Figure 1
Outline of the overall analysis approach. (a) At the voxel level, we used group independent component analysis to extract large-scale canonical resting state networks (cRSNs) from resting state fMRI data and dual regression analysis to identify individual cRSNs. A similarity measure was then estimated for each cRSN relative to a reference RSN. Using clusters derived from cRSNs and the rest of the brain as nodes, we then examined at the node level how the changes observed at the voxel level affected the network properties of the brain using graph analysis. For this, we estimated several whole brain network measures including path length, global efficiency, and degree, among others. Finally, we focused our analysis on the nodes of well-known cRSNs and computed mean connectivity values within cRSNs (within-network functional connectivity, WNFC) and between cRSNs (between-network functional connectivity, BNFC) to investigate changes at the network level. For all estimated measures at all levels, we investigated the measures’ association with age and ACE-R total score. (b) For functional connectivity measures (FCM) that showed significant correlation (p < 0.05, uncorrected) with both age and ACE-R total score, we further performed mediation analysis to elucidate the relationship among the interacting variables. The mediation model shown in (b) was used. Details of the methods used in the analysis are given in the Methods section.
