Sleep Position Detection for Closed-Loop Treatment of Sleep-Related Breathing Disorders.

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Wearable Technology

Background:

• Accurate sleep position detection is crucial for managing position-dependent sleep-related breathing disorders.
• Technical aids require reliable data on patient sleep posture.

Purpose of the Study:

• To compare the effectiveness of personalized versus generalizable sleep position classifiers.
• To evaluate the performance of unobtrusive eight-channel pressure-sensing mats for sleep position and bed occupancy detection.

Main Methods:

• Collected data from six male patients with position-dependent sleep apnea over three nights.
• Utilized leave-one-night-out cross-validation for personalized classifier training.
• Assessed eight-channel pressure-sensing mats for sleep position (supine/non-supine, supine/lateral-left/lateral-right) and in-bed presence classification.

Main Results:

• Personalized classifiers achieved F1-scores of 61.3% (supine/non-supine) and 46.2% (supine/lateral-left/lateral-right).
• Generalizable classifiers achieved F1-scores of 62.1% (supine/non-supine) and 49.1% (supine/lateral-left/lateral-right).
• In-bed presence detection reached a high F1-score of 98.1%.

Conclusions:

• Personalized sleep-position classifiers trained on limited data perform comparably to those trained on larger datasets.
• Eight-channel sensor mattresses accurately detect in-bed presence for closed-loop systems.
• The classification of specific sleep positions using these sensor mats remains limited.