Detection of k-complexes in EEG signals using a multi-domain feature extraction coupled with a least square support...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Accurate sleep scoring using electroencephalogram (EEG) signals is crucial for diagnosing sleep disorders.
• Errors in sleep EEG scoring can lead to misdiagnosis and improper treatment.
• Identifying specific sleep stage features, like k-complexes, is essential for precise sleep analysis.

Purpose of the Study:

• To develop an automated method for detecting k-complexes in sleep stage 2 using multi-domain features from EEG signals.
• To enhance the accuracy and efficiency of sleep scoring for clinical applications.
• To investigate the effectiveness of various signal processing techniques for sleep stage characterization.

Main Methods:

• EEG signals were segmented using a 0.5s sliding window.
• Multi-domain features (statistical, fractal, frequency, non-linear) were extracted from each segment.
• A least square support vector machine (LS-SVM) classifier was trained using 12 selected features to identify k-complexes.
• Performance was compared against K-means and extreme learning machine classifiers.

Main Results:

• The proposed method achieved an average accuracy of 97.7%, sensitivity of 97%, and specificity of 94.2% on the CZ-A1 channel.
• The multi-domain feature-based approach demonstrated superior recognition results compared to other classifiers.
• The LS-SVM classifier effectively identified k-complexes with high precision.

Conclusions:

• The developed automatic method accurately detects k-complexes in EEG signals, crucial for sleep stage 2 classification.
• This technique offers a promising tool for optimizing the diagnosis and treatment of sleep disorders.
• The findings highlight the potential of multi-domain feature analysis in advancing sleep research and clinical practice.