EEG Sleep Stages Classification Based on Time Domain Features and Structural Graph Similarity.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalogram (EEG) signals are crucial for diagnosing sleep disorders.
• Current sleep stage classification methods often rely on time or frequency domain analysis of EEG signals.
• High classification accuracy remains a key challenge in automated sleep staging.

Purpose of the Study:

• To develop and evaluate a novel method for classifying six sleep stages using single-channel EEG signals.
• To combine statistical features, structural graph similarity, and K-means clustering for improved sleep staging.
• To investigate the relationship between sleep stages and time-domain EEG features.

Main Methods:

• EEG segments were partitioned into empirically determined sub-segments.
• Statistical features were extracted from time-domain EEG data.
• A Structural Graph Similarity and K-means (SGSKM) approach was employed for classification.

Main Results:

• The proposed SGSKM method demonstrated superior performance compared to four existing methods and Support Vector Machine (SVM).
• An average classification accuracy of 95.93% was achieved for identifying six sleep stages.
• Relationships between specific sleep stages and time-domain EEG features were explored.

Conclusions:

• The combined SGSKM method offers a highly accurate and effective approach for sleep stage classification from single-channel EEG.
• This method holds potential for improving the diagnosis and treatment of sleep disorders.
• Further research can explore the clinical applicability and validation of this novel technique.