Radar-based sleep stage classification in children undergoing polysomnography: a pilot-study

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Pediatric Health

Background:

• Monitoring sleep in children non-invasively is challenging.
• Polysomnography is the gold standard but can be intrusive for pediatric subjects.

Purpose of the Study:

• To evaluate the feasibility of using Ultra-Wide band (UWB) radar for unobtrusive sleep monitoring in children.
• To assess UWB radar's ability to measure sleep stages and breathing patterns.

Main Methods:

• Thirty-two children (2 months to 14 years) underwent polysomnography with simultaneous UWB radar monitoring.
• UWB radar captured body movements and breathing rate, from which 38 features were extracted.
• An adaptive boosting machine learning classifier was employed to estimate sleep stages.

Main Results:

• The study achieved 89.8% accuracy and a Cohen's Kappa of 0.67 for wake vs. sleep classification.
• Classification accuracy for wake, REM, and non-REM sleep was 72.9% (Kappa=0.47).
• Accuracy for differentiating wake, REM, light, and deep sleep was 58.0% (Kappa=0.43).

Conclusions:

• UWB radar demonstrates potential as a non-contact method for pediatric sleep analysis.
• Current performance levels require further improvement for widespread clinical adoption.
• This technology offers a promising avenue for less intrusive sleep disorder assessment in children.