From Lab to Wrist: The Emerging Role of Wearables in Paediatric Sleep Medicine
Overnight in-laboratory polysomnography (PSG) has long been the gold standard for diagnosing paediatric sleep-disordered breathing. However, long waiting lists, high costs, and the distress of overnight hospital stays for young children have limited patient access.
Recent landmark validation studies highlight that specialized paediatric-trained deep learning algorithms—specifically dual-stream local-global Long Short-Term Memory (LSTM) networks which can process raw wrist movement and heart rate signals to achieve 95% ROC-AUC accuracy in classifying sleep and wakefulness. Unlike traditional adult models or older mathematical scoring methods, these paediatric AI classifiers eliminate proportional bias and estimate total sleep duration within minutes of laboratory gold standards.
Furthermore, systematic evidence across 62 portable sleep monitoring trials reveals that multi-channel home polygraphy and AI-assisted pulse oximetry offer specificities up to 100% for ruling in severe paediatric Obstructive Sleep Apnea (OSA). This allows clinicians to fast-track children with adenotonsillar hypertrophy for urgent surgical adenotonsillectomy without unnecessary laboratory delays.
By combining wrist movement sensors with optical heart rate tracking and paediatric deep learning, we can evaluate children over multiple nights in their natural home environment. This transition from lab to wrist bridges the gap between cutting-edge technology and accessible child healthcare.