Language Outcome After Left Temporal Lobectomy in Patients with Discordant fMRI and Sodium Amobarbital Testing Results

LANGUAGE OUTCOME AFTER LEFT TEMPORAL LOBECTOMY IN PATIENTS WITH DISCORDANT fMRI AND SODIUM AMOBARBITAL TESTING RESULTS Julie K. Janecek, M.S. Marquette University, 2011 The rationale for this study was to examine 1) language lateralization discordance rates between fMRI and the IAT in pre-surgical epilepsy patients and 2) naming outcome after left ATL in a group of patients for whom IAT and fMRI language LIs were discordant. Participants were 229 consecutive pre-surgical epilepsy patients who underwent the IAT and fMRI. IAT LIs (% correct inject right –% correct inject left condition) were calculated based on performance on comprehension, naming, repetition and reading language tasks. The fMRI LIs [(L-R)/(L+R) where L = number of activated left hemisphere voxels and R = number of activated right hemisphere voxels) were calculated for lateral, angular gyrus, temporal, and frontal regions of interest (ROIs) using a published semantic decision task. Discordance was determined using cut scores and difference scores for each method. Regression analyses were performed to investigate predictors of discordance. Additionally, regression formulas developed from a separate sample for predicting language outcome using fMRI and IAT LIs were applied to the discordant cases so that observed and predicted outcome scores could be compared with each method. Discordance rates ranged from 14-17%, depending on ROI. Atypical language dominance on fMRI was most predictive of discordance. Of discordant cases who underwent left ATL, language outcome was more accurately predicted by each method in approximately half the cases. When fMRI indicates left language dominance, IAT LI concordance is high. However, when fMRI indicates atypical language dominance, concordance rates with the IAT decrease. Post-operative language outcome data suggests that the IAT and fMRI each predict outcome in certain cases, suggesting some error variance with each mapping method.