K-complexes Detection in EEG Signals using Fractal and Frequency Features Coupled with an Ensemble Classification...

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• K-complexes are crucial for identifying sleep stage 2.
• Manual detection of k-complexes is time-consuming and requires expert knowledge.
• Automated detection methods are needed to improve efficiency and accuracy.

Purpose of the Study:

• To develop an efficient and accurate method for detecting k-complexes in EEG signals.
• To leverage fractal and frequency features combined with an ensemble classifier.
• To provide a tool for sleep disorder diagnosis and sleep staging.

Main Methods:

• EEG signals were segmented using a sliding window technique.
• Dual-tree complex wavelet transform (DT-CWT) decomposed signals into 10 sub-bands.
• Fractal (Higuchi's algorithm) and frequency features were extracted from high sub-bands.
• An ensemble model combining LS-SVM, k-means, and Naïve Bayes classifiers was used.

Main Results:

• The proposed method achieved an average accuracy of 97.3% in k-complex detection.
• The ensemble classifier outperformed individual classifiers.
• Optimal performance was observed with a 0.5s window size.
• The method demonstrated efficiency compared to existing approaches.

Conclusions:

• The developed method offers an efficient and accurate approach for k-complex detection in EEG.
• This technique can serve as a valuable tool for sleep stage classification.
• The findings support its utility for clinicians in diagnosing sleep disorders.