Stress Detection Using Heart Rate Variability and Respiratory Signals Derived From a Single-Lead ECG.

Area of Science:

• Biomedical Engineering
• Computational Physiology
• Machine Learning in Healthcare

Background:

• Stress detection is crucial for health, but current multimodal methods face hardware and computational limits for wearables.
• Existing single-modality approaches, like those using only heart rate variability (HRV), have limitations in accuracy and efficiency.

Purpose of the Study:

• To develop an efficient and accurate stress detection method using exclusively single-lead electrocardiogram (ECG) signals.
• To investigate the utility of combining ECG-derived respiratory and HRV features with machine learning for real-time stress monitoring.

Main Methods:

• A hybrid methodology was employed, extracting HRV and respiratory signals and their features from single-lead ECG data.
• The XGBoost machine learning model was utilized to evaluate various feature combinations on the ES3 project database.

Main Results:

• Incorporating ECG-derived respiratory features significantly enhanced classification accuracy and computational efficiency over traditional HRV methods.
• Feature importance analysis revealed a minimal set of key features, leading to a highly efficient model with faster inference times than deep learning models.
• The proposed approach demonstrated superior performance and efficiency compared to deep learning models for stress detection.

Conclusions:

• Single-lead ECG-based multimodal analysis, combining feature extraction with machine learning, is feasible for acute stress detection.
• This approach offers a more accessible strategy for biomedical monitoring, providing insights into physiological stress responses.
• The developed method overcomes hardware and computational challenges, paving the way for real-time stress detection in wearable devices.