Functional Covariance Networks: Obtaining Resting-State Networks from Intersubject Variability

In this study, we investigate a new approach for examining the separation of the brain into resting-state networks (RSNs) on a group level using resting-state parameters (amplitude of low-frequency fluctuation [ALFF], fractional ALFF [fALFF], the Hurst exponent, and signal standard deviation). Spatial independent component analysis is used to reveal covariance patterns of the relevant resting-state parameters (not the time series) across subjects that are shown to be related to known, standard RSNs. As part of the analysis, nonresting state parameters are also investigated, such as mean of the blood oxygen level-dependent time series and gray matter volume from anatomical scans. We hypothesize that meaningful RSNs will primarily be elucidated by analysis of the resting-state functional connectivity (RSFC) parameters and not by non-RSFC parameters. First, this shows the presence of a common influence underlying individual RSFC networks revealed through low-frequency fluctation (LFF) parameter properties. Second, this suggests that the LFFs and RSFC networks have neurophysiological origins. Several of the components determined from resting-state parameters in this manner correlate strongly with known resting-state functional maps, and we term these “functional covariance networks”.