Dynamic patterns within the default mode network in schizophrenia subgroups.

Abstract

In this study, resting-state functional magnetic resonance imaging (rs-fMRI) data of 125 schizophrenia (SZ) subjects were analyzed. Based on SZ demographic information and cognitive scores and using an unsupervised clustering method, we identified subgroups of patients and compared DMN dynamic functional connectivity (dFC) between the groups. We captured seven independent subnodes, including anterior cingulate cortex (ACC), posterior cingulate cortex (PCC), and precuneus (PCu), in the DMN by applying group independent component analysis (group-ICA) and estimated dFC between component time courses using a sliding window approach. By using k-means clustering, we separated the dFCs into three reoccurring brain states. Using the statistical method, we compared the state-specific DMN connectivity pattern between two SZ subgroups. In addition, we used a transition probability matrix of a hidden Markov model (HMM) and occupancy rate (OCR) of each state between two SZ subgroups. We found SZ subjects with higher positive and negative syndrome scale (PNASS) showed lower within ACC and lower ACC and PCC connectivity (or ACC/PCC). In addition, we found the transition from state1 to same state is significantly different between two groups, while this result was not significant after multiple comparison tests.