Abstract
Cocaine-use disorder (CUD) affects both structure and function of the brain. A triple network model of large-scale brain networks has been useful for identifying aberrant resting-state functional connectivity (rsFC) associated with mental health disorders including addiction. The present study investigated differences between people with CUD vs. controls (CONs) and whether putative differences were associated with drug-use outcomes. Participants with CUD (n = 38) and CONs (n = 34) completed a resting functional magnetic resonance imaging (fMRI) scan. Participants with CUD completed several mental health measures and participated in an 8-week, drug-use outcomes phase. A classification framework based on the triple network model was built, and triple networks (salience [SN], executive control [ECN], default mode [DMN]) and subcortical (striatum [ST], hippocampus/amygdala) regions were identified with the algorithm of group-information-guided independent components analysis (GIG-ICA) and subsequent support-vector machines. This classifier achieved 77.1% accuracy, 73.8% sensitivity, and 80.0% specificity, with an area under the curve of 0.87 for distinguishing CUD vs. CON. The two groups differed in SN-anterior DMN (aDMN) and ECN-aDMN rsFC, with the CUD group exhibiting stronger rsFC compared to CONs. They also differed in rsFC between several subcortical and triple networks, with CUD generally showing a lack of rsFC. Within the CUD group, ST-aDMN and ST-rECN rsFC were associated with differential drug-use outcomes. Exploratory results suggested SN-aDMN rsFC was associated with anxiety symptoms. These results add to the growing literature showing aberrant triple network and subcortical rsFC associated with substance use disorders. They suggest the aDMN specifically may underlie important differences between people with CUD and CONs and may be a potential target for intervention.