Studies have also shown that abstinent individuals with AUD often manifest poor neuropsychological performance [40,41,42,43,44,45] and heightened impulsivity [46,47,48], and, therefore, it is important to include these domains in the models exploring concomitants and/or determinants of alcohol-related outcomes. In our previous study, we reported that a random forest algorithm was highly useful for classifying AUD individuals from controls using multi-modal measures, including fMRI functional connectivity of the default mode network, neuropsychological performance, and impulsivity. Since AUD and other addiction traits are shown to be associated with reward networks [49,50] and as substance induced neuroadaptation primarily involves reward structures [51], the current study aimed to examine the functional connectivity across the reward network, along with measures of neuropsychological performance and impulsivity, to differentiate abstinent AUD individuals from healthy controls. Similar to our previous study, we computed the predictive power of these multi-domain measures in terms of classification accuracy in a machine learning framework and evaluated the utility of these phenotypic features, especially the reward network connectivity measures. Since recent studies have proposed that resting state fMRI connectivity can potentially serve as
