Connectome-based predictive modeling of early and chronic psychosis symptoms

Abstract

AbstractThe symptoms of psychosis-spectrum disorders, which include positive symptoms (e.g., hallucinations and delusions) and negative symptoms (e.g., memory impairment and disorganized thinking), can cause significant distress and disability. Despite shared symptomatology and converging brain correlates, early (EP) and chronic (CP) psychosis differ in their symptom-related treatment response. At present, the mechanism underlying these differences is unknown, in large part because EP and CP have predominantly been studied and characterized independently or in comparison to control populations. To answer this question, we use connectome-based predictive modeling (CPM) and resting-state functional magnetic resonance imaging to identify biologically-based early (EP, n=107) and chronic (CP, n=123) psychosis symptom networks. We predicted both samples’ total, positive, and negative symptoms from the PANSS. Virtual lesioning analyses highlight the frontoparietal network as a critical component of EP and CP symptom networks, but the specific functional connections used for prediction differ. Finally, group differences compared to healthy controls (n=150) were observed for CP but not EP. These differences broadly overlapped with the symptom model for both EP and CP. Our results encourage using longitudinal studies to track connectivity changes in putative symptom networks during the progression of psychosis, as they may be explicative of EP-CP treatment differences.