Classification and Detection of Cognitive Disorders like Depression and Anxiety Utilizing Deep Convolutional Neural Network (CNN) Centered on EEG Signal

Electroencephalography (EEG) is a test performed to assess the electrical signals spontaneously produced during brain activities. In recent years, it is popularly used for studying both normal and pathological changes occurring in the human brain. With the World Health Organization (WHO) listing psychological disorders as a major health issue faced by the modern society, the current work focuses on this niche. It categorizes cognitive impairment like depression and anxiety using a computer-aided machine learning approach called Convolutional Neural Network. The deep CNN is trained using EEG signals from 30 patients suffering from depression and 30 others suffering from anxiety. Initially, the signal is preprocessed using Fractional Order Butterworth Filter (FOBF). The work considers the occurrence of ultra-damped, hyper-damped, and under-damped poles while designing a FOBF in a composite w-plane (w=sq; where, q is a real number). As usually executed for integer order filters in a composite w-plane, the primary initial fractional Butterworth filter is employed. The characteristics of each electrode's gamma, theta, delta, beta, alpha, and full-band EEG are then analyzed. This results in the removal of 270 nonlinear and linear characteristics. The feature space's dimensions are then reduced using a feature selection approach called Minimal-Redundancy-Maximal-Relevance (MRMR). The EEG characteristics are finally categorized by utilizing the suggested deep CNN, Artificial Neural Network (ANN) and K-Nearest Neighbor (KNN). The accuracy of classification of the proposed approach is evaluated and found to be 97.6%. This shows it is promising for detecting depression and anxiety symptoms accurately and cost-effectively.