Temporal dynamic alterations of regional homogeneity in major depressive disorder: a study integrating machine learning.

Abstract

Previous studies have found alterations in the local regional homogeneity of brain activity in individuals diagnosed with major depressive disorder. However, many studies have failed to consider that even during resting states, brain activity is dynamic and time-varying. The lack of investigation into the dynamic regional homogeneity has hindered the discovery of biomarkers for depression. This study aimed to assess the utility of the dynamic regional homogeneity by a machine learning model (support vector machine). Sixty-five individuals with dynamic regional homogeneity and 57 healthy controls participated in resting-state functional magnetic resonance rescanning and scale estimating. The dynamic regional homogeneity and receiver operating characteristic curve methods were used for analysis of the imaging data. Relative to healthy controls, major depressive disorder patients displayed increased dynamic regional homogeneity values in the left precuneus and right postcentral gyrus. Additionally, receiver operating characteristic curve results of the dynamic regional homogeneity values in the left precuneus and right postcentral gyrus could distinguish major depressive disorder patients from healthy controls; furthermore, changes in the dynamic regional homogeneity were correlated with depression severity.