A Data-Driven Model with Feedback Calibration Embedded Blood Pressure Estimator Using Reflective Photoplethysmography

Area of Science:

• Biomedical Engineering
• Cardiovascular Technology
• Wearable Health Devices

Background:

• Ambulatory blood pressure monitoring (ABPM) is crucial for cardiovascular health assessment, with guidelines recommending out-of-office measurements.
• Traditional ABPM devices often use the oscillometric method, which can disrupt sleep, limiting all-day monitoring capabilities.
• Photoplethysmography (PPG)-based wrist devices offer a noninvasive, convenient alternative for continuous BP estimation.

Purpose of the Study:

• To develop and evaluate a novel, data-driven adaptive calibration model for a PPG-based wrist-type ABPM device.
• To improve the accuracy and reliability of long-term, out-of-office blood pressure measurements.
• To address the limitations of traditional ABPM by minimizing sleep interruption and inter/intra-subject variability.

Main Methods:

• A PPG-based wrist-type device was utilized to estimate blood pressure (BP) using morphological features from PPG waveforms.
• A novel adaptive calibration model, incorporating a 15-second PPG waveform and personal data, was developed.
• The model employed exponential Gaussian process regression for feedback calibration, ensuring accuracy and minimizing variability.

Main Results:

• The adaptive calibration model demonstrated high accuracy in BP estimation, with systolic BP error of -0.1776 ± 4.7361 mmHg and diastolic BP error of -0.3846 ± 6.3688 mmHg.
• Achieved success rates of 99.225% for systolic BP and 98.191% for diastolic BP within the ±15 mmHg criterion.
• Results met the stringent standards of the Association for the Advancement of Medical Instrumentation and the British Hypertension Society Grading criteria.

Conclusions:

• The proposed data-driven adaptive calibration model significantly enhances the accuracy and clinical utility of PPG-based ABPM.
• This wearable technology offers a promising solution for noninvasive, continuous, and comfortable ambulatory blood pressure monitoring.
• Machine learning with feedback calibration is effective for assessing ABPM for clinical applications.