Achieving Accurate Automatic Sleep Staging on Manually Pre-processed EEG Data Through Synchronization Feature...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Sleep staging from electroencephalographic (EEG) signals is complex, time-consuming, and prone to errors due to signal noise.
• Accurate sleep analysis requires effective noise reduction and robust feature extraction methods.
• Existing methods often rely on univariate features, potentially missing complex brain network dynamics.

Purpose of the Study:

• To develop and evaluate a pre-processing and automatic sleep staging pipeline for EEG signals.
• To investigate the efficacy of two novel functional connectivity estimation methods, Synchronization Likelihood (SL) and Relative Wavelet Entropy (RWE), for sleep staging.
• To compare bivariate functional connectivity features against traditional univariate features for improved sleep staging accuracy.

Main Methods:

• A multi-step pre-processing pipeline was implemented to prepare EEG signals for analysis.
• Two novel methods, Synchronization Likelihood (SL) and Relative Wavelet Entropy (RWE), were used to extract bivariate functional connectivity features.
• Classifiers were trained using these extracted features for automatic sleep epoch annotation.
• The methods were validated on EEG data from a European Space Agency bed-rest study.

Main Results:

• The proposed pipeline, utilizing SL and RWE for feature extraction, achieved high accuracy rates exceeding 90% in sleep staging.
• These bivariate functional connectivity features demonstrated superior performance compared to traditional univariate methods.
• The developed methods proved effective in classifying sleep epochs accurately based on ground truth from expert manual staging.

Conclusions:

• Novel functional connectivity methods (SL and RWE) are highly suitable for semi-automatic sleep staging.
• The developed pipeline offers an accurate and efficient approach to analyzing noisy EEG data for sleep research.
• This research contributes to advancing automated sleep analysis techniques, enhancing clinical knowledge extraction from sleep recordings.