EEG Based Machine Learning Models for Automated Depression Detection

Mental disorders are one of the main causes of diseases and have a significant impact on people’s social and economic well-being. This work focuses on the classification of electroencephalogram (EEG) signals of healthy and depressed subjects using different machine learning models. The EEG is an electro-biological measurement tool that captures the electrical activity of the brain. The variations in characteristics of the EEG of healthy and depressed subjects are used to develop machine learning models for automated classification. By leveraging the individual’s recorded EEG, a machine learning model is developed to separate healthy and depressed individuals The EEG signal collected from the open database is preprocessed with the use of notch filters and an Independent Component Analysis (ICA) module. A set of time and frequency domain features are extracted from EEG signals. Using feature ranking algorithm the features are ranked and the significant features are selected. These features signify the prominent brain region associated with depression and are used for classifying the healthy and depressed subjects. Later, to evaluate the model’s effectiveness and to improve reliability and its possibilities of being implemented in real-time, the performance of different machine learning models such as k-nearest neighbour algorithm (KNN), Support Vector Machine (SVM), logistic regression, and Naïve Bayes is compared.