Scholarlink

<p>Automatic sleep staging plays an important role in the diagnosis and long-term management of sleep disorders. However, conventional polysomnography-based approaches usually depend on professional equipment and complex operation, which limits their use in home-based and portable monitoring scenarios. To address the computational constraints of wearable and portable devices, this study proposes MCG-NET, a lightweight model for automatic sleep stage classification from multichannel physiological signals, including electroencephalogram (EEG), electrooculogram (EOG), and electromyogram (EMG). The model consists of three main components: a Multi-Scale Module for extracting features at different temporal scales, a channel- enhanced spatial attention module for strengthening spatial correlations among multichannel features, and a global–local temporal network for capturing both long-range temporal dependencies and local temporal details. Experiments on three public datasets, Sleep-EDF-20, Sleep-EDF-78, and ISRUC-S3, show that MCG-NET achieves a favorable trade-off between classification performance and computational efficiency, indicating its potential for practical deployment in portable sleep monitoring devices.</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

MCG-NET: low-parameter sleep staging model integrating spatiotemporal features

  • 期刊论文

  • 发表日期：

    2026年8月3日

查看全文
  • Jiacheng Xia,
  • Jinlong Yang,
  • Chin Chi Choi

摘要

Automatic sleep staging plays an important role in the diagnosis and long-term management of sleep disorders. However, conventional polysomnography-based approaches usually depend on professional equipment and complex operation, which limits their use in home-based and portable monitoring scenarios. To address the computational constraints of wearable and portable devices, this study proposes MCG-NET, a lightweight model for automatic sleep stage classification from multichannel physiological signals, including electroencephalogram (EEG), electrooculogram (EOG), and electromyogram (EMG). The model consists of three main components: a Multi-Scale Module for extracting features at different temporal scales, a channel- enhanced spatial attention module for strengthening spatial correlations among multichannel features, and a global–local temporal network for capturing both long-range temporal dependencies and local temporal details. Experiments on three public datasets, Sleep-EDF-20, Sleep-EDF-78, and ISRUC-S3, show that MCG-NET achieves a favorable trade-off between classification performance and computational efficiency, indicating its potential for practical deployment in portable sleep monitoring devices.