MULTIVARIATE FUNCTIONAL CONNECTIVITY ESTIMATION IN EEG/MEG WITHOUT SUBJECT-SPECIFIC MRI

Magnetoencephalography (MEG) and electroencephalography (EEG) are powerful techniques for investigating brain dynamics. However, they face some limitations regarding spatial resolution. Accurate functional connectivity (FC) analysis often requires individual structural MRI scans, increasing costs. In this study, we examined whether FC estimates differ between subject-specific MRI and a template anatomy. We acquired simultaneous EEG/MEG data from 11 cognitively healthy participants during a 5-minutes eyes-closed resting-state session. Source reconstruction was performed using individual MRI scans and a surface template based on the New York Head (Huang et al., 2016), adapted to each participant’s head shape and electrode positions. We applied linearly constrained minimum variance beamformer using a scalar and a vector (3D) beamformer solution, preserving 99% of the source variance. FC was quantified using phase-locking value (PLV) and its corrected version (ciPLV). Inter-area FC was estimated using five strategies: three conventional approaches based on a single representative time series per region, and two multivariate methods computing either the mean or the root-mean-square of all pairwise synchronization values. Reliability was assessed by comparing EEG- and MEG-derived FC patterns using Pearson correlation. Preliminary results showed that multivariate approaches achieved the highest correlations, especially when using ciPLV, suggesting improved agreement between EEG and MEG after leakage correction. Moreover, including all source orientations rather than a single projection further improved results, while the absence of individual MRI had minimal impact. Overall, multivariate FC methods with ciPLV and 3D beamformers seemed to yield robust EEG/MEG connectivity estimates with limited dependence on individual MRI data.