A stable feature extraction method in classification epileptic EEG signals.

Area of Science:

• Neurology
• Biomedical Engineering
• Machine Learning

Background:

• Epilepsy is a common neurological disorder requiring accurate diagnosis.
• Electroencephalogram (EEG) signal analysis is crucial for epilepsy diagnosis.
• Feature extraction from EEG signals is a key challenge in accurate diagnosis.

Purpose of the Study:

• To improve local ternary patterns (LTP) for robust feature extraction from epileptic EEG signals.
• To evaluate the effectiveness of the proposed 1D-LTP method for epilepsy diagnosis.
• To compare the performance of various machine learning classifiers using the extracted features.

Main Methods:

• An improved one-dimensional local ternary patterns (1D-LTP) method was developed for EEG feature extraction.
• EEG data from the Bonn University's Department of Etymology was utilized for validation.
• Extracted features were classified using Support Vector Machine, Functional Trees, Bayes Networks, Artificial Neural Network, and Random Forest (RF).

Main Results:

• The proposed 1D-LTP method successfully extracted robust features from epileptic EEG signals.
• Random Forest (RF) achieved the highest classification accuracy among the tested machine learning methods.
• The achieved accuracies were competitive and successful compared to existing literature.

Conclusions:

• The improved 1D-LTP method is a promising technique for robust feature extraction in epileptic EEG analysis.
• Machine learning classifiers, especially RF, are effective in diagnosing epilepsy based on these extracted features.
• This approach contributes to more accurate and efficient epilepsy diagnosis through advanced signal processing.