Scholarlink

<p>Obstructive sleep apnoea (OSA) affects over a billion people globally yet remains underdiagnosed, posing significant health and safety risks. Polysomnography (PSG), the diagnostic gold standard, captures EEG, ECG, airflow, and oxygen saturation—but manual scoring is time-consuming and variable. To review recent artificial intelligence (AI) advances—including machine learning (ML) and deep learning (DL)—for automated sleep staging, OSA detection, and candidate screening in safety-critical industries, we performed a narrative review of studies (2015–2025) applying AI to PSG data, wearable sensors, electronic health records, and patient-reported outcomes. Convolutional neural networks, recurrent models (LSTM/RNN), and hybrid architectures now automate sleep staging and apnoeic-event detection with Cohen’s κ &gt;0.80 and accuracies &gt;95%. Explainable AI tools (LIME, hypnodensity modeling) enhance transparency and clinician trust. Wearable integrations and cloud-based platforms extend diagnostics beyond sleep labs. In aviation, transportation, and military settings, AI-driven screening facilitates early identification of sleep disorders and fatigue risk. Key challenges remain—data privacy, algorithmic bias, and regulatory validation—but federated learning, inclusive global datasets, and precision sleep medicine frameworks offer a path forward. The ethical and transparent deployment of AI-driven PSG can improve diagnostic access, enable personalized therapy, and advance global sleep health equity.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Snore Wars: How Artificial Intelligence is Transforming Obstructive Sleep Apnoea Using Polysomnography

  • 期刊论文

  • 发表日期：

    2025年12月2日

查看全文
  • Mohammad Naksh Kamar,
  • Apurva Anil Jarandikar,
  • Mayur Hemchandra Ingale,
  • Paresh Shahaji Chavan,
  • Vinod Vishwanath Shinde,
  • Gundappa Dhondiappa Mahajan

摘要

Obstructive sleep apnoea (OSA) affects over a billion people globally yet remains underdiagnosed, posing significant health and safety risks. Polysomnography (PSG), the diagnostic gold standard, captures EEG, ECG, airflow, and oxygen saturation—but manual scoring is time-consuming and variable. To review recent artificial intelligence (AI) advances—including machine learning (ML) and deep learning (DL)—for automated sleep staging, OSA detection, and candidate screening in safety-critical industries, we performed a narrative review of studies (2015–2025) applying AI to PSG data, wearable sensors, electronic health records, and patient-reported outcomes. Convolutional neural networks, recurrent models (LSTM/RNN), and hybrid architectures now automate sleep staging and apnoeic-event detection with Cohen’s κ >0.80 and accuracies >95%. Explainable AI tools (LIME, hypnodensity modeling) enhance transparency and clinician trust. Wearable integrations and cloud-based platforms extend diagnostics beyond sleep labs. In aviation, transportation, and military settings, AI-driven screening facilitates early identification of sleep disorders and fatigue risk. Key challenges remain—data privacy, algorithmic bias, and regulatory validation—but federated learning, inclusive global datasets, and precision sleep medicine frameworks offer a path forward. The ethical and transparent deployment of AI-driven PSG can improve diagnostic access, enable personalized therapy, and advance global sleep health equity.