Analysis of EEG-derived brain networks for predicting rTMS treatment outcomes in MDD patients

Major depressive disorder (MDD) is a common and debilitating mental illness. One of the MDD treatments is Repetitive transcranial magnetic stimulation (rTMS) which has shown promise in treating MDD but predicting individual patient response remains a challenge. In this paper, we analyzed EEG signals recorded at four different time points including baseline, and after 3rd, 6th, and 10th rTMS sessions in 18 MDD patients receiving rTMS treatment. Brain networks were constructed using partial transfer entropy of EEG signals in four frequency bands. Graph theory metrics were extracted from these networks, and their correlations with patients' depression levels during treatment were assessed. Furthermore, the ability of these networks' metrics obtained from EEG data of four separate time points in discerning between treatment responders and non-responders to treatment was assessed by classification analysis. Results showed a high correlation between depression severity and certain network metrics, such as node betweenness centrality, diameter, and local efficiency in the delta band, as well as global and local efficiency in alpha frequency bands. Furthermore, based on the results, brain network metrics derived from EEG signals collected at second week of rTMS treatment can predict treatment response with an accuracy of 94.44%. This study investigates the relation between brain network metrics and treatment outcomes for MDD and suggests that analyzing topological changes in brain networks may be a useful approach for predicting patient response to rTMS treatment.