Related Experiment Video
Updated: Aug 29, 2025
07:40
Multi-Modal Home Sleep Monitoring in Older Adults
Published on: January 26, 2019
7.7K
Single-channel EEG sleep staging based on data augmentation and cross-subject discrepancy alleviation
Zhengling He1, Lidong Du2, Peng Wang2
1Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing, China; University of Chinese Academy of Sciences, Beijing, China.
Computers in Biology and Medicine
|September 9, 2022
Summary
This study introduces a novel neural network for automatic sleep staging using electroencephalogram (EEG) data. The approach effectively addresses category imbalance and cross-subject differences, significantly improving classification accuracy.
Area of Science:
- Biomedical Engineering
- Artificial Intelligence
- Sleep Medicine
Background:
- Automatic sleep stage classification using electroencephalogram (EEG) is crucial for sleep analysis.
- Existing machine and deep learning methods struggle with category imbalance and cross-subject variability, limiting accuracy.
Purpose of the Study:
- To propose an innovative end-to-end neural network for automatic sleep staging.
- To overcome limitations of category imbalance and cross-subject discrepancy in EEG-based sleep classification.
Main Methods:
- Developed an end-to-end neural network incorporating four data augmentation techniques for category imbalance.
- Designed domain adaptation modules for feature map distribution alignment and transfer attention mechanism for transferable region identification.
- Utilized Sleep-EDF (2013 & 2018) and Physionet 2018 challenge datasets for validation.
Main Results:
- Achieved Cohen's kappa coefficients of 0.77 (Fpz-Cz) and 0.73 (Pz-Oz) on Sleep-EDF-2013.
- Obtained kappa values of 0.75 (Fpz-Cz) and 0.68 (Pz-Oz) on Sleep-EDF-2018.
- Demonstrated performance improvement on a dataset including individuals with sleep disorders and outperformed similar studies.
Conclusions:
- The proposed EEG data augmentation and domain adaptation approach effectively alleviates cross-subject discrepancy.
- The developed model significantly enhances the performance of automatic sleep staging.
