This is an online service for automated sleep staging of paediatric polysomnography (PSG) in children aged 2-18 years. Machine learning (ML) models are used to classify all stages of sleep (including Stage Wake, N1, N2, N3, and REM) using any combination of EEG, EOG, and Chin EMG. Respiratory (including central and obstructive) events, are classified using a combination of Thoracic and Abdominal Belts, Nasal Flow, Oronasal Flow, and Pulse Oximetry data. Results may be downloaded in various formats for use in commercial PSG software.

Learn more about the machine learning models used, including peer-reviewed validation studies.

Substantial emphasis is placed on explainable AI. Measures of uncertainty, global and local explanations, and data visualization accompany ML-generated predictions. Furthermore, models include clinically relevant features that are familiar to sleep paediatricians.

Accepted uploads are European Data Format (EDF/EDF+) files exported from commercial PSG software. EDF files should be fully anonymized prior to upload. Pre-processing of recording data are handled automatically by this platform.

No suitable PSG data on-hand? Sample open source PSG data are available to try.

This service is currently for academic and non-commercial purposes only. It is not approved for clinical use.

Models are trained and evaluated on PSG from a variety of open source paediatric sleep datasets. Performance cannot be guaranteed in children aged less than two years, complex comorbidities, for poor quality recordings, or in adults.