Group ICA maps were created using all subjects in both groups by FSL MELODIC (Multivariate Exploratory Linear Decomposition into Independent Components) (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/ME-LODIC) with 30 components, the output of which was visually inspected to identify the DMN, SN, LECN, and RECN networks [34]. Dual regression analysis was then performed in FSL to obtain the subject-specific component maps of parameter estimates of functional connectivity, which were then used as the input to non-parametric permutations statistical tests using FSL’s Permutation Analysis of Linear Models (PALM) (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/PALM) for comparison between groups [14,35]. The FSL standard Threshold Free Cluster Enhancement (TFCE) was used to identify statistically significant clusters of voxels while maintaining family-wise-error (FWE) control [36]. We analyzed the regression of the subject-specific functional connectivity on NU scores for all subjects in both groups, and we also compared the functional connectivity of subjects with severe OUD to subjects with mild or moderate OUD. All reported p-values of functional connectivity results are FWE corrected for the number of voxels, the number of contrasts, and the 4 networks examined.
