Abstracts

Rationale: The high spatial and temporal resolution of MEG source imaging makes it an attractive non-invasive tool to lateralize language networks in the presurgical evaluation of patients with drug resistant partial epilepsy. The most extensively studied paradigm for this clinical application estimates dipolar sources of late components of the evoked responses to auditory word presentations, and computes language laterality indices based on the relative abundance of dipoles in the two hemispheres. Single dipole models are however not ideal models of spatially extensive activations in language areas. Furthermore, with auditory stimuli that are extended in time, primary auditory areas continue to be engaged throughout the stimulus presentation. To deal with these limitations, we model evoked responses to auditory word presentations using distributed source imaging techniques and also model responses to control stimuli that are matched to the low-level acoustic properties of the word stimuli. We sought to compare language laterality indices estimated from this approach against laterality assessed from alternative clinical approaches such as fMRI and Wada. Methods: We modified a well studied auditory word-listening task (Continuous Recognition Memory) and randomly interleaved noise stimuli that are matched to the word stimuli in spectral content and duration, thus providing a low-level control for the acoustic content in the word stimuli. Source imaging of the evoked responses to each stimulus category was estimated using L2 minimum norms. The magnitude of the source currents for the noise stimuli was subtracted from those for the word stimuli, and Z-score normalization relative to the pre-trigger baseline was performed. Voxels that reach a criterion threshold within several Freesurfer-generated ROIs in each hemisphere were used to calculate the laterality indices (LI) for different ROIs . Results: We performed this procedure for 10 adult epilepsy patients who had undergone language lateralization based on alternative clinical methods including the Wada procedure, fMRI, or both. In nine patients LIs based on a ROI that combines the supramarginal and inferior parietal 'Desikan-Killiany' ROIs were found to be in good agreement with fMRI or Wada. In one of these patients strong right hemisphere activations in the word condition was noted, but this right dominance was stronger to the noise stimuli, thereby shifting the overall LI to the left when the word-minus-noise responses were examined, consistent with fMRI results. In one patient MEG results favored left dominance, consistent with Wada, but inconsistent with fMRI.Conclusions: Measuring responses to spectrally-matched control stimuli in addition to auditory language stimuli can potentially improve MEG-based language laterality determination by controlling for asymmetries in cortical activations related to low level stimulus properties that are not related to language processing.