Machine Listening for OSA Diagnosis: A Bayesian Meta-Analysis.

The main barrier to diagnosis is the overnight polysomnogram, which requires specialized equipment, skilled technicians, and inpatient beds available only in tertiary sleep centers.

诊断的主要障碍是夜间多导睡眠图，它需要专门的设备、熟练的技术人员，以及仅在三级睡眠中心才有的住院床位。

Recent advances in artificial intelligence (AI) have enabled OSA detection using breathing sound recordings.

人工智能（AI）的最新进展使得通过呼吸声音记录来检测阻塞性睡眠呼吸暂停（OSA）成为可能。

research_question

What is the diagnostic accuracy of and how can we optimize machine listening for OSA?

对于阻塞性睡眠呼吸暂停（OSA），机器听觉的诊断准确性如何，我们又如何优化其检测效果？

study_design_and_methods

PubMed, Embase, Scopus, Web of Science, and IEEE Xplore databases were systematically searched.

系统地检索了PubMed、Embase、Scopus、Web of Science和IEEE Xplore数据库。

Two masked reviewers selected studies comparing the patient-level diagnostic performance of AI approaches using overnight audio recordings vs conventional diagnosis (apnea-hypopnea index) using a train-test split or k-fold cross-validation.

两位审查员在不知情的情况下，选择了使用整夜音频记录比较人工智能方法与传统诊断（呼吸暂停低通气指数）的患者级诊断性能的研究，使用训练-测试分割或k折交叉验证。

Bayesian bivariate meta-analysisand meta-regression were performed.

进行了贝叶斯双变量荟萃分析和荟萃回归。

Publication bias was assessed by using a selection model.

使用选择模型评估了出版偏倚。

Risk of bias and evidence quality were assessed by using the Quality Assessment of Diagnostic Accuracy Studies-2 and the Grading of Recommendations, Assessment, Development, and Evaluation tools.

通过使用诊断准确性研究质量评估工具第二版（Quality Assessment of Diagnostic Accuracy Studies-2）和推荐、评估、发展和评估等级工具（Grading of Recommendations, Assessment, Development, and Evaluation tools），评估了偏倚风险和证据质量。

Results

From 6,254 records, 16 studies (41 models) trained on 4,864 participants and tested on 2,370 participants were included.

从6254条记录中，纳入了16项研究（41个模型），这些研究在4864名参与者上进行了训练，并在2370名参与者上进行了测试。

No study had a high risk of bias.

没有研究存在高风险的偏倚。

Machine listening achieved a pooled sensitivity (95% credible interval) of 90.3% (86.9%-93.1%), a specificity of 86.7% (83.1%-89.7%), a diagnostic OR of 60.8 (39.4-99.9), and positive and negative likelihood ratios of 6.78 (5.34-8.85) and 0.113 (0.079-0.152), respectively.

机器听觉实现了90.3%（95%可信区间：86.9%-93.1%）的汇总敏感性，86.7%（95%可信区间：83.1%-89.7%）的特异性，60.8（39.4-99.9）的诊断比值比，以及6.78（5.34-8.85）的阳性似然比和0.113（0.079-0.152）的阴性似然比。

At apnea-hypopnea index cutoffs of ≥ 5, ≥ 15, and ≥ 30 events per hour, sensitivities were 94.3% (90.3%-96.8%), 86.3% (80.1%-90.9%), and 86.3% (79.2%-91.1%); and specificities were 78.5% (68.0%-86.9%), 87.3% (81.8%-91.3%), and 89.5% (84.8%-93.3%).

在呼吸暂停低通气指数的截断值为≥5、≥15和≥30次/小时时，敏感性分别为94.3%（95%置信区间90.3%-96.8%）、86.3%（80.1%-90.9%）和86.3%（79.2%-91.1%）；特异性分别为78.5%（68.0%-86.9%）、87.3%（81.8%-91.3%）和89.5%（84.8%-93.3%）。

Meta-regression identified increased sensitivity for the following: higher audio sampling frequencies, non-contact microphones, higher OSA prevalence, and train-test split model evaluation.

元回归分析识别出以下因素会增加敏感性：更高的音频采样频率、非接触式麦克风、更高的OSA（阻塞性睡眠呼吸暂停）患病率以及训练-测试分割模型评估。

Accuracy was equal regardless of home smartphone vs in-laboratory professional microphone recordings, deep learning vs traditional machine learning, and variations in age and sex.

无论是在家中使用智能手机还是在实验室使用专业麦克风录制，深度学习还是传统机器学习，以及年龄和性别的变化，准确度都是相等的。

Publication bias was not evident, and the evidence was of high quality.

没有明显的出版偏倚，且证据质量高。

interpretation

In this study, machine listening achieved excellent diagnostic accuracy, superior to the STOP-Bang (snoring, tiredness, observed apnea, BP, BMI, age, neck size, gender) questionnaire and comparable to common home sleep tests.

在这项研究中，机器听觉实现了出色的诊断准确性，优于STOP-Bang（打鼾、疲劳、观察到的呼吸暂停、血压、体质指数、年龄、颈部尺寸、性别）问卷，并与常见的家庭睡眠测试相当。

Digital medicine should be further explored and externally validated for accessible and equitable OSA diagnosis.

数字医学应进一步探索并进行外部验证，以便于普及和公平地诊断OSA。

clinical_trial_registration

PROSPERO database; No.: CRD42024534235; URL: https://www.crd.york.ac.uk/PROSPERO/).

PROSPERO数据库; 编号: CRD42024534235; 网址: https://www.crd.york.ac.uk/PROSPERO/).