Sleep Stage Estimation from Bed Leg Ballistocardiogram Sensors

Area of Science:

• Biomedical Engineering
• Cardiovascular Monitoring
• Sleep Science

Background:

• Electrocardiograms (ECGs) are standard for clinical monitoring, but Ballistocardiograms (BCGs) offer reduced user burden.
• BCG sensors' susceptibility to noise has limited their widespread clinical adoption compared to ECGs.
• Accurate, non-invasive sleep monitoring is crucial for diagnosing sleep disorders and assessing cardiovascular health.

Purpose of the Study:

• To develop a robust BCG measurement method for precise sleep heart rate variability (HRV) detection.
• To create and validate a mathematical model for predicting sleep stages using BCG-derived HRV.
• To enhance the clinical utility of BCG for sleep monitoring, offering an alternative to ECG.

Main Methods:

• A three-step BCG measurement algorithm: preprocessing, heartbeat signal template creation, and template matching for HRV detection.
• Validation of the BCG algorithm using photoplethysmography (PPG) for ground truth HRV on 99 datasets.
• Development and validation of a mathematical model for sleep stage prediction using BCG-derived HRV on 100 datasets.

Main Results:

• The BCG method successfully extracted beat-to-beat intervals with an average detection rate of 86.9% and a mean error of 8.5ms.
• The developed sleep stage prediction model achieved 80% accuracy, surpassing conventional algorithms (max 69%).
• The BCG method demonstrated usability comparable to clinical ECGs for sleep heart rate monitoring.

Conclusions:

• The novel BCG measurement technique provides accurate sleep heart rate monitoring with reduced patient burden.
• The mathematical model offers a highly accurate method for sleep stage prediction, applicable to both BCG and ECG data.
• This research advances non-invasive cardiovascular and sleep monitoring, paving the way for more accessible clinical tools.