Depression Detection using Resting State Three-channel EEG Signal

In universal environment, a patient-friendly inexpensive method is needed to realize the early diagnosis of depression, which is believed to be an effective way to reduce the mortality of depression. The purpose of this study is only to collect EEG signal from three electrodes Fp1, Fpz and Fp2, then the linear and nonlinear features of EEG used to classify depression patients and healthy controls. The EEG recordings were carried out on a group of 18 medication-free depressive patients and 25 gender and age matched controls. In this paper, the selected features include three linear (maximum, mean and center values of the power) and three nonlinear features (correlation dimension, Renyi entropy and C0 complexity). The accuracy and effectiveness of classification model between depressive and control subjects were calculated using leave-one-out cross-validation. The experimental results indicate that selected three channel EEG and features can distinguish the subjects between depression and normal beings, the classification accuracy is 72.25%. It is hoped that the performed results can provide more choices for the early diagnosis of depression in a universal environment.