Nightbeat: Heart Rate Estimation From a Wrist-Worn Accelerometer During Sleep

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Cardiovascular Monitoring

Background:

• Current fitness trackers and smartwatches primarily use optical sensors for heart rate (HR) monitoring.
• Large-scale studies like the U.K. Biobank utilize activity trackers lacking optical sensors, thus missing crucial HR data.
• This data gap limits the potential for discovering population-level health trends from existing longitudinal studies.

Purpose of the Study:

• To present the first dataset of wrist-worn accelerometer recordings with electrocardiogram (ECG) references in uncontrolled, at-home settings.
• To investigate the feasibility of Inertial Measurement Unit (IMU)-only heart rate estimation using ballistocardiograms (BCGs).
• To develop and validate a novel method for extracting HR from IMU data, robust to motion artifacts.

Main Methods:

• Collected a dataset of 310 hours of wrist-worn IMU and ECG recordings from 42 participants during nighttime at home.
• Introduced a novel frequency-based method employing curve tracing to extract HR from IMU signals.
• Developed artifact rejection techniques to improve the accuracy of IMU-based HR estimation.

Main Results:

• The proposed IMU-only method achieved a mean absolute error (MAE) of 0.88 bpm for HR estimation.
• This represents a 76% improvement compared to previous IMU-based approaches.
• The method successfully surpassed established medical standards for heart rate monitors.

Conclusions:

• IMU-only HR estimation using ballistocardiography is a viable and accurate method for cardiac activity monitoring.
• This technology can be integrated into existing longitudinal studies that utilize activity trackers without optical sensors.
• The findings highlight the potential for discovering novel health insights from previously untapped HR data in large populations.