Predicting blood pressure without a cuff using a unique multi-modal wearable device and machine learning algorithm.

Area of Science:

• Biomedical Engineering
• Cardiovascular Health Technology
• Machine Learning in Healthcare

Background:

• Elevated blood pressure is a major risk factor for cardiovascular diseases (CVDs).
• Current cuff-based blood pressure monitoring is inconvenient, uncomfortable, and not continuous.
• Frequent monitoring is crucial for preventing serious cardiovascular complications.

Purpose of the Study:

• To develop a multi-modal wearable device for accurate cuffless blood pressure prediction.
• To utilize a random forest regression (RFR) algorithm for enhanced prediction accuracy.
• To establish a novel human subject study protocol for device validation.

Main Methods:

• A multi-modal wearable device integrating two photoplethysmography (PPG) and two bioimpedance (BioZ) sensors.
• Measurement of pulse wave propagation along the radial artery.
• Application of a random forest regression (RFR) algorithm using multi-modal sensor inputs.

Main Results:

• The RFR model achieved high accuracy in cuffless blood pressure prediction.
• Mean absolute errors for systolic and diastolic blood pressures were below 3.3 mmHg across datasets.
• Multi-modal sensor input demonstrated higher accuracy compared to single-sensor models.

Conclusions:

• The developed multi-modal wearable device and RFR model show potential for robust, continuous blood pressure monitoring.
• This technology offers a practical solution for long-term cardiovascular health management.
• Further validation in diverse populations is needed to establish a universal cuffless blood pressure estimation model.