Abstracts

Rationale: Accurate methods for determining hemispheric language dominance are crucial for assessing neurocognitive risks and planning epilepsy surgery. MEG and fMRI provide non-invasive alternatives to the Wada test. Both task-related evoked responses and beta frequency band power changes are used to determine language dominance with MEG. However, there are various methods for mapping these MEG responses to the cortex and estimating laterality indices (LIs). It is unclear if contrasting language task responses to a control task offers advantages. This study estimated regional LIs from task-related beta power modulations using three methods, with and without contrasts to control task responses, and compared them to fMRI results.

Methods: MEG and fMRI data from 72 temporal lobe epilepsy patients (aged 19-60 years) in the Epilepsy Connectome Project were analyzed. During MEG scans, participants performed a semantic decision (SD) task with animal names ("Animals") and a control task with unfamiliar "Symbols." fMRI sessions included analogous auditory SD and tone decision tasks. Beta power modulations were assessed using a DICS beamformer for contrasted and non-contrasted activation patterns.
LIs were computed using three methods: source magnitude with a half-maximum threshold, counting vertices, and a non-threshold-based bootstrapping approach, from -500 ms to 2 seconds at 300 ms intervals with 10 ms overlaps. We focused on the correlation and concordance between MEG LIs from beta power decrements and fMRI activations in frontal, temporal, angular gyri, and a combined "lateral" ROI (HCP-MMP 1.0 atlas). Given the fast temporal dynamics of MEG responses, we estimated the correlation of regional LIs and ternary classification of language dominance (left, right, symmetric) relative to fMRI results. Concordance values between MEG and fMRI were determined for response intervals with maximal signal-to-noise ratios for individual patients.


Results: MEG and fMRI LIs in language-related regions showed high correlation and concordance. Lateral LIs had a correlation of 0.66 and concordance of 87.50% for the 400-700 ms interval. MEG responses between 450-750 ms indicated pronounced left hemisphere dominance. Using patient-specific optimal response intervals (based on peak latencies of individual LIs) against fMRI showed a maximum correlation of 0.69 and concordance of 91.6%, a 4% improvement over fixed group intervals. LIs from the Animal condition alone had a 78% concordance rate. Compared to methods without task contrasts, LIs showed a 13.6% improvement in concordance rate.

Conclusions: This study confirms the efficacy of MEG beta power modulation in determining hemispheric language dominance. Concordance with fMRI is significantly improved by contrasting responses against those for a control task and using a non-threshold-based bootstrapping method for estimating LIs. Our findings suggest ways to enhance the precision of language dominance assessments using MEG and underscore the importance of task-specific contrasts when mapping language networks.

Funding: Supported by the Epilepsy Connectome Project (U01NS093650; NIH) & MEGIN US LLC (FP00024550).