Faculty Bibliography

OBJECT: Because appropriate patient selection is essential for achieving successful outcomes after epilepsy surgery, the need for more robust methods of predicting postoperative seizure control has been created. Standard multivariate techniques have been only 75 to 80% accurate in this regard. Recent use of artificial intelligence techniques, including neural networks, for analyzing multivariate clinical data has been successful in predicting medical outcome. METHODS: The authors applied neural network techniques to 80 consecutive patients undergoing epilepsy surgery in whom data on demographic, seizure, operative, and clinical variables to predict postoperative seizures were collected. Neural networks could be used to predict postoperative seizures in up to 98% of cases. Student's t-tests or chi-square analysis performed on individual variables revealed that only the preoperative medication index was significantly different (p = 0.02) between the two outcome groups. Six different combinations of input variables were used to train the networks. Neural network accuracies differed in their ability to predict seizures: using all data (96%); all data minus electroencephalography concordance and operative side (93%); all data except intra- or postoperative variables such as tissue pathological category (98%); all data excluding pathological category, intelligence quotient (IQ) data, and Wada results (84%); only demographics and tissue pathological category (65%); and only IQ data (63%). CONCLUSIONS: Analysis of the results reveals that several networks that are trained with the usual accepted variables characterizing the typical evaluation of epilepsy patients can predict postoperative seizures with greater than 95% accuracy

PURPOSE: Posterior, 'Wenicke's,' language areas have a high degree of between-subject variability, as shown by electrical-stimulation mapping. We investigated the possibility of an organized structure in the distribution of posterior language areas. METHODS: Extraoperative subdural grid stimulation was performed on 67 left hemisphere-dominant patients before resective epilepsy surgery during counting, naming, and reading. Intersubject-averaged language maps were generated in which stimulation disrupted only one language function and not the others, or combinations of language functions. RESULTS: Language sites, although highly variable between subjects, were not organized randomly and appeared to be arranged into several focal, non-contiguous, higher probability 'nodes' devoted to different aspects of language processing. Speech-arrest sites were concentrated in classic Wernicke's area. Areas where stimulation induced only reading errors were found in the posterior middle temporal gyrus and the inferior parietal lobule. These regions may correspond with an orthographic input lexicon. Areas eliciting only naming errors were found in the posterior inferior temporal gyrus extending into the mid-middle temporal gyrus and may represent a visual-representation input lexicon. Sites where stimulation elicited errors in both naming and reading were more variable in location than sites devoted to only one function, extended farther anteriorly along the temporal neocortex, and may correspond with a semantic lexicon. CONCLUSIONS: The existence of high-probability nodes in posterior language cortex supports a modality-specific modular architecture and the possibility of a conserved, universal structure

OBJECT: Although it is known that 5 to 10% of patients have language areas anterior to the rolandic cortex, many surgeons still perform standard anterior temporal lobectomies for epilepsy of mesial onset and report minimal long-term dysphasia. The authors examined the importance of language mapping before anterior temporal lobectomy. METHODS: The authors mapped naming, reading, and speech arrest in a series of 67 patients via stimulation of long-term implanted subdural grids before resective epilepsy surgery and correlated the presence of language areas in the anterior temporal lobe with preoperative demographic and neuropsychometric data. Naming (p < 0.03) and reading (p < 0.05) errors were more common than speech arrest in patients undergoing surgery in the anterior temporal lobe. In the approximate region of a standard anterior temporal lobectomy, including 2.5 cm of the superior temporal gyrus and 4.5 cm of both the middle and inferior temporal gyrus, the authors identified language areas in 14.5% of patients tested. Between 1.5 and 3.5 cm from the temporal tip, patients who had seizure onset before 6 years of age had more naming (p < 0.02) and reading (p < 0.01) areas than those in whom seizure onset occurred after age 6 years. Patients with a verbal intelligence quotient (IQ) lower than 90 had more naming (p < 0.05) and reading (p < 0.02) areas than those with an IQ higher than 90. Finally, patients who were either left handed or right hemisphere memory dominant had more naming (p < 0.05) and reading (p < 0.02) areas than right-handed patients with bilateral or left hemisphere memory lateralization. Postoperative neuropsychometric testing showed a trend toward a greater decline in naming ability in patients who were least likely to have anterior language areas, that is, those with higher verbal IQ and later seizure onset. CONCLUSIONS: Preoperative identification of markers of left hemisphere damage, such as early seizure onset, poor verbal IQ, left handedness, and right hemisphere memory dominance should alert neurosurgeons to the possibility of encountering essential language areas in the anterior temporal lobe (1.5-3.5 cm from the temporal tip). Naming and reading tasks are required to identify these areas. Whether removal of these areas necessarily induces long-term impairment in verbal abilities is unknown; however, in patients with a low verbal IQ and early seizure onset, these areas appear to be less critical for language processing

We studied 21 patients who underwent multiple subpial transections (MST) for the surgical treatment of intractable partial-onset seizures in whom the epileptogenic focus overlapped with primary sensorimotor or language cortex. All patients had intracranial EEG localization of seizures and functional mapping with electrical stimulation before surgery. Eighteen patients had cortical resections in addition to MST, At an average follow-up of 21 months, 13 patients either are seizure-free or are experiencing auras or simple partial seizures (SPS) only, 6 patients have had greater than 50% reduction in seizure frequency, 1 patient has had a modest reduction in seizures, and 1 patient is unchanged. MST in the 3 patients who did not undergo cortical resections reduced seizure frequency in all 3, although none are seizure-free. MST, combined with cortical resection, appears to be a safe and effective alternative to subtotal resection of the epileptogenic zone. MST alone, although not curative, reduced seizure frequency in all patients who underwent the procedure. (C) 1997 Elsevier Science Inc

Few studies have examined the clinical features of neocortical temporal lobe epilepsy (NTLE) in carefully selected patients. We reviewed records from 21 patients with NTLE, defined by intracranial electroencephalogram (EEG), who have been seizure free for 1 year or more following temporal lobectomy. The mean age of onset at the time of first seizure was 14 years (range, 1-41 years). Febrile seizures were reported in only 2 patients (9.5%). In contrast to prior mesial temporal lobe epilepsy (MTLE) studies, seizure-free intervals between the initial cerebral insult or first seizure and habitual seizures were uncommon. Possible or known risk factors for epilepsy were reported in 13 of 21 patients (62%). Fifteen (71%) patients reported auras, with experiential phenomena being the most common type. Magnetic resonance imaging was normal or nonspecific in 15 patients, revealed mild hippocampal atrophy in 2, tumors in 2, and heterotopic gray matter and hippocampal atrophy in 1, and cortical dysgenesis in 1. Neuropsychological testing showed deficits consistent with the seizure focus in 13 patients (62%), and Wada test showed ipsilateral memory deficits in 10 (48%). The most common behavioral manifestation was a motionless stare at ictal onset (48%). In contrast to prior studies of MTLE, only 1 NTLE patient had frequent independent, contralateral temporal lobe epileptiform spikes on scalp EEG

Our experience with a very low end interactive image-directed (IIDS) neurosurgical system is presented. The system was developed by the author and consists of a personal desktop computer and a magnetic field digitizer. This low cost solution was pursued as an alternative to available commercial devices which were expensive and not readily modifiable for novel ideas and new applications targeted for Epilepsy surgery. The rationale and description of the system was presented last year at Medicine Meets Virtual Reality III. Included in that detailed report were the fundamental mathematics forming the basics of transformation between the surgical and the digital data spaces. Since then the system has been used in an additional 20 cases now totaling 40 in all. Its advantages and short comings will be described. The theoretical advantages of magnetic field technology over other localization methods is reviewed. Also, our experience with alternative low cost off-the-shelf interfacing devices and other related modifications are described. We have accumulated clinical data to suggest that craniotomy sizes have been reduced, electrode placement has been improved, and that interactive image-directed techniques offer advantages over other common intra-operative localization modalities such as ultrasound. Our conclusion is that interactive image-directed techniques improve neurosurgery and that inexpensive enabling technology is already available providing the technological substrate for low cost devices using virtual reality notions for surgery and medicine. This particular technology offers advantages to traditional surgical techniques demonstrating the attractiveness of rudimentary virtual reality medical applications