Classification of major depressive disorder subjects using Pre-rTMS electroencephalography data with support vector machine approach

The combination of repetitive transcranial magnetic stimulation (rTMS) and electroencephalogram (EEG) is a valuable tool for investigating the functional connectivity in the brain. Using pre-treatment cordance, a relatively new quantitative EEG method combining complementary information from absolute and relative power of EEG spectra, 55 major depression disorder (MDD) subjects were classified into responder or non-responder classes. In order to predict the response of rTMS treatment, support vector machine (SVM) based classification was carried out on pre-treatment cordance and the classification performance was evaluated using 6, 8 and 10-fold cross-validation (CV). Promising findings indicate that it is possible to classify rTMS treatment responders with 85.45% overall accuracy with a sensitivity of 82.35% and 0.925 area under receiver operating characteristics (ROC) curve value.