Abstract— Depression is a common reason for an increase insuicide cases worldwide. Thus, to mitigate the effects of depres-sion, accurate diagnosis and treatment are needed. An electroen-cephalogram (EEG) is an instrument used to measure and recordthe brain’s electrical activities. It can be utilized to produce theexact report on the level of depression. Previous studies provedthe feasibility of the usage of EEG data and deep learning (DL)models for diagnosing mental illness. Therefore, this studyproposes a DL-based convolutional neural network (CNN) calledDeprNet for classifying the EEG data of depressed and normalsubjects. Here, the Patient Health Questionnaire 9 score is usedfor quantifying the level of depression. The performance ofDeprNet in two experiments, namely, the recordwise split and thesubjectwise split, is presented in this study. The results attainedby DeprNet have an accuracy of 0.9937, and the area underthe receiver operating characteristic curve (AUC) of 0.999 isachieved when recordwise split data are considered. On the other hand, an accuracy of 0.914 and the AUC of 0.956 are obtained,while subjectwise split data are employed. These results suggestthat CNN trained on recordwise split data gets overtrained onEEG data with a small number of subjects. The performance ofDeprNet is remarkable compared with the other eight baselinemodels. Furthermore, on visualizing the last CNN layer, it isfound that the values of right electrodes are prominent fordepressed subjects, whereas, for normal subjects, the values ofleft electrodes are prominent.Index Terms— Convolutional neural network (CNN), electroen-cephalography, measurement of depression, pattern classification,visualization.

DeprNet: A Deep Convolution Neural Network Framework for Detecting Depression Using EEG PDF

The list of books you might like