Utilizing Multi-PPG-Sensor Site Information in a Localized Wrist Area for Improving Cuffless Blood Pressure...

Area of Science:

• Biomedical Engineering
• Physiological Measurement
• Wearable Technology

Background:

• Blood pressure (BP) measurement accuracy is sensitive to sensor placement.
• Photoplethysmography (PPG) offers a potential avenue for cuffless BP monitoring.
• Variability in PPG signal features across different sensor sites is not well understood.

Purpose of the Study:

• To investigate the impact of multi-sensor site placement on BP estimation accuracy using PPG.
• To develop and validate a novel deep learning model (3M-BPNet) for enhanced BP estimation.
• To identify optimal sensor configurations for cuffless BP measurement devices.

Main Methods:

• Utilized a 9-channel single-wavelength PPG sensor array on the wrist.
• Analyzed site-specific variations of PPG features.
• Developed 3M-BPNet incorporating signal trimming, Bayesian optimization for channel weighting, and a personalized Random Forest Regression (RFR) model.
• Evaluated the model on 121 subjects.

Main Results:

• 3M-BPNet achieved MAEs of 4.84 mmHg (SBP) and 3.28 mmHg (DBP) across all subjects.
• BP estimation accuracy improved with channel counts up to 5, then declined.
• An optimal 5-channel configuration near the radial artery yielded MAEs of 1.33 mmHg (SBP) and 1.16 mmHg (DBP).
• This 5-channel setup demonstrated significant accuracy gains (72.6% SBP, 73.2% DBP) over single-channel measurement and outperformed prior PPG methods.

Conclusions:

• Sensor location critically influences PPG-based BP estimation accuracy.
• Optimized multi-sensor placement, particularly near the radial artery, enhances cuffless BP monitoring.
• Findings guide the design of accurate wearable BP devices for improved hypertension management.