Automatic Depression Discrimination on fNIRS by using GLM and SVM


Abstract—A method is proposed to distinguish patients with depression from healthy persons using data measured by Functional Near Infrared Spectroscopy (FNIRS) during a cognitive task. Firstly, General Linear Model (GLM) is used to extract features from 52-channel FNIRS data of patients with depression and normal healthy persons. Then a Support Vector Machine (SVM) classifier is designed for classification. The results of experiment show that the method can achieve a satisfactory classification with the accuracy 89.71% for total and 92.59% for patients. Also, the results suggest that FNIRS is a promising clinical technique in the diagnosis and therapy of depression.