On-the-Fly Sleep Scoring Algorithm with Heart Rate, RR Intervals and Accelerometer as Input.

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Sleep Science

Background:

• Real-time sleep stage recognition is crucial for applications like autonomous driving safety systems.
• Current methods often rely on obtrusive electroencephalography (EEG) or specialized equipment.
• There is a need for non-invasive, accessible, and accurate sleep-scoring solutions.

Purpose of the Study:

• To develop and evaluate an on-the-fly sleep-scoring algorithm using data from a smartwatch.
• To assess the algorithm's accuracy in identifying different sleep stages (light, deep, REM) and wakefulness.
• To demonstrate the feasibility of using widely available wearable devices for sleep analysis.

Main Methods:

• An on-the-fly algorithm was developed using Heart Rate (HR), RR intervals, and accelerometer data from a smartwatch.
• Participants wore a commercial smartwatch and a polysomnography (PSG) device for ground truth validation.
• The algorithm's performance was tested against PSG-based sleep scoring.

Main Results:

• The algorithm achieved 88.46% accuracy, 91.42% precision, and 93.52% sensitivity in sleep-wake identification.
• Accuracies for specific sleep stages were: deep sleep (69.38%), light sleep (50.62%), REM sleep (62.02%), and wakefulness (73.48%).
• The developed algorithm shows promise for non-invasive, real-time sleep monitoring.

Conclusions:

• Smartwatch data can be effectively utilized for on-the-fly sleep stage classification.
• The algorithm provides a foundation for developing accessible sleep monitoring tools.
• This technology has potential applications in health monitoring and safety-critical systems.