Abstracts

Rationale: Hemispheric dominance of language is a key variable in planning epilepsy surgery. Non-invasive methods for measuring language dominance using fMRI and MEG have traditionally relied on task-induced responses. Recently, resting-state fMRI connectivity has been shown to correlate with laterality of language networks (eg., Doucet et al., Human Brain Mapping, 2015; Tie et al., Human Brain Mapping, 2014). We sought to determine whether information correlated with language dominance is present in resting-state MEG (rs-MEG) connectivity measures.  Methods: We analyzed MEG data recorded from 18 healthy controls and 15 temporal lobe epilepsy patients as part of the Epilepsy Connectome Project. Recordings were acquired in the awake state using an Elekta 306-channel MEG system at the Medical College of Wisconsin. We analyzed 15 minutes of data from each subject. Noise reduction using spatiotemporal signal space separation and independent component analysis was applied to the data after rejection of bad epochs. Band-limited data were projected to the brain using a cortically-constrained dSPM inverse solution. A time series of source activity was estimated for each Desikan-Killiany parcel in the Freesurfer -segmented brain model. Connectivity between all ROI-pairs was computed using three metrics: coherence (COH), phase lag index (PLI), and debiased weighted phase lag index (d-wPLI) in the theta (4-8 Hz), alpha (8-12 Hz), beta (13-30 Hz), and low-gamma (30-55 Hz) frequency bands. Mean connection strengths within 4 intra-hemispheric ROI-groups (3 language-specific groups, plus the whole hemisphere; see Table 1) were compared across the hemispheres to estimate lateralization indices (LIs). Language laterality based on task-fMRI was available for a subset of subjects (N=21). The fMRI results were based on an auditory semantic decision task contrasted against a tone-decision task (Binder et al., Neurology 1996). Concordance of rs-MEG and fMRI LIs were assessed for both ternary (Left/Symmetric/Right) and binary (Left/Right) classifications. Results: LIs based on the COH measure were not significantly lateralized to either hemisphere in any frequency band, for any ROI-group. LIs based on PLI and the d-wPLI showed strong left lateralization in the 30-55 Hz gamma band, with the latter yielding stronger lateralization. LIs based on gamma-band d-wPLI showed the highest left lateralization (82%) for a frontotemporal ROI-group that included the superior temporal gyrus, pars triangularis, and pars opercularis (Figure 1). Language laterality comparison between fMRI and MEG demonstrated a maximal concordance rate of 71% with the ternary classification, and 76% for the binary classification of LIs (Table 1). Positive correlations were noted between rs-MEG and task-fMRI LIs at the three language-specific ROIs but none reached statistical significance for this sample size. Conclusions: These preliminary findings suggest that a signal predictive of language dominance may be present in rs-MEG connectivity. The specificity of this finding to one frequency band, and the greater left lateralization of connectivity within ROI-groups that are limited to anatomic language areas make it more likely that it is indeed related to the laterality of language networks rather than artefactual. Funding: NIH/NINDS grant U01 NS093650