Identifying major depressive disorder using Hurst exponent of resting-state brain networks

Abstract

Resting-state functional magnetic resonance imaging (fMRI) studies of major depressive disorder (MDD) have revealed abnormalities of functional connectivity within or among the resting-state networks. They provide valuable insight into the pathological mechanisms of depression. However, few reports were involved in the "long-term memory" of fMRI signals. This study was to investigate the "long-term memory" of resting-state networks by calculating their Hurst exponents for identifying depressed patients from healthy controls. Resting-state networks were extracted from fMRI data of 20 MDD and 20 matched healthy control subjects. The Hurst exponent of each network was estimated by Range Scale analysis for further discriminant analysis. 95% of depressed patients and 85% of healthy controls were correctly classified by Support Vector Machine with an accuracy of 90%. The right fronto-parietal and default mode network constructed a deficit network (lower memory and more irregularity in MDD), while the left fronto-parietal, ventromedial prefrontal and salience network belonged to an excess network (longer memory in MDD), suggesting these dysfunctional networks may be related to a portion of the complex of emotional and cognitive disturbances. The abnormal "long-term memory" of resting-state networks associated with depression may provide a new possibility towards the exploration of the pathophysiological mechanisms of MDD.

Keywords: Classification; Functional magnetic resonance imaging (fMRI); Hurst exponent; Major depressive disorder; Resting-state networks.