Abstracts

Rationale: The goal of improving seizures with early surgical treatment is closely tied to improving cognitive and quality of life outcomes. Avoiding or minimizing deficits following surgery depends on identification of critical cortical regions prior to resection. Functional MRI is emerging as a noninvasive alternative for localization of language, but poses challenges in children with cognitive and behavioral impairments and stimulation of subdural grid electrodes remains the gold standard for precise localization of language areas. There have been very few studies that have validated fMRI by comparison with electrical cortical stimulation mapping (ESM), and none in children undergoing epilepsy surgery. We report preliminary data on concordance of fMRI and ESM in children using a novel quantitative approach. Methods: Three children (3 girls, 11-15 years, 2 right-handed) underwent language mapping with fMRI and ESM in the left hemisphere within the first year of our epilepsy surgery program. fMRI language mapping was accomplished using a covert Verb Generation (VG) task performed on a 3-Tesla GE scanner. Children were presented with concrete nouns at different rates (0.2Hz, 0.3 Hz, 0.5Hz) in 30s activation blocks alternating with 30s blocks of finger tapping. ESM was conducted using an Ojemann stimulator with amperages ranging from 4-10 milliamps (50 hz; 5 sec. duration) during expressive language tasks. Fusion of fMRI and post-implantation CT images was performed to allow for direct comparison. Results: fMRI during VG elicited robust activation in the left inferior frontal gyrus in all 3 patients. Bilateral activation was observed in the left-handed patient. ESM identified eloquent cortex in all three patients. A total of 67 contacts were stimulated across patients; 15 produced a positive response. In the whole series, 8/15 positives contacts were located within an active fMRI cluster using a threhold of p <.01 (Bonferroni). Among the 52 negative contacts, 6 were located within an active fMRI cluster and 46 were not. The overall sensitivity of language fMRI was .53 and the specificity was .88. The false positive rate (+fMRI activation; - ESM finding) was .46, and the false negative rate was .11. Conclusions: To our knowledge, we are the first to compare subdural ESM and fMRI in children with intractable epilepsy, and employ a novel co-registration technique allowing for objective analysis of concordance. In our small case series, language fMRI did not activate all critical regions displayed in ESM, and there was significant inter-patient variability (0%-100% for temporal and frontal lobe, respectively). fMRI VG failed to activate left temporal language areas in patients who are at high risk for post-surgical language deficit. Concordance was higher in frontal language areas, but fell short of acceptable levels. Our preliminary results suggest that fMRI may be used as a complementary tool to ESM. Development of comparable fMRI and ESM paradigms may improve concordance.