A Novel Permutation Entropy-Based EEG Channel Selection for Improving Epileptic Seizure Prediction

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computational Biology

Background:

• Epileptic seizure prediction relies on electroencephalography (EEG) signal analysis, with feature extraction and classification being critical.
• Reducing computational load in EEG analysis is essential for developing efficient seizure prediction algorithms.
• Patient-specific approaches are needed to optimize the performance of seizure detection systems.

Purpose of the Study:

• To develop a highly effective and accurate patient-specific algorithm for epileptic seizure prediction.
• To investigate the impact of efficient EEG channel selection on prediction performance and computational efficiency.
• To present a novel method for EEG channel selection using permutation entropy and a genetic algorithm.

Main Methods:

• A patient-specific optimization method for EEG channel selection was developed, utilizing permutation entropy (PE) values.
• K nearest neighbors (KNNs) were employed in conjunction with a genetic algorithm (GA) for channel selection.
• Support vector machine (SVM) was used as the classifier, and the CHB-MIT Scalp EEG Database was utilized for validation.

Main Results:

• The proposed method achieved a high average prediction rate of 92.42% using selected channels, compared to 71.13% with all channels.
• On average, accuracy, sensitivity, and specificity improved by 10.58%, 23.57%, and 5.56%, respectively, with selected channels.
• Four patient cases demonstrated over 90% accuracy, sensitivity, and specificity with a reduced set of channels, showing improved robustness.

Conclusions:

• Patient-specific EEG channel selection is a robust and effective strategy for optimizing epileptic seizure prediction.
• The proposed method significantly enhances prediction accuracy, sensitivity, and specificity while reducing computational demands.
• Tailored channel selection offers a promising approach for developing more efficient and reliable seizure prediction systems.