Maximum a posteriori detection of heartbeats from a chest-worn accelerometer

Area of Science:

• Biomedical Engineering
• Cardiovascular Physiology
• Signal Processing

Background:

• Long-term cardiac monitoring is crucial for assessing illness severity and sleep patterns.
• Current methods can be obtrusive, limiting continuous, unobtrusive monitoring.
• Accelerometers offer a potential solution for wearable, long-term physiological tracking.

Purpose of the Study:

• To evaluate the accuracy and robustness of heartbeat detection using a chest-worn accelerometer.
• To compare a novel signal processing method against a previously described approach.
• To determine the feasibility of using accelerometers for unobtrusive cardiac monitoring.

Main Methods:

• Overnight recordings were conducted on 147 individuals in sleep centers.
• Two heartbeat detection algorithms were compared: local periodicity and a novel maximum a posteriori (MAP) estimation with Markov decision process.
• Inter-beat interval (IBI) and heart rate (HR) were measured and analyzed.

Main Results:

• The novel MAP-based method significantly improved heartbeat detection accuracy.
• Mean absolute error for IBI estimation was 3.5 ms.
• 95% limits of agreement for HR measurement were -1.7 to +1.0 beats per minute, demonstrating high precision.
• Performance remained robust across posture changes and was minimally affected by sleep disorders or demographics.

Conclusions:

• The novel MAP-based method offers highly accurate and robust heartbeat detection using chest-worn accelerometers.
• This technology has the potential to enable unobtrusive ambulatory sleep staging and in-patient cardiac monitoring.
• The findings support the use of accelerometers as a viable tool for long-term cardiac parameter assessment.