Abstracts

Rationale: Task related gamma synchrony and fMRI represent techniques for identifying language cortex based on functional activation. There have been no direct comparisons of fMRI, task-related gamma, and traditional ESM maps based on an identical language task. We present a series of patients where all three methods were employed using the same language task. Methods: We studied three patients who underwent left temporal resective surgery for refractory partial epilepsy. All patients underwent fMRI of language using an auditory definition naming task prior to surgery. The fMRI acquisition and analysis protocols have been previously published. The patients also underwent left hemispheric invasive EEG studies. Language and control tasks used for fMRI were implemented at the bedside during intracranial EEG recording. Multichannel EEG spectral analysis was performed using software written in SCILAB. Log power spectral densities in the 80-100 Hz gamma band during blocks of language and control trials were compared using the Wilcoxon test in order to identify electrodes with significant (p<0.005) task-related gamma synchrony. The patients also underwent ESM of all contacts overlying the left temporal neocortex. Cortical stimulation was performed using biphasic pulse trains of 10 ma and pulse-widths between 0.5 to 1.0 ms. Testing during ESM included auditory definition naming. A thin-slice CT was performed after intracranial electrode placement and co-registered to the T1-weighted anatomical reference images that were acquired for the fMRI study. Subdural electrode overlays were generated by applying appropriate thresholds to the CT volumes and volume rendered on 3-D MRI reconstructions to allow comparison of fMRI, ESM, and gamma-synchrony based maps in the same brain coordinates. Results: Language maps resulting from fMRI, task related gamma synchrony, and ESM are clearly not identical even with an identical language task. fMRI and gamma synchrony maps identify a much broader region of task-related cortical engagement. Direct correlation of fMRI maps to those derived from the other two methods is not possible due to the discrete grid coordinates of ESM and gamma synchrony maps. Based on the anterior extents of the maps, FMRI and gamma synchrony maps recommended resection boundaries significantly anterior to those predicted by ESM in all three patients. Conclusions: Language mapping using fMRI and EEG gamma synchrony differ fundamentally from ESM in that they identify functional activation as opposed to the effects of disrupting cortical processes. Both techniques not only yield more extensive maps of cortical engagement during language tasks than ESM, but these maps also differ significantly from each other. The significance of these differences, the appropriate statistical thresholds that need to be applied to activation maps, and the clinical utility of these alternative mapping methods warrant extensive further study based on language outcome.