Sleep stages classification by fusing the time-related synchronization analysis and brain activations.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Sleep staging is crucial for diagnosing sleep disorders.
• Current methods may overlook temporal patterns within sleep periods.
• Standard sleep scoring uses 30-second intervals, potentially grouping similar brain activity.

Purpose of the Study:

• To propose a novel time-related synchronization analysis framework for sleep scoring.
• To explore potential time-related patterns in sleeping brain activity.
• To improve the robustness and accuracy of sleep staging.

Main Methods:

• Developed the time-related multimodal sleep scoring (TRMSC) model.
• Conducted time-related synchronization analysis on electroencephalogram (EEG) and electrooculogram (EOG) signals.
• Extracted spectral activation features and utilized feature fusion and selection.

Main Results:

• The TRMSC model achieved superior performance compared to existing strategies on the Sleep-EDF dataset.
• Time-related synchronization features compensated for limitations of traditional spectrum-based methods.
• Demonstrated higher classification accuracy in sleep staging.

Conclusions:

• The proposed TRMSC framework offers a new analytical method for clinical sleep research.
• This approach can enhance the accuracy of portable sleep analyzers.
• Time-related synchronization analysis provides valuable insights beyond spectral features for sleep staging.