Deep Learning-Based Classification of the Psychiatric Symptoms Severity
Authors
Ham Jinsil ; Oh Jooyoung
Citation
2023 IEEE 19TH INTERNATIONAL CONFERENCE ON BODY SENSOR NETWORKS, BSN, , 2023-10
Journal Title
2023 IEEE 19TH INTERNATIONAL CONFERENCE ON BODY SENSOR NETWORKS, BSN
Issue Date
2023-10
Abstract
Mental disorders are widely recognized as a major contributor to the global burden of disease. Heart rate variability (HRV) serves as an objective quantitative measure of autonomic nervous system (ANS) dysregulation and can be used to investigate psychiatric symptoms, including depression and anxiety. The aim of this study is to differentiate the severity of depression and anxiety, two prevalent symptoms of mental disorders, based on HRV. The fundamental deep learning architecture, specifically the Multilayer Perceptron (MLP) network, was employed for the classification of severity of both symptoms. By leveraging deep learning network, the classification of depressive symptoms achieved an accuracy of 83.8%, while the classification of anxious symptoms achieved an accuracy of 78.4%, demonstrating superior discrimination power compared to the conventional machine learning models. Clinical Relevance- Using the basic deep learning network called Multilayer Perceptron (MLP), it is possible to classify the severity of depression and anxiety based on heart rate variability (HRV) with accuracies of 83.8% and 78.4% respectively.