Disease-specific alterations of effective connectivity across anti-correlated networks in major depressive disorder and bipolar disorder

Abstract

Major depressive disorder (MDD) and bipolar disorder (BD) share various clinical behaviors and have confounded clinical diagnoses. Converging studies have suggested MDD and BD as disorders with abnormal communication among functional brain networks involved in mental activity and redirection. However, whether MDD and BD show disease-specific alterations in network information interaction remains unclear. This study collected resting-state functional MRI data of 98 patients with MDD, 55 patients with BD, and sex-, age-, and education-matched 95 healthy controls. Spectral dynamic causal model (spDCM) was used to investigate effective connectivities among three large-scale intrinsic functional networks including the default mode network (DMN), salience network (SN), and dorsal attention network (DAN). Effective connectivities showing disease-specific changes were then used as input features of support vector models to predict clinical symptoms and classify individuals with MDD and BD. Compared with healthy controls, both the MDD and BD groups showed increased DAN → SN connectivity. However, within-network connectivities of DMN and DAN showed opposite effects on the diseases. Notably, MDD and BD also showed different alterations on a connectivity loop of SN → DAN → DMN → SN, which could be used to predict the clinical symptom severity of either MDD or BD. Individuals with MDD and BD could be further classified by using connectivities showing opposite disease effects. Our findings reveal common and unique alterations of network interactions in MDD and BD, and further suggest disease-specific neuroimaging markers for clinical diagnosis.