The regions of interest (ROIs) for the reward network were identified from the published literature of review and meta-analyses of reward processing, e.g., [6,70,71]. These included 34 ROIs from 17 bilateral structures involving nine bilateral subcortical structures and eight bilateral cortical regions (Table 2 and Figure 2). The ROIs were marked using ITK-SNAP, an image processing application [72]. The diameters of subcortical and cortical ROIs were 7 mm and 11 mm, respectively, from the MNI coordinates (Table 2). The ROI-to-ROI connectivity [73], the most commonly used method to derive rsFC across brain regions [74], was computed using Pearson correlation coefficients between all unique pairs (N = 561) of BOLD time series data of all 34 ROIs listed in Table 2. The resulting correlation coefficients were Fisher Z-transformed for further statistical analyses. All 561 connections derived from unique combinations of the ROIs were used in the feature selection process (see Section 2.7), which provided the subset of features to be used in the random forest model (see Section 2.8).
