A Holistic Strategy for Classification of Sleep Stages with EEG

Area of Science:

• Neuroscience
• Computer Science
• Biomedical Engineering

Background:

• Manual sleep stage scoring is time-consuming and requires expert analysis.
• Automated systems are crucial for diagnosing sleep disorders efficiently.
• Advancements in machine learning and deep learning offer new possibilities for automated sleep analysis.

Purpose of the Study:

• To propose a holistic strategy, CDFCD, for automated sleep stage classification using EEG signals.
• To introduce novel techniques in clustering, dimensionality reduction, and feature selection.
• To improve the accuracy and efficiency of sleep stage classification.

Main Methods:

• Utilized hierarchical clustering, spectral clustering, and PCA-based subspace clustering.
• Applied SVD-based spectral algorithm and VBMF for dimensionality reduction.
• Employed SGL-DLI and RR-LWS for feature extraction and selection.
• Implemented multiclass Gaussian process classification, RACC, and LSTM-based deep learning for classification.

Main Results:

• The proposed CDFCD methodology achieved a high classification accuracy of 93.51% for six sleep stages.
• The results surpassed previous studies on the Sleep EDF database.
• Demonstrated the effectiveness of the novel techniques integrated within the CDFCD framework.

Conclusions:

• The CDFCD strategy provides a robust and accurate method for automated sleep stage classification.
• This approach can significantly assist clinicians in diagnosing sleep-related disorders.
• Further research can explore the integration of these novel techniques in clinical settings.