Deep ECG-Respiration Network (DeepER Net) for Recognizing Mental Stress.

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Occupational Health

Background:

• Workplace mental stress has significant health and productivity impacts.
• Timely recognition and relief of mental stress are crucial for prevention.
• Existing methods for stress detection may lack accuracy or real-time capabilities.

Purpose of the Study:

• To develop and validate a novel stress detection algorithm using end-to-end deep learning.
• To utilize multiple physiological signals for enhanced stress recognition.
• To assess the algorithm's performance against conventional machine learning models.

Main Methods:

• Recruited 18 subjects to simulate workplace stress using Stroop and math tasks.
• Measured electrocardiogram (ECG) and respiration (RESP) signals using Zephyr BioHarness 3.0.
• Applied a five-fold cross-validation strategy to evaluate the deep learning model.

Main Results:

• The proposed deep learning network achieved an average accuracy of 83.9%.
• The model demonstrated an average F1 score of 0.81 and an average AUC of 0.92.
• Neuron activation patterns correlated with specific ECG and RESP signals, indicating model interpretability.

Conclusions:

• End-to-end deep learning with multiple physiological signals is feasible for workplace mental stress recognition.
• The developed algorithm outperforms traditional machine learning approaches.
• This approach holds promise for improving the quality of life for individuals experiencing work-related stress.