Effect of channel density, inverse solutions and connectivity measures on EEG resting-state networks: a simulation study

Abstract

Along with the study of brain activity evoked by external stimuli, the past two decades witnessed an increased interest in characterizing the spontaneous brain activity occurring during resting conditions. The identification of the connectivity patterns in this so-called "resting-state" has been the subject of a great number of electrophysiology-based studies, using the Electro/Magneto-Encephalography (EEG/MEG) source connectivity method. However, no consensus has been reached yet regarding a unified (if possible) analysis pipeline, and several involved parameters and methods require cautious tuning. This is particularly challenging when different choices induce significant discrepancy in results and drawn conclusions, thereby hindering reproducibility of neuroimaging research. Hence, our objective in this study was to evaluate some of the parameters related to the EEG source connectivity analysis and shed light on their implications on the accuracy of the resulting networks. We simulated, using neural mass models, EEG data corresponding to two of the resting-state networks (RSNs), namely the default mode network (DMN) and the dorsal attentional network (DAN). We investigated the impact of five channel densities (19, 32, 64, 128, 256), three inverse solutions (weighted minimum norm estimate (wMNE), exact low resolution brain electromagnetic tomography (eLORETA), and linearly constrained minimum variance (LCMV) beamforming) and four functional connectivity measures (phase-locking value (PLV), phase-lag index (PLI), and amplitude envelope correlation (AEC) with and without source leakage correction), on the correspondence between reconstructed and reference networks. We showed that, with different analytical choices, a high variability is present in the results. More specifically, our results show that a higher number of EEG channels significantly increased the accuracy of the reconstructed networks. Additionally, our results showed a significant variability in the performance of the tested inverse solutions and connectivity measures. In our specific simulation context, eLORETA and wMNE combined with AEC computed between orthogonalized time series exhibited the highest performance in terms of similarity between reconstructed and reference connectivity matrices. Results were similar for both DMN and DAN. We believe that this work could be useful for the field of electrophysiology connectomics, by shedding light on the challenge of analytical variability and its consequences on the reproducibility of neuroimaging studies.

The time-evolving epileptic brain network: concepts, definitions, accomplishments, perspectives

“…If different recording techniques are used, there is still the problem of an unambiguous fusion of the different scales (Schevon et al, 2019;Driscoll et al, 2021;Presigny and Fallani, 2022). Likewise, suitable strategies to avoid spatial and temporal oversampling of brain dynamics are urgently needed since oversampling can lead to severe misinterpretations of network characteristics (Bialonski et al, 2010;Zalesky et al, 2010;Bialonski et al, 2011;Joudaki et al, 2012;Montes-Restrepo et al, 2014;Porz et al, 2014;Puce and Hämäläinen, 2017;Conrad et al, 2020;Vorderwülbecke et al, 2020;Iivanainen et al, 2021;Allouch et al, 2023). It remains to be shown whether recent developments of edge/vertex removal strategies (Bellingeri et al, 2020; [or, in case of undersampling, techniques to predict edges (Zhou, 2021) and to detect hidden vertices (Su et al, 2012)] can help to avoid misinterpreting characteristics of the timeevolving epileptic brain network.…”