Accurate Cardiac Duration Detection for Remote Blood Pressure Estimation Using mm-Wave Doppler Radar

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiovascular Technology

Background:

• Accurate blood pressure (BP) monitoring is vital for cardiovascular health management.
• Existing BP measurement methods often require physical contact or breath-holding, limiting continuous monitoring.
• Subtle cardiac motions crucial for BP estimation are challenging to detect with radar due to noise and respiration interference.

Purpose of the Study:

• To develop a radar-based model for contactless and continuous blood pressure estimation.
• To enhance the detection of subtle cardiac activities for improved BP monitoring accuracy.
• To overcome limitations of previous radar-based methods, particularly interference from respiration and noise.

Main Methods:

• Implementation of an integrated-spectrum waveform derived from short-time Fourier transform (STFT) for capturing minute cardiac activities.
• Application of a hidden semi-Markov model (HSMM) and a U-net model for cardiac feature extraction from the integrated spectrum.
• Focus on energy changes from high-frequency vibrations, effectively filtering out respiration and noise.

Main Results:

• Significant reduction in root mean square error (RMSE) for interbeat intervals (87.5%), systolic time (88.7%), and diastolic time (73.1%).
• Achieved comparable prediction accuracy during normal breathing, unlike prior methods requiring breath-holding.
• Reported low errors for diastolic BP (3.98±5.81 mmHg) and systolic BP (6.52±7.51 mmHg).

Conclusions:

• The proposed radar-based model enables accurate, touch-free blood pressure monitoring.
• The integrated-spectrum waveform and advanced feature extraction effectively isolate cardiac signals.
• This technology holds promise for continuous, non-invasive cardiovascular health assessment.