Depression classification with MDD (Major Depressive Disorder) using signal processing and machine learning

Abstract

Depression is accorded as one of the leading causes to all the problems related to mental health in the Global disease burden study (GBD). Major depressive disorder (MDD) is when this depression reaches to a larger extent, when depression persists for two weeks or more. Sadly, many individuals of our society tend to neglect depression and refuse to label it as a mental disease and has a tendency to not seek medical help. Not only this, they are being curbed because of the few or very limited biological indicators for MDD and depression identification. Our main objective is to develop a non-intrusive approach that will detect and differentiate brain signals of patients with MDD from healthy patients. We were able to obtain an optimized model with an accuracy of (82%). Primarily we obtained a raw EEG data-set upon research and since it matched our requirements, we performed noise removal on them. Afterwards we extracted relevant features for depression detection, such as one feature was Absolute delta power. Finally, we entered these features into three classification algorithms; Logistic Regression (LR), Support Vector Machine (SVM) and Negative-Bayes (NB). To check the accuracy and precision, we performed a ten-fold cross validation on them. Hopefully, our results will encourage and motivate people suffering from this to seek the proper and effective medical help and to eradicate the negative stigma around it.