A method for blood pressure hydrostatic pressure correction using wearable inertial sensors and deep learning

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Cardiovascular Health

Background:

• Cuffless noninvasive blood pressure (BP) measurement offers potential for early detection of abnormal BP patterns.
• Sensor placement away from heart level, such as on the arm, compromises BP measurement accuracy due to hydrostatic pressure variations.
• Traditional hydrostatic pressure correction methods involve bulky fluid-filled tubes, limiting unobtrusiveness.

Purpose of the Study:

• To introduce IMU-Track, an alternative method for correcting hydrostatic pressure variations in cuffless BP measurements.
• To enable accurate, unobtrusive BP monitoring using wearable sensors placed away from heart level.
• To validate the efficacy of IMU-Track in real-world conditions across diverse participants.

Main Methods:

• Developed IMU-Track, a deep learning model analyzing motion data from wearable inertial sensors.
• Calculated parameterized arm-pose coordinates using IMU-Track for sensors placed on the arm.
• Integrated IMU-Track with BP measurements derived from electrocardiography and finger photoplethysmography.

Main Results:

• IMU-Track significantly reduced mean absolute errors for systolic BP from 13.5 ± 1.1 mmHg to 5.9 ± 0.7 mmHg and for diastolic BP from 15.0 ± 1.0 mmHg to 6.8 ± 0.5 mmHg at 25 cm below heart level.
• Similar error reductions were observed at 25 cm above heart level for both systolic and diastolic BP.
• The arm-tracking inference time on a commercial smartphone was approximately 134 ms, suitable for real-time correction.

Conclusions:

• IMU-Track effectively corrects for hydrostatic pressure variations in cuffless BP measurements.
• This method facilitates accurate passive cuffless BP monitoring at positions away from heart level.
• IMU-Track supports the development of unobtrusive wearable BP monitors for continuous health tracking.