Detecting MDD based on EEG signals: Frontal or Temporal Region

Psychological problems like depression affect a person's growth, including thoughts, feelings, and behaviors. There is no laboratory test for detecting depression, which is the main reason for the wrong diagnosis of depression. Analysis of MDD's underlying neurophysiological functions can improve the detection and treatment of this mental disorder. Increasingly, EEG is used to diagnose and study brain disorders and functions; in this study we introduced a subjective-based method to detect depression with the significance of decreasing the electrode montage required for recording the EEG signals. Features are extracted from the frontal and temporal regions of the brain using eight electrodes. The linear features used are delta, theta, alpha, and beta relative band powers and alpha absolute power. The nonlinear features used are Sample Entropy (sampEn) and Higuchi's fractal dimension (HFD). The classifiers used in this study are Support Vector Machine (SVM), Logistic Regression (LR), and naïve Bayes (NB). The highest classification accuracy of ۹۱.۶۷% with an F۱ score of ۹۴.۱۲% and Roc-Auc score of ۹۸.۴۴% were achieved for detecting depression using NB among the brain's frontal region. On the other hand, the highest classification accuracy among the right hemisphere of the temporal region was ۸۳.۳۴% with a Roc-auc score of ۹۰% and F۱ score of ۸۷.۵%. The analysis found that depression affects the frontal region of the brain and the left hemisphere of the temporal region more significantly with respect to the right hemisphere of the temporal region.

Keywords:

EEG , Major Depressive Disorder , Signal Processing , Machine Learning , Frontal and Temporal Region of the brain