Deep learning in computer aided diagnosis of MDD

Abstract

Electroencephalogram (EEG) is a popular method for diagnosing various neurological diseases. Major Depressive Disorder (MDD) is a mental health disorder that can be diagnosed and treated by making use of EEG. One of the main challenges in using EEG to accurately identify depression is complexity and variation that exist in the EEG of a depressed person. Manually reading EEG and diagnosing depression is very challenging. An efficient computer aided method can be used for this task. Of the many methods that exists, a deep neural network method called Convolutional Neural Networks (CNN) proved to be the most efficient. In this paper a multi-layer deep CNN algorithm is implemented to diagnose depression from EEG of patients. Depression is classified based on a severity index into mild, moderate and major classes. The accuracy, sensitivity and specificity were measured by varying various parameters of the proposed algorithm.