Scholarlink

<p>Obstructive Sleep Apnea Hypopnea Syndrome (OSAHS) is a prevalent systemic disorder affecting approximately 1&#xa0;billion people worldwide, associated with severe outcomes such as sudden death and traffic accidents. Despite its significant impact, OSAHS is frequently underdiagnosed. The current gold standard for assessing OSAHS severity, overnight polysomnography, is both costly and inconvenient. This study aims to develop an automated method for detecting apnea and hypopnea events using a temporal convolutional network (TCN) to improve diagnostic accuracy and reduce computational costs. We introduce a novel Temporal Convolutional Network with a Linearly Scalable Attention Mechanism (ECG-TCN) designed to simultaneously detect both apnea and hypopnea events. The model was trained and validated using the University College Dublin Sleep Apnea Database. The performance of ECG-TCN was evaluated based on per-segment classification accuracy and generalization capacity. The ECG-TCN model achieved an accuracy of 91.6% in per-segment classification, demonstrating superior performance compared to traditional classification models. Additionally, the model exhibited high generalization capacity, indicating its robustness across different datasets. This study presents the first application of a temporal convolutional network with a linearly scalable attention mechanism for the simultaneous detection of apnea and hypopnea events. The ECG-TCN model offers a cost-effective and accurate alternative to traditional diagnostic methods, with the potential to enhance the early detection and management of OSAHS. </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

Automated OSAHS detection from ECG using temporal convolutional network

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月14日

查看全文
  • Lei Cheng,
  • Juan Bai,
  • Aizhu Liu,
  • Suyin Feng,
  • Runfeng Sun

摘要

Obstructive Sleep Apnea Hypopnea Syndrome (OSAHS) is a prevalent systemic disorder affecting approximately 1 billion people worldwide, associated with severe outcomes such as sudden death and traffic accidents. Despite its significant impact, OSAHS is frequently underdiagnosed. The current gold standard for assessing OSAHS severity, overnight polysomnography, is both costly and inconvenient. This study aims to develop an automated method for detecting apnea and hypopnea events using a temporal convolutional network (TCN) to improve diagnostic accuracy and reduce computational costs. We introduce a novel Temporal Convolutional Network with a Linearly Scalable Attention Mechanism (ECG-TCN) designed to simultaneously detect both apnea and hypopnea events. The model was trained and validated using the University College Dublin Sleep Apnea Database. The performance of ECG-TCN was evaluated based on per-segment classification accuracy and generalization capacity. The ECG-TCN model achieved an accuracy of 91.6% in per-segment classification, demonstrating superior performance compared to traditional classification models. Additionally, the model exhibited high generalization capacity, indicating its robustness across different datasets. This study presents the first application of a temporal convolutional network with a linearly scalable attention mechanism for the simultaneous detection of apnea and hypopnea events. The ECG-TCN model offers a cost-effective and accurate alternative to traditional diagnostic methods, with the potential to enhance the early detection and management of OSAHS.