Document resource
Introduction Sleep apnoea in children, though common, often remains underdiagnosed and can lead to significant cognitive, behavioral, and physical consequences, 1 2 However, polysomnography (PSG) remains the clinical gold standard, but its complexity and discomfort make it unsuitable for widespread home use in pediatric populations.1 2 Therefore, this research project aims to develop a convenient, simple, and accessible real-time diagnostic solution that can identify pediatric sleep apnoea in the home environment.Methods We leverage the National Children’s Hospital Sleep Databank, which includes 3,984 pediatric Polysomnography recordings obtained from 3673 patients ages 0–18 years with a mean sleep duration of 10.2 hours. These recordings were accompanied by apnoea events identified by clinicians. 3 SpO2 signal was extracted and partitioned into training (60%), testing (30%), and validation (10%) sets, and was used to train a random forest classifier, a CNN-BiGRU with attention model, and an Ensemble of both to assess for accuracy and overall performance in detecting apnoea events.Results Table 1 shows the results obtained.Discussion Our results suggest that these models, which were just trained on SpO 2, can find apnoea events with up to 85% accuracy. The Ensemble Model performed the best and has great potential as a scalable, non-invasive home monitoring tool for children’s sleep problems.References Bratton TK, Jazayeri M, Senthilvel E, Mendoza MR, Valdes R. Clinical laboratory approaches for diagnoses of sleep-disordered breathing and ADHD-like behavior in children. J. Appl. Lab. Med. May 2023;8(3):568–582.Savini S, et al. Assessment of obstructive sleep apnoea (OSA) in children: an update. Acta Otorhinolaryngol. Ital. 2019;39(5):289.Lee H, et al. A large collection of real-world pediatric sleep studies. Sci. Data vol. Jul. 2022;9(1):421.Abstract P62 Table 1Table of results