Multi-channel EEG-based sleep stage classification with joint collaborative representation and multiple kernel...

Area of Science:

• Neuroscience
• Machine Learning
• Biomedical Engineering

Background:

• Electroencephalography (EEG) is vital for clinical sleep staging.
• Effective feature extraction and representation are critical for accurate EEG-based sleep stage classification.
• Sparse representation (SR) is an advanced unsupervised method for EEG feature learning.

Purpose of the Study:

• To develop a novel two-stage multi-view learning framework for automated sleep staging using EEG signals.
• To leverage collaborative representation (CR) for enhanced EEG feature representation and fusion.
• To evaluate the performance of the proposed framework against existing classification methods.

Main Methods:

• A two-stage multi-view learning framework was constructed for sleep staging.
• Joint collaborative representation (JCR) and joint sparse representation (JSR) algorithms were employed to fuse and learn features from multi-channel EEG signals.
• A multiple kernel extreme learning machine (MK-ELM) with grid search was used for final sleep stage recognition.

Main Results:

• The proposed multi-view learning algorithm achieved superior performance in sleep staging.
• Classification accuracy reached 81.10 ± 0.15% with a K-means clustering dictionary.
• Sensitivity and specificity were reported at 71.42 ± 0.66% and 94.57 ± 0.07%, respectively.

Conclusions:

• The developed multi-view learning framework shows significant potential for sleep staging.
• The framework is applicable to multi-channel or multi-modality polysomnography signals.
• The JCR method demonstrated superiority over JSR in this sleep staging application.