Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Contrastive Learning for Sleep Staging based on Inter Subject Correlation

Tongxu Zhang, Bei Wang

TL;DR

Sleep staging models suffer from cross-subject variability due to EEG differences. This paper introduces MViTime, a MobileViT-based time-series backbone, trained with contrastive learning that incorporates inter-subject correlation (ISC) through self-contrast and cross-subject contrast, plus a cross-subject strategy for deployment to new subjects. The approach achieves state-of-the-art performance on Sleep-EDF datasets and demonstrates that contrastive pre-training and ISC-based cross-subject strategies improve generalization, including improvements in the S1 wake-transition stage. The work provides a practical, data-efficient solution for cross-subject sleep staging using single-channel EEG.

Abstract

In recent years, multitudes of researches have applied deep learning to automatic sleep stage classification. Whereas actually, these works have paid less attention to the issue of cross-subject in sleep staging. At the same time, emerging neuroscience theories on inter-subject correlations can provide new insights for cross-subject analysis. This paper presents the MViTime model that have been used in sleep staging study. And we implement the inter-subject correlation theory through contrastive learning, providing a feasible solution to address the cross-subject problem in sleep stage classification. Finally, experimental results and conclusions are presented, demonstrating that the developed method has achieved state-of-the-art performance on sleep staging. The results of the ablation experiment also demonstrate the effectiveness of the cross-subject approach based on contrastive learning.

Contrastive Learning for Sleep Staging based on Inter Subject Correlation

TL;DR

Sleep staging models suffer from cross-subject variability due to EEG differences. This paper introduces MViTime, a MobileViT-based time-series backbone, trained with contrastive learning that incorporates inter-subject correlation (ISC) through self-contrast and cross-subject contrast, plus a cross-subject strategy for deployment to new subjects. The approach achieves state-of-the-art performance on Sleep-EDF datasets and demonstrates that contrastive pre-training and ISC-based cross-subject strategies improve generalization, including improvements in the S1 wake-transition stage. The work provides a practical, data-efficient solution for cross-subject sleep staging using single-channel EEG.

Abstract

In recent years, multitudes of researches have applied deep learning to automatic sleep stage classification. Whereas actually, these works have paid less attention to the issue of cross-subject in sleep staging. At the same time, emerging neuroscience theories on inter-subject correlations can provide new insights for cross-subject analysis. This paper presents the MViTime model that have been used in sleep staging study. And we implement the inter-subject correlation theory through contrastive learning, providing a feasible solution to address the cross-subject problem in sleep stage classification. Finally, experimental results and conclusions are presented, demonstrating that the developed method has achieved state-of-the-art performance on sleep staging. The results of the ablation experiment also demonstrate the effectiveness of the cross-subject approach based on contrastive learning.
Paper Structure (14 sections, 1 equation, 6 figures, 6 tables)

This paper contains 14 sections, 1 equation, 6 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Contrasting Learning Methods
  4. Sleep Staging Contrastive Learning Model
  5. Sleep Staging Classification Model
  6. Cross-subject Strategy for Sleep Staging Based on Contrastive Learning
  7. Experiments
  8. Dataset
  9. Performance Metrics
  10. Results of Sleep Staging Experiment
  11. Results of Cross-subject Experiments
  12. Ablation
  13. Conclusion
  14. Acknowledgements

Figures (6)

  • Figure 1: Ground truth for contrastive learning of EEG signals.
  • Figure 2: Network structure of MViTime.
  • Figure 3: Unfold and Fold for MViTime.
  • Figure 4: MViTime initializes the network parameters by pre-training with a contrastive learning method and subsequently fine-tunes these parameters.
  • Figure 5: Cross-subject contrastive learning strategy for MViTime.
  • ...and 1 more figures