe interictal spikes prior to surgery were significantly improved following surgery, and 11 of these patients had temporal lobe resection ipsilateral to the location of their interictal spikes. Seventysix percent of those who had shown bilateral temporal lobe interictal spikes were significantly improved. Only 33 percent of patients who had shown unilateral or bilateral extratemporal interictal discharges were significantly improved postoperatively (Barry et al., 1992).

With regard to recording of ictal onset, patients with evidence of unilateral temporal lobe onset fared slightly better postoperatively than did those with evidence of bilateral or indeterminate onset, but this difference did not reach statistical significance. Approximately 70 percent of the patients in these groups improved following surgery. Patients who required depth












electrode placement exhibited slightly less seizure control following surgery than did those patients who did not, however this is not surprising as the necessity of depth recordings suggests a certain degree of ambiguity in focus localization (Barry et al., 1992).

Barry concluded that of all the EEG parameters examined, interictal spikes were the most accurate indicator of seizure focus localization, and that even patients with bilateral interictal spiking would benefit from surgery providing that these spikes were confined to the temporal lobe and localization of onset was confirmed with depth recording. Barry strongly claimed that patients demonstrating unilateral temporal lobe interictal discharges could be adequately assessed without further monitoring, and that these patients should be considered as surgery candidates providing that there is no contradictory data from MRI or physical examination (Barry et al., 1992).

Taken together, the above studies suggest that surface EEG is a valuable diagnostic tool in the localization of seizure foci, independent of the additional benefit that it is a noninvasive procedure. The research also suggests that surface EEG is most accurate and reliable when specific interpretive criteria are developed regarding seizure localization (Risinger et al., 1989; Walczak et al., 1992), and when EEG findings are consistent with those obtained through neuropsychological testing and MRI (Pilcher et al., 1992).


Neuropsychological Testing, MRI, and EEG


Finally, there has been at least one study which involved examination of the independent contributions of EEG, MRI, and neuropsychological data in the prediction of seizure focus. Williamson et al. (1993) retrospectively studied pre-surgical surface EEG's, MRI's, and neuropsychological data from 67 patients who had undergone successful ATL for the treatment of












intractable seizures. Thirty-seven of these patients underwent left temporal lobe resection, while the remaining 30 had right temporal lobe resection.

EEG results showed that interictal activity could be used to correctly lateralize seizure focus in 33 out of 35 patients who exhibited only unilateral temporal lobe interictal activity. In cases of bilateral temporal lobe interictal activity, the side exhibiting the predominance of activity corresponded with side of seizure focus in 21 out of 28 cases. With regard to ictal activity, 54 patients showed EEG evidence of 5-10 Hz rhythmic buildup within 30 seconds of seizure origin. This provided correct lateralization of seizure focus in 47 of these 54 cases.

MRI was available for only 28 patients, whose scans were analyzed qualitatively.

Twenty-three of these were deemed to have unilateral hippocampal atrophy, a finding which was confirmed by the existence of pathology in resected tissue in 21 of these cases, and disconfirmed in two.

Perhaps the most striking findings in the study were with regard to neuropsychological data. The authors used only the lateralization findings provided by the neuropsychological evaluations, and analyzed how well this corresponded to the final decision regarding the lateralization and localization of each patient's seizure focus. Fifty-eight patients had neuropsychological evidence of focus lateralization, however nine of these cases were found to be mislateralized. Interestingly, all nine of the patient's had right temporal lobe seizure foci. Of the 49 patients who were correctly lateralized, 35 were able to be specifically localized to the temporal lobe.

This study again suggests that EEG, MRI, and neuropsychological data are all useful in the prediction of seizure focus, but that each of these techniques has limitations. It is unfortunate that MRI and neuropsychological data were analyzed in only a qualitative manner, and that the three assessment domains were not analyzed with regard to how they correlate with












one another. Correlations among these domains will be a major focus of the current study, which will seek to identify MRI, EEG, and demographic/illness-related sources of variation in the presurgical neuropsychological performance of patients who subsequently undergo anterior temporal lobe resection.


Epilepsy Surgery at the University of Florida


Shands Teaching Hospital at the University of Florida has a Comprehensive Epilepsy Program specifically designed to treat the many patients who present with seizure disorders. Those whose seizures cannot be adequately controlled through the use of standard or experimental medications are evaluated as candidates for resective surgery. The comprehensive evaluation of a patient's suitability for surgery is divided into two phases, and involves the coordination of several disciplines within the hospital.

Phase I of this evaluation consists solely of non-invasive procedures, and includes EEG, MRI, and neuropsychological testing, and, where necessary, additional procedures including Single Photon Emission Computed Tomography (SPECT), Positron Emission Tomography (PET), and Magnetoencephalography (MEG). Ictal and interictal EEG's are obtained by admitting the patient to the Epilepsy Monitoring Unit (EMU), where they undergo Video/EEG telemetry. This EEG is conducted with the use of scalp and sphenoidal electrodes. Each patient is asked to have an accompanying adult stay with them to notify the EEG technicians in the control room when an aura or seizure occurs if the patient is unable to do so. The EEG recorders also have seizure detection capabilities. If necessary, medications are tapered until the patient has had a sufficient number of interpretable events to obtain the maximal amount of information from EEG.












During this hospitalization, each patient undergoes a comprehensive neuropsychological evaluation at bedside. This evaluation involves assessment of intellectual functioning, verbal and nonverbal memory, language, visuospatial/visuoconstructive skills, frontal and executive functions, motor speed and dexterity, and personality. In addition, a thorough interview is conducted to obtain information regarding psychosocial, occupational, and medical history.

Each patient also undergoes MRI. Both TI and T2 weighted images are obtained and used in the evaluation. TI scans are analyzed visually to detect any tissue abnormalities and to assess the relative sizes of the two hippocampi, and quantitative hippocampal volumes are also calculated. Hippocampal T2 relaxation times are calculated from the T2 scan.

Following this evaluation, all of the results from the above disciplines are presented and discussed at a weekly Epilepsy Management Conference (EMC). Based on the informational value of the EEG and convergence or divergence of these data with neuropsychological testing and MRI, decisions are made with regard to the patient's suitability for surgery. If questions remain following Phase I, Phase II evaluation is necessary.

During Phase II, the patient is hospitalized again, and subdural recording strips, grids, or depth electrodes, alone or in combination, are surgically placed on the surface of the brain corresponding to the area of suspected seizure generation. As in Phase I, these patients undergo video/EEG telemetry until they have had a sufficient number of interpretable events. Phase II results are then discussed at the EMC, at which time final treatment decisions are made.

Regardless of whether a patient's seizure disorder is adequately characterized during Phase I, or whether Phase II is required, all patients undergo a Wada procedure (Wada & Rasmussen, 1960) prior to the final treatment decision. The Wada procedure involves the injection of sodium Brevital into the femoral arteries, one at a time, thereby temporarily anesthetizing each hemisphere via the internal carotid artery while the other hemisphere is able to












function relatively normally. During this anesthetization, verbal and visual memory testing is performed, as is language and motor assessment. In some cases, Brevital is introduced into the cerebral vasculature via the posterior cerebral artery. This delivers the anesthetic primarily to the posterior two-thirds of the hippocampus while sparing the majority of temporal cortex, thereby allowing the patient's healthcare team to test hippocampal function more specifically than is possible during the aforementioned Wada technique. The goals of this procedure are to assess hemispheric dominance with regard to language, and to assess each hemisphere's ability to support memory and language function in isolation.


Purpose of the Present Study


As mentioned above, information from noninvasive EEG, MRI, and neuropsychological testing are discussed and compared following Phase I evaluation. Given the medical risk and financial burden associated with invasive Phase II procedures, it is of utmost importance to obtain as much diagnostic information as possible from Phase I data. This requires not only a thorough understanding of the data from each diagnostic discipline, but also of the relative sensitivity and specificity of the data for localizing the seizure focus.

In the ideal situation, data from EEG, MRI, and neuropsychological testing are all

convergent, and suggest a clearly defined epileptogenic focus, removal of which will significantly reduce or eliminate the patient's frequency without leading to significant cognitive impairment. Unfortunately, this is not always the case. In many cases, members of the epilepsy team must deal with conflicting data and must decide which data are most relevant, and whether one of the three types of data appears to be more informative than the others in a particular case. Phase II evaluation is usually conducted in these instances.








55



Neuropsychological data, in particular, often provide less specific indications of seizure lateralization and localization than EEG and MRI. These data often suggest lateralization and localization that is consistent with that obtained with EEG and MRI, but are also indicative of involvement in other brain regions in both the ipsilateral and contralateral hemispheres. The sources of this relative lack of specificity remain unclear. The current study will attempt to identify electroencephalographic, structural-anatomic, and demographic/illness-related sources of variance in neuropsychological performance.














CHAPTER 2
METHODS


Subjects


Subjects were 61 patients with medically intractable unilateral temporal lobe epilepsy

who participated in a comprehensive epilepsy evaluation and subsequently underwent right or left anterior temporal lobectomy (ATL). Thirty-five subjects underwent left ATL and 26 underwent right ATL. All subjects were right-handed, at least 16 years of age, and underwent neuropsychological assessment and Phase I EEG/Video monitoring. All subjects except for two in the right ATL group and one in the left ATL group underwent MRI from which quantitative hippocampal volumes were obtained. Subjects with a history of neurological insult which occurred after the onset of epilepsy or was deemed to be unrelated to their epilepsy were excluded from the study.


Measures


Demographic and Illness-Related Variables


The following demographic and illness-related variables were collected:

Age (in years) at Neuropsychological Assessment, Gender, Education (in years), Seizure Type(s), Age (in years) at Onset of Epilepsy, Age (in years) at Onset of Secondarily Generalized Seizures (if different from Age at Onset), Duration (in years) of Illness from Initial Onset of Epilepsy, and whether or not the subject has a history of febrile seizures and/or secondarily generalized seizures.












Neuropsychological Variables

The following neuropsychological variables were collected and/or calculated from collected values:

Intellectual functioning. Wechsler Adult Intelligence Scale-Revised (WAIS-R): Full Scale IQ (FSIQ), Verbal IQ (VIQ), Performance IQ, (PIQ).

Verbal memory. Wechsler Memory Scale-Revised (WMS-R): Logical Memory I (LMI), Logical Memory II (LMII); California Verbal Learning Test (CVLT): Short Delay Free Recall (SDFR), Short Delay Cued Recall (SDCR), Long Delay Free Recall (LDFR), Long Delay Cued Recall (LDCR), Free Recall Intrusions (FRINT), Cued Recall Intrusions (CRINT). Using zscores, an overall Verbal Memory Score (VMEMSCOR) was calculated using the following equation: VMEMSCOR = [(LDFR + LDCR + LMII) - (FRINT + CRINT)] / 5.

Nonverbal memory. Wechsler Memory Scale-Revised (WMS-R): Visual Reproduction I (VRI), Visual Reproduction II (VRII); Rey-Osterrieth Complex Figure (ROCF): Delayed Recall (RODELAY). Using z-scores, an overall Nonverbal Memory Score (NVMEMSCOR) was calculated using the following equation: NVMEMSCOR = (RODELAY + VRII) / 2.

Language. Boston Naming Test (BNT); Controlled Oral Word Association (COWA); WAIS-R: VIQ. Using z-scores, an overall Language Score (LANGSCOR) was calculated using the following equation: LANGSCOR = (VIQ + COWA + BNT) / 3.

Visuoconstructive. Rey-Osterrieth Complex Figure (ROCF): Copy (ROCOPY): WAISR: Block Design (BD), Object Assembly (OA). Using z-scores, an overall Visuoconstructive Score (VISCONSCOR) was calculated using the following equation: VISCONSCOR = (ROCOPY + BD + OA) / 3.












Motor. Finger Tapping Test (FIT) : (Dominant and non-dominant hands); Grooved Pegboard Test (GPT): (Dominant and non-dominant hands). Using z-scores, an overall Motor Score (MOTORSCOR) was calculated using the following equation: (Right FIT + Right GPT) (Left FTT + Left GPT).


Missing Neuropsychological Data


Following neuropsychological testing on the EMU, it is not uncommon for a given

subject's neuropsychological profile to be missing one or more variables. Data points may go uncollected for a number of reasons including seizure episodes that occur during testing, patient fatigue and illness, and time constraints. As expected, this was found to be the case upon examination of subject neuropsychological profiles in the current study. Missing data was replaced using the method described below.

Using the neuropsychological variables used to generate the aforementioned

neuropsychological domains, along with additional neuropsychological variables collected during Phase I, correlations were calculated among all neuropsychological variables for all subjects. These variables included the following: WAIS-R: FSIQ, VIQ, PIQ; CVLT: Trial I, Trial V, List B, SDFR, SDCR, LDFR, LDCR, Total Recall Trials I-V, Best Recall Trial, Total Intrusions, Semantic Ratio, Serial Ratio, Perseverations, Recognition Hits, Recognition False Alarms; WMSR: LMI, LMII, VRI, VRII; BNT; COWA; ROCF: Copy, Immediate Recall, Delayed Recall; Finger Tapping Test; Grooved Pegboard Test; Trail Making Test: Parts A and B; Wisconsin Card Sorting Test (WCST): Categories, Perseverative Responses, Perseverative Errors.

For each neuropsychological variable used in the study that had at least one missing

value, correlations between that variable and all others were examined. The five variables with which the initial variable shared the highest correlations were then selected and entered into a












multiple regression equation to generate a predicted score to replace the missing value. For example, for a subject missing a BNT score, one would be predicted for him or her using his or her FSIQ, VIQ, COWA, CVLT LDFR, and CVLT SDFR scores, because these were the variables which were shown to correlate most strongly with BNT performance. In cases in which not all of the predictor variables were available, the existing ones were used in the multiple regression equation to predict the missing value. Information contained in the table in the Appendix includes a list of each set of predictors used to generate replacements for missing variables.


MRI Variables


The following MRI variables were collected: Left Hippocampal Volume (LHIPPVOL); Right Hippocampal Volume (RHIPPVOL); Difference in Volume of Hippocampal Formations (DHF).


EEG Variables


For each subject the number of seizures during Phase I with the following lateralization and localization were obtained: Number of seizures with left mesial temporal lobe localization, number of seizures with left global temporal lobe localization, number of seizures with left fronto-temporal lobe localization, number of seizures with left hemisphere lateralization but not specifically localized to the temporal lobe, number of seizures with right mesial temporal lobe localization, number of seizures with right global temporal lobe localization, number of seizures with right fronto-temporal localization, number of seizures with right hemisphere lateralization but not specifically localized to the temporal lobe. Also collected was each subject's number of nonlateralized seizures and the number of seizures which were not lateralized but believed to be localized to one or both temporal lobes.












Each seizure was also coded according to whether or not it secondarily generalized. Seizures which were uninterpretable due to artifact were recorded only for use in descriptive statistics, and were not used in statistical methods which had a bearing on hypotheses.

Following the above data collection, lateralization and localization categories were "collapsed" in order to calculate the following values for each subject: Number of seizures believed to have left hemisphere lateralization, number of seizures believed to have right hemisphere lateralization, number of nonlateralized seizures.

A Seizure Lateralization Index (SLI), a single value which was indicative of a given

subject's distribution of left hemisphere, right hemisphere, and nonlateralized seizures was then calculated for each subject using the following equation where R = number of seizures originating in the right hemisphere, L = number of seizures originating in the left hemisphere and N = number of nonlateralized seizures: SLI = [(R - L) -.5N (R - L / IR - LI)] / R + L + N

This equation yielded SLI values ranging from 1.00 to -1.00 for all subjects. A SLI

value of 1.00 indicated that all of the subject's seizures were lateralized to the right hemisphere, a SLI value of -1.00 indicated that all of the subject's seizures were lateralized to the left, while values closer to 0.00 indicated varying degrees of right or left tendency with regard to lateralization of seizures.

The SLI equation had three major components which will be discussed below. First, (R L) represented the difference in number of seizures lateralized to the right versus left hemisphere for a given patient. This yielded a very basic indicator of the likelihood that a given patient's seizure focus was in the right versus left hemisphere, but did not take into account the occurrence of nonlateralized seizures or the total number of seizures that was recorded.

To address this first issue, [.5N (R - L / IR - LI)] was subtracted from the original (R - L) remainder described above. In the first part of this component of the equation, nonlateralized












seizures were assigned a weight of .5. This weight was chosen because nonlateralized seizures are an important part of the clinical data but are not considered as clinically significant as are seizures with a clear lateralization. In clinical practice for example, a patient may have ten seizures which are all lateralized to the right hemisphere. If the patient then has a seizure which is clearly lateralized to the left hemisphere, this is considered to be a significantly greater contradiction to the previous ten events than would be the case if the patient's eleventh seizure were interpreted as nonlateralized. Nonetheless, the occurrence of the nonlateralized seizure would serve to slightly diminish the epileptologist's confidence that the patient's seizures are truly lateralized to the right hemisphere.

The second part of this component of the equation involved multiplying the .5N by (R L / IR - LI). This served only to correct the mathematical sign assigned to the .5N. If more of the patient's seizures were lateralized to the right (i.e. if R - L yielded a positive number), this part of the equation ensured that .5N would also be assigned a positive value which would then be subtracted from R - L. In the event that more of the patient's seizures were lateralized to the left hemisphere (i.e. if R - L yielded a negative number), this part of the equation ensured that .5N would be assigned a negative value which would then be subtracted from R - L. In this case, subtraction of a negative is the equivalent of adding a positive .5N to R - L. Simply stated, whether R - L was a positive number indicating that most of the patient's seizures were lateralized to the right, or whether R - L was a negative number indicating that most of the patient's seizures were lateralized to the left, the SLI equation was designed in such a way that the presence of nonlateralized seizures served to bring the value of the R-L remainder back toward zero.

Finally, the total number of seizure events was taken into account by dividing the SLI numerator [(R - L) - .5N (R - L / IR - LI)] by the total number of seizures (R + L + N). This












served to put the discrepancy between the number of seizures lateralized to each hemisphere in the context of total number of seizures that occurred. The importance of this can be appreciated by considering the following example. Patient A may have had seven seizures lateralized to the right hemisphere, three lateralized to the left hemisphere, and no nonlateralized seizures. If Patient A's SLI were calculated without dividing by the total number of seizures, it would yield a SLI value of four. Patient B may have had four seizures lateralized to the right hemisphere, none lateralized to the left hemisphere and no nonlateralized seizures. Calculating Patient B's SLI value without dividing by the total number of seizures would also yield a SLI value of four. The two cases, however, are very different from a clinical standpoint. Specifically, Patient A had some seizures lateralized to the right hemisphere and some lateralized to the left, while all of Patient B's seizures were lateralized to the same hemisphere. When division by the total number of seizures was included in the calculation of SLI, Patient A obtained a SLI value of .4, and Patient B obtained a SLI value of 1.0. Thus, it is clear that Patient B's SLI rating is more indicative of right hemisphere seizure lateralization than is Patient A's.


Data Collection Procedures


Demographic and Illness-Related Variables


The demographic and illness-related variables described above were collected from subject files on the EMU and in the Neuropsychology Laboratory.












Neuropsychological Data


These data were gathered from patient presurgical evaluation files in the Psychology Clinic, were entered into a database and were verified following entry. Following is a brief description of each of the measures that was included in the experimental phase of the study:

Wechsler Adult Intelligence Scale-Revised (WAIS-R). The WAIS-R (Wechsler, 1981) is a general measure of intellectual functioning. It consists of a Verbal Scale, which has six subtests, and a Performance Scale, which includes five subtests. The WAIS-R yields a Full Scale IQ, which is a measure of general intellectual functioning, and Verbal and Performance IQ's, which more specifically reflect subject performance on the two scales described above. These IQ scores have a mean of 100 (SD = 15), and allow for comparison of an individual's performance to that of his or her age peers.

Wechsler Memory Scale-Revised (WMS-R). The WMS-R (Wechsler, 1987) is a battery of verbal and visual memory tests, two of which were included in the current study. The Logical Memory subtest of this battery consists of immediate and delayed recall of two short stories which are read to the subject. The Visual Reproduction subtest consists of immediate and delayed reproduction of four geometric figures which the subject is shown for ten seconds each.

California Verbal Learning Test (CVLT). The CVLT (Delis et al., 1987) is a test of

verbal learning and memory. A list of 16 shopping items, consisting of four items from each of four categories, is read to the subject five times. After each reading, he or she is required to recite back as many of the items as possible. The subject is then presented with an interference list, which is read only one time, after which free recall is required. Subject recall of the first list is then assessed after short and long delays, as is ability to recall the list when given categorical












cues. Finally, recognition memory for list items is assessed. Many scores, expressed in standard deviation units, are obtained through computer scoring of this test.

Rey-Osterrieth Complex Figure (ROCF). The ROCF (Rey, 1941; Osterrieth, 1944)

assesses subject ability to copy a complex geometrical design, and reproduce it immediately and after a delay, without having been asked to remember it. ROCF scaled scores (M = 10; SD = 3) used in the current study were derived from the Denman (1984) scoring system and norms, which score each reproduction of the figure based on the presence versus absence of specific details, the accuracy of detail reproduction, and the location of these details within the figure. These

Boston Naming Test (BNT). The BNT (Kaplan, Goodglass, & Weintraub, 1983) is a test of confrontational naming, in which the subject is presented with black and white line drawings of objects that they are then asked to name. Items are arranged in order of difficulty, and the test progresses until either the last item has been completed or the subject gives six consecutive incorrect responses. Semantic and phonemic cues are provided according to specific criteria, but only spontaneous responses and those following semantic cues are eligible to be counted as correct. The total number of correct responses is then transformed into a standard deviation score using norms provided with the test.

Multilingual Aphasia Examination (MAE). The MAE (Benton & Hamsher, 1989) is a battery of tests designed to assess language functioning. The Controlled Oral Word Association (COWA) subtest from this battery was used in the current study. COWA involves presenting the subject with a letter and asking him or her to generate as many words as possible that begin with that letter. After 60 seconds, there is a brief pause, and the examiner gives the subject a new letter with the same instructions. Subjects are asked not to use proper nouns, and not to use the same word with different endings, such as "eat" and "eating." The letters presented to the subject are, in order, "C", "F", and "L". Scores are derived by totaling the number of words generated












across the three trials, subjecting this total to age and education correction, and transforming this corrected raw score into a percentile.

Finger Tapping Test. The Finger Tapping Test is a simple test of motor speed. Subjects use the index finger to alternately depress and release a small tapping device which counts the number of taps performed by the subject. Subjects are instructed to tap as rapidly as possible until told to stop. A subject's score is the average number of taps they are able to perform across four separate trials, each of which lasts ten seconds. The dominant hand is tested first, followed by the non-dominant hand. In the current study, norms developed by Bornstein (1985) were used to convert raw scores into standard deviation units, taking age, gender and education into account.

Grooved Pegboard. The Grooved Pegboard is a test of speeded motor dexterity. The pegboard has a square array of 25 slotted holes and a small well in which 25 grooved pegs are located. The grooves in the pegs are such that the peg will fit only one way; often, rotation of the groove is necessary in order to fit it into the slotted hole. First using only the dominant hand, subjects are instructed to fill all holes as rapidly as possible, completing each row in a left to right manner. The subject must then do the same with the non-dominant hand, filling each row of holes in a right to left manner. The subjects earns two scores on this test, which are the number of seconds required to complete the task with each hand. In the current study, norms developed by Bornstein (1985) were used to convert raw scores into standard deviation units, taking age, gender and education into account.


EEG Data


EEG data were obtained from subject files in the Control Room on the EMU. These files contain information from each subject's hospitalization for video/EEG telemetry, and include a












detailed history of illness and dictations of each recorded seizure event. Following this dictation is a summary of the patient's surface EEG findings.

During EEG, scalp electrodes were placed according to the International 10-20 System (Jasper, 1958), with the frequent additional use of bilateral sphenoidal electrodes. Patients generally have between six and twelve recorded events, but this number may actually range from one to 60 or more. At Shands Hospital, two major criteria must be met in order for a seizure to be determined to be lateralized and or localized to a given brain region. First, the seizure must involve focal rhythmic discharges with a frequency of four to seven Hertz. Second, this focal rhythmic discharge must be evident within the first 30 seconds following initial EEG seizure onset. All other discharges, excluding those containing sharp waves or spikes, are not considered to be localizing unless they develop into a rhythmic pattern of discharge which meets the above criteria for localization. The use of various electrode montages and filters to obtain maximal information when reviewing seizure events is standard practice at Shands Hospital. Any EEG recordings which are obscured by electromyographic or movement artifact are not used for lateralization or localization unless deemed to be valid by the supervising epileptologist.

The investigator read each subject's video/EEG dictation on a seizure by seizure basis.

The electroencephalographer's statement of seizure lateralization, localization, and whether or not the seizure involved secondary generalization was recorded. Several such dictations contained incomplete or ambiguous information. These were discussed in detail with the supervising epileptologist and decisions regarding seizure lateralization and localization were made collaboratively. Seizures which were uninterpretable due to factors such as electromyographic artifact and technical difficulties were counted for descriptive purposes only.












Using the aforementioned procedures, seizure data was collected in two ways. First, all event dictations were read and recorded by the investigator. This was done without regard of the timing of these events in relation to one another.

Subsequently, the investigator reexamined all event dictations and identified all

occurrences when two or more seizures occurred within any three-hour time period. When two or more seizures occurred within a three-hour interval and were of identical lateralization and localization (e.g. were placed into the same lateralization and localization category described above), they were regarded as one event. This mirrors common clinical practice when counting seizure events on the EMU, as it is difficult to ascertain whether two or more temporally-related and similar seizures are actually isolated events. Counting two similar seizures which occur close together in time as different events may artificially inflate the clinical importance of those two events, because in fact, the second may have occurred at least partially as a result of the first. When two or more seizures occurred within a three hour interval but were of different lateralization and or localization, they were regarded as separate events.

After collecting EEG data in both of the aforementioned manners, the data was reviewed with the supervising epileptologist and it was decided to use the data obtained using the method in which temporally-clustered seizures of identical lateralization and localization were regarded as one event. This decision was based on the combined facts that excluding such seizures did not severely limit the amount of available data and that, as mentioned above, this method of collection best represents clinical practice.


MRI Data


All hippocampal volumes were measured by the principal investigator. As noted above, all but five subjects underwent MRI as part of their Phase I evaluation. Imaging was conducted












using magnetic resonance gradient echo scans (MPR3D: TR = 10 mins, TE = 4 mins, 10 degree flip angle, matrix = 130 by 256, 160 mm volume, and section thickness = 1.25 to 1.40 mm). This protocol produces a gapless series of high quality images which can be reformatted into any plane of view. During hippocampal measurement, images were viewed on a computer monitor via optical disk drive. Using the mouse and cursor, each hippocampus was outlined on every section that it appeared in the sagittal plane. In most cases, a single hippocampus appeared on 10 to 15 separate MRI sections. Volumes were then calculated by multiplying hippocampal area for each section by the thickness of that section, and then adding together the resulting volumes to obtain a total volume for that hippocampus. Final volumes were expressed in cubic centimeters. Each subject's right and left hippocampi were measured twice, with the mean of these two values being the final right and left hippocampal volumes.


Experimental Hypotheses


Neuropsychological Domain Scores

1) Subjects who eventually underwent left ATL will have significantly lower Language and Verbal Memory domain scores than will those who eventually underwent right ATL. 2) Subjects who eventually underwent right ATL will have significantly lower scores on the Nonverbal Memory domain.

3) Left and right ATL groups will not differ significantly on Visuoconstructive or Motor domain scores.

Hypotheses 1 through 3 were tested using one-tailed t-tests with an alpha level of .05 for each test.












Seizure Lateralization Index (SLD


4) Subjects who eventually underwent right ATL will have significantly higher SLI scores (indicating a higher proportion of seizures that were lateralized to the right hemisphere) than will those who eventually underwent left ATL.

Hypothesis 4 was tested using a one-tailed t-test with an alpha level of .05. Difference in Hippocampal Formation Volumes (DHF)

5) Subjects who eventually underwent right ATL will have significantly lower DHF scores (indicating a smaller right hippocampus) than will those who eventually underwent left ATL.

Hypothesis 5 was tested using a one-tailed t-test with an alpha level of .05.


Neuropsychological Data, SLI, and DHF

6) SLI, DHF, and neuropsychological data will all share significant correlations with one another.

Hypothesis 6 was tested by calculating Pearson Product-Moment correlations among the three domains of data, with an alpha level of .05. In order to represent each subject's neuropsychological data as a single value, each subject's data were entered into a discriminant function as described below, with the five neuropsychological domains as predictor variables and side of surgery, right or left, being the predicted group. Then these single neuropsychological discriminant function scores were correlated with SLI and DHF values. Relative Abilities of Neuropsychological Domain Scores, SLI, and DHF to Correctly Lateralize
Eventual Side of Surgery


It should be noted and will be shown in the Results section that neuropsychological

domain scores from the original data set and those from the data set with missing values replaced












by predicted values yielded nearly identical correct surgery side prediction rates. Therefore, subsequent analyses were performed using the neuropsychological domain scores with replacement of missing values. In this manner, all subjects received neuropsychological domain scores that were based on the same set of variables. For example, had the original data been used, one subject's Language domain score may have been derived from the BNT, WAIS-R VIQ and COWA, while another subject may have had this score derived only from the BNT and WAIS-R VIQ. Using the complete data set with missing values replaced avoided this situation. 7) SLI will be a somewhat better predictor of side of surgery than DHIF, and both will be significantly better predictors than neuropsychological domain scores.

This hypothesis was based on findings in the existing literature. Although findings regarding the ability of EEG data to correctly lateralize and localize seizure foci have varied greatly from study to study, the methodology and subject population used by Walczak et al. (1992) appears most similar to that used in the current study. These investigators found that surface EEG was 92 to 99 percent accurate in lateralizing temporal lobe seizure foci among patients who later underwent ATL. As in the current study, recordings that were of questionable validity due to artifact were excluded from the study.

The hypothesis that DHF would be somewhat inferior to EEG with regard to correct prediction of surgery side was also based on findings in the literature. Most studies have examined the ability of MRI-based hippocampal volumetrics to correctly lateralize seizure foci among patients with significant hippocampal atrophy. For example, Jack et al. (1989), found that the existence of any degree of hippocampal asymmetry could correctly lateralize seizure foci in 90 percent of cases. However, the subjects in that investigation all had suspected hippocampal sclerosis based on visual MRI analysis. The current study included many subjects who did not












meet criteria for significant hippocampal asymmetry, and therefore surgery side was predicted based upon widely varying DHF values. Studies examining the utility of MRI-based hippocampal volumes in the lateralization of seizure foci only among subjects with known hippocampal asymmetry would likely yield higher correct lateralization rates than the current study.

The hypothesis that neuropsychological data would be inferior to both SLI and DHF in its ability to correctly predict side of surgery was based on two main facts. First, Williamson et al. (1993) found that neuropsychological data, as interpreted by the neuropsychologist, was only 73 percent correct in lateralizing seizure foci among patients who underwent successful ATL. Second, the current study did not include the diagnostic impression of skilled neuropsychologists, it merely subjected the neuropsychological test data to statistical procedures in an attempt to examine the data's ability to predict eventual side of surgery. Although these procedures were based upon those used by neuropsychologists, it cannot be assumed that they precisely mirrored the diagnostic thought processes of, or conclusions that would have been reached by, actual clinicians.

8) The correct surgery side prediction rate obtained when combining DHF and SLI as predictors will be significantly greater than that obtained when using SLI alone. A similar effect will be obtained when comparing correct surgery side prediction rates using neuropsychological domain scores in conjunction with SLI versus using SLI as the sole predictor of lateralization. Maximal correct prediction of side of surgery will be achieved through using neuropsychological domain scores, SLI, and DHF in conjunction with one another. 9) Neuropsychological domain scores will be more accurate in correctly predicting which subjects would undergo left ATL than in predicting those who would undergo right ATL. This was based on the findings of Williamson t al. (1993) which strongly indicated that current












neuropsychological test data are much more sensitive and specific to left hemisphere pathology, creating a situation in which true right hemisphere pathology was often misinterpreted as left hemisphere pathology.

10) SLI and DHF, used individually as predictors of surgery side, will be equally effective in predicting which subjects would undergo left ATL or right ATL. This is based on the fact that there is no reason to assume that either EEG or MRI is more sensitive to pathology in one hemisphere or the other.

Hypotheses 7 through 10 were tested using discriminant analysis and additional statistical procedures described below. Initially, the five neuropsychological domain scores were entered into a discriminant analysis as predictors of eventual side of surgery. The same method was used for SLI, and then for DHF. This revealed the efficacy with which neuropsychological data, SLI and DHF could each predict eventual side of surgery when used independently of other data. It should be noted that when a discriminant analysis is used with only a single predictor variable, as was the case with SLI and DHF, the procedure is essentially a t-test with the added fact that the method also generates a predicted group membership for each subject.

Next, neuropsychological domain scores (as a package), SLI, and DHF were entered into discriminant analyses in all possible combinations of these three domains of data as predictors of each subject's eventual side of surgery.

It should be noted that these discriminant analyses were conducted using the "jackknife" or, "Leave-One-Out" method. This technique is similar to traditional discriminant analysis, however, each subject is removed from the data pool one at a time, discriminant function coefficients are calculated for each predictor variable based on data from all other subjects, and these coefficients are then applied to the data from the subject who had temporarily been removed from the data pool. Finally, prediction regarding that subject's group classification is made based












on his or her discriminant function score. In this manner, each subject was assigned a predicted side of surgery based on discriminant function coefficients that were calculated independently of his or her own data. This is sound statistical methodology in general, but it was particularly important that it was used in the current study. This is due to the fact that the study examined the utility of neuropsychological, EEG, and MRI data to predict eventual side of surgery, among subjects for whom side of surgery decisions were based on the very same neuropsychological, EEG and MRI data. The use of the jackknife method in this situation ensured that each subject's predicted side of surgery was based on a discriminant function developed independently of his or her own data.

In order to determine whether neuropsychological domain scores, DHF, and SLI, used independently, correctly predicted eventual surgery side at significantly differing rates, the proportion of correct predictions generated by each of the three was calculated. These proportions were then tested for equality using the McNemar test for 2 X 2 tables. Comparisons of classification rates obtained using neuropsychological domain scores, DHF, and SLI in varying combinations were made by obtaining the Wilk's Lambda value from each combination of predictors' discriminant analysis, converting this value into a Hotelling's T2, and then conducting an F-test to test for equality between the two Hotelling's T2values in question. An alpha level of .05 was used for these tests.

Three chi-square tests, with alpha levels of .05, were used in order to determine whether significantly different proportions of LATL and RATL subjects were correctly classified using, separately, SLI, DHF, and neuropsychological domain scores as predictors.













Relative Abilities of Neuropsychological Domain Scores, SLI, and DHF to Correctly Lateralize
Actual Side of Seizure Focus

11) Relative to one another, neuropsychological domain scores, SLI, and DHF will show a similar pattern of efficacy in lateralizing seizure onset as when used to lateralize eventual side of surgery as discussed above.

12) Overall, each of these three domains of data will be slightly more effective in predicting actual side of seizure onset than in merely predicting eventual side of surgery.

In order to test Hypotheses 11 and 12, discriminant analyses, McNemar tests for 2 X 2 tables, f-tests conducted on Hotelling's T2 values, and one-tailed t-tests were used in a manner identical to that which was used to test Hypotheses 7 through 10. However, only subjects who met criteria of an Engel classification of I were used in these analyses. These are subjects who were seizure-free for at least one year following surgery and who were noted to be seizure-free during their most recent clinic visit in Neurology.

An example may be useful in illustrating the importance of using this subset of the

experimental population to test these hypotheses. While Hypotheses 7 through 10 concerned the prediction of eventual side of surgery, they did not directly address actual lateralization of seizure foci. For example, it may have been predicted that Patient A would undergo a right ATL, but one can only assume that the right anterior temporal lobe was the actual seizure focus if that individual became seizure-free following surgery. That is to say, for subjects who experienced seizure-freedom following surgery, it can reasonably be assumed that these subjects did have unilateral focal epilepsy and that the side of surgery and side of actual epileptogenic focus were one in the same.

Following these analyses, all subjects were again included in subsequent analyses,

regardless of seizure outcome. In an attempt to identify demographic and illness-related variables












which may serve as sources of variation in presurgical neuropsychological performance, the following procedures were conducted.

As it had been found that a combination of SLI and DHF served as optimal predictors of eventual side of surgery, subjects for whom this combination had correctly predicted side of surgery were selected. Then the predicted side of surgery for each subject using the neuropsychological domain scores was examined. Subjects for whom neuropsychological domain scores yielded a predicted side of surgery which was in agreement with that predicted by SLI and DHF together were placed in one group which was termed the Convergent Neuropsychology Group. Subjects for whom neuropsychological domain scores yielded a predicted side of surgery which was discrepant from that predicted by SLI and DHF together were placed in a second group, which was termed the Divergent Neuropsychology Group. Convergent and Divergent Neuropsychology Group Differences

13) The Divergent Neuropsychology Group will have mean SLI values that are significantly closer to 0.00 than will the Convergent Neuropsychology Group. This hypothesis was based on the fact that subjects who had SLI values at or near a value of 0.00 showed evidence of seizure lateralization in both hemispheres, and the assumption that this bihemispheric seizure involvement would be reflected in the existence of less specifically lateralized and or incorrectly lateralized neuropsychological profiles among these subjects. 14) The Divergent Neuropsychology Group will have a higher proportion of subjects who are not seizure-free following surgery than will the Convergent Neuropsychology Group. This hypothesis was based on the assumption that the Divergent Neuropsychology Group had neuropsychological profiles that were suggestive of neural pathology contralateral to the surgical












hemisphere, and that the continued presence of this suggested contralateral pathology following surgery may give rise to seizures.

15) The Divergent Neuropsychology Group will have a higher proportion of subjects who have a history of secondarily generalized seizures than will the Convergent Neuropsychology Group. This hypothesis was based on the assumption that seizures, particularly those involving secondary generalization, induce cerebral insult both ipsilateral and contralateral to the actual seizure focus, and that this would be reflected in neuropsychological profiles that were more suggestive of contralateral cerebral insult than were EEG and MRI data. 16) The Divergent Neuropsychology Group will have a lower proportion of subjects who meet EMU criteria for hippocampal atrophy (-.45 cm3> DHF or DHF > +.55cm3) (Gilmore et al., 1995) than will the Convergent Neuropsychology Group. This is based on the assumption that the presence of clinically significant and focal hippocampal atrophy would be reflected in neuropsychological profiles that were more specifically convergent with EEG and MRI data. 17) The Divergent Neuropsychology Group will have a higher proportion of subjects who eventually underwent right ATL than the Convergent Neuropsychology Group. This hypothesis was based on the work of Williamson et al. (1993), who reported that all subjects in their study that were incorrectly lateralized on the basis of neuropsychological data had actual right hemisphere seizure foci. Thus, there does not appear to be a demonstrable and consistent neuropsychological "signature" of right temporal lobe complex partial seizures. Exploratory Group Comparisons


Divergent and Convergent Neuropsychology group differences in age at epilepsy onset, duration of illness, proportion of subjects with a history of febrile seizures, and mean years of education will be examined on an exploratory basis.








77



Two-tailed t-tests were used to test these group differences. An alpha level of .05 was used with each test.












CHAPTER 3
RESULTS


Throughout the following text and tables, the group of subjects who eventually

underwent right ATL will be referred to as RATL, and the group that eventually underwent left ATL will be referred to as LATL. When the entire subject population is referred to, the collective group will be referred to as All.


Demographic and Illness-Related Variables


Descriptive statistics for demographic and illness-related variables are shown in Table 1. Two-tailed t-tests revealed no significant differences between the RATL and LATL groups on the following variables: Education, Age at onset of epilepsy, Age at onset of secondarily generalized seizures (if this occurred), and Duration of illness. An alpha level of .05 was used with each test. Chi-square tests with alpha levels of .05 were used to test for group differences on the remaining variables, which were in the form of group percentages. Again, these tests indicated no significant group differences for the tested values.


Neuropsychological Variables


Descriptive statistics for individual neuropsychological variables are shown in Table 2. These data were taken from the original data set, and represent the normed data with some missing values as collected from individual subjects. As can be seen, some data points were missing and therefore, the n for each variable is included. Two-tailed t-tests were conducted and, when necessary, the Welch correction of degrees of freedom was used to account for inequality of variance between the groups. These analyses revealed that the LATL mean scores were 78












Table 1

Descriptive Statistics for Demographic and Illness-Related Variables


Variable All LATL RATL p



Agea 35.61 (10.54) 35.14 (11.44) 36.23 (09.37) ns Educational 12.73 (02.22) 12.51 (02.11) 13.02 (02.37) ns AO, 12.45 (11.72) 12.27 (11.90) 12.68 (11.71) ns AO2a 14.54 (11.88) 14.04 (21.80) 12.20 (10.72) ns Dur. Illness' 24.10 (11.44) 23.23 (11.05) 25.27 (12.05) ns Femaleb 57.40 57.10 57.70 ns Febrile Sz'sb 39.30 31.40 50.00 ns 2'd Gen. Sz'sb 72.10 68.60 76.90 ns



Note. N = 61 for All, 35 for LATL, 25 for RATL. All = all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; "mean number of years (standard deviation in parentheses); AO = age at onset of epilepsy; AO2 = age at onset of secondarily generalized seizures, if this occurred; Dur. Illness = duration of illness; b percentage of each group; Febrile Sz's = percentage of each group with a known history of febrile seizures; 2nd Gen. Sz's = percentage of each group with a known history of secondarily generalized seizures; ns = groups not significantly different at the .05 alpha level.


significantly lower than RATL mean scores for WAIS-R VIQ, t (41.69) = 2.54, p <.05, CVLT SDFR, t (59) = 2.44, p <.05, CVLT LDFR, t (59) = 2.21, p <.05, and CVLT LDCR, t (59) =


2.13, p < .05. An alpha level of .05 was used for each test.













Table 2

Descriptive Statistics for Individual Neuropsychological Variables


All LATL RATL Variable M SD n M SD n M SD n


FSIQ PIQ VIQ

OAZ BDZ LMI LMII

VRI VRII SDFR SDCR LDFR LDCR FRINT CRINT ROCOPY

RODELAY BNT


91.00 92.18

91.82

-.42

-.16

36.54 30.13

52.09 43.81

-1.67

-1.52

-1.72

-1.79

.67 1.26

8.13 7.49

-2.58


9.91 11.69 9.82

.79 .89

26.35 23.89 33.19 36.17 1.71 1.74 1.78 1.81 1.43 2.31 4.08 3.09 2.34


89.03 91.8 89.06

-.44

-.22

31.35

24.63 53.32 44.56

-2.11

-1.80

-2.14

-2.20

.97 1.57 8.23

8.11

-3.14


9.31 12.72 7.68 .83

.90 22.04 17.52 34.66 37.43 1.43 1.66 1.70 1.73 1.58

2.34 3.95 2.98 2.18


93.65 92.69

* 95.54

-.40

-.08

43.60 37.60 50.33 42.75

* -1.08

-1.15

* -1.15

* -1.23

.27 .85

8.00 6.65 -1.94


10.24 26 10.37 26 11.23 26

.75 26 .88 26

30.34 25 29.27 25 31.63 24 35.07 24 1.90 26 1.80 26 1.76 26 1.80 26

1.12 26 2.24 26 4.34 26 3.10 26 2.38 25












Table 2 Continued


All LATL RATL

Variable M SD n M SD n M SD n



COWA 37.32 33.34 60 35.75 33.54 34 39.37 33.62 26

FTITDOM .03 1.43 57 .02 1.64 32 .05 1.13 25

FTENDOM -.26 1.53 57 -.43 1.57 32 -.03 1.48 25

GPTDOM -1.55 2.26 59 -1.55 2.08 33 -1.55 2.52 26

GPTNDOM -1.79 2.69 59 -1.71 2.19 33 -1.89 3.26 26



Note. All = all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; FSIQ = Wechsler Adult Intelligence Scale-Revised (WAIS-R) Full Scale IQ; PIQ = WAIS-R Performance IQ; VIQ = WAIS-R Verbal IQ; OAZ = WAIS-R Object Assembly z-score; BDZ = WAIS-R Block Design z-score; LMI and LMII = Wechsler Memory Scale (WMS-R) Logical Memory I and II percentile scores; VRI and VRII = WMS-R Visual Reproduction I and II percentile scores; SDFR = California Verbal Learning Test (CVLT) Short Delay Free Recall z-score; SDCR = CVLT Short Delay Cued Recall z-score; LDFR = CVLT Long Delay Free Recall z-score; LDCR= CVLT Long Delay Cued Recall z-score; FRINT = CVLT Free Recall Intrusions z-score; CRINT = CVLT Cued Recall Intrusions z-score; ROCOPY and RODELAY = Rey-Osterreith Copy and Delayed Recall scaled scores; BNT = Boston Naming Test z-score; COWA = Controlled Oral Word Association percentile score; FTTDOM and FTTNDOM = Finger Tapping Test z-scores for dominant and nondominant hands, respectively; GPTDOM and GPTNONDOM = Grooved Pegboard Test z-scores for dominant and nondominant hands, respectively. * = LATL and RATL group means were significantly different (p < .05).


Descriptive statistics for individual neuropsychological variables after replacement of

missing values are shown in Table 3. As there were no missing WAIS-R or CVLT scores in the

original data set (see Table 2), two-tailed t-tests revealed an identical pattern of LATL and RATL













Table 3

Descriptive Statistics for Neuropsychological Variables with Replacement of Missing Values



All LATL RATL Variable M SD M SD M SD


9.31 12.72 7.68

.83 .90 21.74 17.26 34.41 37.10 1.43 1.66 1.70 1.73 1.58

2.34 3.95 2.98 2.08


93.65 92.69

* 95.54

-.40

-.08

44.06 38.29 51.14 43.88

* -1.08

-1.15

* -1.15

* -1.23

.27 .85

8.00 6.58

* -1.93


10.24 10.37 11.23

.75 .88

29.82 28.89 30.53 34.63 1.90 1.80 1.76 1.80

1.12 2.24 4.34 3.16 2.34


FSIQ PIQ VIQ

OAZ BDZ LMI LMII VRI VRII

SDFR SDCR LDFR LDCR FRINT CRINT ROCOPY RODELAY BNT


91.00 92.18 91.82

-.42

-.16

36.65 30.43 51.98 43.87

-1.67

-1.52

-1.72

-1.79

.67 1.26

8.13 7.49

-2.59


9.91 11.69

9.82

.79 .89 26.08 23.73 32.56 35.77 1.71

1.74 1.78

1.81 1.43 2.31 4.08 3.09 2.25


89.03 91.80 89.06

-.44

-.22

31.15 24.59 52.61 43.87

-2.11

-1.80

-2.14

-2.20

.97 1.57

8.23 8.11

-3.09












Table 3 Continued


All LATL RATL

Variable M SD M SD M SD



COWA 37.46 33.08 36.04 33.09 39.37 33.62

FTIDOM .00 1.41 -.03 1.61 .04 1.11

FTTNDOM -.29 1.48 -.47 1.50 -.05 1.45

GPTDOM -1.54 2.24 -1.54 2.05 -1.55 2.52

GPTNDOM -1.75 2.65 -1.66 2.14 -1.89 3.26



Note. All = all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; n= 35 for LATL and 26 for RATL for all variables. These data are from the data set in which missing values were replaced with predicted ones as described above. FSIQ = Wechsler Adult Intelligence ScaleRevised (WAIS-R) Full Scale IQ; PIQ = WAIS-R Performance IQ; VIQ = WAIS-R Verbal IQ; OAZ = WAIS-R Object Assembly z-score; BDZ = WAIS-R Block Design z-score; LMI and LMII = Wechsler Memory Scale (WMS-R) Logical Memory I and II percentile scores; VRI and VRII = WMS-R Visual Reproduction I and II percentile scores; SDFR = California Verbal Learning Test (CVLT) Short Delay Free Recall z-score; SDCR = CVLT Short Delay Cued Recall z-score; LDFR = CVLT Long Delay Free Recall z-score; LDCR = CVLT Long Delay Cued Recall z-score; FRINT = CVLT Free Recall Intrusions z-score; CRINT = CVLT Cued Recall Intrusions z-score; ROCOPY and RODELAY = Rey-Osterreith Copy and Delayed Recall scaled scores; BNT = Boston Naming Test z-score; COWA = Controlled Oral Word Association percentile score; FTrDOM and FTTNDOM = Finger Tapping Test z-scores for dominant and nondominant hands, respectively; GPTDOM and GPTNONDOM = Grooved Pegboard Test zscores for dominant and nondominant hands, respectively. * = LATL and RATL group means were significantly different (p < .05).


group differences with regard to these variables. In addition, the LATL group mean was lower

than the RATL group mean on the BNT, with marginal significance, t (59) = 2.05, p < .05. When

this t-test was conducted on the original data it indicated a trend toward significance, t (51) =












1.92, p = .061. A comparison of Tables 2 and 3 shows that the difference in LATL and RATL group means for BNT is actually smaller following the replacement of missing values procedure. It appears that the increased n in the data set with replaced values led to this difference becoming marginally significant. Again, an alpha level of .05 was used with each test.


Hippocampal Volumetrics


Descriptive statistics for hippocampal volumes are shown in Table 4. Two-tailed t-tests indicated that mean right hippocampal volume was significantly smaller for the RATL group, t

(56) = 4.38, p <.001, and that the mean left hippocampal volume was significantly smaller for the LATL group, 1 (56) = 3.94, p < .001. Mean DHF was significantly higher for the LATL group than the RATL group t (56) = 7.83, p < .001. An alpha level of .05 was used with each test.


EEG Variables


Descriptive statistics for number of seizures lateralized and localized to each brain region, number of seizures rendered uninterpretable due to artifact, number of seizures with secondary generalization, and total number of seizures during Phase I evaluation are shown in Table 5.


SLI, DHF, and Neuropsychological Domain Scores (Hypotheses 1 through 5)


Descriptive statistics for SLI, DHF, and neuropsychological domain scores are shown in Table 6. Two-tailed t-tests indicated that among the neuropsychological domain scores, the LATL group mean was significantly lower than the RATL group mean for VMEMSCOR, t (59) = 2.40, p < .05. The LATL group mean was marginally lower than the RATL group mean for LANGSCOR, t (59) = 2.04, p <.05. These analyses partially supported Hypothesis 1, which












Table 4

Descriptive Statistics for Volumetric Variables


Variable All LATL RATL



RHIPPVOLa 3.16 (.70) 3.45 (.53) > 2.74 (.71) LHIPPVOLa 3.17 (.74) 2.88 (.73) < 3.57 (.53) DHF -.01 (.96) .57 (.64) > -.83 (.71) %RHS 31.00 02.90 70.80 %LHS 31.00 52.90 00.00



Note. n = 58 for All, 34 for LATL, 24 for RATL. All = all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; 'values in cubic centimeters; RHIPPVOL = mean right hippocampal volume; LHIPPVOL = mean left hippocampal volume; DHF = mean difference in hippocampal formation volume (right - left, in cm ); %RHS = percentage of each group meeting criteria for significantly smaller right hippocampus (DHF < -.45 cm3); %LHS = percentage of each group meeting criteria for significantly smaller left hippocampus (DHF > +.55 cm3). predicted that both of these group differences would be significant. Hypothesis 2 was not supported by these analyses, as the two groups did not differ significantly on mean NVMEMSCOR, and Hypothesis 3 was supported, as the group means were not significantly different for VISCONSCOR or MOTORSCOR.

Mean SLI was significantly lower for the LATL than for the RATL group, t (58) = 11.91, p <.001, a finding which is consistent with Hypothesis 4. Mean DHF was significantly lower for the RATL group than for the LATL group, t (56) = -7.80, p <.001, which is in support of


Hypothesis 5.













Table 5

Descriptive Statistics for Phase I EEG Variables


All LATL RATL Variable M SD Sum M SD Sum M SD Sum



LHEM .08 .33 5 .14 .43 5 .00 .00 00 LFRTEMP .25 .91 15 .43 1.17 15 .00 .00 00 LTEMP 1.10 1.89 66 1.69 2.19 59 .28 .89 7 LMESTEMP 1.22 1.81 73 1.80 1.94 63 .40 1.26 10 NLAT .35 1.33 21 .37 1.57 13 .32 .90 8 NLATTEMP .32 1.08 19 .40 1.33 14 .20 .58 5 RHEM .10 .57 6 .00 .00 00 .24 .88 6 RFRTEMP .13 .70 8 .14 .85 5 .12 .44 3 RTEMP .58 1.33 35 .03 .17 1 1.36 1.80 34 RMESTEMP 1.35 2.79 81 .14 .55 5 3.04 3.68 76 2'd Gen. .98 1.27 59 .94 1.00 33 1.04 1.59 26 ARTIFACT .68 3.23 41 .37 1.26 13 1.12 4.80 28 TOTAL 5.48 2.98 329 5.14 2.55 180 5.96 3.49 149



Note. All = all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; n = 60 for All, 35 for LATL and 25 for RATL for all variables. LHEM = number of seizures lateralized to the left hemisphere, but not localized more specifically; LFRTEMP = number of seizures localized to the left frontotemporal region; LTEMP = number of seizures localized generally to the left temporal












Table 5 Continued

lobe; LMESTEMP = number of seizures localized specifically to the left mesial temporal lobe; NLAT = number of seizures that were not lateralized; NLATTEMP = number of seizures that were not lateralized but appeared to originate in one or the other temporal lobe; RHEM, RFRTEMP, RTEMP, and RMESTEMP are the right hemisphere counterparts to the left hemisphere variables described above; 2nd Gen. = number of seizures with secondary generalization; ARTIFACT = number of seizures uninterpreted due to artifact; Total = total number of seizures during Phase I (excluding events uninterpretable due to artifact). All seizure lateralization/localization categories above are exclusive of one another.


Correlations Among SLI. DHF, and Neuropsychological Discriminant Function Scores
(Hypothesis 6)


Following are the correlations among SLI, DHF, and neuropsychological discriminant

function scores. These discriminant function scores were obtained by entering into a discriminant analysis subjects' five neuropsychological domain scores (generated from data in Table 3) as predictors of side of surgery. The resulting discriminant function scores represented each subject's five neuropsychological domain scores as a single value. Correlations were as follows: SLI and DHF: r = -.60 (p < .001, n = 57), SLI and neuropsychological discriminant function score: r = .38 (P <.01, n = 60), DHF and neuropsychological discriminant function score: r = -.30 (p< .05, n = 58). These findings are in support of Hypothesis 6.


Correct Surgery-Side Classification Rates (Hypotheses 7 through 10)


The rates at which SLI, DHF, and neuropsychological domain scores, used independently and in combination with one another, were able to correctly predict side of surgery using discriminant analyses are shown in Table 7. These data represent the correct surgery side classification rates using all available data. In keeping with Hypotheses 7 and 8, the following classification rates were compared: SLI vs. DHF, SLI vs. neuropsychology domain scores, DHF vs. neuropsychology domain scores, SLI and DHF combined vs. SLI, neuropsychology domain












Table 6

Descriptive Statistics for SLI, DHF, and Neuropsychological Domain Scores


Variable All LATL RATL M SD M SD M SD



SLI -.19 (.86) -.80 (.38) a -.83 (.71) LANGSCOR -1.27 (1.19) -1.53 (1.17)


Note. All = all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; For SLI: n= 60 for All, 35 for LATL, 25 for RATL; For DHF: n = 58 for All, 34 for LATL, 24 for RATL; For all other variables: N = 61 for All, 35 for LATL, 26 for RATL; SLI = Seizure Lateralization Index; DHF = Difference in Hippocampal Formation Volume (Right - Left, in cm3); LANGSCOR = Neuropsychological Language Domain Score; VMEMSCOR = Neuropsychological Verbal Memory Domain Score; NVMEMSCOR = Neuropsychological Nonverbal Memory Domain Score; VISCONSCOR = Neuropsychological Visuoconstructive Domain Score; MOTORSCOR = Neuropsychological Motor Domain Score; < and > indicate presence and directionality of significant differences in LATL and RATL group means; " p < .01; b p < .05.


scores and SLI combined vs. SLI, and finally, neuropsychology domain scores, SLI, and DHF combined vs. SLI and DHF combined. The results of these tests are presented below. Numbers of subjects here were slightly different from those in Table 7, as it was necessary when












Table 7

Correct Surgery-Side Prediction Rates Using NP, SLI, DHF as Predictors


Predictor LATL RATL All Variables % CORRECT n % CORRECT n % CORRECT n



SLI 88.60 35 84.00 25 86.70 60 DHF 85.30 34 75.00 24 81.03 58 NPO 60.61 33 61.54 26 61.01 59 NP 60.00 35 61.54 26 60.66 61 DHF SLI 97.06 34 91.30 23 94.74 57 NP SLI 91.43 35 88.00 25 90.00 60 NP DHF 85.29 34 75.00 24 81.03 58 NP DHF SLI 97.06 34 91.30 23 94.74 57



Note. All = all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; SLI = Seizure Lateralization Index; DHF = Difference in Hippocampal Formation Volume (Right - Left, in cm 3); NPO = all five neuropsychological domain scores (LANGSCOR, VMEMSCOR, NVMEMSCOR, VISCONSCOR, MOTORSCOR) from the original data set; NP = all five neuropsychological domain scores from the data set in which missing values were replaced with predicted ones; % CORRECT = percentage of each group for which side of surgery was correctly predicted by entering various combinations of predictor variables into discriminant analyses. comparing two sets of predictors' classification rates, to include only those subjects who had data from all predictors involved in the comparison.

McNemar tests for 2 x 2 tables were used to test the first three of the above comparisons. Due to the fact that there were a relatively low number of subjects in some of the cells of these 2












x 2 tables, chi-square statistics could not be calculated reliably for each of the comparisons. Instead, the exact significance of each comparison was obtained using the binomial distribution, and it is this value that is reported below. Results indicated that individual classification rate for SLI (85.96 % correct) and that for DHF (82.46 % correct) were not significantly different (p = .607, n = 57). Classification rate for SLI (86.70 % correct) was significantly better than that for the five neuropsychological domain scores (60.66 % correct) (p = .004, n = 60). Classification rate for DHF (81.03 % correct) was also significantly better than that for the five neuropsychological domain scores (59.65 % correct) (p = .015, n = 58). These findings are in support of Hypothesis 7.

F-tests were used to test the final three comparisons described above. Results indicated that the classification rate for SLI and DHF combined (94.74 % correct) was significantly better than that for SLI alone (85.96 % correct), F (1, 55) = 17.29, p <.01, n = 57). Classification rate for the five neuropsychological domain scores and SLI combined (90.00 % correct) was not significantly different from that for SLI alone (86.70 % correct, n = 60). Finally, classification rate for neuropsychological domain scores, SLI, and DHF combined (94.74 % correct) was identical to that for SLI and DHF combined (n = 57). These analyses partially supported Hypothesis 8, as the classification rate obtained when using SLI and DHF together was a significant improvement upon that obtained when using SLI as the sole predictor of side of surgery. However, it was also hypothesized that combining the five neuropsychological domain scores with SLI would yield a classification rate that was significantly better than that obtained from SLI alone, and that combining all three sets of predictors would yield the most accurate classification rate of all. As can be seen, the data were not supportive of these last two parts of Hypothesis 8.












Hypotheses 9 and 10 were tested by conducting three chi-square tests, each with an alpha level of .05, comparing the proportion of LATL subjects who were correctly classified regarding side of surgery to the proportion of RATL subjects who were correctly classified. Results indicated that in all three cases, the rate at which LATL vs. RATL subjects were correctly classified were not significantly different. These findings do not support Hypothesis 9, which predicted that the five neuropsychological domain scores would be more effective predictors of LATL than RATL, however they are in support of Hypothesis 10 which predicted that SLI and DHF, used independently, would be equally effective in predicting LATL and RATL.

Following is a detailed description of each set of predictors used in the above discriminant analyses, the resulting values ofWilk's Lambda, its significance, and the Standardized Canonical Discriminant Function Coefficients for each predictor variable used in each analysis. In situations in which there was only one value (e.g. SLI or DHF) used as a predictor of side of surgery, only Wilk's Lambda and its significance will be reported, as the Standardized Canonical Discriminant Function Coefficient for a sole predictor in a discriminant function will always be equal to 1.000.

Entering SLI as the sole predictor of side of surgery yielded the following results: Wilk's Lambda = .290 (p < .001), n = 60. Entering DHF as the sole predictor of side of surgery yielded the following results: Wilk's Lambda = .477 (p < .001), n = 58. Entering the five neuropsychological domain scores from the original data set as predictors of side of surgery yielded the following results: Wilk's Lambda = .841, (p > .05), Standardized Canonical Discriminant Function Coefficients: LANGSCOR = .380, NVMEMSCOR = -.715, VMEMSCOR = .897, VISCONSCOR= .124, MOTORSCOR= -.089, n= 59. Entering the five neuropsychological domain scores from the data set with replacement of missing values yielded the following results: Wilk's Lambda = .819, (p < .05), Standardized Canonical Discriminant












Function Coefficients: LANGSCOR = .521, NVMEMSCOR = -.753, VMEMSCOR = .797, VISCONSCOR =.071, MOTORSCOR = -.113, N = 61.

Entering DHF and SLI as predictors of side of surgery yielded the following results: Wilk's Lambda= .231 (p <.001), Standardized Canonical Discriminant Function Coefficients: SLI= .826, DHF= -.560, n = 57.

Entering the five neuropsychological domain scores and SLI as predictors of side of

surgery yielded the following results: Wilk's Lambda = .263 (p < .001), Standardized Canonical Discriminant Function Coefficients: LANGSCOR = .324, NVMEMSCOR = -.118, VMEMSCOR = .108, VISCONSCOR= .011, MOTORSCOR =.053, SLI= .981 n = 60.

Entering the five neuropsychological domain scores and DHF as predictors of side of surgery yielded the following results: Wilk's Lambda = .428 (p < .001), Standardized Canonical Discriminant Function Coefficients: LANGSCOR = -.073, NVMEMSCOR = .460, VMEMSCOR = -.322, VISCONSCOR = -.124, MOTORSCOR = .036, DHF = .937, n = 57.

Entering the five neuropsychological domain scores, DHF and SLI together as predictors of side of surgery yielded the following results: Wilk's Lambda= .217 (p <.001), Standardized Canonical Discriminant Function Coefficients: LANGSCOR = .183, NVMEMSCOR = -.223, VMEMSCOR= .058, VISCONSCOR = .127, MOTORSCOR= .051, DHF = -.541, SLI = .827, n = 57.


Correct Seizure Focus Lateralization Rates (Hypotheses 11 and 12)


In order to examine correct seizure focus lateralization rates, as predicted by various combinations of SLI, DHF and neuropsychological domain scores, it was necessary to include only subjects who were seizure-free (those with Engel Classification = I) following surgery in these analyses. As described above, for these subjects it can be assumed that side of surgery and












side of seizure focus are one and the same. Descriptive statistics for this subset of the subject population are shown in Table 8.

Two-tailed t-tests were conducted, again using the Welch correction of degrees of freedom where necessary to account for inequality of variance between the groups. An alpha level of .05 was used with each test. These analyses indicated that LATL group mean for SLI was significantly lower than that for the RATL group, t (25.16) = 9.21, p <.001. Mean DHF was significantly lower for the RATL group than for the LATL group, t (40) = -7.34, P < .001. Among the neuropsychological domain scores, the LATL group mean was significantly lower than the RATL group mean for VMEMSCOR, t (42) = 2.22, p <.05. The LATL group mean was significantly lower than the RATL group mean for LANGSCOR, 1t (42) = 2.07, p <.05.

The rates at which SLI, DHF, and neuropsychological domain scores, used independently and in combination with one another, were able to correctly predict seizure focus lateralization using discriminant analyses are shown in Table 9. In keeping with Hypotheses 11 and 12, the following classification rates were compared: SLI vs. DHF, SLI vs. the five neuropsychology domain scores, DHF vs. the five neuropsychology domain scores, SLI and DHF combined vs. SLI, the five neuropsychology domain scores and SLI combined vs. SLI, and finally, the five neuropsychology domain scores, SLI, and DHF combined vs. SLI and DHF combined. The results of these tests are presented below. Number of subjects in each analysis here were slightly different from those in Table 9, as it was necessary when comparing two sets of predictors' classification rates, to include only those subjects who had data from all predictors involved in the comparison.

McNemar tests for 2 x 2 tables were used to test the first three of the above comparisons. Again, the exact significance of each comparison was obtained from the binomial distribution,












Table 8


Descriptive Statistics for SLI, DHF, and Neuropsychological Domain Scores for Subjects who had Engel Classification = I Following Surgery


Variable All LATL RATL M SD M SD M SD



SLI -.23 (.87) -.83 (.36) < .63 (.60) DHF -.02 (.97) .58 (.62) >a -.90 (.66) LANGSCOR -1.20 (1.19) -1.50 (1.11)


Note. All = all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; For DHF: n = 42 for All, 25 for LATL, 17 for RATL; For all other variables: n = 44 for All, 26 for LATL, 18 for RATL; SLI = Seizure Lateralization Index; DHF = Difference in Hippocampal Formation Volume (Right - Left, in cm3); LANGSCOR = Neuropsychological Language Domain Score; VMEMSCOR = Neuropsychological Verbal Memory Domain Score; NVMEMSCOR = Neuropsychological Nonverbal Memory Domain Score; VISCONSCOR = Neuropsychological Visuoconstructive Domain Score; MOTORSCOR = Neuropsychological Motor Domain Score; < and > indicate presence and directionality of significant differences in LATL and RATL group means; a p <.01; b p <.05.


and it is this value that is reported below. Results indicated that classification rate for SLI (88.10 % correct) and that for DHF (85.71 % correct) were not significantly different (p = 1.00, n = 42). Classification rate for SLI (86.67 % correct) was significantly better than that for the five neuropsychological domain scores (65.91 % correct) (p = .019, n = 44). Classification rate for




Full Text

PAGE 1

BIOBEHAVIORAL SOURCES OF VARIANCE IN PRESURGICAL NEUROPSYCHOLOGICAL PERFORMANCE AMONG PATIENTS WITH TEMPORAL LOBE EPILEPSY By DAVID J. MOSER 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 1997

PAGE 2

ACKNOWLEDGEMENTS Completion of this study would not have been possible were it not for the important contributions of many individuals who deserve my recognition and gratitude. First and foremost, I would like to extend my sincere thanks to my dissertation chairperson. Dr. Russell Bauer, for the enormous amount of knowledge, time, effort, patience and humor that he brought to this project. I would also like to thank Dr. Eileen Fennell and Dr. Duane Dede for their insightful comments, support and valuable input during all phases of the study. The assistance and guidance of Dr. Robin Gilmore is greatly appreciated, as she was very giving of her knowledge and expertise regarding epilepsy and also allowed me access to Epilepsy Monitoring Unit resources. I would also like to extend my sincere gratitude to Dr. James Algina, who went to great lengths to provide valuable and timely statistical guidance. Dr. Christiana Leonard also deserves recognition and thanks for her assistance and generosity in allowing me access to the Image Processing Laboratory. I would also like to thank Dr. Tara Spevack for her valuable and thoughtfiil input throughout the study. Dr. Steven Roper, Mr. Timothy Lucas, Ms. Arlene Frank, Ms. Candace Lariz, Dr. Rita Jakus, Ms. Donna Lilly, Ms. Christie Snively, the graduate students in Dr. Russell Bauer's Neuropsychology Laboratory, and the Epilepsy Monitoring Unit technicians also played important roles in the formation and completion of the study. Finally I would like to thank my wife, Becky, and our daughter, Amelia for their constant and unwavering love, support and good humor. ii

PAGE 3

TABLE OF CONTENTS page ACKNOWLEDGEMENTS ii ABSTRACT v CHAPTERS 1 OVERVIEW OF RELEVANT LITERATURE I Epilepsy: Description, Incidence, and Classification 1 Neuropsychological Assessment of Epilepsy Patients 4 Hippocampal Pathology and Neuropsychological Performance 25 Quantitative MRI Volumetric Studies in Epilepsy 29 Hippocampal Volumes, Pathology, and Neuropsychological Performance 34 EEG Localization of Seizure Foci 37 Neuropsychological Testing, MRI, and EEG 50 Epilepsy Surgery at the University of Florida 52 Purpose of the Present Study 54 2 METHODS 56 Subjects 56 Measures 56 Missing Neuropsychological Data 58 MRI Variables 59 EEG Variables 59 Data Collection Procedures 62 Experimental Hypotheses 68 3 RESULTS 78 Demographic and Illness-Related Variables 78 Neuropsychological Variables 78 Hippocampal Volumetrics 84 EEG Variables 84 4 DISCUSSION 100 Discussion of Experimental and Exploratory Hypotheses 1 00 General Discussion 113 iii

PAGE 4

Directions for Future Research 117 APPENDIX 120 REFERENCES 124 BIOGRAPHICAL SKETCH 131 iv

PAGE 5

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 BIOBEHAVIORAL SOURCES OF VARIANCE IN PRESURGICAL NEUROPSYCHOLOGICAL PERFORMANCE AMONG PATIENTS WITH TEMPORAL LOBE EPILEPSY By David J. Moser December, 1997 Chairman: Russell M. Bauer, Ph.D. Major Department: Clinical and Health Psychology Most epilepsy centers use electroencephalographic (EEG), structural-anatomic (MRI), and neuropsychological (NP) data to identify potentially resectable regions of epileptogenic tissue in epilepsy surgery candidates. In ideal cases, these data are convergent and identify a discrete region to be resected. Sometimes, however, NP data are less specific than EEG and MRI data in lateralizing and localizing seizure onset. In the current study, discriminant function analyses (DFA) were used to evaluate the statistical efficacy of EEG, MRI, and NP in predicting side of seizure focus and eventual surgery in a group of complex partial seizure patients undergoing unilateral anterior temporal lobectomy (ATL). Subjects were then divided into two groups depending upon whether their NP data were convergent or divergent with EEG and MRI regarding prediction of eventual surgical hemisphere. These groups were then compared regarding demographic and illness-related variables. Subjects were 61 surgical candidates with complex partial seizures who eventually underwent either right (n = 26) or left (n = 35) ATL. For each subject, scalp EEG onset was v

PAGE 6

coded and transformed to produce a single value reflecting degree of seizure onset laterality (SLI). Likewise, quantitative MRI data were used to produce a value reflecting amount of volumetric asymmetry between the two hippocampi (DHF). Neuropsychological performance was characterized by calculating average z-scores in language, verbal memory, nonverbal memory, motor and visuoconstructive domains. Then, SLI, DHF, and the five NP domain scores were entered as predictors of side of surgery and of side of epileptogenic focus using DFA. Results indicated that SLI was a slightly better predictor of side of surgery and side of seizure focus than DHF, and that both were significantly superior to NP domain scores in this regard. Correct prediction rates generally improved when these SLI, DHF, and NP domain scores were used in combination with one another rather than in isolation. Group comparisons made between those subjects for whom NP data were convergent with EEG and MRI and those for whom these data were divergent yielded no significant differences on demographic or illnessrelated variables. Implications for making treatment decisions in epilepsy programs, and directions for future research, were discussed. vi

PAGE 7

CHAPTER 1 OVERVIEW OF RELEVANT LITERATURE Epilepsy: Description. Incidence, and Classification Epilepsy is a neurological disorder characterized by recurrent seizures (Locharemkul, Primrose, Pilcher, Ojemann, & Ojemann, 1992). A seizure occurs when a group of neurons becomes "irritable" and fires in repeated bursts (Engel, 1989), and may lead to clinically observable changes in behavior if sufficient numbers of neurons are involved. Epilepsy is the most common chronic neurological disorder in the United States, affecting more than 2 million adults and children (Mcintosh, 1992). Although widely used, the term "epilepsy" is not a diagnosis in itself, but is a general term, which subsumes many different syndromes (Kuzniecky & Jackson, 1995). In 1964, the Commission on Classification and Terminology (CCT) of the International League against Epilepsy (ILAE) sought to categorize these various seizure types in such a way that professionals worldwide could treat, research, and discuss these events using consistent criteria and terminology. Prior to that time, many separate classification systems had been developed for different purposes, a situation which gave rise to varying and confusing terminology (Wyllie & Luders, 1993). The efforts of the CCT were successful, and the classification system that was developed remained in use until 1981, when it was revised into the system which is currently in use. A brief description of the current system follows. Currently, all seizures are initially placed in one of two large categories. A seizure is defined as "generalized" if its onset involves activation of a large number of neurons in both hemispheres. "Partial" or "focal" seizures, on the other hand, are those that begin with activation 1

PAGE 8

2 of neurons in a limited part of one hemisphere. A second major classification concerns the presence or absence of alterations of consciousness during the seizure. Those accompanied by no such alterations are termed "simple", while those that produce changes in consciousness are defined as "complex." For the purposes of the present study, the following will concern only partial seizures. According to the 1981 classification system, partial seizures are fiirther subdivided into three categories. As can be inferred fi-om the above, "simple partial seizures" are those that begin focally and involve no alteration of consciousness, while "complex partial seizures" (CPS) begin focally and do involve such changes either immediately or shortly after onset. Finally, there are seizures which are described as "secondarily generalized", which begin as simple or complex partial events and propagate until large areas of both hemispheres are involved. Simple and complex partial seizures are then described more specifically according to the presence or absence of concomitant events such as motor involvement, somatosensory experiences, autonomic signs, and cognitive and emotional disturbances (Commission on Classification and Terminology of the International League Against Epilepsy, 1981). Approximately 40 percent of all epileptic individuals suffer from CPS (Gastaut et al., 1975), and approximately 60 percent of these patients have seizure disorders which are refractory to medication (Rodin, 1968). Thus, finding an alternative treatment for patients with CPS is essential. An increasingly available and effective treatment for patients whose epilepsy cannot be controlled with medication is surgical resection of an identified "epileptogenic focus", which has been defined as that brain location from which a patient's habitual seizures arise, and removal of which will theoretically result in complete cessation of seizures (Risinger, 1991). It has recently been estimated that approximately 1,000 of these resective seizure surgeries are performed each

PAGE 9

year (Pilcher, Locharemkul, Primrose, Ojemann, & Ojemann, 1992), and that 80 to 90 percent of these patients benefit significantly fi-om this intervention. In order to provide healthcare professionals with a usefiil and consistent indicator of surgical outcome, Engel (1987) developed a classification system whereby each postsurgical patient receives a score of I through IV. Criteria for each classification are as follows: Class I (Seizure-free): A) Completely seizure-free since surgery, excluding any seizures which may have occurred during the first few postoperative days; B) Auras only since surgery; C) Some seizures after surgery, but seizure-free for at least two years; or D) Atypical generalized convulsions with antiepileptic drug withdrawal only. Class II (Rare Seizures or "ahnost seizure-free"): A) Initially seizure-free following surgery but has rare seizures now; B) Rare seizures since surgery; C) More than rare seizures after surgery, but rare seizures for at least two years; or D) Nocturnal seizures only, which cause no disability. Class III (Worthwhile Improvement): A) Worthwhile seizure reduction; or B) Prolonged seizure-free intervals amounting to greater than half the follow-up period, but not less than two years. Class FV (No Worthwhile Improvement): Significant seizure reduction; B) No appreciable change; or C) Seizures worse. As can be inferred from these criteria, a given patient's Engel Classification may change across the course of time following surgery. Resective surgery has been shown to be particularly successfiil when the seizure focus and subsequent target of resection is in the temporal lobe (Chelune, 1981; Engel, Van Ness, Rasmussen, & Ojemann, 1993). Although the aforementioned rates of surgical success are impressive, it is important to note that the effectiveness of surgery relies heavily on the selection of appropriate surgery candidates, and must be judged in the context of several factors in addition to a given patient's degree of potential postoperative seizure relief

PAGE 10

4 When evaluating an individual for seizure surgery, the potential for seizure control and confidence with which a given seizure focus is identified must be estimated and compared with the potential for the serious cognitive and psychosocial impairments that may result from temporal lobe resection. This type of "risk-benefit" assessment allows both the physician and patient to make well-informed decisions regarding the likely effects of surgery (Ivnik, Sharbrough, & Laws, 1988). Because such a determination is complex and multifactorial, data in several domains is usually collected to assist in making surgery decisions. Those to be discussed in this study include neuropsychological evaluation, magnetic resonance imaging (MRI), and electroencephalographic (EEG) investigation. Neuropsychological Assessment of Epilepsy Patients Neuropsychological testing plays a unique role in the assessment of the epileptic patient, as it provides information which is complementary to physiological and anatomical measures such as EEG and MRI. Although EEG and MRI provide valuable information about the state of the brain, only neuropsychological testing can reveal the brain's actual behavioral capacities. In fact, properly administered and interpreted neuropsychological testing may reveal brain dysfunction before structural abnormalities are evident using other measures (Jones-Gotman, 1991). One goal of neuropsychological assessment in the study of epilepsy is to identify impairments that may suggest that seizure onset is lateralized to a particular hemisphere, or even more specifically, localized within that hemisphere. Such information, when combined with data obtained through MRI, EEG, and clinical interview, may reveal an epileptogenic zone (Pilcher et al., 1992). In addition to identification of this area of seizure onset, neuropsychological data also plays an important role in determining how likely it is that removal of brain tissue will result in a

PAGE 11

5 significant degree of seizure control (Dodrill, 1986). Factors which have been associated with greater postoperative seizure control include fewer neuropsychological impairments demonstrated on preoperative testing (Wannamaker & Matthews, 1976), and resection of tissue in the hemisphere or temporal lobe in which impairment is implicated by neuropsychological testing (Bengzon, Rasmussen, Gloor, Dussault, & Stevens, 1968). Finally, even though an epileptogenic zone may be identified, and prognosis for seizure control may be positive, it is essential to examine the neuropsychological data with regard to the potential cognitive and psychosocial impairments which may result fi^om tissue resection (Rausch, 1987). In order for neuropsychological data to be properly interpreted, the results of each measure employed must be considered within the context of other test scores as well as psychosocial information, as it is patterns within the data that are most informational, rather than a patient's score on one particular test (Rausch, 1987). Thus, it is important that neuropsychological batteries assess a wide range of brain functions. Although batteries vary widely across epilepsy centers, most thorough batteries include measures of general intellectual functioning, attention, mental control, language, visual perception, and auditory and visual memory (Rausch, 1987). Due to the highly epileptogenic nature of the temporal lobes and the fact that 80 percent of all cortical excisions are made from this region (Jones-Gotman, 1987), it is particularly important that temporal lobe and hippocampal function be assessed using both sensitive and specific measures. The role of the temporal lobe and hippocampus in language and memory, and the resultant risk of amnesia that is posed by resection in these areas have been well documented (Bauer, Tobias, & Valenstein, 1994). Therefore, such assessment typically includes measures of verbal and visuospatial memory functioning and language (Jones-Gotman, 1991). Typically, verbal memory impairment has been associated with language-dominant hemisphere

PAGE 12

6 involvement, while visuospatial memory impairment has been associated with non-dominant hemisphere involvement (Milner, 1967). This pattern of differing performance associated with the language-dominant and non-dominant hemispheres is referred to as material-specific memory impairment. There have been many investigations of temporal lobe functioning in epilepsy, including studies of both non-surgical and surgical epilepsy populations. Memory Functioning in Non-surgical and Surgical Epileptic Populations In 1969, Fedio and Mirsky compared the performance of children with left temporal lobe seizure foci, right temporal lobe seizure foci, and a group with bilateral EEG abnormalities and no specific seizure foci on a battery of intelligence, memory and attention measures. Results showed that children with left temporal lobe epilepsy (TLE) demonstrated lower Verbal than Performance IQ scores on the Wechsler Intelligence Scale for Children, while those with right TLE showed the opposite pattern of performance. Although these groups did not differ in verbal or visual memory span, supraspan tests of verbal memory revealed deficits in the left TLE group, characterized by difficulty learning a word list and recalling the list after a delay. Supraspan tests of nonverbal memory revealed impairment in the right TLE group, and this group also showed significant impairment while attempting to reproduce the Rey-Osterrieth Complex Figure after a short delay. The third patient group showed no significant verbal or visual learning impairment, although these subjects did tend to have difficulty on measures of sustained attention, a deficit which was not evident in either the left or right TLE groups (Fedio & Mirsky, 1969). In 1979, Ladavas and co-investigators sought to determine whether seizure foci in the left temporal lobe, right temporal lobe, left fi-ontal lobe, or right frontal lobe would have different effects on visual and verbal tests of short and long term memory (Ladavas, Umilta, & Provinciali,

PAGE 13

1979). The authors found that patients with right and left temporal lobe seizure foci did not show material-specific memory impairments on measures of memory span, such as Digit Span from the Wechsler-Bellvue Form I and the Corsi Block Tapping Test. These groups did, however, show material-specific effects on measures designed to assess long term memory including verbal tests such as a task which required subjects to learn supraspan strings of digits and another involving recall of a word list after a ten minute delay, and purportedly nonverbal tests such as learning supraspan Corsi block tapping sequences and reproducing the Rey-Osterrieth Complex Figure ten minutes after copying it. It was concluded that temporal lobe abnormalities appear to cause deficits in long term memory, while leaving short term memory tasks such as immediate recall of presented stimuli relatively unaffected. Patients with frontal lobe seizure foci were impaired in a material-specific manner on tests of recency judgment, but were unimpaired on all other measures. These findings are consistent with Baddeley and Warrington's (1970) data concerning the performance of temporal lobe amnesics on tests of short and long term memory and with the findings of Fedio and Mirsky (1969) mentioned eariier. Another interesting finding fi-om Ladavas' study was that the degree of the material-specific long term memory impairments exhibited by the patients appeared to be directly related to length of illness (Ladavas et al., 1979). Delaney (Delaney, Rosen, Mattson, & Novelly, 1980) surmised that some of the variation in memory performance demonstrated across studies may be due to differences in diagnostic criteria, seizure severity and duration, psychological measures, age of seizure onset, duration of epilepsy, seizure fi-equency, and types and levels of medication. These investigators carefiilly matched four groups of epilepsy patients on all of the above factors, and examined their performance on verbal and visual memory tasks relative to that of normal controls. Results revealed that, consistent with most previous studies, patients with a left temporal lobe seizure focus demonstrated marked impairment on verbal memory tasks, while those with a right

PAGE 14

g temporal lobe focus had greater difficulty with visual memory tasks. The performance of patients with either left or right frontal lobe seizure foci did not differ from the normal controls, or from one another. A closer look at the data revealed that left and right temporal lobe patients did not differ significantly in their ability to immediately recall the stories and figures from the WMS, but after a delay, left temporal lobe patients had forgotten a significantly greater amount of the verbal material, while their counterparts had lost a significantly greater amount of the visual material. The authors concluded that the results of studies demonstrating the material-specific memory deficits among left and right temporal lobe epilepsy patients require delayed recall measures in order to reveal reliable differences (Delaney et al., 1980). Mungas and co-investigators examined the performance of patients with left hemisphere TLE, right TLE, and normal controls on a verbal learning task (Mungas, Ehlers, Walton, & McCutchen, 1985) which was modeled after the Rey Auditory Verbal Learning Test. Consistent with the previous study, the groups did not differ significantly in their ability to immediately recall verbal material. However, subjects with left TLE were impaired in their ability to recall the list after a delay. This group's performance was most impaired when the subjects were asked to recall the list after given phonemic/graphemic cues, mildly impaired on delayed free recall, and relatively intact when provided with semantic cues. The right TLE patients and normal controls did not differ on this measure. These data are consistent with those found in earlier investigations (Delaney et al., 1980), in that performance on immediate recall measures did not discriminate the left and right TLE patients from one another, or from controls, and that the left hemisphere group did show impairment in verbal memory across a delay. This impairment further emphasizes the role of the left medial temporal lobe in verbal memory. With regard to the left TLE patients' impaired ability to utilize phonemic/graphemic cues, the authors suggest that temporal cortex

PAGE 15

9 may perform relatively non-semantic language processing, and that the hippocampus' role in verbal memory may be purely learning and retrieval, and predominantly nonlinguistic (Mungas et al., 1985). In another study involving a list-learning task, Hermam, Wyler, Richey, and Rea (1987) compared the performance of 15 patients with left TLE, fifteen with right TLE and fifteen healthy controls on the California Verbal Learning Test (CVLT). No significant differences existed between the groups with regard to demographic variables, and the two patient groups did not differ in terms of duration of illness, age of seizure onset, or level of intellectual fimctioning. Furthermore, 23 of the patients imderwent intracarotid sodium amytal testing to determine which hemisphere was dominant for speech. In all of those cases, left hemisphere dominance was determined. Patients with a femily history of left-handedness were excluded fi"om the study. Results showed that left TLE patients demonstrated significantly impaired verbal learning across the five CVLT trials and recalled significantly fewer total words on immediate fi^ee recall than did the right TLE and control groups, which were indistinguishable fi-om one another on these variables. Furthermore, patients with left TLE also showed a less sophisticated strategy for organizing to-be-remembered words, in that they did not cluster the word list semantically as much as the other groups. No significant group effects were found on variables concerning recognition of list words, delayed recall, or the effects of an interfering word list. The ability of an immediate recall measure to discriminate right and left TLE patients and the inability of a delayed recall task to do so is in contrast to earlier findings described above (Delaney et al., 1980; Mungas et al., 1985). This inconsistency was attributed to carefiil patient selection with regard to language dominance, the fact that patients in this study were being evaluated for surgery and therefore may have had more severe seizure disorders, and the fact that EEG assessment of actual seizure onset was used to determine seizure lateralization and

PAGE 16

10 localization (Hermann et al., 1987). Absence of impairment of recognition memory is consistent with earlier findings (Delaney et al., 1980) Despite that a preponderance of studies have demonstrated that patients with a right hemisphere seizure focus show more severe visual than verbal memory deficits, and that patients with left hemisphere foci tend to show the opposite, it is important to note that not all studies have shown this. In 1973, Glowinski compared the performance of patients with left and right seizure foci on three subtests of the WMS. Glowinski's study revealed that patients with TLE demonstrated a significantly more severe memory deficit than did epileptic patients who did not have temporal lobe foci. Further analysis revealed that, although the expected pattern of performance was borne out among patients with right and left temporal lobe foci, the deficits exhibited by the two groups were not significantly different. Taken together, the aforementioned studies have generally upheld the concept of material-specific memory deficits with regard to memory performance in patients with left and right temporal lobe seizure foci. Most notably, material-specificity was shown most clearly on tests which included measures of long term memory, that is, memory after a delay. Tests of immediate recall yielded less consistent results. In their discussion of the differing memory performance among temporal lobe epileptics, Hermann et al. (1987) implied that patients being considered for surgery and those actually qualifying for this treatment, may have seizure disorders which differ in severity and along other dimensions fi-om those of nonsurgical populations. Subsequently, surgical populations may perform differently on neuropsychological tests, and this may reveal additional information about memory performance in TLE. In 1987, Rausch and Babb examined the performance of 26 patients who had undergone left anterior temporal lobectomy (ATL) and 45 who had undergone right ATL on the Logical

PAGE 17

11 Memory, Verbal Paired Associates, and Visual Reproduction subtests of the Wechsler Memory Scale. Results showed that, as a group, patients who underwent left ATL performed significantly more poorly on immediate and delayed recall of semantically unrelated word pairs and on a test of prose recall. In contrast, patients who underwent right ATL, performed significantly worse on immediate and delayed recall of geometric figures. It is important to note the existence of immediate memory impairment in surgical patients, as this effect was not shown in the aforementioned studies involving nonsurgical populations. Using the same subtests of the Wechsler Memory Scale, Jones-Gotman et al. (1989) compared the verbal memory performance of epilepsy patients with left, right, and bilateral temporal lobe seizure foci. Results showed that patients with a left seizure focus and those with bilateral abnormaUties with a predominance of left hemisphere involvement demonstrated significantly worse performance prior to resective surgery than their right-sided counterparts, demonstrated further impairment 14 days post-surgery, and ultimately returned to baseline at long-term follow-up. Patients with a right seizure focus and those with predominantly right-sided bilateral abnormalities demonstrated superior performance relative to their left-sided counterparts prior to surgery, and demonstrated increasing improvement during their two post-surgery testing sessions. Thus, all four groups of patients were distinguishable from one another using the verbal subtests of the WMS, with unilateral right hemisphere patients showing the least impairment, followed by those with predominantly right hemisphere bilateral abnormalities, those with unilateral left hemisphere seizure foci, and finally, those with predominantly left hemisphere bilateral abnormalities. With regard to performance on the Visual Reproduction subtest, results showed that the group with unilateral left hemisphere seizure foci performed significantly better than all other groups both preand post-operatively, and that the remaining groups could not be distinguished from one another using this measure. Thus, material-specific memory impairment

PAGE 18

12 was shown with regard to both preand post-operative verbal and visual memory (Jones-Gotman et al., 1989). More detailed studies of neuropsychological performance among surgical populations will be discussed later. Language Functioning in Epilepsy In an effort to explain the sources of variation in the fmdings concerning verbal memory in TLE populations, Mayeux and co-investigators (Mayeux, Brandt, Rosen, & Benson, 1980) hypothesized that this ambiguity may be partially due to the varying degrees of language function and dysfunction experienced among these patients. The authors reasoned that few studies performed language assessments, as compared to the relatively large emphasis placed on verbal memory deficits. They also noted signs of anomia, circumlocution, and paraphasic errors among their patients, whose most common complaint was one of verbal memory. The experimental population consisted of 14 patients with well-documented left TLE, seven with right TLE, and eight patients whose seizures were non-focal and tended to secondarily generalize. The groups, which were matched for age, duration of illness, and seizure frequency, were administered a thorough battery including the WAIS-R, the WMS with delayed recall of prose passages, auditory and visual trigrams test, the Benton Visual Retention Test, copy and delayed recall of the Rey-Osterrieth Complex Figure, the Boston Naming Test and a measure of verbal fluency. Data analyses revealed that the only significant differences among group performance was on the Boston Naming Test, on which the left TLE group performed significantly worse than the other groups which did not differ from one another. This difference remained significant even after the effects of individual variations in vocabulary were statistically factored out. The authors concluded that subtle language impairment, particularly a naming impairment, could be the result

PAGE 19

13 of epileptogenic foci in the left inferior temporal lobe. This "anomia" could be perceived by both the patient and physician to be a memory impairment (Mayeux et al., 1980). It should be noted that these data do not strongly indicate that naming deficits account for all apparent memory impairment in temporal lobe epilepsy. Furthermore, including tests of confrontational naming and verbal fluency mark an improvement on earlier studies, however, these do not constitute a thorough language battery. In a similar study, Stanulius and Valentine (1988) examined the preand postoperative performance of patients undergoing left or right ATL on tasks of verbal fluency, confi-ontational naming (Boston Naming Test), and verbal learning (Rey Auditory Verbal Learning Test). Although the groups did not differ on the verbal fluency task, left ATL patients performed significantly poorer than right ATL patients, both preand postoperatively, on the confrontational naming task. Furthermore, left ATL patients showed a mild decline in postoperative performance, whereas the right ATL group improved significantly relative to their preoperative scores. The authors indicated that, among the left ATL patients, performance on the Boston Naming Test and RAVLT correlated at a value of .5 1, which they interpreted to mean that language and memory fimctions have been confounded in studies which have claimed to reveal material-specific memory deficits after ATL. The "anomia" observed in the above studies has also been documented earher in the literature (Heilman, Wilder, & Malzone, 1972). These investigators studied 10 patients who underwent dominant ATL, and found that four of them demonstrated signs of anomic aphasia on a test of confi-ontational naming of real objects. This deficit was persistent beyond six months postoperatively, and was not related to episodes of seizure, impairment of verbal fluency or comprehension.

PAGE 20

14 Despite the existence of data suggesting that tests of verbal memory are confounded by language dysfunction, Hermann and Wyler (1988) provided evidence to the contrary when they compared preand post-operative performance of left and right ATL patients on the Multilingual Aphasia Exam. This is a comprehensive language battery which includes tests of Visual Naming, Sentence Repetition, Controlled Oral Word Association, Oral Spelling, Token Test (comprehension). Aural Comprehension of Words and Phrases, and Reading Comprehension of Words and Phrases. Patients were tested pre-operatively, and six months post-surgery, with alternate forms administered for four of the above subtests. Right ATL resections involved the removal of, on average, 5.1 cm of ATL, while left resections were sUghtly smaller and averaged 4.7 cm. The left hemisphere resections were tailored according to intraoperative electrocorticography and speech-mapping procedures. Results showed that the two groups did not differ with regard to seizure outcome. Nine of 15 left ATL patients, and 10 of 14 right ATL patients became seizure-free following surgery, while all remaining subjects experienced a decrease in seizure frequency of at least 75 percent. The two groups did not differ significantly on any of the individual MAE subtests administered preoperatively, however, the left ATL group did perform more poorly on average. Postoperative performance also revealed no significant differences between the groups, although the left ATL group showed a nonsignificant trend toward demonstrating a greater pre-postoperative decline. Despite this trend, the left ATL group showed significant pre-postoperative improvement on the Token Test and Controlled Oral Word Association. These data strongly suggest that left and right ATL resection can be result in excellent seizure control without a significant loss in language ability, and that material-specific memory deficits observed in TLE are not likely the result of language dysfiinction. The improvement shown by the left ATL group was attributed to a reduction in "neural noise", resulting in more effective ipsilateral brain

PAGE 21

15 function. A similar hypothesis, with regard to contralateral functioning, has previously been proposed (Novelly, Augustine, & Mattson, 1984). Chelune, Awad, and Luders (1989) generated consistent data, using relatively thorough memory and language battery to test 19 right ATL and 23 left ATL patients. Subjects were administered the RAVLT, Wechsler Memory Scale-Revised (WMS-R), Aphasia Screening Exam, Boston Naming Test and measures of Verbal IQ and verbal fluency. Testing was conducted prior to surgery, as well as six months postoperatively. As a group, patients imdergoing right ATL performed better than left ATL patients on Verbal IQ, Boston Naming Test, WMS-R verbal subtests, and the RAVLT. With regard to prepostoperative performance changes, the only changes reaching significance involved the WMS-R verbal memory index and delayed recall of prose from the WMS-R. In these cases, right ATL patients showed significant improvement postoperatively, while left ATL patients did not. It was also noted that a similar, but nonsignificant trend was observed with regard to performance on the BNT. The authors concluded that the general lack of change in performance on language tests suggests that the material-specific deficits observed before and after left ATL appear to be independent, not confounded, with language function. Obviously this is in marked contrast to earlier findings (Mayeux et al., 1980; Stanulius & Valentine, 1988). The degree to which language impairment is associated with TLE and ATL remains unclear. It appears, however, that those studies which utilized more thorough language batteries have shown that language dysfunction may be associated with temporal lobe involvement, but does not likely significantly confound the important memory findings yielded in other studies (Herman & Wyler, 1988; Chelune, Awad & Luders, 1989). The mixed findings of the above studies indicate the need for more thorough language assessment in epilepsy research studies.

PAGE 22

16 Executive Functions in Epilepsy Due to the fact that most seizure disorders and resective surgeries involve the temporal lobes, much of the research examining neuropsychological performance among epilepsy populations have focused most specifically on tests of memory and language. Recent data, however, suggests that patients with seizures of temporal lobe origin often demonstrate impairment on tests of executive functions which have traditionally been thought to be reflective of frontal lobe functioning. The precise nature of this impairment is imclear. At least two theories have been proposed, and these will be discussed below. A measure of executive function which has been shown to be sensitive to temporal lobe seizure involvement is the Wisconsin Card Sorting Test (WCST). The WCST is a commonly used neuropsychological test which is thought to assess subject ability to create and test strategies in a problem-solving context, and to engage in abstract thinking (Grant 1948; Grant & Berg, 1948). The literature has traditionally suggested that patients with frontal lobe pathology, particularly in the dorsolateral prefrontal cortex, perform poorly on this test relative to normal controls and patients with pathology in other brain regions (Mikier, 1963; Heaton, 1981). More recently, however, the specificity of the WCST performance with regard to detection of frontal lobe pathology has come into question. In 1991, Anderson and co-investigators examined WCST performance in 91 patients with "stable focal brain lesions" detected on CT or MRI. The authors found no significant differences in WCST performance among patients with frontal as compared to non-frontal lobe lesions. Furthermore, some patients with extensive frontal lobe pathology performed within normal limits on this measure, while some with non-frontal lobe involvement failed outright. Overall, the authors were unable to identify specific frontal lobe regions which appeared to be specifically

PAGE 23

17 involved in WCST performance, and were able to correctly categorize patients as frontal or nonfrontal in only sixty-two percent of their cases (Anderson, Damasio, Jones, & Tranel, 1991). Previously, Hermann, Wyler, and Richey (1988) examined WCST performance among 35 subjects with complex partial seizures, 16 of who had language-dominant hemisphere temporal lobe onset, and the remainder having had non-dominant temporal lobe onset. A control group of five patients with primary generalized epilepsy and one with a parietal lobe seizure focus was also included in the study. Results showed that patients with non-dominant temporal lobe seizures made significantly more perseverative errors than did the other two groups, and that the patients with dominant temporal lobe seizures made more of these errors than did the controls. The groups did not difiTer on number of categories achieved. The authors attributed this, not to a temporal lobe role in WCST performance, but to an interference in frontal lobe functioning caused by "neural noise" generated by temporal lobe dysfiinction. In support of this theory, the authors found that 17 patients showed improved WCST performance after undergoing ATL, which theoretically removed the temporal lobe source of "neural noise" which had been interfering with frontal lobe functioning. To further examine the relative efifects of frontal versus temporal lobe involvement on WCST , Corcoran and Upton (1993) administered a modified version of the task to 47 patients with unilateral seizure foci. The Stroop Test and a test of verbal fluency were also administered, as traditional tests of frontal lobe functioning. Sixteen of the 47 patients had documented hippocampal sclerosis as detected by MRI, with the cases equally divided between right and left pathology. Thirteen patients had evidence of unilateral temporal lobe seizure foci without apparent hippocampal sclerosis. Seven of these patients had left sided seizure foci, and six had right-sided foci. The remaining 18 patients had unilateral frontal lobe seizure foci, with 10 of these demonstrating left-sided onset and eight having right-sided onset.

PAGE 24

18 The authors found that patients with hippocampal sclerosis took longer to complete the WCST and made more perseverative errors than did members of the other groups. They also achieved fewer categories than the temporal lobe patients without sclerosis, but not the frontal lobe group. Patients with frontal lobe sei2aire foci made more errors on the Stroop Test, and similar to those with non-sclerotic temporal lobe foci, performed more poorly on the test of verbal fluency than did patients with hippocampal sclerosis (Corcoran & Upton, 1993). Looking at the performance of the patients with hippocampal sclerosis, those with right temporal lobe pathology made significantly more categorical errors and took longer to complete the task than did those with left temporal lobe pathology. These lateralization fmdings are consistent with those of Hermann et al. (1988). The authors interpreted their results within the context of Gray's (1982) model of the hippocampus as a "comparator" which compares incoming sensory information to internally stored information about previous responses. Damage to the hippocampus would theoretically lead to dysfunction in this aspect of "working memory", and lead to perseverative responses and failure to maintain set on the WCST (Corcoran & Upton, 1993). Lateralization fmdings contradictory to those above were provided by Strauss et al., who examined WCST performance in 77 patients with complex partial seizures of temporal lobe origin, 35 of whom had left-sided seizure onset, and 42 of whom had right-sided onset (Strauss, Hunter, & Wada, 1993). The authors found that patients with left-sided onset, and an age of onset of less than one year, showed increased perseverative errors and an inability to shift set. Patients with right-sided onset showed more perseverative tendencies than would normals, but this was not as marked as in the left-sided group, and did not appear to be affected by age of seizure onset. Additionally, no relationships were shown between lateralization of language or general intellectual fimctioning and WCST performance for either group (Strauss et al., 1993).

PAGE 25

19 Shigaki et al. (1995) found that left hippocampal volume was significantly correlated with WCST performance, as patients with decreased volumes achieved fewer categories, and made more perseverative responses and perseverative errors on the WCST, than did patients with decreased right hippocampal volumes. These results are more consistent with those of Strauss et al. (1993), than with the previous studies described above (Hermann et al., 1988; Upton & Corcoran). As can be seen, the studies conducted in this area have produced mixed results, leaving WCST performance among patients with temporal lobe seizures an issue which will require further research to clarify. Neuropsychological Test Data as a Predictor of Surgical Outcome Rausch (1987) has pointed out that, although many studies have used surgical resection of the temporal lobe and hippocampus as an independent variable in an effort to reveal the effects of temporal lobe resection on neuropsychological functioning, relatively few have explored how to use test performance to predict surgical outcome and select the most appropriate candidates for surgery. In 1968, Bengzon and co-investigators examined data from 547 patients who underwent temporal lobe excision between 1953 and 1960 (Bengzon et al., 1968). The investigators selected all patients (N_= 50) who were either seizure-free following surgery or who had occasional seizures for 12 to 24 months post-operatively, followed by a cessation of seizures. Placed in a second group were those patients who experienced little or no change in seizure activity following surgery (N = 54). Patients who showed mild improvement were excluded from the study in order to accentuate the differences between the above two groups. Once the groups were formed, they were compared on seven domains of assessment including a) demographics; b) general seizure history; c) seizure pattern analysis and typology; d) skull X-rays,

PAGE 26

20 pneumoencephalograms and angiograms, e) "operative and post-operative features"; f) neuropsychological exams; and g) preand postoperative EEG's. Findings in the neuropsychological domain were as follows. The neuropsychological battery used was that recommended by Milner (1954). In the study described here, Bengzon examined IQ, normal vs. erratic performance patterns, presence vs. absence of performance deterioration, and speech lateralization and localization. Data analyses showed that, as mentioned above, the majority of patients in the "surgical success" group demonstrated neuropsychological exam performance consistent with the existence of a temporal lobe seizure focus. This was not true of the "surgical failure" group, most of whom showed evidence of extratemporal involvement, with or without concomitant temporal lobe abnormalities. With regard to lateralization, Bengzon et al. found that 40 percent of the "surgical success group" showed neuropsychological evidence of right hemisphere lateralization, 33 percent showed left laterahzation, while the remaining patients were unlateralized, bilateral, or normal. Of the "surgical feilure" group, 62 percent were either lateralized to the left hemisphere or bilateral, only 20 percent lateralized to the right hemisphere, and the remaining patients were unlateralized or normal. No other neuropsychological factors were significant (Bengzon et al., 1968). The implications of this early study are important and deserve review. Characteristics of the surgical success group included neuropsychological lateralization to the hemisphere upon which surgery was performed, with greatest success occurring in those patients whose seizures were localized to the temporal lobe. Those with evidence of more diffuse involvement did not fare nearly as well postoperatively. The authors attribute the fact that more patients in the "surgery success" group had right hemisphere foci and resection to the fact that this likely limited the effects of disease and surgery on speech and language (Bengzon et al., 1968).

PAGE 27

21 Wannamaker and Matthews (1976) examined the prognostic value of neuropsychological testing in a population of 14 patients who underwent seizure surgery between 1960 and 1973. Patients were divided into three groups as follows: five who were seizure fi"ee postoperatively, six who showed a clear reduction in seizure activity, and three who showed no benefit of surgery. The neuropsychological battery used in the assessment of these patients was not described in detail, but included the Wechsler Adult Intelligence Scale, tests developed by Halstead, and additional tests aimed at the assessment of motor and sensory abilities. An overall preand postoperative impairment index was generated for each patient based on performance on this battery. Results showed that the patients who were most impaired on preoperative neuropsychological measures were less likely to experience a postoperative decrease in seizure activity, and were also at greatest risk for exacerbated neuropsychological impairment (Wannamaker & Matthews, 1976). The authors also noted that patients undergoing right hemisphere surgery were more likely to show a postoperative decrease in seizure activity and less likely to show a decline in neuropsychological test performance following surgery. Although this is consistent with the results of Bengzon's (Bengzon et al., 1968) study, the authors not only point out that neuropsychological tests may be more sensitive to left hemisphere abnormalities, but also the fact that the patients undergoing right hemisphere surgery typically had lower preoperative seizure frequencies than their left hemisphere counterparts. Unfortunately, laterality and localization of seizures were not discussed in greater detail. Studying a sample of 142 patients who underwent anterior temporal lobectomy, Ivnik, Sharbrough, and Uws (1988) sought to determine effective ways of estimating potential surgical nsks and benefits in order to counsel patients regarding surgery. Patients included in the study were considered medically intractable, had ictal EEG documentation of anterior temporal lobe

PAGE 28

22 seizure focus, and clinical history consistent with TLE. Seizure foci were further confirmed intraoperatively through the use of electrocorticography with both surfece and depth electrodes. All patients were administered neuropsychological batteries preand postoperatively, which consisted of the age-appropriate Wechsler Intelligence Scale, the Wechsler Memory Scale, and the Auditory Verbal Learning Test. Data analysis revealed that right lobectomy patients earned higher Verbal IQ and Verbal Capacity factor scores prior to surgery than did the left lobectomy group, and that this difference increased after surgery. With regard to Performance IQ and Perceptual Organization factor scores, both groups improved postoperatively, with the right lobectomy group showing the greater improvement. The authors carefully pointed out that the vast majority of patients in the study performed within normal limits on the measure of general intellectual functioning both preand postoperatively (Ivnik et al., 1988). With regard to memory measures, the left lobectomy group performed significantly worse than the right group preoperatively on tests of verbal memory. Following surgery, the left temporal lobectomy group showed a significant deterioration in performance on immediate verbal recall, but did not show a relative decrement in delayed verbal recall. Interestingly, surgery did not appear to impair visual memory in either surgical group, a finding which the authors attribute to the possibility that a delayed visual recall measure may have been necessary to reveal impairment. Regarding learning, the two surgical groups did not differ significantly preoperatively on trials one through five of the AVLT or on a measure of "total amount learned", however the right hemisphere group appeared to be less vulnerable to the effects of interference, and performed slightly better on delayed measures office recall and recognition. Following surgery, the right lobectomy group showed improvement on every AVLT variable, while the left group showed a decline on all AVLT variables, resulting in significant group differences for each value (Ivnik et al., 1988).

PAGE 29

23 The authors concluded that right and left anterior temporal lobectomy clearly had differential postoperative effects, particularly involving learning and memory. It was noted that the degree of postoperative decline in learning and memory abilities was usually related to preoperative functioning, in that those subjects who go into surgery with these abilities intact are most likely to experience a significant postoperative decline. Patients with only mild preoperative impairments are more likely to experience a less severe postoperative decline (Ivnik et al., 1988). These findings are in contrast to those of Wannamaker and Matthews (1976). In another, more thorough examination of preand postoperative neuropsychological fimctioning, Bauer and co-investigators studied 2 1 patients undergoing left temporal lobe resection and twelve undergoing right temporal lobe resection for the treatment of intractable epilepsy (Bauer et al., 1995). The neuropsychological battery, which was administered three months prior to surgery and approximately five months after surgery, consisted of the Wechsler Adult Intelligence Scale-Revised, the Logical Memory and Visual Reproduction subtests of the Wechsler Memory Scale-Revised, the California Verbal Learning Test, the Rey-Osterrieth Complex Figure, tests of confrontational naming and verbal fluency, the Benton Test of Facial Recognition, the Judgment of Line Orientation Test, and tests of executive fimctions including the Wisconsin Card Sorting Test. Data analyses revealed that in general, this population scored below the mean on the majority of tests, reflecting preoperative general neuropsychological impairment. A slight trend toward material-specific memory neuropsychological impairment, with left hemisphere patients showing impairment on verbal tasks and right hemisphere patients showing impairment on nonverbal tasks, was noted. However, the authors were carefiil to point out that this did not reach statistical significance (Bauer et al., 1995), and that this is consistent with previous findings (Delaney et al., 1980; Glowinski, 1973).

PAGE 30

24 With regard to post-operative performance, patients undergoing left temporal lobe excision performed significantly worse than those undergoing right hemisphere surgery on tests of verbal memory and language. The opposite effect was shown on tests of nonverbal memory and frontal functioning, however the significance of these data were not as robust as those concerning verbal measures. The authors characterized these results as a "widening" of differences that existed preoperatively between patients with left and right seizure foci, consistent with the findings detailed above (Ivnik, Sharbrough, & Laws, 1988). Also consistent with previous studies (Bengzon et al., 1968; Wannamaker & Matthews, 1976), patients undergoing right temporal lobe resection were more likely to show no change or improve on neuropsychological measures including tests of memory, language and frontal functions. This is partly attributed to the important role that the right hemisphere is believed to play in general attentional processes. It was also noted that resection of the left temporal lobe was associated with the greatest postoperative decline in test performance, although these patients did improve on some measures of nonverbal memory such as the Rey-Osterrieth Complex Figure. With regard to general neuropsychological fimctioning, those patients undergoing left temporal lobe resection who showed the best preoperative performance tended to show the greatest postoperative decrements, while patients undergoing right temporal lobe resection who had the lowest preoperative scores tended to show the greatest postoperative improvement (Bauer et al., 1995). Although somewhat mixed, the results of the above studies suggest that neuropsychological testing provides useful information regarding surgical prognosis. This information includes not only the likelihood that effective seizure control will be attained, but also is indicative of the potential for cognitive decline among different patient groups. Additional

PAGE 31

25 inquiry in this area will be important for maximizing the utility of information obtained through noninvasive pre-operative assessment. Hi ppocampal Pathology and Neuropsychological Performance In a pioneering investigation of the neuropathological correlates of epilepsy, Margerison and Corsellis (1966) studied 55 epilepsy patients using clinical observation, EEG, and subsequent post-mortem histological study of the brain. Subsequent to clinical and EEG assessment, 26 of these subjects received a firm diagnosis of TLE. Of these 26, 22 were found to have hippocampal sclerosis. Eight of 13 subjects who were firmly diagnosed as not having TLE did not show any evidence of hippocampal sclerosis. Among the cases with pathology, the most commonly observed hippocampal abnormalities consisted of neuronal loss and fibrous gliosis in the Ammon's horn, also affecting field HI (Sommer sector), H3 (end folium), and the dentate gyrus. Margerison and Corsellis surmised that the hippocampus has a "selective vulnerability" to the metabolic changes which occur during temporal lobe seizures, and that these changes lead to the development of sclerosis. In turn, sclerosis may maintain, or even increase, the likelihood of seizures occurring, thereby creating a "vicious circle." This important study gave rise to subsequent investigations of hippocampal pathology in epilepsy, its association with neuropsychological functioning and EEG, and how to assess it most accurately. Due to the highly epileptogenic nature of the hippocampus (Lothman, 1991), and the association between epilepsy and hippocampal sclerosis (Margerison & Corsellis, 1966), it is essential that any neuropsychological assessment of the epileptic patient include specific measures of hippocampal fiinction. Although the aforementioned retrospective and prognostic studies demonstrated the differential neuropsychological effects of left, right, and bilateral temporal lobe abnormalities and resection, measures of temporal lobe fiinctioning vary

PAGE 32

26 considerably in their sensitivity to hippocampal pathology. In addition to the differences among neuropsychological batteries used to assess these patients, Jones-Gotman (1987) has pointed out that the differences observed across studies examining the effects of temporal lobe resection may well be the result the varying amounts of tissue that is resected in different treatment centers. At the Montreal Neurological Hospital, she has observed three general patterns of performance on memory tasks used in the assessment of patients who have undergone minor or extensive hippocampal lobe resection. There are tasks which do not appear to be sensitive to the degree of hippocampal resection, such as recall of prose passages and paired associates from the WMS (Mihier, 1967) and recall of the Rey-Osterrieth Complex Figure (Jones-Gotman, 1986). Other tasks, such as repetition of supraspan digit sequences (Milner, 1971) and recall of a complex figure which has been presented to the subject in a gradual, piecemeal manner (JonesGotman, 1986) appear to reflect significant impairment after small resections and significantly greater impairment after large resections. Finally, there are tasks such as subject-ordered pointing to abstract words and designs (Petrides and Milner, 1982), that reveal impairment only in patients that have had a large amount of the hippocampus removed (Jones-Gotman, 1987). Following is a closer look at some of the important studies in this area. In an effort to reveal the important role of the hippocampus in visuospatial learning and memory, Jones-Gotman (1986), examined the abilities of normal controls and patients who had undergone temporal lobe resection to learn to copy a series of 13 abstract designs. The designs were presented to each subject as many times as necessary until he or she was able to correctly reproduce 12 of them from memory. Results showed that patients who underwent left temporal lobectomy did not differ significantly fi-om controls in their performance, except for the fact that surgical patients required more trials to reach criterion than did the controls. Likewise, the performance of patients who had right temporal lobe resection with less than 1.5 cm of the

PAGE 33

27 hippocampus removed did not differ from that of the control subjects. Right temporal lobectomy patients who had more than 1 .5 cm of hippocampal tissue removed showed significant impairment across the first five learning trials, and thereafter did not differ significantly from controls. All subjects were required to recall as many of the designs as possible after a 24 hour interval. Again, only subjects with large right hippocampal resections were significantly impaired, with the remaining groups performing similarly to one another. Jones-Gotman (1986) concluded that patients with extensive right hippocampal resections demonstrated a significant visuospatial learning impairment, in that they were not as able other subjects to benefit from repeated exposure to the stimuli. With regard to the fact that patients who had undergone left temporal lobe resection had difficulty reaching criterion across the learning trials, Jones-Gotman (1986) surmised that this may be due to their having difficulty verbally encoding the figures, whereas their right hemisphere counterparts likely had difficulty encoding them in a spatial manner. The authors also explored primacy and recency effects among their subjects and found that patients with large right hippocampal lesions recalled almost no items from the primary part of the list, significantly fewer items from the middle parts of the list than other groups, and performed similarly to the other groups on items from the last part of the list. All groups showed a significant recency effect. In an analogous study. Frisk and Milner (1990) studied patients with left and right temporal resections and normal controls and their respective abilities to learn a short story. As in the previous study, the to-be-learned material was presented repeatedly until each subject was able to answer correctly a series of questions regarding the story's content. Results showed that patients with left temporal lobe resections evidenced slower rates of learning than the other groups, with the most significant impairment shown by patients whose left temporal lobe

PAGE 34

28 resection had included part of the body of the hippocampus. The same pattern of performance emerged among the groups when subjects were asked to repeat the story and answer questions about the story after a twenty minute delay. The investigators concluded that excision of left temporal lobe tissue impaired learning and retention of a verbal passage, and that the severity of this impairment was strongly related to extent of hippocampal damage. It is surmised that these deficits arise from an abnormally high rate of forgetting among the patients with extensive hippocampal resections (Frisk & Milner, 1990). In an attempt to reveal more specifically the specialization of memory roles within different temporal lobe areas, Rausch and Babb (1987) examined resected tissue from 10 patients who had undergone left ATL. Pre-operatively, all of these patients had been administered the Wechsler Adult Intelligence Scale, and subtests of the Wechsler Memory Scale which required them to recall word pairs, prose, and geometric figures. Two independent examiners calculated cell densities for resected regions of the temporal lobe including the upper and lower dentate fascia, CAM, Prosubiculum, Subiculum, Presubiculum, and the hippocampal, fiisiform, inferior temporal, and middle temporal gyri. Correlational analyses demonstrated that lower cell counts in the hippocampus, particularly CAl, were significantly associated with simple recall of word pairs, whereas low counts in the hippocampal gyrus were associated with the more complex task of immediate and delayed prose recall. An interesting, yet nonsignificant association was also demonstrated between low cell counts in the prosubiculum and recall of word pairs, while low cell counts in the subiculum appeared to associate inversely with prose recall. As expected given the fact that these patients underwent left ATL, none of the regional cell counts were significantly correlated with performance on the nonverbal memory task (Rausch & Babb, 1987).

PAGE 35

29 The authors concluded that anatomically simpler areas, such as CAl and the prosubiculum appear to perform relatively simpler verbal memory tasks, whereas the hippocampal gyrus, with its widespread connections to association cortex, may support more complex verbal memory processes. Thus, evidence is provided that the temporal lobe regions are at least somewhat functionally divided (Rausch & Babb, 1987). The aforementioned studies strongly suggest that the temporal lobe, and hippocampus in particular, may be at least somewhat functionally divided. More detailed studies will be needed in order to define more specifically the roles played by different temporal lobe and hippocampal regions. From a clinical standpoint, what is needed are ways to detect hippocampal sclerosis and to assess hippocampal functioning more effectively prior to surgery, rather than through the examination of resected tissue, hi addition to ongoing progress in the neuropsychological testing of hippocampal function, great strides have recently been made in the use of MRI as it relates to epilepsy. Quantitative MRI Volumetric Studies in Epilepsy Magnetic Resonance Imaging refers to a variety of techniques used to image tissue. Unlike computed tomography (CT) and X-ray, which rely on ionizing radiation to produce an image, MRI is noninvasive and relies upon the inherent magnetic and electrical properties of the nuclei of tissue particles. It is of no known danger to the subject, and has very few contraindications, which include pacemakers, the presence of internal metal objects such as aneurysm clips, and severe claustrophobia (Andreasen, 1988). The basic principle of MRI relies upon the fact that certain nuclei, when placed in a magnetic field, will absorb energy within specific radio frequencies, and subsequently remit this energy as they return to their original state (Kuzniecky & Jackson, 1995).

PAGE 36

30 In the case of brain MRI, the patient is placed within a magnetic field which causes hydrogen ions to line up with one another according to their magnetic properties. Once this alignment has occurred, the brain is exposed to a radio frequency pulse. This pulse rotates the ions from their position, and when the pulse is discontinued, the ions remit energy as they return to their alignment. The MR image is produced by a radio frequency receiver, which records the amount of energy or "signal" remitted from each small piece of brain tissue, each of which is referred to as a "voxel." These voxels are then arranged in image form in pixels, each of which will have a shade of grey corresponding to the amount of signal that was remitted from a given brain area. Different types of tissue contain varying degrees of hydrogen atoms, and therefore remit varying amounts of signal. Thus, these areas will be distinct from one another on the MR image (Andreasen, 1988). As MRI involves the assignation of quantitative values to the signal remitted by many small brain areas, these values can be easily stored in the scanner and reformatted to produce fine images in coronal, sagittal, or horizontal planes (Kuzniecky & Jackson, 1995). Many MRI studies of epilepsy patients have involved quantitative measurement of the temporal lobes, hippocampi, or both, in an attempt to determine the existence of atrophy in one hemisphere relative to the other, or relative to previously collected norms. The simultaneous improvement of MRI resolution and development of new methods of quantitative measurement of brain structures has led to greatly advanced MRI detection of such atrophy since 1990. Abnormalities that were once evident only upon examination of resected tissue are now detectable in the pre-operative patient (Spencer, 1994). Quantitative MRI studies are considered preferable to visual assessment of MRI, as the former are reproducible, can correct for improper head rotation in the scanner, and are slightly more sensitive to atrophy (Bronen et al., 1994). As will be discussed below, quantitative MRI-based measures of hippocampal atrophy have been

PAGE 37

31 shown to correlate with hippocampal sclerosis, lateralization of seizures by EEC, degree of hippocampal neuronal loss, performance on verbal memory measures, and post-operative seizure outcome. In 1991, Cascino et al. investigated the correlation between MRJ-based hippocampal volumes, and histopathology observed in the temporal lobe tissue that was removed from 24 patients with intractable partial seizures. All of these patients' seizures were lateralizeable using noninvasive methods such as scalp EEG. The investigators quantified neuronal cell loss in the subiculum, prosubiculum, and sectors C1-C4 of the hippocampus. Of the 24 patients, 15 met criteria for mesial temporal lobe sclerosis (greater than 50 percent neuronal loss) upon histological examination of the resected tissue. Fourteen of those 15 had significant hippocampal atrophy on the same side as seizure origin, based on MRI. The authors added that MRI-based volume measurements were 93 percent sensitive and 100 percent specific in determination of seizure lateralization. The severity of pathology appeared to correlate with the hippocampal volumes, although the volumes were not able to predict pathology in cases of moderate to severe neuronal loss. In order to establish normative values for MRI-based anterior temporal lobe and hippocampal volumes in nonepileptic subjects. Jack et al. (1989) obtained these measurements from 52 healthy subjects between the ages of 20 and 40. After normalizing these measurements with regard to total intracranial volume, the authors found no significant effects of age or sex on volume of either the right or left anterior temporal lobes. A small, but significant difference in this measurement was found in right-handers, with the right anterior temporal lobe being slightly larger than the left. Age, sex, and handedness did not have a significant effect on right or left hippocampal volumes, however for the group of subjects it was found that the right hippocampus

PAGE 38

32 was, on average, significantly larger than the left (2.8 versus 2.5 cm\ respectively.) Reasons underlying this asymmetry remain unclear. In a retrospective study of 41 right-handed patients who underwent surgery for medically intractable complex partial seizures. Jack et al. (1990) compared the sensitivity and specificity of five measures in their abilities to correctly lateralize seizure onset. Seizure onset, of course, had been previously lateralized through the use video-recorded surface EEG monitoring, clinical observation, and intraoperative electrocorticography. MRI was administered preoperatively, but was not used to determine seizure lateralization. The following techniques were compared and were found to lateralize seizure onset in order of decreasing usefiilness: MRI hippocampal volume measurements, visual grading of MRI to determine hippocampal atrophy, anterior temporal lobe volume measurements, visual grading of MRI to determine anterior temporal lobe atrophy, and unilateral signal intensity of MRI images with long repetition time. Quantitative MRI-based measurements of hippocampal volume correctly lateralized seizure onset in 76 percent of the cases, with no false lateralizations. Decisions regarding the significance of hippocampal atrophy were based on the norms collected previously (Jack et al., 1989). The authors considered atrophy to be significant if the difference between the two hippocampi were at least two standard deviations greater than is seen in normal individuals. A more liberal criterion of 0 cm^ difference between hippocampi would have correctly lateralized 90 percent of cases, with no false lateralizations, however the authors point out that such a liberal criteria could lead to important errors on a case by case basis, as it does not allow for measurement error. It was concluded that quantitative MRI evidence of hippocampal atrophy, combined with EEG data may render invasive recording unnecessary, and that disagreement between these two measures strongly indicates the need for invasive monitoring. It was also noted that visual grading of hippocampal atrophy was particularly difficult in cases of right hippocampal disease, as the right

PAGE 39

33 hippocampus is typically larger than the left, and any atrophy tends to make the two appear more similar in size. In a more recent investigation of the sensitivity and specificity of quantitative MRI volumetrics in the study of epilepsy. Spencer et al. (1993) studied 56 patients with intractable epilepsy. Twenty-nine had intracranial ictal EEC's suggestive of unilateral medial temporal lobe seizure onset. Of these, 21 had quantitatively measured hippocampal atrophy in the same hemisphere, and the remainder did not. Of 21 patients whose intracranial EEC's did not suggest temporal lobe onset, seven had unilateral hippocampal atrophy and 14 did not. Thus the authors concluded that their hippocampal volume measurements were 75 percent sensitive, and 64 percent specific in the detection of unilateral medial temporal lobe seizure onset. In this study, right and left hippocampal volume differences were expressed as a proportion of summed right and left hippocampal volumes, a strategy which the authors admit is not sensitive to bilateral atrophy. The authors also pointed out that only patients whose seizures were not adequately lateralized and localized with scalp EEC and other noninvasive measures were including in this study, suggesting that their measures of sensitivity and specificity may be low, as they were obtained through the study of their more challenging cases. Spencer (1994) noted that the varying MRI protocols and methods of temporal lobe and hippocampal measurement across epilepsy research centers makes it difficult to assess the sensitivity and specificity of qualitative MRI assessment with regard to its correlation with EEC localization of seizure foci. Upon reviewing the literature. Spencer determined that of 809 patients who underwent EEC localization, 337 showed temporal lobe abnormalities on MRI. Of these, qualitative MRI assessment of temporal lobe abnormalities was 55 percent sensitive and 78 percent specific in its agreement with EEC localization of seizure foci. Across studies that

PAGE 40

34 compared quantitative MRI measurements of hippocampal volume with EEG localization, MRI was shown to be 71 percent sensitive and specific in its correspondence with EEG. Spencer (1994) further noted that studies comparing abnormalities detected through qualitative MRI assessment to histological validation of pathology in resected tissue showed MRI to be 69 percent sensitive and 68 percent specific in the detection of temporal lobe abnormalities. Quantitative MRI measurements of hippocampal volume revealed 65 percent sensitivity and 80 percent specificity. Spencer cautions that her combination of the above studies did not take into account method of EEG localization, quality of MRI resolution or interpretation, or bias in patient selection. Hi ppocampal Volumes. Patholoev. and Neuropsychological Performance In 1990 Sass administered the verbal Selective Reminding Test pre-operatively to 35 patients diagnosed with intractable temporal lobe epilepsy, and compared scores to neuronal cell densities found in areas CAl, CA2, CA3, the hilar area, and the granule cell layer of the dentate gyrus (Sass et al., 1990). Histological examination of the resected tissue revealed significant neuronal loss in all examined areas, relative to autopsy controls. Cell densities among patients with left seizure foci did not differ from those with right foci, or between those whose EEG lateralization and localization required invasive procedures compared to those whose did not. This latter finding, along with reportedly excellent post-operative seizure control, suggests that noninvasive methods were just as effective as invasive ones in terms of identifying the damaged hippocampus. Both left and right temporal lobe epilepsy patients performed significantly below the norm for healthy individuals on the Verbal Selective Reminding Test, however patients with a left seizure focus performed significantly worse than those with a right hemisphere focus. Significant correlations were found between neuronal densities in the CAS and hilar areas, and

PAGE 41

35 long-term memory retrieval scores on the Verbal Selective Reminding Task, but only for patients with a left hemisphere seizure focus. In a study of 25 patients with medically intractable TLE and 14 right-handed control subjects, Lencz et al. (1990) investigated the relationships between MRI-based measurements, tissue pathology, and neuropsychological functioning. Results showed that, in control subjects, right and left hippocampal size was not significantly different, although the right temporal lobe was, on average, larger than the left. Among the patients, however, the hippocampus and temporal lobe ipsilateral to the seizure focus was smaller than in the contralateral hemisphere. Histological examination of resected hippocampal tissue revealed that neuronal densities were significantly and positively correlated with MRI-based hippocampal measurements for areas CAl, CA3, the Hilar area, and the granule cell layer of the dentate gyrus. Comparisons of temporal and hippocampal measurements and pre-operative performance on memory tests revealed significant relationships between left temporal lobe size and performance on the verbal Selective Reminding Task, particularly when left/right temporal lobe size ratios among patients with left hemisphere seizure foci were correlated with test scores. A significant correlation was also found when comparing hippocampal measurements of patients with left hemisphere seizure foci with percent retention scores from the Wechsler Memory Scale Logical Memory subtest. Right hemisphere hippocampal and temporal lobe measurements did not correlate significantly with performance on any of the verbal or nonverbal memory tests, indicating the need for more sensitive and specific neuropsychological tests of right hippocampal function. Trenerry et al. (1993) examined the correlation between preoperative hippocampal volumes and preand post-operative performance on memory tests among patients who had temporal lobe resection, including the hippocampus and amygdala, in the treatment of intractable seizures. Hippocampal volumes were expressed in proportion to total intracranial volume. The

PAGE 42

36 authors found that more severe pathology in the left hippocampus is associated with improved verbal and nonverbal memory performance subsequent to resection, and that resection of a relatively spared left hippocampus is associated with poorer memory performance in both domains. With regard to resection of the right hippocampus, this was associated only with a decline in visual learning, but not retention, when the resected hippocampus was relatively healthy. The authors concluded that right hippocampal volume, and the difference in right and left volumes, are more usefiil in predicting postoperative memory performance than is left hippocampal volume alone. Their results also suggested that a healthy left hippocampus may be responsible for aspects of both verbal and nonverbal learning, but that the right hippocampus may be more able to subsume the verbal memory responsibilities of a non-healthy left hippocampus, than a healthy left hippocampus can subsume the visual memory responsibilities of a non-healthy right hippocampus. On the whole, the investigators reported that the correlations between preoperative hippocampal volumes and pre-operative performance on memory tests were very small, suggesting that the tests they used (Wechsler Memory Scale-Revised, Rey Auditory Verbal Learning Test, and the Visual Spatial Learning Test) were not very sensitive to the memory fimctions specific to the hippocampal formation. Pre-operative hippocampal volumes were, as expected, smaller ipsilateral to the side of surgery. In 1995, Shigaki et al. examined the relationships between hippocampal volumes and neuropsychological performance among a well-defined population of 27 epilepsy patients who were candidates for temporal lobe resection. Hippocampal volumes were expressed relative to total intracranial volume so as to adjust for individual differences in overall brain volume. Results showed significant relationships between left hippocampal volume and measures of verbal memory, such as the CVLT. Percent retained at long delay, delayed recall intrusions; WMS-R: Delayed recall performance, and percent of immediately recalled material retained after

PAGE 43

37 a delay. Also, as described above, left hippocampal volume was shown to be significantly correlated with performance on the WCST, as patients with decreased volumes achieved fewer categories and made more perseverative responses and errors than did patients with decreased right hippocampal volumes. The authors identified the importance of analyzing hippocampal volumes, not with regard to relative differences in right and left values, but as ratios which take total intracranial volume into account. It was also noted that several neuropsychological measures were highly correlated with WAIS-R Verbal IQ or Performance IQ, and statistical methods were utilized to partial out the effects of intellectual fiinctioning in such cases. Several correlations which had appeared to be significant became nonsignificant after controlling for intellectual fiinctioning (Shigaki et al., 1995). These studies strongly suggest that MRI-based quantitative hippocampal volumes are an easily obtained, reliable, and usefiil diagnostic tool in the evaluation of hippocampal sclerosis. Further research will be necessary to determine precisely which parameters of hippocampal volume and asymmetry are the most sensitive and specific predictors of seizure lateralization and localization. MRI resolution is increasing dramatically with improved technology. This is expected to lead to more accurate and reliable measures of hippocampal volume. EEG Localization of Seizure Foci Although neuropsychological data and MRI studies reveal valuable information regarding possible seizure foci, the electroencephalogram (EEG) has long been the traditional foundation for localization (Risinger, 1991). EEG allows for both non-invasive and invasive detailed analysis of electrical neural activity. Initially, EEG consisted of recordings made fi-om electrodes attached to the scalp and was termed "surfece EEG." Much research and development

PAGE 44

38 led to the advent of invasive EEG, which is carried out by placing recording grids and strips directly on the brain surface, or stereotactically inserting depth electrodes into the brain. While both surface and invasive EEG have unique advantages and limitations, it is widely agreed upon that invasive EEG should be avoided whenever possible. Any invasive procedure, including that required to place subdural strips, grids, and depth electrodes, carries with it a risk of serious infection, discomfort, and added financial burden to the patient (Barry, Sussmann, O'Connor, & Hamer, 1992). Even when intracranial recording is a necessity, it is important to glean as much information as possible fi-om surface EEG, as it can provide information which cannot be obtained fi-om invasive procedures (Risinger, 1991). Information obtained through the use of surface electrodes has several limitations which are important to understand. First, it has been shown that electrical signals are attenuated or weakened as they travel the distance from discharging neurons to the scalp electrode. In addition, brain tissue, the meninges, skull, and scalp tissue serve to impede the electrical signal. The angle produced by the surface electrode's orientation to the active neurons may also affect its ability to accurately measure neuronal activity. All of these factors serve to reduce the amplitude of the actual neuronal discharge as measured by surface electrodes, and also causes these electrodes to be sensitive mainly to lower frequency discharges, as these are less likely to be attenuated by distance and impeding materials (Risinger, 1991). Other confounding factors which affect the utility of surface EEG include the existence of artifacts both from within the patient and in the external environment. Electrical activity associated with eye movements, the heartbeat, and musculature of the head and neck is often recorded by surface electrodes, as is activity produced by nearby electrical equipment. The development of improved electrodes has reduced the amount of such activity that is recorded, and additional progress in EEG technology has enabled physicians to filter out some degree of the

PAGE 45

39 artifact which is detected. Despite this, surface electrodes continue to record a sufBcient amount of artifact to render some EEG recordings unreadable (Ray, 1990). Intracranial recording obviously reduces the distance between neurons and recording electrodes, places less impeding material between the two, and is less sensitive to artifact from muscle activity, however, this technique also has its limitations. Because intracranial electrodes are placed in such close proximity to discharging neurons, it is necessary to reduce their sensitivity to obtain an accurate measurement of neural activity. This reduction in "gain" on the EEG machine may cause weaker electrical signals from neurons that are distant from the recording electrode to be undetected. Furthermore, unlike surface electrodes, only a limited number of intracranial electrodes may be used, as their placement requires the creation of a bunhole or skull flap (Risinger, 1991). The most commonly used array of surface electrodes is called the International 10-20 System, developed by Jasper (1958). This system is so named as electrodes are placed at 10 and 20 percent deviations from the following four cranial landmarks: a) the bridge of the nose or "nasion"; b) the bump on the back of the skull or "inion"; and c & d) the left and right preauricular areas, the small depressions above the cheekbones and anterior to the ears. Electrodes are labeled according to their locations and lateralization in the following manner: "F" = frontal, "P" = parietal, "C" = central, 'T" = temporal, and "O" = occipital. Sphenoidal electrodes, implanted in the cheek, are labeled "SP". Electrode lateralization is also coded, with odd numbers referring to the left hemisphere , even numbers corresponding to the right hemisphere, and "Z" indicating a midline placement. For example, electrode T3 is placed on the left temporal lobe, while "FZ" is placed on the midline of the frontal lobe (Ray, 1990). EEG recordings are actually the result of comparisons made between electrodes which are linked together and placed separately on the scalp. What emerges as one electrode's

PAGE 46

40 recording is that electrical signal which is not common to both placements. An electrical signal which is identical at two sites is referred to as "isoelectric." Deciding which comparisons to make can be difficult, as different strategies have their advantages and limitations. One common technique is to link a number of electrodes together across the scalp and average the activity which they record. That averaged electrical activity is then used as the reference to which subsequent recordings are compared (Ray, 1990). Using the International 10-20 System, (Jasper, 1958), Risinger (1991) has reported that the most reliable localizing information with regard to surface EEG is obtained through recording of actual seizure activity. EEC's of temporal lobe seizure onset typically begin with a desynchronization or flattening of background neural activity, with approximately half of all seizures subsequently revealing a pattern of rhythmic neuronal activity in the four to eight Hertz or theta frequency range. An electrode indicating an area of maximum amplitude in the theta frequency range within thirty seconds of seizure onset is highly suggestive of seizure focus. This information is thought to be most reliable when changes are noted on EEG prior to the patient's demonstrating any physical of "clinical" signs of seizure activity. When clinical signs precede EEG change, it is suspected that the seizure has spread from its place of origin, and that the electrode showing maximum amplitude may provide false localizing information. Other seizures, even those originating from a temporal lobe focus, may produce more ambiguous EEG patterns which provide little lateralizing or localizing information (Risinger, 1991). In order to predict the lateralization and location of seizure onset most reliably, Risinger (1991) has suggested several criteria that must be met. As stated above, the optimal situation is one in which EEG change precedes clinical seizure semiology. Also, lateralization and localization suggested by ictal and interictal abnormalities should be consistent, or at least not contradictory to one another. Finally, surface EEG should suggest lateralization and localization

PAGE 47

41 corroborative with that suggested by neuropsychological data and MRI. In instances in which these criteria are not met, or an extratemporal focus is suggested, invasive procedures are reconunended. Ictal and Interictal Surface EEG and Seizure Localization Given the importance of maximizing the information provided by surface EEG, many investigators have sought to estimate the sensitivity and specificity of this focus-lateralizing strategy, as well as to improve the existing methods of doing so. In order to obtain a reference point against which to compare surface EEG's ability to localize seizure foci, some researchers have used invasive recording as the "gold standard" with which to compare surface EEG (e.g. Lieb, Walsh, Babb, Walter, & Crandall, 1976; Spencer et al., 1985; Risinger, 1989), while others have conducted retrospective studies of patients who underwent successful ATL and determined whether their seizure foci were correctly localized by surface EEG (e.g. Barry et al., 1992). A major comparison between surface and depth EEG was conducted by Lieb (Lieb et al., 1976), who studied 34 patients with medically intractable seizures who underwent simultaneous surface and depth recordings as potential candidates for resective surgery. Surface electrodes were placed according to the International 10-20 system (Jasper, 1958), while depth electrodes were placed in the anterior, mid, and posterior portions of the hippocampal gyrus, in the anterior, mid, and posterior regions of the pes hippocampus, and occasionally in the amygdala and thalamus. Only seizures which originated unilaterally in the depths of one temporal lobe were considered. Bilateral surface and depth electrodes were recording during all events included in the study. It is important to note that the number of potential recording sites was greater than the number of available channels, and the areas of greatest electrical activity were sampled more heavily than others. Therefore, surface montages were incomplete during most events.

PAGE 48

42 Lieb found that, of 161 total seizures originating unilaterally in the depth of the temporal lobe, 68.2 percent did not register on surface electrodes, and that in cases in which surface electrodes did pick up activity, it was bilateral and synchronous in 16.9 percent of events. In only 14.8 percent of the studied seizures was the event lateralized by ipsilateral surface electrodes, with no incidents of incorrect lateralization. With regard to seizures with varying concomitants, 86 percent of seizures accompanied by clinical behavior change produced alterations in surface electrodes, while 18 percent of those accompanied only by auras produced were registered on surface EEC. Only 10 percent of subclinical seizures produced changes on surface EEG (Lieb et al., 1976). Lieb suggested that depth electrodes are significantly more useful than surface electrodes in localizing the onset of temporal lobe seizures and in estimating the time of seizure onset. The authors surmised that this is at least partially due to the fact that higher electrical frequencies are filtered out by the biological materials between neurons and surface electrodes, and that mostly incomplete surface montages were used in the study. It was also noted that surface electrodes were able to be used for seizure lateralization in some cases when several events were considered together (Lieb et al., 1976). In a similar investigation using non-simultaneous depth and surface EEG, Spencer and co-investigators (Spencer et al., 1985) compared the surface and depth electrode recordings of 54 patients with complex partial seizures of temporal and frontal lobe origin. Three experienced electroencephalographers independently attempted to lateralize and localize seizure onset using at least three ictal surfece EEC's from each patient. A team of four physicians, working together, attempted to do the same using at least three ictal depth recordings. Of the 54 patients, 35 had seizure foci that were able to be localized using depth recordings. Twenty-seven of these patients showed temporal lobe onset, seven had frontal

PAGE 49

43 seizure foci, and one patient had a parietal focus. With regard to the reliability of surface EEG, the independent raters showed 46 to 59 percent agreement with one another with regard to seizure lateralization and 28 to 35 percent concordance with one another concerning localization to a specific lobe. These values were corrected for chance. Although most disagreements on lateralization were the result of one or two raters being imable to lateralize a given seizure, raters lateralized seizures to opposite hemispheres two to four percent of the time. With regard to the concordance of surface EEG with depth recordings, levels of agreement on lateralization ranged from 46 to 49 percent, while that for localization to specific lobe ranged from 21 to 38 percent (Spencer et al., 1985). Despite these rather low values, when the investigators analyzed data concerning only patients whose seizures had been depth localized to the temporal lobe, reliability estimates increased to 66 to 75 percent, while accuracy values increased to 57 to 60 percent. It is important to note, however, that surfece EEG resulted in incorrect lateralization in 3 to 17 percent of these cases. Patients with frontal seizures were most difficult to assess (Spencer et al., 1985). Important implications of this study include the authors' assertion that interpretation of surface EEG is very difficult even for experienced electroencephalographers. The findings that individuals rating surface EEG agreed more with one another than with depth localization, and that at least one third of diagnosed temporal lobe epilepsy patients were unlateralizeable using surface EEG were particulariy troubling. It should also be noted, however, that seizure onset could not be determined by depth recording in all cases, and that the population studied in this investigation was inherently difficult to lateralize and localize due to the fact that they all showed sufficient surface EEG ambiguity as to require depth recordings. The authors surmised that development of a set of specific and consistent interpretation criteria may increase the accuracy and reliability of surface EEG (Spencer et al., 1985).

PAGE 50

44 In an effort to create and evaluate such criteria, Risinger and co-investigators studied 1 10 patients with medically intractable complex partial seizures. All of these patients were suspected to have temporal lobe onset on the basis of seizure history and semiology, and all underwent surface and depth EEG. Risinger's hypothesis was that, at ictal onset, scalp electrodes would agree with subsequent depth recordings with regard to seizure lateralization and localization (Risinger, Engel, Van Ness, Henry, & Crandall, 1989). Surfece recording was done using the International 10-20 System (Jasper, 1958), with additional bilateral sphenoidal electrodes. Based on clinical experience, the authors developed the following set of EEG lateralization and localization criteria: a) A seizure showing a unilateral temporal or sphenoidal rhythmic discharge of five Hz or greater within 30 seconds of EEG ictal onset was defined as being "focal"; b) Ictal recordings that lacked a clear temporal or sphenoidal rhythm of five Hz or greater were defined as "nonlocalizing", even when some other lateralized or focal EEG change was apparent. The authors also sought to evaluate variations in timing of seizure onset, and developed the following criteria: a) Focal EEG recordings which showed a rhythmic pattern of five Hz or greater as the first EEG change (other than suppression of background activity), or those which began as slower temporal/sphenoidal fi-equencies but reached five Hz or greater within 30 seconds of EEG onset were collectively referred to as having "Initial Focal Onset"; b) A focal pattern of five Hz or greater which appeared after some other type of lateralized or diffiise electrical discharge was defined as having "Delayed Focal Onset." Patients whose ictal recordings showed evidence of both focal and nonlocalizing features were considered to have focal seizures, and those whose recordings showed features of both initial and delayed onset were considered to have initial onset. Overall, an initial or delayed focal ictal rhythmic discharge maximal at one temporal or sphenoidal electrode was used to localize onset and predict depth localization (Risinger et al., 1989).

PAGE 51

45 Two independent raters analyzed 706 ictal recordings using the above system. Ictal recordings made with depth electrodes were analyzed more generally and were defmed as lateralized and localized to one or the other temporal lobe, extratemporal regions, or were determined to be nonlocalizing. Of the 1 10 patients, 57 were determined to have focal seizure onsets using surface EEC Of these 47 (82 percent) were correctly lateralized and localized when compared to their depth recordings. Patients with uniformly focal seizures were correctly localized 94 percent of the time, while those whose EEC's reflected a combination of focal and nonfocal seizures were correctly localized 67 percent of the time. Two patients were incorrectly lateralized, three were nonlocalizeable, and five showed evidence of extratemporal onset. Raters had' 'major disagreements" on five to 10 percent of selected recordings, however there were no disagreements regarding lateralization in instances in which localization was agreed upon. Of the remaining 53 patients, 34 were localized to one temporal lobe with depth recordings, and the remaining 19 patients were unable to be lateralized or localized (Risinger et al., 1989). The authors concluded that distinguishing between initial and delayed seizure onset did not increase accuracy of seizure localization. However, they were able to show that experienced electroencephalographers could in fact predict seizure lateralization and localization reliably and accurately using a consistent set of interpretation criteria, particularly in patients whose seizures consistently met criteria for having focal onset. It was also noted that false localization occurred more commonly than felse lateralization, and that complex partial seizures originating in extratemporal regions may show EEC characteristics which make them appear to have temporal lobe onset (Risinger et al., 1989). The use of depth electrode localization as a "gold standard" against which to measure the accuracy and reliability of surface EEC was criticized by Walczak (Walczak, Radtke, & Lewis, 1992), who pointed out that depth recording is typically limited to a two week period, during

PAGE 52

46 which a given patient may not experience all of his or her seizure types. Furthermore, it has been estimated that up to half of patients who undergo cortical resection following depth recordings have poor surgical outcome (Walczak et al., 1992). Instead of depth recording, Walczak considered freedom from seizures following resection of a given area to be the best indicator that the resected tissue was the seizure focus. An early study which used both surface and depth electrodes for seizure localization was conducted by Lieb, Engel, Gevins, and Crandall (1981), who studied 52 patients who underwent surface and depth EEG prior to having ATL for medically intractable seizures. Surface recordings were made using the hitemational 10-20 System (Jasper, 1958), while depth electrodes were placed bilaterally in the medical temporal region including the anterior, mid, and posterior hippocampal gyrus, anterior, mid, and posterior pes hippocampus, the amygdala, and occasionally in the uncus, thalamus, and supplementary motor cortex. As with Lieb's eariier study described above (Lieb et al., 1976), recording montages favored depth sites more heavily than surface sites. Surgical outcome was categorized into four groups, ranging from total seizure relief to no worthwhile benefit from surgery. Two raters assessed between two and five interictal recordings and between one and six from each patient. A total of 455 seizures were assessed. With regard to surgery outcome, 73 percent of the patients fell into one of the three "improved" groups, meaning that they were either seizure free, had less than one seizure per year, or less than one seizure per month (Lieb et al., 1981). Results showed that several interictal variables were associated with poor surgical outcome, including the presence of bilaterally synchronous surface spikes, sharp waves, and diffuse background slowing. Those associated with surgical success included the presence of multiple independent deep spike patterns. With regard to ictal recording, a high proportion of bilaterally synchronous onsets (surface, deep, or both) and the existence of variable onset

PAGE 53

47 locations were associated with poor outcome. A high proportion of seizure onsets (surface, deep, or both) from the hemisphere later chosen for resective surgery, and a high proportion of deep focal onsets were associated with surgical success (Lieb et al., 1981). Although this study did not directly compare the utility of surface and depth electrodes in the localization of seizure foci, it was shown that these two methods yield unique information with regard to surgical outcome. Using surface and depth recording together yielded significantly more accurate prognoses than using either one in isolation (Lieb et al., 198 1). Subsequent studies evaluated the accuracy and reliability of these two techniques in greater detail. Also employing the strategy of using surgical outcome as an indicator of the accuracy of preoperative surface EEG localization, Walczak and co-investigators attempted to determine the accuracy and reliability of surface EEG with regard to its ability to lateralize seizure foci and to determine whether some EEG features are more lateralizing than others (Walczak et al., 1992). Thirty-five patients were included in the study, all of whom had undergone ATL for the treatment of medically intractable complex partial seizures, and all of whom had been seizure free for a period of at least two years after surgery and prior to the study. Three independent raters evaluated 137 total seizures with regard to activity at onset, the presence or absence of rhythmic theta/alpha activity, topography of theta/alpha activity, and post-ictal findings. In order for rhythmic theta/alpha activity to be considered present, this activity must have occurred for at least ten seconds and have begun within forty seconds of EEG seizure onset. Overall, surface EGG was determined to be 76 to 83 percent accurate with regard to lateralization of temporal lobe seizures, and only 47 to 65 percent accurate with regard to extratemporal lateralization. When seizures that were uninterpretable due to artifact were excluded, these ranges of accuracy increased to 92 to 99 percent, and 73 to 100 percent, respectively. With regard to reliability, two raters agreed on lateralization of temporal lobe

PAGE 54

48 seizures 79 to 8 1 percent of the time, and 47 to 65 percent of the time on lateralization of extratemporal seizures. Seizure foci were incorrectly lateralized in 0 to 6 percent of temporal cases, and not at all in extratemporal cases (Walczak et al., 1992). Analyses of specific EEG features revealed that activity at onset was indicative of onset 33 to 59 percent of the time in temporal lobe seizures, however, these values increased to 92 to 97 percent when uninterpretable or generalized events were excluded. The presence of rhythmic theta/alpha activity led to lateralization accuracy of 64 to 76 percent, which increased to 97 to 99 percent when uninterpretable or generalized events were excluded. Post-ictal features were slightly less accurate in predicting lateralization (Walczak et al., 1992). Important implications of this study include the finding that surface EEG is able to provide lateralizing information with moderately good accuracy and reliability with regard to temporal lobe seizures, particularly when uninterpretable events are ignored. The presence of rhythmic alpha/theta activity was the best predictor of seizure lateralization when all events were considered, however onset activity and post-ictal features were also usefiil when assessing only interpretable events. Lateralization errors were found to be rare, particularly when more than one of the above features were used in lateralization decisions. It was strongly shown that rhythmic theta/alpha activity is highly correlated with temporal lobe localization, and that extratemporal seizures are quite difficult to assess with surface EEG. Similar to Risinger et al. (1989), Walczak was able to improve upon the accuracy/reliability findings of Spencer (Spencer et al., 1985) by employing a consistent set of EEG interpretation criteria (Walczak et al., 1992). Studying 48 patients who underwent ATL for the treatment of medically intractable seizures, Barry examined pre-operative ictal and interictal scalp EEG's in an attempt to reveal electroencephalographic indicators of surgery outcome. Each patient had undergone at least two 30 minute periods of interictal scalp recording, using the International 10-20 System of electrode

PAGE 55

49 placement (Jasper, 1958). Patients also underwent extended EEG monitoring and a tapering of medication levels, in an attempt to obtain recordings of ictal onset. Forty-five of the patients had seizures during this recording, and nineteen of them ultimately received depth electrode placement, due to bilateral or indeterminate seizure activity (Barry et al., 1992). Two independent raters examined the interictal recordings, and using the site of maximum voltage as a location indicator, coded interictal activity as temporal or extratemporal. Patients with 95 percent or more of their interictal spikes occurring in one hemisphere were defined as "unilateral", while the others were defined as "bilateral." Ictal onset was coded in a similar manner, with 80 percent being the cut-off between unilateral and bilateral/indeterminate seizure onset. In making both interictal and ictal localization decisions, electrodes NP1&2, F7&8, and T3-6 were defined as "temporal" while all others were defined as "extratemporal" (Barry et al., 1992). Results showed that, of the 48 patients, 73 percent experienced fewer than three seizures per year following surgery, while 65 percent became seizure fi-ee. Twenty-seven percent continued to have fi-equent seizures. All 12 patients who had shown unilateral temporal lobe interictal spikes prior to surgery were significantly improved following surgery, and 1 1 of these patients had temporal lobe resection ipsilateral to the location of their interictal spikes. Seventysix percent of those who had shown bilateral temporal lobe interictal spikes were significantly improved. Only 33 percent of patients who had shown unilateral or bilateral extratemporal interictal discharges were significantly improved postoperatively (Barry et al., 1992). With regard to recording of ictal onset, patients with evidence of unilateral temporal lobe onset fared slightly better postoperatively than did those with evidence of bilateral or indeterminate onset, but this difference did not reach statistical significance. Approximately 70 percent of the patients in these groups improved follovwng surgery. Patients who required depth

PAGE 56

50 electrode placement exhibited slightly less seizure control following surgery than did those patients who did not, however this is not surprising as the necessity of depth recordings suggests a certain degree of ambiguity in focus localization (Barry et al., 1992). Barry concluded that of all the EEC parameters examined, interictal spikes were the most accurate indicator of seizure focus localization, and that even patients with bilateral interictal spiking would benefit fi"om surgery providing that these spikes were confined to the temporal lobe and localization of onset was confirmed with depth recording. Barry strongly claimed that patients demonstrating unilateral temporal lobe interictal discharges could be adequately assessed without further monitoring, and that these patients should be considered as surgery candidates providing that there is no contradictory data from MRI or physical examination (Barry et al., 1992). Taken together, the above studies suggest that surface EEG is a valuable diagnostic tool in the localization of seizure foci, independent of the additional benefit that it is a noninvasive procedure. The research also suggests that surfece EEG is most accurate and reliable when specific interpretive criteria are developed regarding seizure localization (Risinger et al., 1989; Walczak et al., 1992), and when EEG findings are consistent with those obtained through neuropsychological testing and MRI (Pilcher et al., 1992). Neuropsychological Testing. MRI. and EEG Finally, there has been at least one study which involved examination of the independent contributions of EEG, MRI, and neuropsychological data in the prediction of seizure focus. Williamson et al. (1993) retrospectively studied pre-surgical surface EEG's, MRI's, and neuropsychological data from 67 patients who had undergone successfiil ATL for the treatment of

PAGE 57

51 intractable seizures. Thirty-seven of these patients underwent left temporal lobe resection, while the remaining 30 had right temporal lobe resection. EEG results showed that interictal activity could be used to correctly lateralize seizure focus in 33 out of 35 patients who exhibited only unilateral temporal lobe interictal activity. In cases of bilateral temporal lobe interictal activity, the side exhibiting the predominance of activity corresponded with side of seizure focus in 2 1 out of 28 cases. With regard to ictal activity, 54 patients showed EEG evidence of 5-10 Hz rhythmic buildup within 30 seconds of seizure origin. This provided correct lateralization of seizure focus in 47 of these 54 cases. MRI was available for only 28 patients, whose scans were analyzed qualitatively. Twenty-three of these were deemed to have unilateral hippocampal atrophy, a finding which was confirmed by the existence of pathology in resected tissue in 21 of these cases, and disconfirmed in two. Perhaps the most striking findings in the study were with regard to neuropsychological data. The authors used only the lateralization findings provided by the neuropsychological evaluations, and analyzed how well this corresponded to the final decision regarding the lateralization and localization of each patient's seizure focus. Fifty-eight patients had neuropsychological evidence of focus lateralization, however nine of these cases were found to be mislateralized. Interestingly, all nine of the patient's had right temporal lobe seizure foci. Of the 49 patients who were correctly lateralized, 35 were able to be specifically localized to the temporal lobe. This study again suggests that EEG, MRI, and neuropsychological data are all usefiil in the prediction of seizure focus, but that each of these techniques has limitations. It is unfortunate that MRI and neuropsychological data were analyzed in only a qualitative manner, and that the three assessment domains were not analyzed with regard to how they correlate with

PAGE 58

52 one another. Correlations among these domains will be a major focus of the current study, which will seek to identify MRI, EEG, and demographic/illness-related sources of variation in the presurgical neuropsychological performance of patients who subsequently undergo anterior temporal lobe resection. Epilepsy Surgery at the University of Florida Shands Teaching Hospital at the University of Florida has a Comprehensive Epilepsy Program specifically designed to treat the many patients who present with seizure disorders. Those whose seizures cannot be adequately controlled through the use of standard or experimental medications are evaluated as candidates for resective surgery. The comprehensive evaluation of a patient's suitability for surgery is divided into two phases, and involves the coordination of several disciplines within the hospital. Phase I of this evaluation consists solely of non-invasive procedures, and includes EEG, MRI, and neuropsychological testing, and, where necessary, additional procedures including Single Photon Emission Computed Tomography (SPECT), Positron Emission Tomography (PET), and Magnetoencephalography (MEG). Ictal and interictal EEG's are obtained by admitting the patient to the Epilepsy Monitoring Unit (EMU), where they undergo Video/EEG telemetry. This EEG is conducted with the use of scalp and sphenoidal electrodes. Each patient is asked to have an accompanying adult stay with them to notify the EEG technicians in the control room when an aura or seizure occurs if the patient is unable to do so. The EEG recorders also have seizure detection capabilities. If necessary, medications are tapered until the patient has had a sufiBcient number of interpretable events to obtain the majtimal amount of information from EEG.

PAGE 59

During this hospitalization, each patient undergoes a comprehensive neuropsychological evaluation at bedside. This evaluation involves assessment of intellectual fiinctioning, verbal and nonverbal memory, language, visuospatial/visuoconstructive skills, frontal and executive functions, motor speed and dexterity, and personality. In addition, a thorough interview is conducted to obtain information regarding psychosocial, occupational, and medical history. Each patient also undergoes MRI. Both Tl and T2 weighted images are obtained and used in the evaluation. Tl scans are analyzed visually to detect any tissue abnormalities and to assess the relative sizes of the two hippocampi, and quantitative hippocampal volumes are also calculated. Hippocampal T2 relaxation times are calculated from the T2 scan. Following this evaluation, all of the results from the above disciplines are presented and discussed at a weekly Epilepsy Management Conference (EMC). Based on the informational value of the EEG and convergence or divergence of these data with neuropsychological testing and MRI, decisions are made with regard to the patient's suitability for surgery. If questions remain following Phase I, Phase II evaluation is necessary. During Phase II, the patient is hospitalized again, and subdural recording strips, grids, or depth electrodes, alone or in combination, are surgically placed on the surface of the brain corresponding to the area of suspected seizure generation. As in Phase I, these patients undergo video/EEG telemetry until they have had a sufficient number of interpretable events. Phase II results are then discussed at the EMC, at which time final treatment decisions are made. Regardless of whether a patient's seizure disorder is adequately characterized during Phase I, or whether Phase II is required, all patients undergo a Wada procedure (Wada & Rasmussen, 1960) prior to the fmal treatment decision. The Wada procedure involves the injection of sodium Brevital into the femoral arteries, one at a time, thereby temporarily anesthetizing each hemisphere via the internal carotid artery while the other hemisphere is able to

PAGE 60

54 function relatively normally. During this anesthetization, verbal and visual memory testing is performed, as is language and motor assessment. In some cases, Brevital is introduced into the cerebral vasculature via the posterior cerebral artery. This delivers the anesthetic primarily to the posterior two-thirds of the hippocampus while sparing the majority of temporal cortex, thereby allowing the patient's healthcare team to test hippocampal function more specifically than is possible during the aforementioned Wada technique. The goals of this procedure are to assess hemispheric dominance with regard to language, and to assess each hemisphere's ability to support memory and language function in isolation. Purpose of the Present Study As mentioned above, information fi-om noninvasive EEG, MRI, and neuropsychological testing are discussed and compared following Phase I evaluation. Given the medical risk and financial burden associated with invasive Phase II procedures, it is of utmost importance to obtain as much diagnostic information as possible from Phase I data. This requires not only a thorough understanding of the data from each diagnostic discipline, but also of the relative sensitivity and specificity of the data for localizing the seizure focus. In the ideal situation, data from EEG, MRI, and neuropsychological testing are all convergent, and suggest a clearly defined epileptogenic focus, removal of which will significantly reduce or eliminate the patient's frequency without leading to significant cognitive impairment. Unfortunately, this is not always the case. In many cases, members of the epilepsy team must deal with conflicting data and must decide which data are most relevant, and whether one of the three types of data appears to be more informative than the others in a particular case. Phase II evaluation is usually conducted in these instances.

PAGE 61

55 Neuropsychological data, in particular, often provide less specific indications of seizure lateralization and localization than EEG and MRI. These data often suggest lateralization and localization that is consistent with that obtained with EEG and MRI, but are also indicative of involvement in other brain regions in both the ipsilateral and contralateral hemispheres. The sources of this relative lack of specificity remain unclear. The current study will attempt to identify electroencephalographic, structural-anatomic, and demographic/ilhiess-related sources of variance in neuropsychological performance.

PAGE 62

CHAPTER 2 METHODS Subjects Subjects were 61 patients with medically intractable unilateral temporal lobe epilepsy who participated in a comprehensive epilepsy evaluation and subsequently underwent right or left anterior temporal lobectomy (ATL). Thirty-five subjects underwent left ATL and 26 underwent right ATL. All subjects were right-handed, at least 16 years of age, and underwent neuropsychological assessment and Phase I EEGA^ideo monitoring. All subjects except for two in the right ATL group and one in the left ATL group underwent MRI from which quantitative hippocampal volumes were obtained. Subjects with a history of neurological insult which occurred after the onset of epilepsy or was deemed to be unrelated to their epilepsy were excluded from the study. Measures Demographic and Illness-Related Variables The following demographic and illness-related variables were collected: Age (in years) at Neuropsychological Assessment, Gender, Education (in years). Seizure Type(s), Age (in years) at Onset of Epilepsy, Age (in years) at Onset of Secondarily Generalized Seizures (if different from Age at Onset), Duration (in years) of Illness from Initial Onset of Epilepsy, and whether or not the subject has a history of febrile seizures and/or secondarily generalized seizures. 36

PAGE 63

57 Neuropsychological Variables The following neuropsychological variables were collected and/or calculated from collected values: Intellectual functioning . Wechsler Adult Intelligence Scale-Revised (WAIS-R): Full Scale IQ (FSIQ), Verbal IQ (VIQ), Performance IQ, (PIQ). Verbal memory . Wechsler Memory Scale-Revised (WMS-R): Logical Memory I (LMI), Logical Memory II (LMII); California Verbal Learning Test (CVLT): Short Delay Free Recall (SDFR), Short Delay Cued Recall (SDCR), Long Delay Free Recall (LDFR), Long Delay Cued Recall (LDCR), Free Recall Intrusions (PRINT), Cued Recall Intrusions (CRINT). Using zscores, an overall Verbal Memory Score (VMEMSCOR) was calculated using the following equation: VMEMSCOR = [(LDFR + LDCR + LMII) (PRINT + CRINT)] / 5. Nonverbal memory . Wechsler Memory Scale-Revised (WMS-R): Visual Reproduction I (VRI), Visual Reproduction II (VRII); Rey-Osterrieth Complex Figure (ROCF): Delayed Recall (RODELAY). Using z-scores, an overall Nonverbal Memory Score (NVMEMSCOR) was calculated using the following equation: NVMEMSCOR = (RODELAY + VRII) / 2. Language . Boston Naming Test (BNT); Controlled Oral Word Association (COWA); WAIS-R: VIQ. Using z-scores, an overall Language Score (LANGSCOR) was calculated using the following equation: LANGSCOR (VIQ + COWA + BNT) / 3. Visuoconstructive . Rey-Osterrieth Complex Figure (ROCF): Copy (ROCOPY): WAISR: Block Design (BD), Object Assembly (OA). Using z-scores, an overall Visuoconstructive Score (VISCONSCOR) was calculated using the following equation: VISCONSCOR = (ROCOPY + BD + OA) / 3.

PAGE 64

58 Motor . Finger Tapping Test (FTT) : (Dominant and non-dominant hands); Grooved Pegboard Test (GPT): (Dominant and non-dominant hands). Using z-scores, an overall Motor Score (MOTORSCOR) was calculated using the following equation: (Right FTT + Right GPT) (Left FTT + Left GPT). Missing Neuropsychological Data Following neuropsychological testing on the EMU, it is not uncommon for a given subject's neuropsychological profile to be missing one or more variables. Data points may go uncollected for a number of reasons including seizure episodes that occur during testing, patient fatigue and ilhiess, and time constraints. As expected, this was found to be the case upon examination of subject neuropsychological profiles in the current study. Missing data was replaced using the method described below. Using the neuropsychological variables used to generate the aforementioned neuropsychological domains, along with additional neuropsychological variables collected during Phase I, correlations were calculated among all neuropsychological variables for all subjects. These variables included the following: WAIS-R: FSIQ, VIQ, PIQ; CVLT: Trial I, Trial V, List B, SDFR, SDCR, LDFR, LDCR, Total Recall Trials I-V, Best Recall Trial, Total Intrusions, Semantic Ratio, Serial Ratio, Perseverations, Recognition Hits, Recognition False Alarms; WMSR: LMI, LMII, VRI, VRII; BNT; COWA; ROCF: Copy, hnmediate Recall, Delayed Recall; Finger Tapping Test; Grooved Pegboard Test; Trail Making Test: Parts A and B; Wisconsin Card Sorting Test (WCST): Categories, Perseverative Responses, Perseverative Errors. For each neuropsychological variable used in the study that had at least one missing value, correlations between that variable and all others were examined. The five variables with which the initial variable shared the highest correlations were then selected and entered into a

PAGE 65

59 multiple regression equation to generate a predicted score to replace the missing value. For example, for a subject missing a BNT score, one would be predicted for him or her using his or her FSIQ, VIQ, COWA, CVLT LDFR, and CVLT SDFR scores, because these were the variables which were shown to correlate most strongly with BNT performance. In cases in which not all of the predictor variables were available, the existing ones were used in the multiple regression equation to predict the missing value. Information contained in the table in the Appendix includes a list of each set of predictors used to generate replacements for missing variables. MRI Variables The following MRI variables were collected: Left Hippocampal Volume (LHIPPVOL); Right Hippocampal Volume (RHIPPVOL); Difference in Volume of Hippocampal Formations (DHF). EEG Variables For each subject the number of seizures during Phase I with the following lateralization and localization were obtained: Number of seizures with left mesial temporal lobe localization, number of seizures with left global temporal lobe localization, number of seizures with left fi-onto-temporal lobe localization, number of seizures with left hemisphere lateralization but not specifically localized to the temporal lobe, number of seizures with right mesial temporal lobe localization, number of seizures with right global temporal lobe localization, number of seizures with right fronto-temporal localization, number of seizures with right hemisphere lateralization but not specifically localized to the temporal lobe. Also collected was each subject's number of nonlateralized seizures and the number of seizures which were not lateralized but believed to be localized to one or both temporal lobes.

PAGE 66

60 Each seizure was also coded according to whether or not it secondarily generalized. Seizures which were uninterpretable due to artifact were recorded only for use in descriptive statistics, and were not used in statistical methods which had a bearing on hypotheses. Following the above data collection, lateralization and localization categories were "collapsed" in order to calculate the following values for each subject: Number of seizures believed to have left hemisphere lateralization, number of seizures believed to have right hemisphere lateralization, number of nonlateralized seizures. A Seizure Lateralization Index (SLI), a single value which was indicative of a given subject's distribution of left hemisphere, right hemisphere, and nonlateralized seizures was then calculated for each subject using the following equation where R = number of seizures originating in the right hemisphere, L = number of seizures originating in the left hemisphere and N = number of nonlateralized seizures: SLI = [(R L) .5N (R L / |R L|)] / R + L + N This equation yielded SLI values ranging from 1 .00 to -1 .00 for all subjects. A SLI value of 1.00 indicated that all of the subject's seizures were lateralized to the right hemisphere, a SLI value of -1 .00 indicated that all of the subject's seizures were lateralized to the left, while values closer to 0.00 indicated varying degrees of right or left tendency with regard to lateralization of seizures. The SLI equation had three major components which will be discussed below. First, (R L) represented the difference in number of seizures lateralized to the right versus left hemisphere for a given patient. This yielded a very basic indicator of the likelihood that a given patient's seizure focus was in the right versus left hemisphere, but did not take into account the occurrence of nonlateralized seizures or the total number of seizures that was recorded. To address this first issue, [.5N (R L / |R L|)] was subtracted from the original (R L) remainder described above. In the first part of this component of the equation, nonlateralized

PAGE 67

61 seizures were assigned a weight of .5. This weight was chosen because nonlateralized seizures are an important part of the clinical data but are not considered as clinically significant as are seizures with a clear lateralization. In clinical practice for example, a patient may have ten seizures which are all lateralized to the right hemisphere. If the patient then has a seizure which is clearly lateralized to the left hemisphere, this is considered to be a significantly greater contradiction to the previous ten events than would be the case if the patient's eleventh seizure were interpreted as nonlateralized. Nonetheless, the occurrence of the nonlateralized seizure would serve to slightly diminish the epileptologist's confidence that the patient's seizures are truly lateralized to the right hemisphere. The second part of this component of the equation involved multiplying the .5N by (R L / |R L|). This served only to correct the mathematical sign assigned to the .5N. If more of the patient's seizures were lateralized to the right (i.e. if R L yielded a positive number), this part of the equation ensured that .5N would also be assigned a positive value which would then be subtracted fi-om R L. In the event that more of the patient's seizures were lateralized to the left hemisphere (i.e. if R L yielded a negative number), this part of the equation ensured that .5N would be assigned a negative value which would then be subtracted fi-om R L. In this case, subtraction of a negative is the equivalent of adding a positive .5N to R L. Simply stated, whether R L was a positive number indicating that most of the patient's seizures were lateralized to the right, or whether R L was a negative number indicating that most of the patient's seizures were lateralized to the left, the SLI equation was designed in such a way that the presence of nonlateralized seizures served to bring the value of the R-L remainder back toward zero. Finally, the total number of seizure events was taken into account by dividing the SLI numerator [(R L) .5N (R L / |R L|)] by the total number of seizures (R + L + N). This

PAGE 68

served to put the discrepancy between the number of seizures lateralized to each hemisphere in the context of total number of seizures that occurred. The importance of this can be appreciated by considering the following example. Patient A may have had seven seizures lateralized to the right hemisphere, three lateralized to the left hemisphere, and no nonlateralized seizures. If Patient A's SLI were calculated without dividing by the total number of seizures, it would yield a SLI value of four. Patient B may have had four seizures lateralized to the right hemisphere, none lateralized to the left hemisphere and no nonlateralized seizures. Calculating Patient B's SLI value without dividing by the total number of seizures would also yield a SLI value of four. The two cases, however, are very different from a clinical standpoint. Specifically, Patient A had some seizures lateralized to the right hemisphere and some lateralized to the left, while all of Patient B's seizures were lateralized to the same hemisphere. When division by the total number of seizures was included in the calculation of SLI, Patient A obtained a SLI value of .4, and Patient B obtained a SLI value of 1.0. Thus, it is clear that Patient B's SLI rating is more indicative of right hemisphere seizure lateralization than is Patient A's. Data Collection Procedures Demographic and Illness-Related Variables The demographic and illness-related variables described above were collected from subject files on the EMU and in the Neuropsychology Laboratory.

PAGE 69

63 Neuropsychological Data These data were gathered from patient presurgical evaluation files in the Psychology Clinic, were entered into a database and were verified following entry. Following is a brief description of each of the measures that was included in the experimental phase of the study: Wechsler Adult Intelligence Scale-Revised (WAIS-R) . The WAIS-R (Wechsler, 1981) is a general measure of intellectual fianctioning. It consists of a Verbal Scale, which has six subtests, and a Performance Scale, which includes five subtests. The WAIS-R yields a Full Scale IQ, which is a measure of general intellectual functioning, and Verbal and Performance IQ's, which more specifically reflect subject performance on the two scales described above. These IQ scores have a mean of 100 (SD =15), and allow for comparison of an individual's performance to that of his or her age peers. Wechsler Memory Scale-Revised (WMS-R) . The WMS-R (Wechsler, 1987) is a battery of verbal and visual memory tests, two of which were included in the current study. The Logical Memory subtest of this battery consists of immediate and delayed recall of two short stories which are read to the subject. The Visual Reproduction subtest consists of immediate and delayed reproduction of four geometric figures which the subject is shown for ten seconds each. California Verbal Learning Test fCVLT) . The CVLT (Delis et al., 1987) is a test of verbal learning and memory. A list of 16 shopping items, consisting of four items from each of four categories, is read to the subject five times. After each reading, he or she is required to recite back as many of the items as possible. The subject is then presented with an interference list, which is read only one time, after which free recall is required. Subject recall of the fu^ list is then assessed after short and long delays, as is ability to recall the list when given categorical

PAGE 70

64 cues. Finally, recognition memory for list items is assessed. Many scores, expressed in standard deviation units, are obtained through computer scoring of this test. Rev-Osterrieth Complex Figure (ROCF) . The ROCF (Rey, 1941; Osterrieth, 1944) assesses subject ability to copy a complex geometrical design, and reproduce it immediately and after a delay, without having been asked to remember it. ROCF scaled scores (M = 10; SD = 3) used in the current study were derived from the Denman (1984) scoring system and norms, which score each reproduction of the figure based on the presence versus absence of specific details, the accuracy of detail reproduction, and the location of these details within the figure. These Boston Naming Test (BNT) . The BNT (Kaplan, Goodglass, & Weintraub, 1983) is a test of confrontational naming, in which the subject is presented with black and white line drawings of objects that they are then asked to name. Items are arranged in order of difficulty, and the test progresses until either the last item has been completed or the subject gives six consecutive incorrect responses. Semantic and phonemic cues are provided according to specific criteria, but only spontaneous responses and those following semantic cues are eligible to be counted as correct. The total number of correct responses is then transformed into a standard deviation score using norms provided with the test. Multilingual Aphasia Examination (MAE) . The MAE (Benton & Hamsher, 1989) is a battery of tests designed to assess language functioning. The Controlled Oral Word Association (COWA) subtest from this battery was used in the current study. COWA involves presenting the subject with a letter and asking him or her to generate as many words as possible that begin with that letter. After 60 seconds, there is a brief pause, and the examiner gives the subject a new letter with the same instructions. Subjects are asked not to use proper nouns, and not to use the same word with different endings, such as "eat" and "eating." The letters presented to the subject are, in order, "C", "F", and "L". Scores are derived by totaling the number of words generated

PAGE 71

65 across the three trials, subjecting this total to age and education correction, and transforming this corrected raw score into a percentile. Finger Tapping Test . The Finger Tapping Test is a simple test of motor speed. Subjects use the index finger to alternately depress and release a small tapping device which counts the number of taps performed by the subject. Subjects are instructed to tap as rapidly as possible until told to stop. A subject's score is the average number of taps they are able to perform across four separate trials, each of which lasts ten seconds. The dominant hand is tested first, followed by the non-dominant hand. In the current study, norms developed by Bomstein (1985) were used to convert raw scores into standard deviation imits, taking age, gender and education into account. Grooved Pegboard . The Grooved Pegboard is a test of speeded motor dexterity. The pegboard has a square array of 25 slotted holes and a small well in which 25 grooved pegs are located. The grooves in the pegs are such that the peg will fit only one way; often, rotation of the groove is necessary in order to fit it into the slotted hole. First using only the dominant hand, subjects are instructed to fill all holes as rapidly as possible, completing each row in a left to right manner. The subject must then do the same with the non-dominant hand, filling each row of holes in a right to left manner. The subjects earns two scores on this test, which are the number of seconds required to complete the task with each hand. In the current study, norms developed by Bomstein (1985) were used to convert raw scores into standard deviation units, taking age, gender and education into account. EEG Data EEG data were obtained fi-om subject files in the Control Room on the EMU. These files contain information fi-om each subject's hospitalization for video/EEG telemetry, and include a

PAGE 72

66 detailed history of illness and dictations of each recorded seizure event. Following this dictation is a summary of the patient's surface EEG findings. During EEG, scalp electrodes were placed according to the International 10-20 System (Jasper, 1958), with the frequent additional use of bilateral sphenoidal electrodes. Patients generally have between six and twelve recorded events, but this number may actually range from one to 60 or more. At Shands Hospital, two major criteria must be met in order for a seizure to be determined to be lateralized and or localized to a given brain region. First, the seizure must involve focal rhythmic discharges with a frequency of four to seven Hertz. Second, this focal rhythmic discharge must be evident within the first 30 seconds following initial EEG seizure onset. All other discharges, excluding those containing sharp waves or spikes, are not considered to be localizing unless they develop into a rhythmic pattern of discharge which meets the above criteria for localization. The use of various electrode montages and filters to obtain maximal information when reviewing seizure events is standard practice at Shands Hospital. Any EEG recordings which are obscured by electromyographic or movement artifact are not used for lateralization or localization unless deemed to be valid by the supervising epileptologist. The investigator read each subject's video/EEG dictation on a seizure by seizure basis. The electroencephalographer's statement of seizure lateralization, localization, and whether or not the seizure involved secondary generalization was recorded. Several such dictations contained incomplete or ambiguous information. These were discussed in detail with the supervising epileptologist and decisions regarding seizure lateralization and localization were made collaboratively. Seizures which were uninterpretable due to factors such as electromyographic artifact and technical difficulties were counted for descriptive purposes only.

PAGE 73

67 Using the aforementioned procedures, seizure data was collected in two ways. First, all event dictations were read and recorded by the investigator. This was done without regard of the timing of these events in relation to one another. Subsequently, the investigator reexamined all event dictations and identified all occurrences when two or more seizures occurred within any three-hour time period. When two or more seizures occurred within a three-hour interval and were of identical lateralization and localization (e.g. were placed into the same lateralization and localization category described above), they were regarded as one event. This mirrors common clinical practice when counting seizure events on the EMU, as it is difficult to ascertain whether two or more temporally-related and similar seizures are actually isolated events. Counting two similar seizures which occur close together in time as different events may artificially inflate the clinical importance of those two events, because in fact, the second may have occurred at least partially as a result of the first. When two or more seizures occurred within a three hour interval but were of different lateralization and or localization, they were regarded as separate events. After collecting EEG data in both of the aforementioned manners, the data was reviewed with the supervising epileptologist and it was decided to use the data obtained using the method in which temporally-clustered seizures of identical lateralization and localization were regarded as one event. This decision was based on the combined facts that excluding such seizures did not severely limit the amount of available data and that, as mentioned above, this method of collection best represents clinical practice. MRIData All hippocampal volumes were measured by the principal investigator. As noted above, all but five subjects imderwent MRI as part of their Phase I evaluation. Imaging was conducted

PAGE 74

68 using magnetic resonance gradient echo scans (MPR3D: TR =10 ms, TE = 4 ms, 10 degree flip angle, matrix = 130 by 256, 160 mm volume, and section thickness = 1.25 to 1.40 mm). This protocol produces a gapless series of high quality images which can be reformatted into any plane of view. During hippocampal measurement, images were viewed on a computer monitor via optical disk drive. Using the mouse and cursor, each hippocampus was outlined on every section that it appeared in the sagittal plane. In most cases, a single hippocampus appeared on 10 to 15 separate MRI sections. Volumes were then calculated by multiplying hippocampal area for each section by the thickness of that section, and then adding together the resulting volumes to obtain a total volume for that hippocampus. Final volumes were expressed in cubic centimeters. Each subject's right and left hippocampi were measured twice, with the mean of these two values being the final right and left hippocampal volumes. Experimental Hypotheses Neuropsychological Domain Scores 1) Subjects who eventually imderwent left ATL will have significantly lower Language and Verbal Memory domain scores than will those who eventually underwent right ATL. 2) Subjects who eventually underwent right ATL will have significantly lower scores on the Nonverbal Memory domain. 3) Left and right ATL groups will not differ significantly on Visuoconstructive or Motor domain scores. Hypotheses 1 through 3 were tested using one-tailed t-tests with an alpha level of .05 for each test.

PAGE 75

69 Seizure Lateralization Index (SLI) 4) Subjects who eventually underwent right ATL will have significantly higher SLI scores (indicating a higher proportion of seizures that were lateralized to the right hemisphere) than will those who eventually underwent left ATL. Hypothesis 4 was tested using a one-tailed t-test with an alpha level of .05. Difference in Hippocampal Formation Volumes (DHF) 5) Subjects who eventually underwent right ATL will have significantly lower DHF scores (indicating a smaller right hippocampus) than will those who eventually underwent left ATL. Hypothesis 5 was tested using a one-tailed t-test with an alpha level of .05. Neuropsychological Data. SLI. and DHF 6) SLI, DHF, and neuropsychological data will all share significant correlations with one another. Hypothesis 6 was tested by calculating Pearson Product-Moment correlations among the three domains of data, with an alpha level of .05. In order to represent each subject's neuropsychological data as a single value, each subject's data were entered into a discriminant function as described below, with the five neuropsychological domains as predictor variables and side of surgery, right or left, being the predicted group. Then these single neuropsychological discriminant fiinction scores were correlated with SLI and DHF values. Relative Abilities of Neuropsychological Domain Scores. SLI. and DHF to Correctly Lateralize Eventual Side of Surgery It should be noted and will be shown in the Results section that neuropsychological domain scores fi-om the original data set and those fi-om the data set with missing values replaced

PAGE 76

70 by predicted values yielded nearly identical correct surgery side prediction rates. Therefore, subsequent analyses were performed using the neuropsychological domain scores with replacement of missing values. In this manner, all subjects received neuropsychological domain scores that were based on the same set of variables. For example, had the original data been used, one subject's Language domain score may have been derived from the BNT, WAIS-R VIQ and COW A, while another subject may have had this score derived only from the BNT and WAIS-R VIQ. Using the complete data set with missing values replaced avoided this situation. 7) SLI will be a somewhat better predictor of side of surgery than DHF, and both will be significantly better predictors than neuropsychological domain scores. This hypothesis was based on findings in the existing literature. Although findings regarding the ability of EEG data to correctly lateralize and localize seizure foci have varied greatly from study to study, the methodology and subject population used by Walczak et al. (1992) appears most similar to that used in the current study. These investigators found that surface EEG was 92 to 99 percent accurate in lateralizing temporal lobe seizure foci among patients who later underwent ATL. As in the current study, recordings that were of questionable validity due to artifect were excluded from the study. The hypothesis that DHF would be somewhat inferior to EEG with regard to correct prediction of surgery side was also based on findings in the literature. Most studies have examined the ability of MRI-based hippocampal volumetrics to correctly lateralize seizure foci among patients with significant hippocampal atrophy. For example. Jack et al. (1989), found that the existence of any degree of hippocampal asymmetry could correctly lateralize seizure foci in 90 percent of cases. However, the subjects in that investigation all had suspected hippocampal sclerosis based on visual MRI analysis. The current study included many subjects who did not

PAGE 77

71 meet criteria for significant hippocampal asymmetry, and therefore surgery side was predicted based upon widely varying DHF values. Studies examining the utility of MRI-based hippocampal volumes in the lateralization of seizure foci only among subjects with known hippocampal asymmetry would likely yield higher correct lateralization rates than the current study. The hypothesis that neuropsychological data would be inferior to both SLI and DHF in its ability to correctly predict side of surgery was based on two main facts. First, Williamson et al. (1993) found that neuropsychological data, as interpreted by the neuropsychologist, was only 73 percent correct in lateralizing seizure foci among patients who underwent successful ATL. Second, the current study did not include the diagnostic impression of skilled neuropsychologists, it merely subjected the neuropsychological test data to statistical procedures in an attempt to examine the data's ability to predict eventual side of surgery. Although these procedures were based upon those used by neuropsychologists, it cannot be assumed that they precisely mirrored the diagnostic thought processes of, or conclusions that would have been reached by, actual clinicians. 8) The correct surgery side prediction rate obtained when combining DHF and SLI as predictors will be significantly greater than that obtained when using SLI alone. A similar effect will be obtained when comparing correct surgery side prediction rates using neuropsychological domain scores in conjunction with SLI versus using SLI as the sole predictor of lateralization. Maximal correct prediction of side of surgery will be achieved through using neuropsychological domain scores, SLI, and DHF in conjunction with one another. 9) Neuropsychological domain scores will be more accurate in correctly predicting which subjects would undergo left ATL than in predicting those who would undergo right ATL. This was based on the findings of Williamson et al. (1993) which strongly indicated that current

PAGE 78

72 neuropsychological test data are much more sensitive and specific to left hemisphere pathology, creating a situation in which true right hemisphere pathology was often misinterpreted as left hemisphere pathology. 10) SLI and DHF, used individually as predictors of surgery side, will be equally effective in predicting which subjects would undergo left ATL or right ATL. This is based on the fact that there is no reason to assume that either EEG or MRI is more sensitive to pathology in one hemisphere or the other. Hypotheses 7 through 10 were tested using discriminant analysis and additional statistical procedures described below. Initially, the five neuropsychological domain scores were entered into a discriminant analysis as predictors of eventual side of surgery. The same method was used for SLI, and then for DHF. This revealed the efficacy with which neuropsychological data, SLI and DHF could each predict eventual side of surgery when used independently of other data. It should be noted that when a discriminant analysis is used with only a single predictor variable, as was the case with SLI and DHF, the procedure is essentially a t-test with the added fact that the method also generates a predicted group membership for each subject. Next, neuropsychological domain scores (as a package), SLI, and DHF were entered into discriminant analyses in all possible combinations of these three domains of data as predictors of each subject's eventual side of surgery. It should be noted that these discriminant analyses were conducted using the "jackknife" or, "Leave-One-Out" method. This technique is similar to traditional discriminant analysis, however, each subject is removed fi^om the data pool one at a time, discriminant fiinction coefficients are calculated for each predictor variable based on data from all other subjects, and these coefficients are then applied to the data from the subject who had temporarily been removed from the data pool. Finally, prediction regarding that subject's group classification is made based

PAGE 79

73 on his or her discriminant function score, hi this manner, each subject was assigned a predicted side of surgery based on discriminant function coefficients that were calculated independently of his or her own data. This is sound statistical methodology in general, but it was particularly important that it was used in the current study. This is due to the &ct that the study examined the utility of neuropsychological, EEG, and MRI data to predict eventual side of surgery, among subjects for whom side of surgery decisions were based on the very same neuropsychological, EEG and MRI data. The use of the jackknife method in this situation ensured that each subject's predicted side of surgery was based on a discriminant fimction developed independently of his or her own data. In order to determine whether neuropsychological domain scores, DHF, and SLI, used independently, correctly predicted eventual surgery side at significantly differing rates, the proportion of correct predictions generated by each of the three was calculated. These proportions were then tested for equality using the McNemar test for 2 X 2 tables. Comparisons of classification rates obtained using neuropsychological domain scores, DHF, and SLI in varying combinations were made by obtaining the Wilk's Lambda value fi-om each combination of predictors' discriminant analysis, converting this value into a Hotelling's T^, and then conducting an F-test to test for equality between the two Hotelling's T\alues in question. An alpha level of .05 was used for these tests. Three chi-square tests, with alpha levels of .05, were used in order to determine whether significantly different proportions of LATL and RATL subjects were correctly classified using, separately, SLI, DHF, and neuropsychological domain scores as predictors.

PAGE 80

74 Relative Abilities of Neuropsychological Domain Scores. SLI. and DHF to Correctly Lateralize Actual Side of Seizure Focus 1 1) Relative to one another, neuropsychological domain scores, SLI, and DHF will show a similar pattern of efficacy in lateralizing seizure onset as when used to lateralize eventual side of surgery as discussed above. 12) Overall, each of these three domains of data will be slightly more effective in predicting actual side of seizure onset than in merely predicting eventual side of surgery. In order to test Hypotheses 1 1 and 12, discriminant analyses, McNemar tests for 2 X 2 tables, f-tests conducted on Hotelling's values, and one-tailed t-tests were used in a maimer identical to that which was used to test Hypotheses 7 through 10. However, only subjects who met criteria of an Engel classification of I were used in these analyses. These are subjects who were seizure-free for at least one year following surgery and who were noted to be seizure-free during their most recent clinic visit in Neurology. An example may be useful in illustrating the importance of using this subset of the experimental population to test these hypotheses. While Hypotheses 7 through 10 concerned the prediction of eventual side of surgery, they did not directly address actual lateralization of seizure foci. For example, it may have been predicted that Patient A would undergo a right ATL, but one can only assume that the right anterior temporal lobe was the actual seizure focus if that individual became seizure-free following surgery. That is to say, for subjects who experienced seizure-freedom following surgery, it can reasonably be assumed that these subjects did have unilateral focal epilepsy and that the side of surgery and side of actual epileptogenic focus were one in the same. Following these analyses, all subjects were again included in subsequent analyses, regardless of seizure outcome. In an attempt to identify demographic and illness-related variables

PAGE 81

75 which may serve as sources of variation in presurgical neuropsychological performance, the following procedures were conducted. As it had been found that a combination of SLI and DHF served as optimal predictors of eventual side of surgery, subjects for whom this combination had correctly predicted side of surgery were selected. Then the predicted side of surgery for each subject using the neuropsychological domain scores was examined. Subjects for whom neuropsychological domain scores yielded a predicted side of surgery which was in agreement with that predicted by SLI and DHF together were placed in one group which was termed the Convergent Neuropsychology Group. Subjects for whom neuropsychological domain scores yielded a predicted side of surgery which was discrepant from that predicted by SLI and DHF together were placed in a second group, which was termed the Divergent Neuropsychology Group. Convergent and Divergent Neuropsychology Group Differences 13) The Divergent Neuropsychology Group will have mean SLI values that are significantly closer to 0.00 than will the Convergent Neuropsychology Group. This hypothesis was based on the fact that subjects who had SLI values at or near a value of 0.00 showed evidence of seizure lateralization in both hemispheres, and the assumption that this bihemispheric seizure involvement would be reflected in the existence of less specifically lateralized and or incorrectly lateralized neuropsychological profiles among these subjects. 14) The Divergent Neuropsychology Group will have a higher proportion of subjects who are not seizure-free following surgery than will the Convergent Neuropsychology Group. This hypothesis was based on the assumption that the Divergent Neuropsychology Group had neuropsychological profiles that were suggestive of neural pathology contralateral to the surgical

PAGE 82

76 hemisphere, and that the continued presence of this suggested contralateral pathology following surgery may give rise to seizures. 1 5) The Divergent Neuropsychology Group will have a higher proportion of subjects who have a history of secondarily generalized seizures than will the Convergent Neuropsychology Group. This hypothesis was based on the assumption that seizures, particularly those involving secondary generalization, induce cerebral insult both ipsilateral and contralateral to the actual seizure focus, and that this would be reflected in neuropsychological profiles that were more suggestive of contralateral cerebral insult than were EEG and MRI data. 16) The Divergent Neuropsychology Group will have a lower proportion of subjects who meet EMU criteria for hippocampal atrophy (-.45 cm^> DHF or DHF > +.55cm^) (Gilmore et al., 1995) than will the Convergent Neuropsychology Group. This is based on the assumption that the presence of clinically significant and focal hippocampal atrophy would be reflected in neuropsychological profiles that were more specifically convergent with EEG and MRI data. 17) The Divergent Neuropsychology Group will have a higher proportion of subjects who eventually underwent right ATL than the Convergent Neuropsychology Group. This hypothesis was based on the work of Williamson et al. (1993), who reported that all subjects in their study that were incorrectly lateralized on the basis of neuropsychological data had actual right hemisphere seizure foci. Thus, there does not appear to be a demonstrable and consistent neuropsychological "signature" of right temporal lobe complex partial seizures. Exploratory Group Comparisons Divergent and Convergent Neuropsychology group differences in age at epilepsy onset, duration of illness, proportion of subjects with a history of febrile seizures, and mean years of education will be examined on an exploratory basis.

PAGE 83

77 Two-tailed t-tests were used to test these group differences. An alpha level of .05 was used with each test.

PAGE 84

CHAPTERS RESULTS Throughout the following text and tables, the group of subjects who eventually underwent right ATL will be referred to as RATL, and the group that eventually underwent left ATL will be referred to as LATL. When the entire subject population is referred to, the collective group will be referred to as All. Demographic and Illness-Related Variables Descriptive statistics for demographic and illness-related variables are shown in Table 1. Two-tailed t-tests revealed no significant differences between the RATL and LATL groups on the following variables: Education, Age at onset of epilepsy. Age at onset of secondarily generalized seizures (if this occurred), and Duration of illness. An alpha level of .05 was used with each test. Chi-square tests with alpha levels of .05 were used to test for group differences on the remaining variables, which were in the form of group percentages. Again, these tests indicated no significant group differences for the tested values. Neuropsvchological Variables Descriptive statistics for individual neuropsychological variables are shown in Table 2. These data were taken fi-om the original data set, and represent the normed data with some missing values as collected fi-om individual subjects. As can be seen, some data points were missing and therefore, the n for each variable is included. Two-tailed t-tests were conducted and, when necessary, the Welch correction of degrees of fi-eedom was used to account for inequality of variance between the groups. These analyses revealed that the LATL mean scores were 78

PAGE 85

79 Table 1 Descriptive Statistics for Demographic and Illness-Related Variables Variable All LATL RATL B Age* 35.61 (10.54) 35.14(11.44) 36.23 (09.37) ns Education* 12.73 (02.22) 12.51 (02.11) 13.02 (02.37) ns AO* 12.45(11.72) 12.27(11.90) 12.68(11.71) ns A02* 14.54(11.88) 14.04 (21.80) 12.20 (10.72) ns Dur. Illness* 24.10(11.44) 23.23 (11.05) 25.27 (12.05) ns Female'' 57.40 57.10 57.70 ns Febrile Sz's"" 39.30 31.40 50.00 ns Gen. Sz's*" 72.10 68.60 76.90 ns Note. N = 61 for All, 35 for LATL, 25 for RATL. All all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; * mean number of years (standard deviation in parentheses); AO age at onset of epilepsy; A02 = age at onset of secondarily generalized seizures, if this occurred; Dur. Uhiess = duration of illness; percentage of each group; Febrile Sz's = percentage of each group with a known history of febrile seizures; 2"'' Gen. Sz's = percentage of each group with a known history of secondarily generalized seizures; ns = groups not significantly different at the .05 alpha level. significantly lower than RATL mean scores for WAIS-R VIQ, t (41.69) = 2.54, g < .05, CVLT SDFR, t (59) = 2.44, p < .05, CVLT LDFR, t (59) = 2.21, p < .05, and CVLT LDCR, t (59) = 2. 13, g < .05. An alpha level of .05 was used for each test.

PAGE 86

80 Table 2 Descriptive Statistics for Individual Neuropsychological Variables All LATL RATL Variable MSDn MSDa MSDn FSIQ PIQ VIQ OAZ BDZ LMI LMII VRI VRII SDFR SDCR LDFR LDCR PRINT CRINT ROCOPY 91.00 9.91 61 92.18 11.69 61 91.82 9.82 61 -.42 .79 61 -.16 .89 61 36.54 26.35 59 30.13 23.89 59 52.09 33.19 58 43.81 36.17 58 -1.67 1.71 61 -1.52 1.74 61 -1.72 1.78 61 -1.79 1.81 61 .67 1.43 61 1.26 2.31 61 8.13 4.08 61 RODELAY 7.49 3.09 61 BNT -2.58 2.34 53 89.03 9.31 35 91.8 12.72 35 89.06 7.68 35 -.44 .83 35 -.22 .90 35 31.35 22.04 34 24.63 17.52 34 53.32 34.66 34 44.56 37.43 34 -2.11 1.43 35 -1.80 1.66 35 -2.14 1.70 35 -2.20 1.73 35 .97 1.58 35 1.57 2.34 35 8.23 3.95 35 8.11 2.98 35 -3.14 2.18 28 93.65 10.24 26 92.69 10.37 26 * 95.54 11.23 26 -.40 -.08 .75 26 .88 26 43.60 30.34 25 37.60 29.27 25 50.33 31.63 24 42.75 35.07 24 -1.08 1.90 26 -1.15 1.80 26 -1.15 1.76 26 -1.23 1.80 26 .27 1.12 26 .85 2.24 26 8.00 4.34 26 6.65 3.10 26 -1.94 2.38 25

PAGE 87

81 Table 2 Continued All LATL RATL Variable M SD n M ss fi M SD n COWA 37.32 33.34 60 35.75 33.54 34 39.37 33.62 26 FTTDOM .03 1.43 57 .02 1.64 32 .05 1.13 25 FTTNDOM -.26 1.53 57 -.43 1.57 32 -.03 1.48 25 GPTDOM -1.55 2.26 59 -1.55 2.08 33 -1.55 2.52 26 GPTNDOM -1.79 2.69 59 -1.71 2.19 33 -1.89 3.26 26 Note. All = all subjects; LATL subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; FSIQ Wechsler Adult Intelligence Scale-Revised (WAIS-R) Full Scale IQ; PIQ = WAIS-R Performance IQ; VIQ = WAIS-R Verbal IQ; OAZ = WAIS-R Object Assembly z-score; BDZ = WAIS-R Block Design z-score; LMl and LMII = Wechsler Memory Scale (WMS-R) Logical Memory I and II percentile scores; VRI and VRII = WMS-R Visual Reproduction I and II percentile scores; SDFR = California Verbal Learning Test (CVLT) Short Delay Free Recall z-score; SDCR = CVLT Short Delay Cued Recall z-score; LDFR = CVLT Long Delay Free Recall z-score; LDCR = CVLT Long Delay Cued Recall z-score; FRINT = CVLT Free Recall Intrusions z-score; CRINT = CVLT Cued Recall Intrusions z-score; ROCOPY and RODELAY = Rey-Osterreith Copy and Delayed Recall scaled scores; BNT Boston Naming Test z-score; COWA = Controlled Oral Word Association percentile score; FTTDOM and FTTNDOM = Finger Tapping Test z-scores for dominant and nondominant hands, respectively; GPTDOM and GPTNONDOM = Grooved Pegboard Test z-scores for dominant and nondominant hands, respectively. * = LATL and RATL group means were significantly different < .05). Descriptive statistics for individual neuropsychological variables after replacement of missing values are shown in Table 3. As there were no missing WAIS-R or CVLT scores in the original data set (see Table 2), two-tailed t-tests revealed an identical pattern of LATL and RATL

PAGE 88

82 Table 3 Descriptive Statistics for Neuropsychological Variables with Replacement of Missing Values All LATL RATL Variable M SD M SD M SD FSIQ 91.00 9.91 89.03 9.31 93.65 10.24 PIQ 92.18 11.69 91.80 12.72 92.69 10.37 VIQ 91.82 9.82 89.06 7.68 * 95.54 11.23 OAZ -.42 .79 -.44 .83 -.40 .75 BDZ -.16 .89 -.22 .90 -.08 .88 LMI 36.65 26.08 31.15 21.74 44.06 29.82 LMII 30.43 23.73 24.59 17.26 38.29 28.89 VRI 51.98 32.56 52.61 34.41 51.14 30.53 VRII 43.87 35.77 43.87 37.10 43.88 34.63 SDFR -1.67 1.71 -2.11 1.43 * -1.08 1.90 SDCR -1.52 1.74 -1.80 1.66 -1.15 1.80 LDFR -1.72 1.78 -2.14 1.70 * -1.15 1.76 LDCR -1.79 1.81 -2.20 1.73 * -1.23 1.80 PRINT .67 1.43 .97 1.58 .27 1.12 CRINT 1.26 2.31 1.57 2.34 .85 2.24 ROCOPY 8.13 4.08 8.23 3.95 8.00 4.34 RODELAY 7.49 3.09 8.11 2.98 6.58 3.16 BNT -2.59 2.25 -3.09 2.08 * -1.93 2.34

PAGE 89

83 Table 3 Continued All LATL RATL Variable M SD M SD M SD COWA 37.46 33.08 36.04 33.09 39.37 33.62 FTTDOM .00 1.41 -.03 1.61 .04 1.11 FTTNDOM -.29 1.48 -.47 1.50 -.05 1.45 GPTDOM -1.54 2.24 -1.54 2.05 -1.55 2.52 GPTNDOM -1.75 2.65 -1.66 2.14 -1.89 3.26 Note. All = all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; n = 35 for LATL and 26 for RATL for all variables. These data are from the data set in which missing values were replaced with predicted ones as described above. FSIQ = Wechsler Adult Intelligence ScaleRevised (WAIS-R) Full Scale IQ; PIQ = WAIS-R Performance IQ; VIQ WAIS-R Verbal IQ; OAZ = WAIS-R Object Assembly z-score; BDZ = WAIS-R Block Design z-score; LMI and LMU = Wechsler Memory Scale (WMS-R) Logical Memory I and II percentile scores; VRI and VRII = WMS-R Visual Reproduction I and II percentile scores; SDFR = California Verbal Learning Test (CVLT) Short Delay Free Recall z-score; SDCR = CVLT Short Delay Cued Recall z-score; LDFR = CVLT Long Delay Free Recall z-score; LDCR = CVLT Long Delay Cued Recall z-score; FRINT = CVLT Free Recall Intrusions z-score; CRINT = CVLT Cued Recall Intrusions z-score; ROCOPY and RODELAY ^ Rey-Osterreith Copy and Delayed Recall scaled scores; BNT = Boston Naming Test z-score; COWA = Controlled Oral Word Association percentile score; FTTDOM and FTTNDOM = Finger Taping Test z-scores for dominant and nondominant hands, respectively; GPTDOM and GPTNONDOM = Grooved Pegboard Test zscores for dominant and nondominant hands, respectively. * = LATL and RATL group means were significantly different (p < .05). group differences with regard to these variables. In addition, the LATL group mean was lower than the RATL group mean on the BNT, with marginal significance, t (59) = 2.05, p < .05. When this t-test was conducted on the original data it indicated a trend toward significance, t (5 1) =

PAGE 90

84 1.92, p = .061. A comparison of Tables 2 and 3 shows that the difference in LATL and RATL group means for BNT is actually smaller following the replacement of missing values procedure. It appears that the increased n in the data set with replaced values led to this difference becoming marginally significant. Again, an alpha level of .05 was used with each test. Hi ppocampal Volumetrics Descriptive statistics for hippocampal volumes are shown in Table 4. Two-tailed t-tests indicated that mean right hippocampal volume was significantly smaller for the RATL group, t (56) = 4.38, g < .001, and that the mean left hippocampal volume was significantly smaller for the LATL group, t (56) = 3.94, p < .001. Mean DHF was significantly higher for the LATL group than the RATL group t (56) = 7.83, g < .001. An alpha level of .05 was used with each test. EEG Variables Descriptive statistics for number of seizures lateralized and localized to each brain region, number of seizures rendered uninterpretable due to artifact, number of seizures with secondary generalization, and total number of seizures during Phase I evaluation are shown in Table 5. SLI. DHF. and Neuropsychological Domain Scores (Hypotheses 1 through 5) Descriptive statistics for SLI, DHF, and neuropsychological domain scores are shown in Table 6. Two-tailed t-tests indicated that among the neuropsychological domain scores, the LATL group mean was significantly lower than the RATL group mean for VMEMSCOR, t (59) = 2.40, E < .05. The LATL group mean was marginally lower than the RATL group mean for LANGSCOR, t (59) = 2.04, p < .05. These analyses partially supported Hypothesis 1, which I J

PAGE 91

85 Table 4 Descriptive Statistics for Volumetric Variables Variable All LATL RATL RHIPPVOL' 3.16 (.70) 3.45 (.53) > 2.74 (.71) LHIPPVOL" 3.17 (.74) 2.88 (.73) < 3.57 (.53) DHF* -.01 (.96) .57 (.64) > -.83 (.71) %RHS 31.00 02.90 70.80 %LHS 31.00 52.90 00.00 Note, n = 58 for All, 34 for LATL, 24 for RATL. All = all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; ' values in cubic centimeters; RHIPPVOL = mean right hippocampal volume; LMPPVOL = mean left hippocampal volume; DHF mean difference in hippocampal formation volume (right left, in cm^); %RHS = percentage of each group meeting criteria for significantly smaller right hippocampus (DHF < -.45 cm^); %LHS = percentage of each group meeting criteria for significantly smaller left hippocampus (DHF > +.55 cm^). predicted that both of these group differences would be significant. Hypothesis 2 was not supported by these analyses, as the two groups did not differ significantly on mean NVMEMSCOR, and Hypothesis 3 was supported, as the group means were not significantly different for VISCONSCOR or MOTORSCOR. Mean SLI was significantly lower for the LATL than for the RATL group, t (58) = 1 1.91, g < .001, a finding which is consistent with Hypothesis 4. Mean DHF was significantly lower for the RATL group than for the LATL group, t (56) = -7.80, p < .001, which is in support of Hypothesis 5.

PAGE 92

86 Table 5 Descriptive Statistics for Phase I EEG Variables All LATL RATL Variable M SD Sum M SD Sum M SD Sum LHEM .08 .33 5 .14 .43 5 .00 .00 00 LFRTEMP .25 .91 15 .43 1.17 15 .00 .00 00 LTEMP 1.10 1.89 66 1.69 2.19 59 .28 .89 7 LMESTEMP 1.22 1.81 73 1.80 1.94 63 .40 1.26 10 NLAT .35 1.33 21 .37 1.57 13 .32 .90 8 INLiAl ItJVlr 1 1 .Uo 1 o .'fU 1 .J J If .JO c J RHEM .10 .57 6 .00 .00 00 .24 .88 6 RFRTEMP .13 .70 8 .14 .85 5 .12 .44 3 RTEMP .58 1.33 35 .03 .17 1 1.36 1.80 34 RMESTEMP 1.35 2.79 81 .14 .55 5 3.04 3.68 76 2'^ Gen. .98 1.27 59 .94 1.00 33 1.04 1.59 26 ARTIFACT .68 3.23 41 .37 1.26 13 1.12 4.80 28 TOTAL 5.48 2.98 329 5.14 2.55 180 5.96 3.49 149 Note. All = all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; n = 60 for All, 35 for LATL and 25 for RATL for all variables. LHEM = number of seizures lateralized to the left hemisphere, but not localized more specifically; LFRTEMP = number of seizures localized to the left frontotemporal region; LTEMP = number of seizures localized generally to the left temporal

PAGE 93

87 Table 5 Continued lobe; LMESTEMP = number of seizures localized specifically to the left mesial temporal lobe; NLAT = number of seizures that were not lateralized; NLATTEMP number of seizures that were not lateralized but appeared to originate in one or the other temporal lobe; RHEM, RFRTEMP, RTEMP, and RMESTEMP are the right hemisphere counterparts to the left hemisphere variables described above; 2"^ Gen. = number of seizures with secondary generalization; ARTIFACT = number of seizures uninterpreted due to artifact; Total = total number of seizures during Phase I (excluding events uninterpretable due to artifact). All seizure lateralization/localization categories above are exclusive of one another. Correlations Among SLI. DHF. and Neuropsychological Discriminant Function Scores (Hypothesis 6) Following are the correlations among SLI, DHF, and neuropsychological discriminant fixnction scores. These discriminant ftinction scores were obtained by entering into a discriminant analysis subjects' five neuropsychological domain scores (generated from data in Table 3) as predictors of side of surgery. The resulting discriminant fimction scores represented each subject's five neuropsychological domain scores as a single value. Correlations were as follows: SLI and DHF: r = -.60 (g < .001, n = 57), SLI and neuropsychological discriminant fimction score: r = .38 (p < .01, n 60), DHF and neuropsychological discriminant fimction score: r = -.30 (p < .05, n = 58). These findings are in support of Hypothesis 6. Correct Surgery-Side Classification Rates (Hypotheses 7 through 10) The rates at which SLI, DHF, and neuropsychological domain scores, used independently and in combination with one another, were able to correctly predict side of surgery using discriminant analyses are shown in Table 7. These data represent the correct surgery side classification rates using all available data. In keeping with Hypotheses 7 and 8, the following classification rates were compared: SLI vs. DHF, SLI vs. neuropsychology domain scores, DHF vs. neuropsychology domain scores, SLI and DHF combined vs. SLI, neuropsychology domain

PAGE 94

88 Table 6 Descriptive Statistics for SLI. DHF. and Neuropsychological Domain Scores Variable AU LATL RATL M SD M SD M SD SLI -.19 (.86) -.80 ( 38) .66 (.57) DHF -.01 (.96) .57 ( ,64) >» -.83 (.71) LANGSCOR -1.27 (1.19) -1.53 (1 .17) <" -.92 (1.15) VMEMSCOR -1.22 (1.28) -1.54 (1 .17) <" -.78 (1.32) NVMEMSCOR -.58 (1.07) -.49 (1 .06) -.71 (1.09) VISCONSCOR -.40 (.81) -.42 ( .83) -.38 (.78) MOTORSCOR .50 (2.25) .55 (1 .99) .42 (2.61) Note. All = all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; For SLL n = 60 for All, 35 for LATL, 25 for RATL; For DHF: n 58 for All, 34 for LATL, 24 for RATL; For all other variables: N = 61 for All, 35 for LATL, 26 for RATL; SLI = Seizure Lateralization Index; DHF = Difference in Hippocampal Formation Volume (Right Left, in cm^); LANGSCOR = Neuropsychological Language Domain Score; VMEMSCOR = Neuropsychological Verbal Memory Domain Score; NVMEMSCOR = Neuropsychological Nonverbal Memory Domain Score; VISCONSCOR = Neuropsychological Visuoconstructive Domain Score; MOTORSCOR = Neuropsychological Motor Domain Score; < and > indicate presence and directionality of significant differences in LATL and RATL group means; * g < .01; g < .05. scores and SLI combined vs. SLI, and finally, neuropsychology domain scores, SLI, and DHF combined vs. SLI and DHF combined. The results of these tests are presented below. Numbers of subjects here were slightly different fi-om those in Table 7, as it was necessary when

PAGE 95

89 Table 7 Correct Surgery-Side Prediction Rates Using NP. SLI. DHF as Predictors Predictor LATL RATL All Variables % CORRECT n % CORRECT n % CORRECT n SLI 88.60 35 84.00 25 86.70 60 DHF 85.30 34 75.00 24 81.03 58 NPO 60.61 33 61.54 26 61.01 59 NP 60.00 35 61.54 26 60.66 61 DHF SLI 97.06 34 91.30 23 94.74 57 NP SLI 91.43 35 88.00 25 90.00 60 NP DHF 85.29 34 75.00 24 81.03 58 NP DHF SLI 97.06 34 91.30 23 94.74 57 Note. All = all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; SLI = Seizure Lateralization Index; DHF = Difference in Hippocampal Formation Volume (Right Left, in cm^); NPO = all five neuropsychological domain scores (LANGSCOR, VMEMSCOR, NVMEMSCOR, VISCONSCOR, MOTORSCOR) fi-om the original data set; NP = all five neuropsychological domain scores from the data set in which missing values were replaced with predicted ones; % CORRECT = percentage of each group for which side of surgery was correctly predicted by entering various combinations of predictor variables into discriminant analyses. comparing two sets of predictors' classification rates, to include only those subjects who had data from all predictors involved in the comparison. McNemar tests for 2 x 2 tables were used to test the first three of the above comparisons. Due to the fact that there were a relatively low number of subjects in some of the cells of these 2

PAGE 96

90 X 2 tables, chi-square statistics could not be calculated reliably for each of the comparisons. Instead, the exact significance of each comparison was obtained using the binomial distribution, and it is this value that is reported below. Results indicated that individual classification rate for SLI (85.96 % correct) and that for DHF (82.46 % correct) were not significantly different (p = .607, n = 57). Classification rate for SLI (86.70 % correct) was significantly better than that for the five neuropsychological domain scores (60.66 % correct) (g = .004, n = 60). Classification rate for DHF (8 1 .03 % correct) was also significantly better than that for the five neuropsychological domain scores (59.65 % correct) (g = .015, n = 58). These findings are in support of Hypothesis 7. F-tests were used to test the final three comparisons described above. Results indicated that the classification rate for SLI and DHF combined (94.74 % correct) was significantly better than that for SLI alone (85.96 % correct), F (1, 55) = 17.29, p < .01, n = 57). Classification rate for the five neuropsychological domain scores and SLI combined (90.00 % correct) was not significantly different fi-om that for SLI alone (86.70 % correct, n = 60). Finally, classification rate for neuropsychological domain scores, SLI, and DHF combined (94.74 % correct) was identical to that for SLI and DHF combined (n = 57). These analyses partially supported Hypothesis 8, as the classification rate obtained when using SLI and DHF together was a significant improvement upon that obtained when using SLI as the sole predictor of side of surgery. However, it was also hypothesized that combining the five neuropsychological domain scores with SLI would yield a classification rate that was significantly better than that obtained fi-om SLI alone, and that combining all three sets of predictors would yield the most accurate classification rate of all. As can be seen, the data were not supportive of these last two parts of Hypothesis 8.

PAGE 97

91 Hypotheses 9 and 10 were tested by conducting three chi -square tests, each with an alpha level of .05, comparing the proportion of LATL subjects who were correctly classified regarding side of surgery to the proportion of RATL subjects who were correctly classified. Results indicated that in all three cases, the rate at which LATL vs. RATL subjects were correctly classified were not significantly different. These findings do not support Hypothesis 9, which predicted that the five neuropsychological domain scores would be more effective predictors of LATL than RATL, however they are in support of Hypothesis 10 which predicted that SLI and DHF, used independently, would be equally effective in predicting LATL and RATL. Following is a detailed description of each set of predictors used in the above discriminant analyses, the resulting values of Wilk's Lambda, its significance, and the Standardized Canonical Discriminant Function Coefficients for each predictor variable used in each analysis. In situations in which there was only one value (e.g. SLI or DHF) used as a predictor of side of surgery, only Wilk's Lambda and its significance will be reported, as the Standardized Canonical Discriminant Function Coefficient for a sole predictor in a discriminant function will always be equal to 1 .000. Entering SLI as the sole predictor of side of surgery yielded the following results: Wilk's Lambda = .290 (g < .001), n = 60. Entering DHF as the sole predictor of side of surgery yielded the following results: Wilk's Lambda = .477 (p < .001), n = 58. Entering the five neuropsychological domain scores from the original data set as predictors of side of surgery yielded the following results: Wilk's Lambda = .841, (g > .05), Standardized Canonical Discriminant Function Coefficients: LANGSCOR = .380, NVMEMSCOR = -.715, VMEMSCOR = .897, VISCONSCOR = . 124, MOTORSCOR = -.089, n = 59. Entering the five neuropsychological domain scores from the data set with replacement of missing values yielded the following results: Wilk's Lambda = .819, (p < .05), Standardized Canonical Discriminant

PAGE 98

92 Function Coefficients: LANGSCOR =521, NVMEMSCOR = -.753, VMEMSCOR = .797, VISCONSCOR = .071, MOTORSCOR = -. 1 13, N = 61. Entering DHF and SLI as predictors of side of surgery yielded the following results: Wilk's Lambda = .231 (g < .001), Standardized Canonical Discriminant Function Coefficients: SLI = .826, DHF = -.560, n = 57. Entering the five neuropsychological domain scores and SLI as predictors of side of surgery yielded the following results: Wilk's Lambda = .263 (g < .001), Standardized Canonical Discriminant Function Coefficients: LANGSCOR = .324, NVMEMSCOR = -.118, VMEMSCOR = .108, VISCONSCOR = .01 1, MOTORSCOR = 053, SLI = .981, n = 60. Entering the five neuropsychological domain scores and DHF as predictors of side of surgery yielded the following results: Wilk's Lambda = .428 (p < .001), Standardized Canonical Discriminant Function Coefficients: LANGSCOR = -.073, NVMEMSCOR = .460, VMEMSCOR = -.322, VISCONSCOR = -. 124, MOTORSCOR = .036, DHF = .937, n = 57. Entering the five neuropsychological domain scores, DHF and SLI together as predictors of side of surgery yielded the following results: Wilk's Lambda = .217 (p < .001), Standardized Canonical Discriminant Function Coefficients: LANGSCOR = .183, NVMEMSCOR = -.223, VMEMSCOR = .058, VISCONSCOR = . 127, MOTORSCOR = .051, DHF = -.541, SLI = .827, n = 57. Correct Seizure Focus Lateralization Rates (Hypotheses 1 1 and 12) In order to examine correct seizure focus lateralization rates, as predicted by various combinations of SLI, DHF and neuropsychological domain scores, it was necessary to include only subjects who were seizure-fi-ee (those with Engel Classification = I) following surgery in these analyses. As described above, for these subjects it can be assumed that side of surgery and

PAGE 99

93 side of seizure focus are one and the same. Descriptive statistics for this subset of the subject population are shown in Table 8. Two-tailed t-tests were conducted, again using the Welch correction of degrees of freedom where necessary to account for inequality of variance between the groups. An alpha level of .05 was used with each test. These analyses indicated that LATL group mean for SLI was significantly lower than that for the RATL group, t (25.16) = 9.21, p < .001. Mean DHF was significantly lower for the RATL group than for the LATL group, t (40) = -7.34, g < .001. Among the neuropsychological domain scores, the LATL group mean was significantly lower than the RATL group mean for VMEMSCOR, t (42) = 2.22, p < .05. The LATL group mean was significantly lower than the RATL group mean for LANGSCOR, t (42) = 2.07, p < .05. The rates at which SLI, DHF, and neuropsychological domain scores, used independently and in combination with one another, were able to correctly predict seizure focus lateralization using discriminant analyses are shown in Table 9. In keeping with Hypotheses 1 1 and 12, the following classification rates were compared: SLI vs. DHF, SLI vs. the five neuropsychology domain scores, DHF vs. the five neuropsychology domain scores, SLI and DHF combined vs. SLI, the five neuropsychology domain scores and SLI combined vs. SLI, and finally, the five neuropsychology domain scores, SLI, and DHF combined vs. SLI and DHF combined. The results of these tests are presented below. Number of subjects in each analysis here were slightly different fi"om those in Table 9, as it was necessary when comparing two sets of predictors' classification rates, to include only those subjects who had data fi-om all predictors involved in the comparison. McNemar tests for 2 x 2 tables were used to test the first three of the above comparisons. Again, the exact significance of each comparison was obtained fi-om the binomial distribution.

PAGE 100

94 Table 8 Descriptive Statistics for SLI. DHF. and Neuropsychological Domain Scores for S ubjects who had Engel Classification = I Following Surgery Variable All LATL RATL M SD M SD M SD SLI -.23 ( 87) -.83 (.36) <• .63 ( 60) DHF -.02 ( 97) .58 (.62) >» -.90 ( 66) LANGSCOR -1.20 (1 19) -1.50 (1.11) <•> -.78 (1 19) VMEMSCOR -1.26 (1 .29) -1.60 (1.26) <^ -.76 (1 .19) NVMEMSCOR -.44 (1 .03) -.33 (.92) -.61 (1 .18) VISCONSCOR -.35 ( .82) -.29 (.85) -.44 ( .79) MOTORSCOR .45 (2 .36) .39 (1.94) .54 (2 .92) Note. All = all subjects; LATL = subjects who eyentually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; For DHF: n = 42 for All, 25 for LATL, 17 for RATL; For all other variables: n = 44 for All, 26 for LATL, 18 for RATL; SLI = Seizure Lateralization Index; DHF Difference in Hippocampal Formation Volume (Right Left, in cm^); LANGSCOR = Neuropsychological Language Domain Score; VMEMSCOR = Neuropsychological Verbal Memory Domain Score; NVMEMSCOR = Neuropsychological Nonverbal Memory Domain Score; VISCONSCOR = Neuropsychological Visuoconstructive Domain Score; MOTORSCOR = Neuropsychological Motor Domain Score; < and > indicate presence and directionality of significant differences in LATL and RATL group means;*E<.0I;''g<.05. and it is this value that is reported below. Results indicated that classification rate for SLI (88.10 % correct) and that for DHF (85.71 % correct) were not significantly different (p = 1.00, n = 42). Classification rate for SLI (86.67 % correct) was significantly better than that for the five neuropsychological domain scores (65.91 % correct) (2 = .019, n = 44). Classification rate for

PAGE 101

95 Table 9 Correct Seizure Focus Lateralization Rates Using NP. SLI. DHF as Predictors Predictor LATL RATL All Variables % CORRECT n % CORRECT n % CORRECT n SLI DHF NP DHF SLI NP SLI NP DHF NP DHF SLI 92.31 26 88.00 25 69.23 26 92.00 25 92.31 26 84.00 25 96.00 25 83.33 18 82.35 17 61.11 18 94.12 17 88.89 18 76.47 17 94.12 17 86.67 44 85.71 42 65.91 44 92.86 42 90.91 44 80.95 42 95.24 42 Note. All = all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; SLI = Seizure Lateralization Index; DHF = Difference in Hippocampal Formation Volume (Right Left, in cm^); NP = all five neuropsychological domain scores fi-om the data set in which missing values were replaced with predicted ones (LANGSCOR, VMEMSCOR, NVMEMSCOR, VISCONSCOR, MOTOSCOR); % CORRECT = percentage of each group for which side of seizure focus was correctly predicted by entering various combinations of predictor variables into discriminant analyses. DHF (85.71 % correct) was also significantly better than that for the five neuropsychological domain scores (61.90 % correct) (p = .041, n = 42). These findings are in support of Hypothesis 11. F-tests were used to test the final three comparisons described above. Results indicated that classification rate for SLI and DHF combined (92.86 % correct) was significantly better than

PAGE 102

96 that for SLI alone (88. 10 % correct), F (1, 40) = 13.76, E < .01, n = 42. Classification rate for the five neuropsychological domain scores and SLI combined (90.91 % correct) was not significantly different fi-om that for SLI alone (86.67 % correct) (n = 44). Finally, classification rate for neuropsychological domain scores, SLI, and DHF combined (95.24 % correct) was not significantly different fi-om that for SLI and DHF combined (92.86 % correct) (n = 42). These analyses partially supported Hypothesis 1 1, as the classification rate obtained when using SLI and DHF together was a significant improvement upon that obtained when using SLI as the sole predictor of lateralization of seizure focus. They also support Hypothesis 1 1, in that the best classification rate was obtained when using SLI, DHF, and the five neuropsychological domain scores together as predictors. It should be noted, however, that it was also predicted that combining the five neuropsychological domain scores and SLI would yield a classification rate that was superior to that obtained when using SLI as the sole predictor of lateralization of seizure focus. These findings are also in partial support of Hypothesis 12, in that classification rate for the five neuropsychological variables and that for DHF were slight improvements over the rates obtained when using these as predictors as side of surgery. Despite this, there was no such improvement with regard to using SLI as a predictor of lateralization of seizure focus, and this classification rate was nearly identical to that obtained when using SLI as a predictor of side of surgery. Following is a detailed description of each set of predictors used in these discriminant analyses, the resulting values of Wilk's Lambda, its significance, and the Standardized Canonical Discriminant Function Coefficients for each predictor in each analysis. Again, in situations in which there was only one value (e.g. SLI or DHF) used as a predictor of side of surgery, only Wilk's Lambda and its significance will be reported.

PAGE 103

97 Entering SLI as the sole predictor of seizure focus lateralization yielded the following results: Wilk's Lambda = .292 (p < .001), n = 44. Entering DHF as the sole predictor of seizure focus lateralization yielded the following results: Wilk's Lambda = .426 (g < .001), n = 42. Entering the five neuropsychological domain scores generated fi-om data shown in Table 3 yielded the following results: Wilk's Lambda = .766, (g > .05), Standardized Canonical Discriminant Function Coefficients: LANGSCOR = .758, NVMEMSCOR = -.424, VMEMSCOR = .627, VISCONSCOR = -.475, MOTORSCOR = -.1 19, n = 44. Entering DHF and SLI as predictors of seizure focus lateralization yielded the following results: Wilk's Lambda = .228 (p < .001), Standardized Canonical Discriminant Function Coefficients: SLI = .778, DHF = -.578, n = 42. Entering the five neuropsychological domain scores and SLI as predictors seizure focus lateralization yielded the following results: Wilk's Lambda = .262 (p < .001), Standardized Canonical Discriminant Function Coefficients: LANGSCOR = . 174, NVMEMSCOR = -.206, VMEMSCOR = .260, VISCONSCOR = .081, MOTORSCOR = 064, SLI = .971, n = 44. Entering the five neuropsychological domain scores and DHF as predictors of seizure focus lateralization yielded the following results: Wilk's Lambda = .384 (g < .001), Standardized Canonical Discriminant Function Coefficients: LANGSCOR = -.248, NVMEMSCOR =.311, VMEMSCOR = -.227, VISCONSCOR = .070, MOTORSCOR = .064, DHF = .922, n = 42. Entering the five neuropsychological domain scores, DHF and SLI together as predictors of seizure focus lateralization yielded the following results: Wilk's Lambda = .210 (p < .001), Standardized Canonical Discriminant Function Coefficients: LANGSCOR = .132, NVMEMSCOR = -.233, VMEMSCOR = .198, VISCONSCOR = .121, MOTORSCOR = .026, DHF = -.551, SLI = .779, n = 42.

PAGE 104

98 Convergent vs. Divergent Neuropsychology Group Comparisons (Hypoth eses 13 through 17 and Exploratory Comparisons) As described above, the Convergent Neuropsychology Group represented that portion of the subject population for whom using SLI and DHF combined as predictors in a discriminant analysis yielded the correct predicted side of surgery and the five neuropsychological domains used as predictors in another discriminant analysis were in concordance with this. The Divergent Neuropsychology Group represented that portion of the population for whom a combination of SLI and DHF yielded the correct predicted side of surgery, but the five neuropsychological domains yielded a prediction which was not in agreement with this. Descriptive statistics for these two groups' values on demographic and illness-related variables are shown in Table 10. In keeping with Hypotheses 14 through 18 and the two exploratory hypotheses described above, group means for the following variables were compared using two-tailed t-tests, with an alpha level of .05 for each test: Age at onset of epilepsy. Duration of illness. Education, SLI for the LATL group, SLI for the RATL group, and Engel Classification. Additionally, the remaining variables, which were in the form of group percentages, were tested using chi-square tests, again with an alpha level of .05 for each test. None of the aforementioned tests indicated that there were any significant group differences on any of these demographic and illness-related variables. Therefore, Hypotheses 13 through 17 were not supported by these findings, and the aforementioned exploratory group comparisons yielded nonsignificant differences.

PAGE 105

99 Table 10 Descriptive Statistics for Demographic and Illness-Related Variables: Co nvergent Versus Divergent Neuropsychology Groups. Variable Neuropsychology Neuropsychology Convergent Group Divergent Group S & AO' 11.69 (12.05) 32 12.44 (11.67) 22 Dur. Illness* 22.89 (11.26) 32 27.14 (12.21) 22 Education* 12.53 (2.24) 32 12.82 (2.36) 22 SLl LATL* -.84 (.33) 19 -.81 (.36) 14 SLI RAIL* .78 (.39) 13 .77 (.50) 8 Engel Class.' 1.28 (.52) 32 1.36 (.73) 22 % Engel Class. > l" 25.00 32 27.27 22 % 2""* Gen. Sz's' 75.00 32 59.09 22 % Febrile Sz's" 43.75 32 36.36 22 %HA'' 62.50 32 68.18 22 Note. AO = age at onset of epilepsy; Dur. Illness = duration of illness; SLI LATL = Seizure Lateralization Index (SLI) for subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); SLI RATL = SLI for subjects who eventually underwent right ATL; Engel Class. = Engel Classification following surgery; % Engel Class. > 1 = percentage of each group that had Engel Classification of > 1 following surgery; % 2"^ Gen. Sz's = percentage of each group with a history of secondarily generalized seizures; % Febrile Sz's = percentage of each group with a known history of febrile seizures; % HA = percentage of each group meeting criteria for hippocampal asymmetry (-.45 cm^ > DHF or DHF > +.55 cm^). ' mean group value (standard deviation in parentheses); percentage of each group.

PAGE 106

CHAPTER 4 DISCUSSION Discussion of Experimental and Exploratory Hypotheses Neuropsychological Domain Scores (Hypotheses 1 through 3) Hypothesis 1 was partially supported by the data, in that the LATL group had a significantly lower mean VMEMSCOR than did the RATL group. This is not surprising, as the LATL group demonstrated lower mean scores on CVLT: LDFR and LDCR, two of the five variables included in the calculation of VMEMSCOR. The remaining three variables included in that calculation, LMII, PRINT and CRINT did not have significantly different group means, but the LATL and RATL group means for these variables did differ in the expected direction. The difference in LATL and RATL group means for VMEMSCOR is very consistent with the concept of material specificity, which states that subjects with left hemisphere pathology tend to perform more poorly on verbal memory tasks than do subjects with right hemisphere pathology (Milner, 1967). Hypothesis 1 also included the prediction that the LATL group would have a significantly lower mean LANGSCOR than the RATL group. Although the group means differed in the expected direction, the discrepancy was of only marginal significance. In examining the constituents of LANGSCOR, the LATL group had significantly lower group means for WAIS-R VIQ and BNT, two of the three variables from which LANGSCOR was calculated. There was almost no group difference for COWA, the third variable included in the calculation of LANGSCOR. This is likely due to the fact that there are reasons other than language impairment that may lead to poor COWA performance. This test is often considered to be primarily a test of 100

PAGE 107

101 behavioral fluency, rather than a language test per se. Viewed in this manner, one might expect subjects with frontal pathology to perform poorly on this test in addition to those with left temporal lobe involvement. In essence, it is probable that COWA is sensitive to left temporal lobe seizure involvement, but is not specific to pathology in this area. The fact that the LATL group earned a marginally lower mean LANGSCOR is again consistent with the concept of material specificity described above (Milner, 1967). Hypothesis 2 was not supported by the data in that the RATL group did not earn a significantly lower NVMEMSCOR than the LATL group. Examining the constituents of NVMEMSCOR on a univariate basis reveals that the RATL group earned a slightly lower mean score for RODELAY, and that the two groups performed nearly identically on VRII. This finding is not supportive of the concept of material specificity (Milner, 1967), however, it is consistent with previous findings (Williamson et al., 1993) and is a good example of the fact that there appear to be few neuropsychological tests that are sensitive and specific to right temporal lobe pathology. The data supported Hypothesis 3, as the LATL and RATL groups did not differ significantly on either VISCONSCOR or MOTORSCOR. In the case of VISCONSCOR, the LATL group earned a slightly lower score on two of the three constituents of this score (OA and BD z-scores), while the RATL group earned a lower mean score for the remaining contributor, ROCOPY. These three tests have been traditionally considered to place a premium on the subject's visuoconstructive ability. Although visuoconstructive skills are typically localized to the posterior cortex, and particularly to the parietal lobes, pathology in these regions of either hemisphere may impair performance in this domain (Black & Strub, 1976). Within this context, it is not surprising that a population of subjects with right or left temporal lobe seizure foci would not differ significantly on these tests.

PAGE 108

102 The LATL and RATL groups performed as expected on MOTORSCOR and the tests used to calculate this value. The two groups performed very similarly on both GPT and FTT, for both dominant and nondominant hands. As a whole, the subject population appeared to be significantly more impaired on GPT, a test of motor dexterity, than on FTT, a test of motor speed. As primary motor cortex is located superior to and outside of the temporal lobe, it is not surprising that the LATL and RATL groups did not differ significantly on these tests. Seizure Lateralization Index (SLI) and Difference in Hippocampal Formation (DHF) Values (Hypotheses 4 and 5) As would obviously be expected, the RATL group had a significantly higher mean SLI score, indicating a higher proportion of seizures which were lateraUzed to the right hemisphere, than did the LATL group. Table six reveals that both groups had mean SLI scores which strongly indicated that the majority of seizures were lateralized to the expected hemisphere, and that the distribution of SLI scores was bimodal. Likewise, the data supported Hypothesis 5, which predicted that the RATL group would have a significantly lower mean value for DHF than would the LATL group. As would be expected, this reflects the fact that the RATL group tended to have significantly smaller right than left hippocampi, while the opposite was true of the LATL group. As was shown in Table 6, RATL and LATL mean DHF scores were in the range that would be considered to be significant hippocampal asymmetry, according to criteria used at EMC meetings (-.45 cm^ > DHF indicates a significantly smaller right hippocampus; DHF > +.55cm^ indicates a significantly smaller left hippocampus) (Gilmore et al., 1995).

PAGE 109

103 Correlations Among Seizure Lateralization Index (SLD. Difference in Hi ppocampal Formation (DHF) Values and Neuropsychological Domain Scores (Hypothesis 6) As was predicted in Hypothesis 6, SLI, DHF, and Neuropsychological Domain Scores (represented as a single value by using each subjects' discriminant function score as described above) were significantly correlated with one another, suggesting that all three domains converged with regard to whether subjects demonstrated right or left hemisphere pathology. This correlation was strongest between SLI and DHF, which is not surprising, as these two domains of data are not likely as affected by potential demographic and illness-related sources of variance as are neuropsychological data. The correlations between SLI and neuropsychological domain scores and that between DHF and neuropsychological domain scores were similar and significant, but were much lower than that observed between SLI and DHF. Again, it is surmised that this is due to the fact that neuropsychological data are affected by variations in demography and illnessrelated variables that may be unrelated to neural pathology. Correct Surgery Side Classification Rates (Hypotheses 7 through 10) With regard to the efficacy with which each domain of data and various combinations thereof, were able to be used as accurate predictors of eventual side of surgery, the data were strongly supportive of Hypothesis 7. Used independently to predict side of surgery, SLI and DHF shared similar correct classification rates, with SLI being slightly superior to DHF. These rates were fairly accurate with those reported in past studies with similar subject populations and methods (Walczak et al., 1992; Jack et al., 1989). Both SLI and DHF yielded correct classification rates that were significantly superior to those obtained when neuropsychological domain scores as a predictor of eventual side of surgery.

PAGE 110

104 The fact that neuropsychological domain scores were weaker predictors of eventual side of surgery is not surprising for several reasons. First, as noted above, the current study evaluated the empirical validity of neuropsychological measures rather than how clinicians utilize this information in making diagnostic decisions. The neuropsychological domain scores discussed above were developed according to techniques generally used by neuropsychologists, but it was not possible to create these scores in such a way that they would capture and characterize the individual subject's complex pattern of neuropsychological test scores as effectively as would be done by a skilled neuropsychologist. Furthermore, findings by Williamson et al. (1993) suggested that, even with the added benefit of an actual neuropsychologist's interpretation, neuropsychological data were approximately 73 percent successfiil in lateralizing seizure foci. This figure represents an increase of thirteen percentage points in classification rate over that obtained in the current study. Finally, as will be discussed below, there was the possibility that the neuropsychological data were less effective in predicting side of surgery than were SLI and DHF, due to the fact that neuropsychological data are subject to demographic and illness-related sources of variance which do not interfere with EEG and MRI-based measures. A closer examination of the neuropsychological domain scores and the relative contribution of each to the prediction of a given subject's eventual side of surgery can be carried out by examining the standardized canonical discriminant function coefficients for each domain score. These are the standardized weights that the analysis placed on each predictor variable in order to yield the maximal correct surgery side classification rate. As would be expected in light of the above discussion, VMEMSCOR was assigned the greatest weight, as the LATL and RATL group mean scores for this value were significantly different from one another. LANGSCOR was assigned a moderate weight, and VISCONSCOR and MOTORSCOR were assigned very low weights. Surprisingly, however, the second highest weight, just slightly below that assigned to

PAGE 111

105 VMEMSCOR, was assigned to NVMEMSCOR. This domain score did not show significant LATL and RATL group differences when it, and its constituents, were examined univariately. From that standpoint, NVMEMSCOR would appear to be relatively meaningless with regard to one's attempt to predict a subject's eventual side of surgery. When placed in the context of the rest of the neuropsychological data, however, it became an important contributor to this prediction. This reflects nicely the fact that neuropsychological test scores must be interpreted within the context of one another, and not in isolation from other variables. Hypothesis 8 was partially supported by the data. As expected, combining SLI and DHF as predictors of side of surgery significantly increased the correct classification rate that had been obtained by using SLI alone. The standardized canonical discriminant function coefficients obtained from the combined SLI/DHF analysis indicate, as would be expected from earlier analyses, that a stronger weight was assigned to SLI than DHF, but both predictors were of significant importance in the analysis. The finding that using a combination of neuropsychological domain scores and SLI did not significantly improve upon the classification rate obtained when SLI was used alone was not expected. In this analysis, the standardized canonical discriminant function coefficients indicated that SLI had, by far, the strongest weight assigned to it. The neuropsychology LANGSCOR was assigned a moderate weight while the other domain scores played minimal roles in the prediction of side of surgery. Although the inability of neuropsychological data to add significantly to the classification rate of SLI was not hypothesized, it is consistent with the findings above which suggest that these data were a weaker predictor of side of surgery. It was also hypothesized that using neuropsychological domain scores, SLI, and DHF together would yield the maximal correct classification rate. However, this combination of predictors produced a classification rate that was identical to that obtained by using just SLI and

PAGE 112

106 DHF together. Possible reasons for the inability of neuropsychological data to add to SLI and DHF were addressed later in the study and will be discussed below. Hypothesis 9 was not supported by the data, as neuropsychological domain scores, when used as predictors of eventual side of surgery, yielded nearly identical correct classification rates for both LATL and RATL groups. This result conflicts with the findings of Williamson et al. (1993) who reported that of the nine cases that were incorrectly lateralized in their study, all involved right hemisphere seizure foci. Again, that study included the actual interpretation of skilled neuropsychologists, and did not rely solely upon the data to predict side of surgery or seizure focus. As predicted in Hypothesis 10, when SLI was used as a predictor of side of surgery, classification rates were nearly identical for the LATL and RATL groups. This is not surprising, as there is no known reason to assume that EEG is more sensitive or specific to seizure activity in either the left or right hemisphere. There was also no significant difference between LATL and RATL correct classification rates when DHF was used as the sole predictor of side of surgery, although the rate was slightly lower for the RATL group than for the LATL group. As mentioned above, the EMU criteria for significant hippocampal asymmetry takes into account the assumption that, on average, the right hippocampus will be approximately .10 cm^ larger than the left in non-neurological subjects. As was shown in Table 4, the left and right mean hippocampal volumes in this population were nearly identical. A closer examination of these data shows, however, that the LATL group tended to have slightly larger hippocampi in the surgical hemisphere than did the RATL group. Furthermore, examining the hippocampus contralateral to surgery, or the presumably "healthy hippocampus" a similar pattern is seen. This would suggest that it may be ill-advised to apply a DHF distribution obtained fi-om non-neurological subjects that is skewed toward larger values for

PAGE 113

107 the right hippocampal volume, to a population of subjects with temporal lobe epilepsy. If the assumption that the right hippocampus is normally larger than the left in epilepsy patients was unfounded yet was continued to be used, the current criteria for significant hippocampal asymmetry (-.45 cm^> DHF or DHF > +.55cm^) (Gilmore et al., 1995) would lead to an erroneously inflated rate of suspected left hippocampal sclerosis and a decrease in the number of cases meeting criteria for suspected right hippocampal sclerosis. Correct Seizure Focus Lateralization Classification Rates (Hypotheses 1 1 and 12) To this point, the discussion has focused mainly on the usefulness of neuropsychological, EEG, and MRI data in predicting eventual side of surgery. While this proved to be an interesting investigation, confining the study to this topic would limit its scientific utility for two important reasons. First, it cannot necessarily be assumed that predicted side of surgery and actual side of seizure focus are one and the same. Although great care is taken to select only the best candidates for surgery, and the rate of surgical success is high, it would be naive to assume that some percentage of patients do not have epileptogenic regions in addition to, or separate fi-om, the suspected epileptogenic region identified during Phase I evaluation. Second, using EEG, MRI, and neuropsychological data only to predict eventual side of surgery would limit the study to an investigation of which sources of data the involved clinicians rely upon most strongly when making decisions regarding side of surgery. Hypothetically, Data Domain A could be one hundred percent accurate in predicting side of seizure focus. However, if the involved clinicians commonly placed little confidence in Data Domain A and instead looked to other predictors to make surgical decisions. Data Domain A would remain, in fact, a valuable indicator of seizure focus but would be a very poor predictor of side of surgery.

PAGE 114

108 For these reasons, it was of utmost importance to select out only those subjects who had Engel Classifications of I following surgery and to re-run the above discriminant analyses on this subset of the population. Again, these subjects were seizure-free at the time of outcome determination, and it can therefore be reasonably assumed that their actual seizure foci had been resected during surgery. Thus it was surmised that using various data to predict side of surgery for these subjects would be tantamount to using the data to actually lateralize seizure foci. The data were supportive of Hypothesis 1 1, which predicted that relative to one another, neuropsychological domain scores, SLI and DHF would show a similar pattern of efficacy in lateralizing seizure foci as when used to lateralize side of surgery. The correct seizure lateralization classification rate obtained by using SLI as the sole predictor was slightly better than that obtained by using DHF as the sole predictor, and both were significantly superior to the rate yielded when entering neuropsychological domain scores as predictors. This pattern of performance and significance was identical to that found when using these data to predict side of surgery. As had been found earlier, DHF and SLI used in combination yielded a significantly higher correct lateralization classification rate than that obtained when using SLI alone. Neuropsychological domain scores used in combination with SLI improved upon the rate obtained when SLI was used alone, but not significantly. Finally, using neuropsychological domain scores in combination with SLI and DHF as predictors added very slightly to the classification rate obtained by SLI and DHF together, but this difference was nonsignificant. With regard to Hypothesis 12, it was found that using SLI as the sole predictor of lateralization of seizure focus yielded a correct classification rate which was nearly identical to that obtained when using SLI to predict side of surgery. In the case of DHF and the neuropsychological domain scores, however, there was a slight improvement in classification rate

PAGE 115

109 when these data were used independently to lateraHze seizure foci. This was expected, as the subset of subjects selected for these analyses presumably had relatively discrete seizure foci, and it was hypothesized that they would therefore be lateralized more readily by the data. It is not clear why the classification rate obtained using SLI as the sole predictor of seizure focus lateralization was not slightly improved during this phase of the study. Convergent Versus Divergent Neuropsychology Group Comparisons (Hypotheses 13 through 17) The data were not supportive of Hypothesis 13, which predicted that RATL subjects in the Divergent Neuropsychology Group would have a mean SLI value that was significantly closer to zero than the RATL subjects in the Convergent Neuropsychology Group. The same was predicted for the LATL subjects in each group. This was based on the assumption that subjects in the Divergent Neuropsychology Group would have a pattern of neuropsychological data that was suggestive of a relatively widespread pattern of cerebral seizure involvement. It was also predicted that these subjects would have SLI values that were closer to zero, indicating the presence of nonlateralized seizures and/or some events lateralized to each hemisphere. Although the Divergent Neuropsychology Group did have mean SLI values that were closer to zero than the Convergent Neuropsychology Group, this difference was minimal and probably of no scientific import. The specific reason for the nonsupport of Hypothesis 13 is unclear, but it would appear that there are several possible alternatives. First, although SLI appears to be a useful indicator of the lateralization of a given seizure, its value does not speak to the degree of neural spread following seizure onset. For example, a seizure may be very clearly lateralized to the left hemisphere according to the EEG readout and subsequent dictation, however, this seizure may have also involved secondary generalization which resulted in widespread seizure activity. In this

PAGE 116

110 example, the seizure event would be coded as having been lateralized to the left hemisphere and it would serve to push that subject's SLI value toward -1 .00, despite the fact that a widespread brain area was truly involved in the seizure. On the contrary, there were almost certainly seizure events which were interpreted as nonlateralized, but were in actuality had focal onset and remained relatively focal which became distorted through the various forms of EEG interference discussed earlier. Such an event would serve to push that subject's SLI value toward zero, when in fact the event was quite focal. Thus, even if the goal of dividing subjects into Neuropsychology Convergent and Divergent groups was met and this difference was reflected in the degree to which each group's seizure activity was widespread in the brain, it may have been incorrect to assume that this would have been reflected as a difference in mean SLI values between the groups. Hypothesis 14 predicted that a higher proportion of subjects in the Divergent Neuropsychology Group would have less than optimal surgery outcome (Engel Classification > I), based on the assumption that their divergent neuropsychology data was indicative of more widespread seizure involvement than were the same data from the Neuropsychology Convergent Group. This hypothesis was not supported by the data, as the two groups had nearly identical proportions of subjects who had Engel Classifications that were greater than I. Although the Divergent Neuropsychology Group may have had more widespread patterns of seizure involvement, it may have been incorrect to assume that this was the equivalent of having widespread areas of seizure genesis. That is, simply because a subject may have diffuse seizure involvement which may or may not be reflected in a similarly diffuse pattern of neuropsychological impairment, it does not necessarily follow that the subject would fare more poorly from a relatively small resection such as that involved in ATL.

PAGE 117

Ill Hypothesis 15 addressed the potential role of secondarily generalized seizures in the production of divergent neuropsychological data. Again, it was predicted that subjects with a history of secondarily generalized seizures would have an associated pattern of neuropsychological performance that was suggestive of relatively widespread neural impairment, and that this would result in an increased incidence of mislateralization when neuropsychological data was used to predict side of surgery. Surprisingly, this hypothesis was not supported by the data, and in fact, the Convergent Neuropsychology Group had a higher proportion of subjects with a history of secondarily generalized seizures than did the Divergent Neuropsychology Group. Although it appears unlikely that different methodology would lead to the support of this hypothesis, it should be noted that the methodology used to investigate this hypothesis in the current study was not ideal. Subjects were coded only with regard to whether or not they had a history of secondarily generalized seizures. Unfortunately, this methodology does not address the number or frequency of secondarily generalized seizures that a given subject may have experienced. Should that information become available in the future, it may be useful to investigate the relationship between number and frequency of secondarily generalized seizures and the existence of divergent neuropsychological data. Finally, it also possible that neuropsychological data are not sensitive to the presumed widespread cerebral insuh that is suffered during a secondarily generalized seizure. Interestingly, Hypothesis 16 was also not supported by the data. It was predicted that subjects with no evidence of hippocampal asymmetry would be more likely to be in the Divergent Neuropsychology Group, because those with significant hippocampal asymmetry would be more likely to have concomitant focal findings on both EEG and neuropsychological tests. Although the proportions were very similar, the Neuropsychology Divergent Group actually had a slightly higher proportion of subjects with significant hippocampal asymmetry. Apparently, the degree to

PAGE 118

112 which neuropsychological domain scores used in the study were convergent or divergent with other data was independent of the existence of hippocampal asymmetry. Finally, Hypothesis 17 was also not supported by the data. As described above (Williamson et al., 1993), it has been reported that it is generally more difficult to use neuropsychological data to identify right hemisphere seizure involvement than that occurring in the left hemisphere. Therefore, it was hypothesized in the current study that the Divergent Neuropsychology Group would contain a higher proportion of subjects who received RATL than the Convergent Neuropsychology Group. This was not the case, as both groups shared nearly identical proportions of subjects who received RATL and LATL. Exploratory Comparisons of Convergent vs. Divergent Neuropsychology Groups The Convergent and Divergent Neuropsychology Groups shared very similar mean values for age at onset of epilepsy and duration of illness. In addition, the proportions of each group having had a known history of febrile seizures were not significantly different. It was decided to approach these comparisons in an exploratory manner, as it was unclear whether there was reason to believe that these variables would affect neuropsychological performance in any consistent manner across subjects. Assuming that seizures damage the brain, it was suspected that an earlier age at onset and febrile seizures may lead to more focal and specific neuropsychological profiles, as the subject's brain would have suffered this neuropsychological insult during brain, and particularly hippocampal, development. On the contrary, however, it could be argued that having an early age at onset and/or febrile seizures would lead to unusual functional organization of the brain, and neuropathology in a given area such as the left hippocampus may not lead to the neuropsychological deficits that one would typically expect of a subject with such pathology.

PAGE 119

113 It was generally assumed that subjects with a longer duration of illness would have had more seizures, and that these individuals would therefore tend to have more difftise neuropsychological impairment. While this led the examiners to suspect that perhaps the Divergent Neuropsychology Group would have a greater mean duration of illness, this variable is obviously highly correlated with age at onset of epilepsy. Thus, this again presented the confusing situation described above, in which it could be argued that early age at onset could produce either more focal or more diffuse neuropsychological profiles for the aforementioned reasons. Nonetheless, it would appear that age at onset, history of febrile seizures, and duration of illness were not significant factors in producing neuropsychological profiles that were either convergent or divergent with EEG and MRI data. Finally, the Convergent and Divergent Neuropsychology Groups shared very similar values for mean years of education. It was unclear whether or not this variable would have an effect on the convergence or divergence of neuropsychological data with that fi^om other domains, and these data would suggest that it does not. General Discussion Taken together, the findings in the current study suggest several points which may be misinterpreted if not considered carefully. First, the findings tend to depict neuropsychological data as having little value in the lateralization of seizure foci. Second, EEG data appear to be extremely useful and accurate in this pursuit. Finally, MRI data appear to be of moderate value. In the case of the neuropsychological data, the current study's methodology must be taken into account before assuming that these data are of little use. Although the constituents of each domain score were chosen carefully and with regard to actual clinical procedure, the study did not seek to continuously reformulate and evaluate each domain score until it produced the

PAGE 120

114 highest correct lateralization rate possible. Thus, it is possible and even probable that domain scores may have been created from the existing data that would have yielded higher correct lateralization rates than were obtained in the current study. Assuming for a moment that the domain scores that were created were ideal, they did not take into account the full complexity that is involved when a neuropsychologist conceptualizes a given subject's pattern of test scores. Such conceptualization is both an art and a science, and the methodology used in the current study removed virtually all of the art from this process. The Brunswdk (1955) Lens Model of Probabilistic Functionalism may help to illustrate this point. This model discriminates between the empirical validity of data (i.e. to what degree data alone can indicate a correct clinical decision), from the manner in which clinicians actually consider the data and use judgement and expertise to render an opinion based upon it. The current study was more the former, as it investigated the empirical validity of neuropsychological measures, and not how, or to what degree of success these values are used by actual chnicians. It must also be considered that the current methodology forced the neuropsychological data to make a left or right decision regarding seizure focus. In actuality, it is almost certain that some portion of the cases that were mislateralized using neuropsychological data would have been considered bilateral or nonlateralized by an actual neuropsychologist, and would therefore not have been truly mislateralized. While the current methodology has the advantage of having treated each subject's data in a very consistent manner, it ahnost certainly resulted in an underestimation of the value of neuropsychological data in lateralizing seizure foci. The manner in which the Convergent and Divergent Neuropsychology Groups were formed must also be considered. Although the methodology used in the current study effectively separated those cases in which neuropsychological data was in agreement with the other domains of data with regard to side of surgery from those in which neuropsychological data was divergent.

PAGE 121

115 there is at least one other way in which this could have been done. Instead of simply selecting out those subjects for whom neuropsychological data actually mislateralized the side of surgery relative to the other data domains, it may have been more effective to examine the discriminant function scores obtained through the use of neuropsychological domain scores as predictors, and to explore the degree to which they were convergent or divergent with SLI and DHF. The degree of convergence or divergence among these domains of data may have varied significantly even in cases in which they ultimately predicted the same side of surgery. For example, Subject A's neuropsychological data may have yielded a discriminant function score that just minimally met criteria for lateralization that was divergent with that obtained from SLI and DHF. Subject B's data, however, may have been in strong disagreement with SLI and DHF with regard to lateralization. The current methodology did not capitalize on this potential variance in the formation of the Convergent and Divergent Neuropsychology Groups as convergence and divergence were treated as a dichotomous, rather than continuous variable. Representing the concept of convergence and divergence as a continuous variable may allow the investigator to identify the sources of variance in neuropsychological data, a pursuit which led to disappointing results in the current study. Perhaps in future studies it would also be useful to form a third group for those subjects whose neuropsychological data was neither truly convergent or divergent with SLI and DHF. Despite the negative findings described above, it would be naive to assume that demographic and illness-related variables do not serve to increase the variance in neuropsychological performance. It is possible that such factors did play such a role in the current study, but not in a manner that was consistent enough across subjects to emerge as a main effect. Perhaps the use of a technique such as path analysis, along with the concept of convergence/divergence represented as a continuous variable would shed more light on the

PAGE 122

116 relationship between demographic and illness-related variables and neuropsychological performance. As is the case in neuropsychology, the clinical interpretation of EEG data can be accurately characterized as a combination of art and science. The neurologist must make sense of differing patterns of neural activity recorded through many electrodes arranged in several different montages. He or she must also contend with the varying degrees of distortion and artifect caused by factors described above. It is important to note that the methodology used to process EEG data in the current study included the art, science and fmal interpretation of actual neurologists, as dictations of seizure events were used to lateralize seizures. Raw EEG data were not used independently of clinical interpretation as was the case with neuropsychological data. This fact likely led to a situation in which the usefiilness of EEG data was somewhat inflated relative to that of neuropsychological data. With regard to the methodology used to process MRJ data in the current study, it is probable that the findings described above do not fully characterize the value of hippocampal volumetrics in the lateralization and localization of seizure foci. In actual clinical practice, the neurologist considers whether the subject's MRI does or does not contain evidence of significant hippocampal asymmetry. The current study merely examined any degree of asymmetry between the two hippocampi and forced the data to predict seizure focus lateralization in many cases in which the degree of asymmetry was so small that it would not have been used to make lateralization decisions in actual practice. Including only those subjects whose MRI's revealed evidence of hippocampal asymmetry would certainly have increased the correct lateralization rate obtained when using DHF as a predictor. In all likelihood, this correct lateralization rate would have approached 100 percent.

PAGE 123

117 Directions for Future Research The current study was somewhat broad and involved a relatively large combination of variables including neuropsychological, EEG, MRI, demographic, and illness-related data. Although seventeen experimental hypotheses and four exploratory comparisons were addressed, these certainly did not constitute an exhaustive list of the questions that could have been asked of these data. It is planned that future investigations will involve these and additional data in an effort to clarify some of the issues that the current study left unresolved. It would be particularly interesting to conduct further investigation of the current neuropsychological domain scores in order to determine the best possible combination of constituents to predict seizure focus lateralization. An idea of which variables may be most helpful could be obtained by examining more closely the standardized canonical discriminant function coefficients outlined in the above Results section. Using various combinations of those variables that appear most useful in predicting lateralization and then testing the resulting correct classification rates would be a relatively straightforward endeavor. In order to obtain a more ecologically valid estimate of the value of neuropsychological data in an epilepsy population, it would be necessary to have neuropsychologists examine each subject's data and render an interpretation of it with regard to seizure lateralization. This could be easily accomplished by forming summary sheets that contained each subject's neuropsychological data and a brief paragraph containing his or her medical history. These sheets would be circulated among three or four neuropsychologists, with each indicating his or her conceptualization of seizure focus lateralization by drawing an X along a line ranging from strong right lateralization to strong left lateralization. This would yield a neuropsychological

PAGE 124

118 lateralization index, a continuous variable that would be analogous to SLI and DHF, that would allow for investigation of the value of neuropsychological data as used by the clinician. As the current study focused solely on the laterali2ation of seizure foci, it would also be possible to examine the utility of neuropsychological, EEG, and MRI data in the determination of seizure focus localization. Regarding localization, the epilepsy management team is first concerned with whether the seizure focus is temporal or extratemporal. Second, if the focus is temporal, an attempt is made to determine whether it is medially or laterally situated within the temporal lobe. It would be necessary to obtain additional subjects, specifically some with extratemporal seizure foci, in order to obtain a sufficiently diverse research population for this investigation, but the actual methodology used in the current study would not have to be altered significantly. The main differences would be the development of a seizure localization index based on EEG data that would be indicative of a given subject's pattern of temporal versus extratemporal seizures. The neuropsychological domain scores used in the current study may require some altering in order to make them as specific as possible, but in general, the memory and language scores would be sensitive mainly to temporal lobe involvement, while the motor and visuoconstructive scores would be sensitive to extratemporal pathology. The hippocampal volumetric data could likely be kept in its original form. It remains the principal investigator's strong opinion that neuropsychology plays an important role in epilepsy programs. Neuropsychological data are not only useful for lateralization and localization of seizure foci, they also aid in the prediction of post-surgical cognitive fiinctioning. Furthermore, the neuropsychologist can provide effective cognitive rehabilitation and psychosocial counseling both preand post-surgery. It is essential that rigorous scientific investigations continue to focus on neuropsychological functioning in epilepsy so that

PAGE 125

we may continue to provide ever-improving care to the millions who suffer from seizure disorders.

PAGE 126

APPENDIX Missing Variables and the Predictors used in Multiple Regression Equations to Generate Predicted Values to Replace them Missing Number Missing Predictors ^ Variable LATL RATL LATL RATL BNT 7 1 FSIQ .47 .32 VIQ COWA LDFR SDFR COWA 0 1 FSIQ .52 VIQ PIQ BNT LDFR LMI 1 I FSIQ .70 .45 VIQ TOTAL SDCR LDFR 120

PAGE 127

121 Appendix Table Continued Missing Number Missing Variable LATL RATL Predictors LATL RATL LMII VRI VRII FTTDOM VIQ TOTAL SDCR LDFR LDCR FSIQ VIQ PIQ ROIMM RODELAY FSIQ PIQ TOTAL ROIMM RODELAY FSIQ TOTAL LDFR LMI GPTDOM .24 .61 .28 .40 .23 .60 .70 .62

PAGE 128

122 Appendix Table Continued Missing Number Missing Predictors Variable LATL RAIL LATL RATX FTTNDOM 3 1 COWA .15 .46 TOTAL TRIAL 5 SDCR LDFR GPTDOM 2 0 FSIQ .45 PIQ TMTA TMTB FTTDOM GPTNDOM 2 0 PIQ .46 BEST TOTINT VRI TMTA Note. All = all subjects; LATL = subjects who eventually underwent left Anterior Temporal Lobectomy (ATL); RATL = subjects who eventually underwent right ATL; BNT = Boston Naming Test z-score; COWA = Controlled Oral Word Association percentile score; LMI and LMII = Wechsler Memory Scale (WMS-R) Logical Memory I and II percentile scores; VRI and VRII = WMS-R Visual Reproduction I and II percentile scores; FTTDOM and FTTNDOM = Finger Tapping Test z-scores for dominant and nondominant hands, respectively; GPTDOM and GPTNONDOM = Grooved Pegboard Test z-scores for dominant and nondominant hands.

PAGE 129

123 Appendix Table Continued respectively; FSIQ = Wechsler Adult Intelligence Scale-Revised (WAIS-R) Full Scale IQ; PIQ = WAIS-R Performance IQ; VIQ = WAIS-R Verbal IQ; TRIAL 5 = California Verbal Learning Test (CVLT) Trial 5 z-score; TOTAL = CVLT Total T-score; SDFR = CVLT Short Delay Free Recall z-score; SDCR = CVLT Short Delay Cued Recall z-score; LDFR = CVLT Long Delay Free Recall z-score; LDCR = CVLT Long Delay Cued Recall z-score; TOTINT = CVLT Total Intrusions z-score; ROIMM and RODELAY = Rey-Osterreith Immediate and Delayed Recall scaled scores, respectively; TMT A and TMT B = Trailmaking Test (TMT) Part A and B zscores, respectively; Predictors = variables entered into multiple regression equations to generate predicted values to replace missing values; R^ = multiple correlation squared for these multiple regression equations.

PAGE 130

REFERENCES Anderson, S. W., Damasio, H., Jones, R. D., & Tranel, D. (1991). Wisconsin card sorting test performance as a measure of frontal lobe damage. Journal of Clinical and Experimental Neuropsychology. 13 (6), 909-922. Andreasen, N. C. (1988). Nuclear magnetic resonance imaging. In N. C. Andreasen (Ed.), Brain imaging: Applications in psychiatry (pp. 67-121). Washington, D.C.: American Psychiatric Press, Inc. Baddeley, A. D., & Warrington, E. K. (1970). Amnesia and the distinction between longand short-term memory. Journal of Verbal Learning and Verbal Behavior. 9. 176-189. Barry, E., Sussman, N. M., O'Connor, M. J., & Hamer, R. N. (1992). Presurgical electroencephalographic patterns and outcome from anterior temporal lobectomy. Archives of Neurology. 49. 21-27. Bauer, R. M., Breier, J. I., Gilmore, R., Crosson, B., Fennell, E. B., & Roper, S. (1995). Neuropsychological functioning before and after unilateral temporal lobectomy for intractable epilepsy. Manuscript submitted for publication. Bauer, R. M., Tobias, B., and Valenstein, E. (1993). Amnesic Disorders. In K. M. Heilman and E. Valenstein (Eds.), Clinical neuropsychology (3rd ed., pp. 523-602). New York: Oxford University Press. Bengzon, A. R. A., Rasmussen, T., Gloor, P., Dussault, J., & Stephens, M. (1968). Prognostic factors in the surgical treatment of temporal lobe epileptics. Neurology. 18 (8), 717731. Benton, A., & Hamsher, K. (1989). Multilingual aphasia examination: Second edition . Iowa City: A.J.A. Associates. Berg, E. A. (1948). A simple, objective technique for measuring flexibility in thinking. Journal of General Psychology. 39. 15-22. Black D. W. and Strub, R. L. (1976). Constructional apraxia in patients with discrete missile wounds of the brain. Cortex. 12. 212-220. Bomstein, R.A. (1985). Normative data on selected neuropsychological measures from a nonclinical sample. Journal of Clinical Psychology. 41. 651-659. Bronen, R. A., Anderson, A. W., & Spencer, D. D. (1994). Quantitative MRI for epilepsy: A clinical and research tool? American Journal of Neuroradiology. 15. 1 157-1 160. 124

PAGE 131

125 Brunswik, E. (1955). Representative design in probabilistic theory. Psychological Review. 62. 236-242. Cascino, G. D., Jack, C. R., Parisi, J. E., Sharbrough, F. W., Hirschom, K. A., Meyer, F. B., Marsh, W. R., & O'Brien, P. C. (1991). Magnetic resonance imaging-based volume studies in temporal lobe epilepsy: Pathological correlations. Aimals of Neurologv. 30 (1), 31-36. Chelune, G. J. (1991). Using neuropsychological data to forecast postsurgical outcome. In H. Luders (Ed ), Epilepsy surgery (pp. 477-485). New York: Raven Press. Chelune, G. C, Awad, I., & Luders, H. (1989). Verbal memory deficits after temporal lobectomy: Independent or confounded by language [Abstract]. Epilepsia. 30 (5), 712. Commission on Classification and Terminology of the International League Against Epilepsy. (1964). A proposed international classification of epileptic seizures. Epilepsia. 5. 297306. Commission on Classification and Terminology of the International League Against Epilepsy. (1981). Proposal for revised clinical and electroencephalographic classification of epileptic seizures. Epilepsia. 22. 489-501. Corcoran, R., & Upton, D. (1993). A role for the hippocampus in card sorting? Cortex. 29 (2), 293-304. Delaney, R. C, Rosen, A. J., Mattson, R. H., & Novelly, R. A. (1980). Memory fimction in focal epilepsy: A comparison of non-surgical, unilateral temporal lobe and fi-ontal lobe samples. Cortex. 16. 103-117. Delis, D., Kramer, J., Kaplan, E., Ober, B. A., & Fridlund, A. (1987). The California verbal learning test . New York: Psychological Corporation. Dodrill, C. B., Wilkus, R. J., Ojemann, G. A., Ward, A. A., Wyler, A. R., van Belle, G., & Tamas, L. (1986). Multidisciplinary prediction of seizure relief from cortical resection surgery. Aimals of Neurology. 20. 2-12. Engel, J. (1987). Outcome with respect to epileptic seizures. In J. Engel (Ed.), Surgical treatment of the epilepsies (pp.553-571). New York: Raven Press. Engel, J. (1989). Seizures and epilepsy . Philadelphia: FA Davis Company. Engel, J., Van Ness, P. C, Rasmussen, T. B., & Ojemann, L. M. (1993). Outcome with respect to epileptic seizures. In J. Engel (Ed ), Surgical treatment of the epilepsies (2nd ed., pp. 609-621). New York: Raven Press. Fedio, P., & Mirsky, A. F. (1969). Selective intellectual deficits in children with temporal lobe or centrencephalic epilepsy. Neuropsychologia. 7. 287-300.

PAGE 132

126 Frisk, v., & Milner, B. (1990). The role of the left hippocampal region in the acquisition and retention of story content. Neuropsychologia. 28 (4), 349-359. Gastaut, H., Gastaut, J. L., Goncalves e Silva, G. E., & Fernandez-Sanchez, G. R. (1975). Relative frequency of different types of epilepsy: A study employing the classification of the International League Against Epilepsy. Epilepsia. 16 (3), 457-461. Gilmore, R. L., Childress, M. D., Leonard, C, Quisling, R., Roper, S., Eisenschenk, S. & Mahoney, M. (1995). Hippocampal volumetrics differentiate patients with temporal lobe epilepsy and extratemporal lobe epilepsy. Archives of Neurology. 52 (8), 819-824. Glowinski, H. (1973). Cognitive deficits in temporal lobe epilepsy: An investigation of memory fimctions. The Journal of Nervous and Mental Disease. 157 (2), 129-137. Grant, D. A. & Berg, E. A. (1948). A behavioural analysis of degree of reinforcement and ease of shifting to new responses in a Weigl-type card sorting problem. Journal of Experimental Psvcholoev. 38. 404-41 1. Gray, J. A. (1982). The neuropsychology of anxiety . Oxford: Oxford University Press. Heaton, R. K. (1981). Wisconsin card sorting test manual . Odessa, FL: Psychological Assessment Resources, Inc. Heilman, K. M., Wilder, B. J., & Malzone, W. F. (1972). Anomic aphasia following anterior temporal lobectomy. Transactions of the American Neurological Association. 97. 291293. Hermaim, B. P., & Wyler, A. R. (1988). Effects of anterior temporal lobectomy on language fiinction: A controlled study. Annals of Neurology. 23 (6), 585-588. Hermann, B. P., Wyler, A. R., & Richey, E. T. (1988). Wisconsin card sorting test performance in patients with complex partial seizures of temporal lobe origin. Journal of Clinical and Experimental Neuropsychology. 10 (4), 467-476. Hermann, B. P., Wyler, A. R, Richey, E. T., & Rea, J. M. (1987). Memory fimction and verbal learning ability in patients with complex partial seizures of temporal lobe origin. Epilepsia. 28 (5), 547-554. Ivnik, R. J., Sharbrough, F. W., & Laws, E. R. (1988). Anterior temporal lobectomy for the control of partial complex seizures: Information for counseling patients. Mayo Clinic Proceedings. 63. 783-793. Jack, C. R., Sharbrough, F. W., Twomey, C. K., Cascino, G. D., Hirschom, K. A., Marsh W. R, Zinsmeister, A. R., & Scheithauer, B. (1990). Temporal lobe seizures: Lateralization with MR volume measurements of the hippocampal formation. Radiology. 175. 423-429.

PAGE 133

127 Jack, C. R., Twomey, C. K., Zinsmeister, A. R., Sharbrough, F. W., Petersen, R. C, & Cascino, G. D. (1989). Anterior temporal lobes and hippocampal formations: Normative volumetric measurements in young adults. Radiology. 172 (2), 549-554. Jasper, H. (1958). Report on the committee on methods of clinical examination in electroencephalography. Electroencephalography and Clinical Neurophysiology. 10. 370-375. Jones-Gotman, M. (1986). Memory for designs: The hippocampal contribution. Neuropsychologia. 24. 193-203. Jones-Gotman, M. (1986). Right hippocampal excision impairs learning and recall of a list of abstract designs. Neuropsychologia. 24. 659-670. Jones-Gotman, M. (1987). Commentary: Psychological evaluation-testing hippocampal function. In J. Engel, Jr. (Ed.), Surgical treatment of the epilepsies (pp. 203-21 1). New York: Raven Press. Jones-Gotman, M. (1991). Localization of lesions by neuropsychological testing. Epilepsia. 32 (Suppl. 5), S41-S52. Jones-Gotman, M., So, N., Andermann, F., Gloor, P., Olivier, A., & Quesney, L. F. (1989). Memory and cognition in bitemporal epileptic patients undergoing depth electrode studies [Abstract]. Epilepsia. 30. 713. Kaplan, E., Goodglass, H., & Weintraub, S. (1983). The Boston naming test . Philadelphia: Lea & Febiger. Kuzniecky, R. I., & Jackson, G. D. (1995). Magnetic resonance in epilepsy . New York: Raven Press. Ladavas, E., Umilta, C, & Provinciali, L. (1979). Hemisphere-dependent cognitive performances in epileptic patients. Epilepsia. 20. 493-502. Lencz, T., McCarthy, G., Bronen, R. A., Scott, T. M., Insemi, J. A., Sass, K. J., Novelly, R. A., Kim, J. H., & Spencer, D. D. (1992). Annals of Neurology. 31 (6), 629-637. Lieb, J. P., Walsh, G. 0., Babb, T. L., Walter, R. D., & Crandall, P. H. (1976). A comparison of EEG seizure patterns recorded with surface and depth electrodes in patients with temporal lobe epilepsy. Epilepsia. 17. 137-160. Locharemkul, C, Primrose, D., Pilcher, W., Ojemann, L. M., & Ojemann, G. A. (1992). Update in epilepsy: Part I: Diagnosis and treatment of epilepsy. New York State Journal of Medicine. 92 m. 14-17. Lothman, E. W. (1991). Functional anatomy: A challenge for the decade of the brain Epilepsia. 32 (Suppl. 5), S3-S13.

PAGE 134

128 Margerison, J. H., & Corsellis, J. A. N. (1966). Epilepsy and the temporal lobes: A clinical, electroencephalographic and neuropathological study of the brain in epilepsy, with particular reference to the temporal lobes. Brain. 89. 499-530. Mayeux, R. Brandt, J., Rosen, J., & Benson, D. F. (1980). Interictal memory and language impairment in temporal lobe epilepsy. Neurology. 30. 120-125. Mcintosh, G. C. (1992). Neurological conceptualizations of epilepsy. In T. L. Bennett (Ed.) The neuropsychology of epilepsy (pp. 17-37). New York: Plenum Press. Milner, B. (1967). Brain mechanisms suggested by studies of the temporal lobe. In F. Darly (Ed ), Brain mechajiisms underlying speech and language (pp. 122-145). New York: Grune and Stratton. Milner, B. (1971). Interhemispheric differences in the localization of psychological processes in man. Brain Medical Bulletin. 27. 272-277. Mungas, D., Ehlers, C, Walton, N., & McCutchen, C. B. (1985). Verbal learning differences in epileptic patients with left and right temporal lobe foci. Epilepsia. 26 (4), 340-345. Novelly, R. A., Augustine, E. A., & Mattson, R. H. (1984). Selective memory improvement and impairment in temporal lobectomy for epilepsy. Aimals of Neurology. 15 (1), 64-67. Osterrieth, P. A. (1944). Le test de copie d'une figure complexe. Archives de Psychologic. 30. 206-356. Petrides, M., & Mibier, B. (1982). Deficits on subject-ordered tasks after fi-ontaland temporal-lobe lesions in man. Neuropsvchologia. 20. 249-262. Pilcher, W. H., Locharemkul, C, Primrose, D., Ojemann, L. M., & Ojemann, G. A. (1992). Update in epilepsy. III. Surgical therapy of intractable epilepsy. New York State Journal of Medicine. 92. 92-96. Rausch, R. (1987). Psychological evaluation. In J. Engel, Jr. (Ed.), Surgical treatment of the epilepsies (pp. 18 1-195). New York: Raven Press. Rausch, R., & Babb, T. L. (1987). Evidence for memory specialization within the mesial temporal lobe in man. In J. Engel, Jr. (Ed.), Fundamental mechanisms of human brain fimction (pp. 103-109). New York: Raven Press. Ray, W. (1990). The electrocortical system. In J. T. Cacioppo & L. G. Tassinary (Eds.), Principles of psvchophysiology: Physical, social, and inferential elements (pp. 385-4 12). New York: Cambridge University Press. Rey, A. (1941). L'examen psychologique dans les cas d'encephalopathie traumatique. Archives de Psychologic. 28 (1 12), 286-340.

PAGE 135

129 Risinger, M. W. (1991). Electroencephalographic strategies for determining the epileptogenic zone. In H. Luders (Ed.), Epilepsy surgery (pp. 337-349). New York: Raven Press. Risinger, M. W., Engel, J., Van Ness, P. C, Henry, T. R., & Crandall, P. H. (1989). Ictal localization of temporal lobe seizures with scalp/sphenoidal recordings. Neurology. 39. 12881293. Rodin, E. H. (1968). The prognosis of patients with epilepsy . Springfield, Illinois: C.C. Thomas. Sass, K. J., Spencer, D. D., Kim, J. H., Westerveld, M., Novelly, R. A., & Lencz, T. (1990). Verbal memory impairment correlates with hippocampal pyramidal cell density. Neurology. 40. 1694-1697. Shigaki, C. S., Crosson, B. A., Leonard, C. M., Bauer, R. M., Gilmore, R., Roper, S. N., Fennell, E. B., and Sadek, J. R. (1995). The relationship between hippocampal volume differences and performance on neuropsychological assessment measures in epileptic patients. Manuscript submitted for publication. Spencer, S. S. (1994). The relative contributions of MRI, SPECT, and PET imaging in epilepsy. Epilepsia. 35 (Suppl. 6), S72-S89. Spencer, S. S. (1993). Diagnosis of medial temporal lobe seizure onset: Relative specificity and sensitivity of quantitative MRI. Neurology. 43. 21 17-2124. Spencer, S. S., Williamson, P. D., Bridgers, S. L., Mattson, R. H., Cicchetti, D. V., & Spencer, D. D. (1985). Reliability and accuracy of localization by scalp ictal EEC Neurology. 35, 1567-1575. Stanulius, R. G., & Valentine, R. J. (1988). Material-specific memory deficits-Language or memory? [Abstract]. Epilepsia. 29 (5), 681. Strauss, E., Hunter, M., & Wada, J. (1993). Wisconsin card sorting performance: Effects of age of onset of damage and laterality of dysfunction. Journal of Clinical and Experimental Neuropsychology. 15 (6), 896-902. Trenerry, M. R., Jack, C. R., Ivnik, R. J., Sharbrough, F. W., Cascino, G. D., Hirschhom, K. A., Marsh, W. R., Kelly, P. J., & Meyer, F. B. (1993). MRI hippocampal volumes and memory function before and after temporal lobectomy. Neurology. 43. 1800-1805. Wada, J., & Rasmussen, T. (1960). Intracarotid injection of sodium amytal for the lateralization of cerebral speech dominance: Experimental and clinical observations. Journal of Neurosurgery. 17. 266-282. Walczak, T. S., Radtke, R. A., & Lewis, D. V. (1992). Accuracy and reliability of scalp ictal EEG. Neurology. 42. 2279-2285.

PAGE 136

130 Wannamaker, B. B., & Matthews, C. G. (1976). Prognostic implications of neuropsychological test performance for surgical treatment of epilepsy. The Journal of Nervous and Mental Disease. 163 (1), 29-34. Wechsler, D. (1981). Wechsler adult intelligence scale-revised: Manual . New York: Psychological Corporation. Wechsler, D. (1987). Wechsler memory scale-revised: Manual . New York: Psychological Corporation. Williamson, P. D., French, J. A., Thadani, V. M., Kim, J. H., Novelly, R. A., Spencer, S. S., Spencer, D. D., & Mattson, R. H. (1993). Characteristics of medial temporal lobe epilepsy: II. Interictal and ictal scalp electroencephalography, neuropsychological testing, neuroimaging, surgical results, and pathology. Annals of Neurology. 34 (6), 781-787. Wyllie, E. & Luders, H. (1993). Classification of seizures: Section A: Epileptic seizures: EEG and clinical manifestations. In E. Wyllie (Ed.), The treatment of epilepsy: Principles and practices (pp. 359-361). Philadelphia: Lea and Febiger.

PAGE 137

BIOGRAPHICAL SKETCH David Moser was bom in Danvers, Massachusetts in 1967. After earning a B.A. from Colby College in 1989, he began graduate study in the Department of Clinical and Health Psychology at the University of Florida in 1991. Mr. Moser earned his M.S. degree in 1995 with a thesis entitled, "Creation of a Drawing Bias through Priming: Evidence of Conceptually-Driven Processes in Implicit Memory." Mr. Moser then completed the current study as his dissertation research, and earned his doctoral degree in 1997. He and his family have recently moved to Providence, RI, so that he could begin a neuropsychology postdoctoral fellowship at Brown University. 131

PAGE 138

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Russell M. Bauer, Chair Professor of CUnical and Health Psychology I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Eileen B. Fennell Professor of Clinical and Health Psychology I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in s^pe and quality, as a dissertation for the degree of Doctor of Philosophy. Duane E. Dede Assistant Professor of Clinical and Health Psychology I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a/lissertaticui for the degree of Doctor of Philosophy. // Robin L. Gilmore Professor of Neuroscience I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. 2^ J. Algina of Founi f Education