A Wearable Device for Depression Detection with Machine Learning Using Physiological Biomarkers

Depression is a pervasive mental health disorder with increasing global prevalence, but traditional diagnostic methods often lack real-time monitoring of this disorder. This study presents the development of a wearable device that continuously tracks physiological biomarkers, specifically heart rate (HR), oxygen saturation (SpO2), and body temperature to predict the mental health of the user. Using an ESP32 microcontroller, the device processes data from MAX30100 and LM35 sensors and applies machine learning to classify the user's depression stage (no, mild, moderate, or severe) based on these parameters. Device data is transmitted via Bluetooth to a laptop where a Logistic Regression (LR) based machine learning model assesses depression severity. From all the models trained using various ML algorithms like K Nearest Neighbor, Random Forest, Gradient Boosting, etc, LR and Support Vector Machine (SVM) demonstrated the highest predictive accuracy, with Logistic Regression achieving 84.5% accuracy, 85.2% precision, 84.6% recall, and 84.8% F1-score, making it optimal for real-time monitoring. The wearable device displays results on an OLED screen and a mobile app, alerting caregivers in cases of moderate or severe depression. This system offers a proactive, cost-effective solution for mental health management, supporting both individual users and healthcare providers in tracking and intervening in depression.