Identification of depression using support vector machine with different connectivity

Abstract

Patients with depression have shown attention bias and inhibition deficits of negative emotional valance. However, it is not clear how the distributed brain networks support the function of inhibition and whether the modulation altered in depression. Thirty-seven patients with depression and 37 matched controls were undertook the audio-visual emotion task-fMRI and whole-brain psychophysiological interaction analysis was employed to obtain three different types of connectivity features, including task-modulated connectivity (TMC), task-independent connectivity (TMC) and task-functional connectivity (TFC). Support vector machine method was used to classify depression and explore the relation of reaction time prediction. Results indicated decreased modulation related to frontoparietal cortex, increased in temporal lobe and sensorimotor system in depression. Moreover, TMC performed better in predicting RT, while TIC and TFC had better classification performance. This study reveals that aberrant modulation of neural response is widely associated with the inhibitory dysfunction in depression and support that different connectivity features provide supplementary information for underpinning the functional integration and its alterations of brain networks.