L-Tetrolet Pattern-Based Sleep Stage Classification Model Using Balanced EEG Datasets.

Area of Science:

• Neuroscience
• Computational Biology
• Biomedical Engineering

Background:

• Manual sleep stage classification from electroencephalogram (EEG) is time-consuming and prone to errors.
• Automated methods using machine learning are needed to standardize sleep analysis for diagnosing sleep disorders.
• Existing methods require improvement in accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate a novel L-tetrolet pattern-based machine learning model for automated sleep stage classification.
• To assess the model's performance across normal sleep, insomnia, and fused sleep pattern datasets.
• To enhance the accuracy and reliability of sleep stage identification in clinical settings.

Main Methods:

• Utilized the cyclic alternating pattern (CAP) sleep dataset for training and testing.
• Developed a feature generation method using L-tetrolet functions and pooling decomposition.
• Implemented a hybrid feature selection technique, threshold selection-based ReliefF and INCA (TSRFINCA), fused with a cubic support vector machine classifier.

Main Results:

• Achieved high classification accuracies: 95.43% for Insomnia, 91.05% for Normal, and 92.31% for Fused cases.
• The L-tetrolet pattern and TSRFINCA model demonstrated robust performance across different sleep conditions.
• Successfully identified six sleep stages within the analyzed datasets.

Conclusions:

• The proposed L-tetrolet pattern and TSRFINCA model represents a significant advancement in automated sleep stage classification.
• This model accurately classifies sleep stages, even in the presence of sleep disorders, offering a more reliable diagnostic tool.
• The findings contribute to the field of knowledge engineering for sleep medicine.