Enterprise AI Analysis
Contactless Depression Screening via Facial Video-Derived Heart Rate Variability
Depression is a prevalent mental health condition that frequently remains undiagnosed, highlighting the need for objective and scalable screening tools. Heart rate variability (HRV) has emerged as a potential physiological marker of depression, and facial video-based HRV measurement offers a novel, contactless approach that could facilitate widespread, non-invasive depression screening. We analyzed data from 1453 individuals who completed facial video recordings and the Patient Health Questionnaire-9 (PHQ-9). A stacking ensemble classifier was developed using HRV features and basic demographic information to classify individuals with depressive symptoms. The ensemble incorporated four base learners (logistic regression, gradient boosting, XGBoost, and SVM) with an SVM meta-learner. Model performance was evaluated using 5-fold cross-validation. The stacking model achieved its best discrimination of AUROC 0.64 (AUPRC 0.45 and MCC 0.21). Incorporating demographic features alongside HRV improved performance over HRV alone. Feature importance analysis revealed that smoking status, sex, and medical comorbidities were the strongest contributors to the predictions. Facial video-derived HRV, combined with simple demographic factors, can moderately distinguish individuals with depressive symptoms in a contactless manner. Although predictive performance was modest, this non-invasive approach shows promise for accessible, large-scale depression screening.
Key Executive Impact Metrics
Our analysis distills the core performance indicators and potential benefits for your enterprise from this cutting-edge research.
0.00
Stacking Model AUROC
0.00
Stacking Model MCC
0
Participants Analyzed
0.00
MCC in Obese Subgroup
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Methodology Overview
This study outlines the methodology for developing a contactless depression screening tool. It involved collecting facial video data from 1453 participants, alongside Patient Health Questionnaire-9 (PHQ-9) scores. A stacking ensemble machine learning model, integrating logistic regression, gradient boosting, XGBoost, and SVM, was developed and evaluated using 5-fold cross-validation. The model used both HRV features derived from facial videos and basic demographic information to classify individuals with depressive symptoms. The performance was assessed using AUROC, AUPRC, and MCC, with hyperparameters optimized for MCC.
Key Findings
The stacking model achieved an AUROC of 0.64, AUPRC of 0.45, and MCC of 0.21. Performance was improved by combining demographic features with HRV, rather than using HRV alone. Smoking status, sex, and medical comorbidities were identified as the strongest predictors. While predictive performance was modest overall, the model performed better in specific subgroups, such as participants with obesity (MCC: 0.65) and current smokers (MCC: 0.51), suggesting enhanced discrimination where autonomic, metabolic, and behavioral influences overlap. The approach demonstrates potential for accessible, large-scale depression screening, despite limitations in absolute accuracy.
Limitations & Future Work
Limitations include susceptibility of facial video-derived HRV to environmental artifacts (lighting, movement), reliance on self-reported PHQ-9 scores (though HADS-D was also used), and a sample primarily from clinical settings in South Korea. Future work should focus on improving artifact correction, adaptive signal normalization, validating the model across diverse populations, integrating it into a stepped-care model (e.g., as an initial filter), and exploring additional data streams for enhanced performance.
Enterprise Process Flow
2,021 participants screened
→
Initial Psychiatric Diagnosis Categorization (N=2,021)
→
568 Exclusions (Missing HRV/Information)
→
1,453 Participants Included for Analysis
→
PHQ-9 Based Classification: No Depressive (N=896) vs. Depressive Symptoms (N=557)
0.64
AUROC for Stacking Model (Combined features)
The stacking ensemble model achieved an Area Under the Receiver Operating Characteristic (AUROC) of 0.64 when combining HRV and demographic features, indicating its overall discriminative ability for identifying depressive symptoms.
| Feature Set | AUROC (Avg) | MCC (Avg) | Key Takeaway |
|---|---|---|---|
| HRV Alone | 0.45-0.55 | 0.05-0.10 | Limited predictive power, not very informative for discrimination. |
| Demographic Alone | 0.55-0.60 | 0.10-0.15 | Stronger predictive power than HRV alone, highly relevant factors. |
| Combined (HRV + Demographic) | 0.64 | 0.21 | Best performance, modest boost over demographics alone. |
Enhanced Prediction for Specific Subgroups
The model demonstrated substantially higher predictive performance in certain subgroups. For participants with obesity (BMI ≥ 30), the MCC reached 0.65. Similarly, for current smokers, the MCC was 0.51. This suggests that where autonomic, metabolic, and behavioral influences significantly overlap, the model's discrimination is enhanced, indicating stronger physiological differences.
Advanced ROI Calculator
Quantify the potential savings and efficiency gains for your organization by integrating our AI-powered contactless mental health screening solution.
Estimated Annual Savings
$0
Annual Hours Reclaimed
0
Phased Implementation Roadmap
A structured approach to integrating contactless AI screening into your enterprise, ensuring seamless adoption and measurable impact.
Phase 1: Pilot Program & Data Integration (Weeks 1-4)
Establish a small-scale pilot with a target employee group. Integrate existing HR and health data systems for baseline comparison and initial model training. Configure facial video collection in a controlled environment.
Phase 2: Model Customization & Validation (Weeks 5-12)
Refine and customize the AI model using your organization's specific data, focusing on local population demographics and health profiles. Conduct internal validation studies to ensure accuracy and fairness across subgroups.
Phase 3: Scaled Deployment & Training (Months 3-6)
Roll out the contactless screening solution across relevant departments or locations. Provide comprehensive training for HR, health, and management teams on interpreting results and privacy protocols. Establish support channels for users.
Phase 4: Continuous Monitoring & Optimization (Ongoing)
Implement a continuous feedback loop for model performance monitoring and recalibration. Regularly review screening data, user feedback, and mental health outcomes to refine the system and maximize its effectiveness.
Ready to Transform Your Enterprise with AI?
Schedule a personalized consultation with our AI specialists today to explore how contactless depression screening can benefit your organization.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat