Mann-Whitney U Test Based Identification of Significant ECG Features for Improving Accuracy of Major Depressive Disorder Classification

Major Depressive Disorder (MDD) is a composite mental state manifested through continuous negative thoughts, mood disturbances, and frequent thoughts of self-harm. Existing MDD diagnostic methods rely exclusively on mental health assessment questionnaires. There is a pressing need for an AI-driven approach to automate the detection of MDD, assisting healthcare practitioners. Since MDD has been shown to affect cardiovascular systems, this study aims to identify significant Electrocardiogram (ECG) features and automatically detect MDD using them. The study acquires ECG signals from 30 subjects, including 15 healthy subjects (HS) and 15 MDD patients. A total of 21 features are derived from ECG. Out of 21 features, significant ECG features are identified using 3 methods: Area Under the Receiver Operating Characteristic curve (AUC-ROC), Least Absolute Shrinkage and Selection Operator (LASSO), and Mann-Whitney U test (ρ < 0.001). Four Machine Learning (ML) classifiers: Random Forest (RF), Support Vector Machine (SVM), K Nearest Neighbor (KNN), Decision Tree (DT) with Grid Search optimization, are used to test classification accuracy obtained using features selected by 3 methods. The best classification accuracy (92.99%) is achieved by 16 significant features selected using the Mann-Whitney U test with an SVM-radial basis function (rbf) classifier. This is almost 2% more than the accuracy obtained using all 21 features.