Radar HRV Monitoring With Physiological Prior Inspired Deep Neural Networks.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Cardiovascular Physiology

Background:

• Heart Rate Variability (HRV) is a key indicator of cardiovascular and autonomic nervous system health.
• Existing contactless HRV monitoring methods using radar are limited by signal noise and variations in complex physiological conditions, restricting them to lab settings.
• Robust, real-world HRV monitoring is needed for diverse patient populations.

Purpose of the Study:

• To develop a robust, physiological prior-inspired deep learning framework for contactless HRV monitoring using radar.
• To address limitations of existing methods in real-world scenarios with complex physiological conditions.
• To validate the framework's performance on a large-scale dataset of outpatients.

Main Methods:

• A hybrid deep neural network was designed to model spatio-temporal relationships between full-body radar reflections and heartbeats, leveraging the prior that internal heartbeats drive torso movements.
• A signal augmentation strategy was developed based on the cardiac motion's self-similarity prior to improve HRV distribution and performance across diverse conditions.
• The framework was validated on a large-scale dataset of 7,150 outpatients with complex physiological conditions in real-world settings.

Main Results:

• The proposed method achieved low mean errors for key HRV metrics: IBI (19.21 ms), RMSSD (16.23 ms), SDSD (16.70 ms), and pNN50 (7.28%).
• Cardiac condition classification based on radar-derived HRV demonstrated performance comparable to electrocardiogram (ECG)-based methods.
• The framework showed significant potential for accurate, contactless HRV monitoring in complex, real-world patient scenarios.

Conclusions:

• The physiological prior-inspired deep learning framework offers a robust solution for contactless HRV monitoring using radar.
• This approach overcomes limitations of previous methods, enabling reliable monitoring in diverse and complex physiological conditions.
• The study highlights the potential for widespread clinical application of radar-based HRV monitoring.