Blood Pressure Monitoring System Using a Two-Channel Ballistocardiogram and Convolutional Neural Networks

Area of Science:

• Biomedical Engineering
• Cardiovascular Health

Background:

• Hypertension is a leading global cause of mortality, necessitating accurate and continuous blood pressure monitoring.
• Current methods for blood pressure monitoring can be invasive or inconvenient for long-term use.

Purpose of the Study:

• To develop and validate a non-invasive, sensor-free system for continuous blood pressure estimation using ballistocardiography (BCG).
• To assess the system's accuracy in estimating systolic and diastolic blood pressures (SBP and DBP) under resting and post-exercise recovery conditions.

Main Methods:

• A chair-shaped ballistocardiogram (BCG) system was developed to record two-channel BCG signals without attached sensors.
• Empirical mode decomposition and Hilbert transform were used for BCG signal noise reduction and instantaneous phase calculation.
• A convolutional neural network regression model was trained to predict SBP and DBP from BCG phase data.

Main Results:

• The developed system demonstrated accurate blood pressure estimation, meeting Association for the Advancement of Medical Instrumentation (AAMI) international standards during rest.
• The model successfully estimated rapidly rising blood pressure in the recovery state after treadmill exercise.
• The standard deviation for SBP in the recovery session exceeded 0.7, indicating potential areas for further refinement.

Conclusions:

• A non-invasive, chair-based BCG system can accurately estimate blood pressure, offering a promising tool for hypertension diagnosis and management.
• The system shows potential for continuous monitoring, particularly in assessing blood pressure dynamics during physiological stress and recovery.
• Further research may focus on improving accuracy in dynamic states like post-exercise recovery.