Paediatric sleep apnea event prediction using nasal air pressure and machine learning

• DOI: 10.1111/jsr.13851
• Publication Date: 2023-02-21T06:52:52Z
• Authors: Matthew G. Crowson; Kevin S. Gipson; Orna Katz Kadosh; Elizabeth Hartnick; Ellen Grealish; Donald G. Keamy; Thomas Bernard Ki...; Christopher J. Ha...
• Keywords: Sleep

Summary Sleep‐disordered breathing is an important health issue for children. The objective of this study was to develop a machine learning classifier model the identification sleep apnea events taken exclusively from nasal air pressure measurements acquired during overnight polysomnography paediatric patients. A secondary differentiate site obstruction hypopnea event data using model. Computer vision classifiers were developed via transfer either normal while asleep, obstructive hypopnea, or central apnea. separate trained identify as adeno‐tonsillar tongue base. In addition, survey board‐certified and board‐eligible physicians completed compare clinician versus classification performance events, indicated very good our relative human raters. sample database available modelling comprised 417 normal, 266 122 131 derived 28 four‐way achieved mean prediction accuracy 70.0% (95% confidence interval [67.1–72.9]). Clinician raters correctly identified tracings 53.8% time, whereas local 77.5% accurate. 75.0% [68.7–81.3]). Machine applied feasible may exceed diagnostic expert clinicians. Nasal hypopneas “encode” information regarding obstruction, which only be discernable by learning.