A Radar-Based Contactless System for Joint Phonocardiogram Reconstruction and Cardiac State Segmentation Using a...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Contactless vital signs monitoring is crucial where traditional sensors are not feasible.
• Radar systems offer a promising avenue for non-invasive physiological monitoring.

Purpose of the Study:

• To develop a radar-based system for contactless phonocardiogram (PCG) waveform reconstruction and cardiac state segmentation.
• To evaluate the system's performance using a public dataset and compare it with conventional methods.

Main Methods:

• Utilized a self-attention one-dimensional (1D) U-Net model for processing radar-derived signals.
• Estimated PCG-like waveforms, their envelopes, and segmented cardiac phases (S1, systole, S2, diastole).
• Validated the approach on a synchronized radar-PCG dataset using a 24 GHz Doppler radar and digital stethoscope.

Main Results:

• Achieved a 13.4885 dB reduction in log-spectral distance, indicating enhanced waveform fidelity.
• Improved cardiac phase segmentation with Micro-F1 increasing from 74.41% to 84.17% and Macro-F1 from 68.40% to 80.43%.
• Demonstrated the potential for real-time, low-power embedded deployment for continuous cardiac monitoring.

Conclusions:

• The radar-based system effectively reconstructs PCG waveforms and segments cardiac states contactlessly.
• Challenges remain in S2 detection due to respiratory interference, low signal amplitude, and subject motion.
• The technology is viable for contactless auscultation and continuous cardiac monitoring applications.