Abstracts

Rationale: Determination of language dominance remains a key component of the presurgical evaluation of patients with drug-resistant epilepsy. We compared MEG regional language laterality measures based on distributed source modeling of evoked fields during auditory word recognition to those derived from a well-characterized fMRI protocol. Methods: We analyzed data from 45 adult patients with drug-resistant partial epilepsy who performed an auditory word recognition task during MEG recording and also completed an fMRI language study as part of their presurgical evaluation for epilepsy surgery. Source imaging of the evoked responses was performed using dynamic statistical parametric mapping (dSPM). Language laterality indices (LIs) were calculated for four regions of interest (ROIs) by counting the number of unit dipoles with above-threshold activation within a 300-600 ms time window after stimulus onset. Language laterality classifications at each ROI were compared to corresponding classifications based on fMRI. Results: The most lateralized MEG responses to language stimuli were observed in an ROI that included the angular and supramarginal gyri (AngSmg). For this ROI, using a half-maximal threshold, language-evoked responses were left dominant in 32 (71%) patients, right dominant in 8 (18%), and symmetric in 5 patients (11%). The best agreement between MEG and fMRI on the ternary classification of regional language dominance into left, right, or symmetric groups was also found at the AngSmg ROI (69%). This was followed by the whole-hemisphere and temporal ROIs (both 62%). The frontal ROIs showed the least agreement with fMRI (51%). Gross discordances between the two imaging modalities in the form of language lateralization to opposite hemispheres was observed in 7 patients (16%) at the AngSmg ROI, 9 patients (20%) for the whole-hemisphere, 13 patients (29%) at the frontal ROI, and 14 patients (31%) at the temporal ROI. Gross discordances were disproportionately of the type where MEG indicated right hemisphere dominant language, and were significantly associated with right hemispheric seizure origin (p < 0.05 at three of the four ROIs). Conclusions: dSPM of language-evoked neuromagnetic fields using an auditory word recognition task shows the best agreement with fMRI in a posterior brain region that includes the inferior parietal lobe. The concordance between MEG and fMRI in this ROI is similar to what has been reported in prior comparisons of MEG and the Wada test using this task. However, our data also suggest that epileptic networks can bias the estimated laterality of language representations toward the hemisphere of seizure origin. This has not been recognized previously and deserves further study. Funding: None