中国听力语言康复科学杂志2025，Vol.23Issue(3)：276-280,5.DOI:10.3969/j.issn.1672-4933.2025.03.012
阻塞性睡眠呼吸暂停儿童听功能评估及模型构建
Assessment of Auditory Function and Predictive Model Construction in Children with Obstructive Sleep Apnea
邹怡 1陆颖霞 2宇文奇 2张旭3
作者信息
- 1. 首都医科大学生物医学工程学院 北京 100069||首都医科大学附属首都儿童医学中心 北京 100020
- 2. 首都医科大学附属首都儿童医学中心 北京 100020
- 3. 首都医科大学生物医学工程学院 北京 100069
- 折叠
摘要
Abstract
Objective This study aims to assess the auditory function in children with obstructive sleep apnea(OSA)and to construct a machine learning-based predictive model for hearing loss in pediatric OSA patients.Methods Children diagnosed with OSA through polysomnography(PSG)were enrolled as study subjects,categorized into a mild group(64 ears)and a moderate-to-severe group(96 ears).A control group of 94 ears from healthy children undergoing routine physical examinations during the same period was also included.Comprehensive auditory function assessments were conducted on all the participants.Three machine learning methods:Support Vector Machine,Logistic Regression,and Naive Bayes were employed to construct predictive models for hearing loss in children with OSA.The optimal model was selected based on performance metrics.Results ①No significant differences were observed in hearing thresholds at 0.25-2 kHz across the three groups.However,significant differences were noted at frequencies≥4 kHz and in the extended high-frequency range(P<0.05).②Distortion product otoacoustic emission(DPOAE)amplitudes showed no significant differences at 0.75-3 kHz but exhibited significant variations at 4,6 and 8 kHz(P<0.05).③Auditory brainstem response(ABR)measurements revealed no significant differences in Wave Ⅰ and Ⅲ latencies or the Ⅰ-Ⅲ interpeak interval.However,significant differences were observed in Wave V latency and the Ⅲ-V and Ⅰ-Ⅴ interpeak intervals(P<0.05).④The Naive Bayes-based model demonstrated superior performance in predicting hearing loss in children with OSA.⑤Feature importance analysis identified the lowest oxygen saturation and oxygen desaturation index as the most influential predictors of hearing loss.Conclusion This study establishes a significant association between OSA and hearing loss in children,with the severity of hypoxemia emerging as a critical determinant.The Naive Bayes-based predictive model offers a robust tool for identifying hearing loss risk in pediatric OSA patients,highlighting the importance of early intervention and monitoring.关键词
阻塞性睡眠呼吸暂停/儿童/听力损失/机器学习Key words
Obstructive sleep apnea/Children/Hearing loss/Machine learning分类
医药卫生引用本文复制引用
邹怡,陆颖霞,宇文奇,张旭..阻塞性睡眠呼吸暂停儿童听功能评估及模型构建[J].中国听力语言康复科学杂志,2025,23(3):276-280,5.
