Massie Frederik, Vits Steven, Verbraecken Johan, Bergmann Jeroen
Natural Interaction Lab, Department of Engineering, University of Oxford, Oxford, United Kingdom.
Faculty of Medicine and Health Sciences, University of Antwerp, Antwerp, Belgium.
J Clin Sleep Med. 2025 May 1;21(5):789-804. doi: 10.5664/jcsm.11534.
Home sleep apnea testing based on peripheral arterial tonometry is increasingly being deployed because of its ability to test for multiple nights. However, home sleep apnea tests based on peripheral arterial tonometry do not have access to modalities such as airflow and cortical arousals and instead rely on alternative sources of information to detect respiratory events. This results in an a priori performance disadvantage. In this study, we describe the Panorama algorithm, which aims to reduce this disadvantage by incorporating information from characteristically repetitive sequences in physiological changes associated with respiratory events. These include changes in peripheral arterial tone, pulse rate, and oxygen saturation. The method was designed to facilitate manual review by providing the scoring rationale for each respiratory event.
The method was developed and evaluated using a dataset of 266 participants from a multicentric cohort suspected of having obstructive sleep apnea. All participants underwent simultaneous polysomnography and home sleep apnea testing based on peripheral arterial tonometry, and all polysomnography data were double-scored. Scoring was performed according to the 3% and 4% rules for hypopnea scoring. Clinical endpoint parameters, including the obstructive sleep apnea severity categorization accuracy and Cohen's kappa, were selected to compare the algorithm to a conventional context-unaware algorithm. Data analysis and reporting followed the TRIPOD+AI reporting guidance for prediction models that use machine learning.
Regarding obstructive sleep apnea severity categorization accuracy, the Panorama algorithm significantly outperformed context-unaware autoscoring by 9% using 3% rule scoring and 7% using 4% rule scoring.
The context-aware method significantly improves the performance of home sleep apnea tests based on peripheral arterial tonometry while still facilitating scoring review by providing event-specific scoring rationale.
Registry: ClinicalTrials.gov; Name: A Validation Study of the NightOwl PAT-based Home Sleep Apnea Test; URL: https://clinicaltrials.gov/ct2/show/NCT04191668; Identifier: NCT04191668.
Massie F, Vits S, Verbraecken J, Bergmann J. Context-aware analysis enhances autoscoring accuracy of home sleep apnea testing. 2025;21(5):789-804.
基于外周动脉张力测量的家庭睡眠呼吸暂停检测因其能够进行多晚检测而越来越多地被采用。然而，基于外周动脉张力测量的家庭睡眠呼吸暂停检测无法获取气流和皮层觉醒等模式，而是依赖其他信息来源来检测呼吸事件。这导致了先验性能劣势。在本研究中，我们描述了全景算法，该算法旨在通过纳入与呼吸事件相关的生理变化中特征性重复序列的信息来减少这一劣势。这些变化包括外周动脉张力、脉搏率和血氧饱和度的变化。该方法旨在通过为每个呼吸事件提供评分依据来便于人工审核。
使用来自一个多中心队列的266名疑似阻塞性睡眠呼吸暂停参与者的数据集开发并评估该方法。所有参与者同时接受多导睡眠图检查和基于外周动脉张力测量的家庭睡眠呼吸暂停检测，所有多导睡眠图数据都进行了双重评分。根据呼吸浅慢的3%和4%规则进行评分。选择临床终点参数，包括阻塞性睡眠呼吸暂停严重程度分类准确性和科恩kappa系数，将该算法与传统的无上下文感知算法进行比较。数据分析和报告遵循使用机器学习的预测模型的TRIPOD+AI报告指南。
关于阻塞性睡眠呼吸暂停严重程度分类准确性，全景算法在使用3%规则评分时比无上下文感知自动评分显著高出9%，在使用4%规则评分时高出7%。
上下文感知方法显著提高了基于外周动脉张力测量的家庭睡眠呼吸暂停检测的性能，同时仍通过提供特定事件的评分依据来便于评分审核。
注册机构：ClinicalTrials.gov；名称：基于NightOwl PAT的家庭睡眠呼吸暂停检测的验证研究；网址：https://clinicaltrials.gov/ct2/show/NCT04191668；标识符：NCT04191668。
Massie F, Vits S, Verbraecken J, Bergmann J. 上下文感知分析提高了家庭睡眠呼吸暂停检测的自动评分准确性。2025；21(5):789 - 804。
