Abstracts

Rationale: Mapping of eloquent cortex plays an essential role in the pre-surgical evaluation of patients with mass lesions or intractable epilepsy. Functional magnetic resonance imaging (fMRI) and magnetoencephalography (MEG) are two non-invasive methods that are increasingly employed for these purposes. Both modalities have been compared to the “gold standards” of the intra-carotid amobarbital procedure and cortical mapping, however few studies have compared the results from these techniques directly. In this study, we obtained fMRI and MEG data on 16 healthy subjects during a semantic decision-making task to assess similarities and differences between techniques. Methods: Novel and repeated word stimuli were presented visually at an ISI of 600-1500ms while participants were instructed to either respond to repeating words or to infrequent target words (animals). Experimental paradigms were designed to be analogous between MEG and fMRI modalities with the exception that a block design was used for fMRI and an event-related design for MEG to optimize the signal-to-noise ratio. Consonant strings and false font sequences were used as non-language control stimuli and contrasted to the novel word activations to reveal language-specific cortex. MEG data were analyzed using a distributed source solution (dSPM) that constrains results to the cortical surface. Single-subject fMRI data were analyzed using general linear modeling. Group activation maps for fMRI and MEG were calculated based on alignment of major anatomical landmarks with a spherical morphing algorithm. Mean activations for MEG and fMRI were determined within temporal and frontal lobe regions of interest that were derived from an automated gyral-based labeling system.Results: MEG results showed the strongest activity in temporal language areas, including the superior and middle temporal gyri. In contrast, fMRI activated these areas inconsistently and revealed the strongest and most consistent activation in inferior frontal language regions, including the pars triangularis and pars opercularis. This indicates a differential functional-anatomical preference of these neuroimaging techniques for detecting language processing in the human brain.Conclusions: Our findings suggest that used in combination, fMRI and MEG yield a more comprehensive and robust characterization of the cortical language network and thereby increases the predictive validity of non-invasive language mapping in pre-surgical patients. Supported by NS18741, NS44623, NS056091 & FACES