Using a Bodily Weight-Fat Scale for Cuffless Blood Pressure Measurement Based on the Edge Computing System.

Area of Science:

• Biomedical Engineering
• Health Informatics
• Machine Learning

Background:

• Accurate blood pressure (BP) monitoring is crucial for managing cardiovascular diseases, diabetes, and kidney disease, especially in elders.
• Cuffless BP measurement technologies offer enhanced user comfort and convenience.
• Ballistocardiogram (BCG) and impedance plethysmogram (IPG) are emerging physiological signals for cuffless BP estimation.

Purpose of the Study:

• To develop and validate an edge computing system for real-time cuffless blood pressure measurement.
• To integrate BCG and IPG signal acquisition, processing, feature extraction, and machine learning-based BP estimation.
• To implement the system on an embedded platform (STM32F756ZG NUCLEO) for mobile health applications.

Main Methods:

• BCG and IPG signals were acquired using a custom bodily weight-fat scale.
• Signal processing involved filtering, segmentation, and extraction of pulse transit time (PTT).
• An XGBoost machine learning model, with optimized hyperparameters for edge deployment, estimated BP using calibration-based and calibration-free features.

Main Results:

• The edge computing system demonstrated low error rates for systolic blood pressure (SBP) and diastolic blood pressure (DBP) estimation: 2.2 ± 10.9 mmHg and 1.87 ± 6.79 mmHg, respectively.
• These results are comparable to server-based computing, indicating the feasibility of on-device BP measurement.
• The system's performance validates its potential for integration into mobile health devices.

Conclusions:

• Edge computing enables real-time, cuffless blood pressure measurement using BCG and IPG signals on embedded systems.
• The proposed method enhances the practicality of bodily weight-fat scales for continuous BP monitoring.
• This technology holds significant promise for advancing mobile health and remote patient care.