Applying nonlinear measures to the brain rhythms: an effective method for epilepsy diagnosis

Area of Science:

• Neurology
• Biomedical Engineering
• Signal Processing

Background:

• Epilepsy affects nearly 50 million people globally, highlighting the critical need for accurate diagnosis.
• Electroencephalogram (EEG) signal analysis is a primary method for characterizing epilepsy.
• Developing effective strategies for classifying epileptic EEGs is crucial for patient care.

Purpose of the Study:

• To evaluate the effectiveness of nonlinear features for classifying epileptic EEG signals.
• To identify the most informative features and frequency sub-bands for epilepsy detection.
• To assess the performance of different machine learning classifiers in epilepsy diagnosis.

Main Methods:

• Extracted four nonlinear features: Higuchi Fractal Dimension (HFD), Katz Fractal Dimension (KFD), Hurst exponent, and L-Z complexity.
• Analyzed features from EEG signals and their frequency sub-bands.
• Ranked features using the Relieff algorithm and applied them to MLPNN, Linear SVM, and RBF SVM classifiers.

Main Results:

• The Katz Fractal Dimension (KFD) emerged as the highest-ranking feature.
• EEG beta and theta sub-bands yielded the most important features (KFDs).
• High classification accuracies (up to 100%) were achieved using only KFD from beta and theta bands.

Conclusions:

• KFD extracted from beta and theta sub-bands are the most effective indicators for epilepsy classification.
• Using these top features significantly reduces classification complexity while maintaining high accuracy.
• This approach offers a simplified yet powerful method for epilepsy diagnosis using EEG data.