Fiche publication
Date publication
mai 2025
Journal
PloS one
Auteurs
Membres identifiés du Cancéropôle Est :
Pr BASTOGNE Thierry
,
Pr CHENUEL Bruno
Tous les auteurs :
Guyot P, Eveilleau M, Bastogne T, Ayav C, Carpentier N, Chenuel B
Lien Pubmed
Résumé
Obstructive sleep apnea-hypopnea syndrome (OSAHS) is one of the most common sleep disorders affecting nearly one billion of the global adult population, making it a major public health issue. Even if in-lab polysomnography (PSG) remains the gold standard to diagnose OSAHS, there is a growing interest to develop new solutions with more convenient at home devices enhanced with AI-based algorithms for the detection of sleep apnea. This retrospective study aimed to assess the performances of a new method based on nocturnal long-term electrocardiogram signal to detect apneas and hypopneas, in patients who performed attended in-lab PSG. After assessing the quality of the ECG signal, the new method automatically detected apneas and hypopneas using dedicated machine learning algorithm. The agreement between the new ECG-based detection method and the standard interpretation of PSG by a sleep clinician was determined in a blind manner. Eighty-five exams were included into the study with a mean bias between the proposed method and the scorer of 3.5 apneas-hypopneas/hour (/h) (95% CI -48.1 to 55.1). At a threshold of 15/h, sensibility and specificity were 93.3% and 66.7% respectively, and positive and negative predictive values were 87.5% and 80%, respectively. The proposed method using nocturnal long-term electrocardiogram signals showed very high performances to detect apneas and hypopneas. Its implementation in a simple ECG-based device would offer a promising opportunity for preliminary evaluation of patients suspected or at-risk of OSAHS.
Mots clés
Humans, Pilot Projects, Electrocardiography, methods, Algorithms, Female, Male, Middle Aged, Polysomnography, methods, Retrospective Studies, Adult, Sleep Apnea Syndromes, diagnosis, Sleep Apnea, Obstructive, diagnosis, Aged
Référence
PLoS One. 2025 05 16;20(5):e0318622
