Distinguishing depression from healthy controls using brain network features: A fNIRS and machine learning approach

Objective This exploratory study aimed to investigate potential brain network biomarkers of depression by examining local connectivity features using functional near-infrared spectroscopy (fNIRS). Method 31 depressed students and 32 health controls were recruited. Data was collected from both groups during resting-state and verbal fluency task (VFT). Results In the frontopolar region, depressed participants exhibited increased network connectivity during the resting state, while decreased connectivity was observed during the VFT. The AUC values for all classifiers exceeded 0.7, with the random forest model showing the highest AUC value and exhibiting strong specificity and sensitivity in VFT. Conclusions The identified local brain network features with group differences suggest potential biomarkers for distinguishing depressed students, providing valuable insights for understanding depression.