Towards precision EEG connectomics: Evaluating the benefits of dense sampling

EEG connectomics research offers the potential to better understand human neurodevelopment, brain disorders, and brain-behavior associations. Several functional connectivity (FC) measures are widely used but few studies have directly compared metrics of reliability across measures and how data quantity influences these properties. Here, we collected a densely sampled dataset from 25 parent-child pairs, with 80 minutes of passive viewing EEG data collected per participant over 4 sessions, and calculated connectomes using 9 popular phase- and envelope-based measures (coherence, COH; phase locking value, PLV; corrected imaginary phase locking value, CIPLV; imaginary coherence, IMCOH; phase-lag index, PLI; weighted phase-lag index, WPLI; envelope correlation with pairwise orthogonalization, ECPWO; and envelope correlation with symmetric orthogonalization, ECSO), including one effective connectivity measure (phase slope index; PSI). We used connectome individualization, derived from fingerprinting-style analyses, as a multivariate reliability metric, as used in fMRI-FC studies. We used simulations with individual head geometry but no "true" connectivity to assess whether identifiability was influenced by volume conduction. We found that COH and PLV were vulnerable to volume conduction influence on identifiability; CIPLV, ECSO, and ECPWO were semi-vulnerable; and IMCOH, WPLI, PLI, and PSI were minimally vulnerable. Next, we considered individualization and reliability of age-group effects with increasing time of data collection. IMCOH had the overall best performance among minimally vulnerable measures, considering individualization and reliability of group effects. We further found that reliability of IMCOH along with other volume conduction-corrected EEG-FC measures continued to improve with up to 25-30 minutes of data. Together, our findings can support study design decisions in EEG connectomics research.