Systematic evaluation of an exhaustive set of connectivity estimators in bivariate and multivariate modes for an improved virtual source MEG connectivity analysis for the study of resting state recordings

Abstract

Brain activity is measured using noninvasive electrophysiological techniques, such as electroencephalography (EEG) and magnetoencephalography (M/EEG), recorded from sensors outside the skull and it is regularly transformed into a virtual source space which is parcellated into anatomical brain areas using an atlas. Then, functional connectivity (FC) is estimated between pairs of regions, with their brain activity characterized by a representative time series extracted from multiple voxel time series (multidimensional), using various techniques. Alternatively to bivariate FC estimators and these techniques, their multivariate representative-free extension has been proposed. Multivariate FC estimators can also be considered. The impact of these different strategies on FC estimation is still unclear. An appropriate framework for systematically evaluating FC estimators in the virtual MEG space and across multiple processing steps for brain network construction is missing. Here, we compared an exhaustive set of bivariate FC estimators paired with techniques for extracting representative time series, their multivariate extensions, and multivariate estimators under two scenarios: 1) on the discrimination of MCI subjects versus healthy controls, using a k-NN classifier and an appropriate graph distance metric, and 2) on the repeatability of brain network topologies in a test-retest study. Our results demonstrate that the multivariate extension of bivariate FC estimators (representative-free approach), which summarizes pairwise FC strength across all voxels of two brain areas, and accurate multivariate estimators that consider pairs of region-wise voxel time series at once, clearly outperform bivariate FC estimators paired with representative time series in both scenarios. Given the remarkable differences among the three approaches across frequency bands and FC estimators, caution is warranted when comparing results across studies that apply different pairs of extraction methods and FC estimators.