A comparative study on classification of sleep stage based on EEG signals using feature selection and classification...

Area of Science:

• Neuroscience
• Computer Science
• Medical Diagnostics

Background:

• Sleep scoring is crucial for diagnosing neurological and psychiatric disorders.
• Manual sleep staging by experts is time-consuming and labor-intensive.
• Automated methods are needed to improve efficiency and accuracy in sleep analysis.

Purpose of the Study:

• To develop an accurate and automated method for sleep stage classification.
• To assist sleep experts by reducing the burden of manual sleep scoring.
• To enhance diagnostic capabilities in psychiatry and neurology through intelligent systems.

Main Methods:

• Utilized electroencephalogram (EEG) signals for sleep stage classification.
• Employed feature extraction using 20 algorithms, yielding 41 parameters.
• Applied feature selection techniques (mRMR, FCBF, ReliefF, t-test, Fisher score) to identify optimal features.
• Implemented classification algorithms including Random Forest, FFNN, DT, SVM, and RBF.

Main Results:

• Achieved a classification accuracy of 97.03% with the proposed automated method.
• Demonstrated the effectiveness of selected features in representing EEG signals for accurate classification.
• Provided a comparative analysis of computation times and accuracy rates across different classification algorithms.

Conclusions:

• The proposed automated sleep scoring method demonstrates high accuracy and efficiency.
• This intelligent system has the potential to significantly aid sleep experts in clinical practice.
• The findings support the development of advanced, AI-driven sleep analysis tools.