Non-invasive EEG based Feature Extraction framework for Major Depressive Disorder analysis

Abstract

Depression and several other behavioral health disorders are serious public health concerns worldwide. Persistent behavioral health issues have a wide range of consequences that affect people personally, culturally and socially. Major depressive disorder (MDD) is a psychiatric ailment that affects people of all ages worldwide. It has grown into a major global health issue as well as an economic burden. Clinicians are using several medications to limit the growth of this disease at an early stage in young people. The goal of this research is to improve the depression diagnosis by altering Electroencephalogram (EEG) signals and extracting the Differential Entropy (DE) and Power Spectral Density (PSD), using machine learning and deep learning techniques. This study analyzed the EEG signals of 30 healthy people and 34 people with Major Depressive Disorder (MDD). K-nearest neighbors (KNN) had the highest accuracy among machine learning algorithms of 99.7%, while Support vector machine (SVM) had acquired 95.7% accuracy. The developed Deep Learning approach, convolution neural network (CNN), achieved 99.6% accuracy. With these promising results, this study establishes the viability of an Electroencephalogram based diagnosis of MDD.