Predicting depression severity using effective and functional brain connectivity of the electroencephalography...

Depression, also known as major depressive disorder (MDD), is a mental health condition that can lead to self-injury and suicide with significant effects on individuals and communities. Recent studies suggest that analysing functional connectivity (FC) from electroencephalography (EEG) signals provides insights into brain network integration in depressive states. Effective connectivity (EC) assesses the directional influence between brain regions, offering deeper insights into neural circuit dynamics. This study aimed to capture the subtle changes in brain dynamics, identify predictive biomarkers of MDD, and elucidate its neurophysiological basis. Resting-state EEG signals from 44 subjects with MDD were used to extract connectivity features. Graph-theoretical-based EC features from the phase slope index (PSI) and FC features from the weighted phase lag index (WPLI) were analysed. Correlation analysis showed significant associations between EC features (diameter) and depression severity, as well as between FC features (global efficiency) and severity, both in the alpha frequency band. These significant features were fed into several machine learning regression models, which demonstrated comparable performance in predicting depression scores. FC features performed slightly better (4.71 root mean square error and 3.93 mean absolute error) than EC features (5.01 root mean square error and 4.29 mean absolute error). These findings indicate that alterations in functional and effective connectivity are linked to depression severity and could improve diagnostic accuracy and therapeutic strategies, while offering new avenues for research into brain connectivity in MDD.