Multi-frequency Dynamic Weighted Functional Connectivity Networks for Schizophrenia Diagnosis

Abstract

Frequency-specific functional connectivity (FC) networks based on resting-state functional magnetic resonance imaging (rs-fMRI) have been successfully applied to the analysis and diagnosis of various mental illnesses, such as schizophrenia. However, most of the existing frequency-specific FC studies just focus on investigating the static temporal properties of FC networks, ignoring the important dynamic characteristics and spatial properties of FC networks. To address these issues, we proposed novel dynamic weighted FC networks to investigate the interactions among distributed brain regions. To take full advantage of the dynamic characteristics of the networks, temporal, spatial and spatio-temporal variabilities of dynamic networks were extracted as the classification features. To validate the effectiveness of our proposed method, we performed experiments on subjects with baseline rs-fMRI data from SchizConnect database. Experimental results demonstrated that the proposed method outperforms the state-of-the-art approaches in schizophrenia identification. In addition, we found most of the discriminative features distributed in frontal and subcortical area, which coincide with the pathological regions of cognitive progressing in schizophrenia patients.