Establishment of Effective Biomarkers for Depression Diagnosis With Fusion of Multiple Resting-State Connectivity Measures

Major depressive disorder (MDD) is a severe mental disorder and is lacking in biomarkers for clinical diagnosis. Previous studies have demonstrated that functional abnormalities of the unifying triple networks are the underlying basis of the neuropathology of depression. However, whether the functional properties of the triple network are effective biomarkers for the diagnosis of depression remains unclear. In our study, we used independent component analysis to define the triple networks, and resting-state functional connectivities (RSFCs), effective connectivities (EC) measured with dynamic causal modeling (DCM), and dynamic functional connectivity (dFC) measured with the sliding window method were applied to map the functional interactions between subcomponents of triple networks. Two-sample t-tests with p < 0.05 with Bonferroni correction were used to identify the significant differences between healthy controls (HCs) and MDD. Compared with HCs, the MDD showed significantly increased intrinsic FC between the left central executive network (CEN) and salience network (SAL), increased EC from the right CEN to left CEN, decreased EC from the right CEN to the default mode network (DMN), and decreased dFC between the right CEN and SAL, DMN. Moreover, by fusion of the changed RSFC, EC, and dFC as features, support vector classification could effectively distinguish the MDD from HCs. Our results demonstrated that fusion of the multiple functional connectivities measures of the triple networks is an effective way to reveal functional disruptions for MDD, which may facilitate establishing the clinical diagnosis biomarkers for depression.