Prescreening depression using wearable electrocardiogram and photoplethysmogram data from a psycholinguistic...

Area of Science:

• Cardiology
• Psychiatry
• Biomedical Engineering

Background:

• Depression is a widespread mental health disorder impacting quality of life.
• Current diagnostic methods can be subjective and time-consuming.
• Objective biomarkers for depression prescreening are needed.

Purpose of the Study:

• To investigate the relationship between depression and cardiovascular function.
• To explore time-series features from ECG and PPG as potential depression biomarkers.
• To develop machine learning models for depression risk classification and severity prediction.

Main Methods:

• Collected ECG and PPG data from 60 participants with varying depression levels.
• Assessed depression using Beck Depression Inventory-II (BDI-II) and Patient Health Questionnaire-9 (PHQ-9).
• Developed machine learning models (Random Forest, XGBoost, Logistic Regression, SVM) for classification and regression.

Main Results:

• Specific ECG (RR interval variability) and PPG features differed significantly between healthy and depressed individuals.
• Support Vector Machine (SVM) achieved high classification performance (AUROC 0.83 for BDI-II, 0.78 for PHQ-9).
• SHAP analysis identified systolic-SD1 and RR-SD1 as key predictors; regression models predicted depression scores with low error.

Conclusions:

• Wearable ECG and PPG technologies show feasibility for depression prescreening.
• Cardiac activity-based biomarkers offer a cost-effective, objective, and non-invasive approach to mental health assessment.
• These findings support the development of novel tools complementing traditional depression diagnostics.