Classifying depression patients and normal subjects using machine learning techniques

Depression is one of the most common mental disorder that at its worst can lead to suicide. Diagnosing depression in the early curable stage is very important. In this paper we study performance of different classification techniques for classifying depression patients from normal subjects. For this aim, power spectrum of three frequency band (alpha, beta, theta) and the whole bands of EEG are used as features. We have shown that Support Vector Machine (SVM) classifier using Genetic algorithm for feature selection can achieve accuracy of 88.6% on classifying depression patients.