Performance Analysis of Entropy Methods in Detecting Epileptic Seizure from Surface Electroencephalograms

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Signal Processing

Background:

• Physiological signals such as Electrocardiography (ECG) and Electroencephalography (EEG) are inherently complex and nonlinear.
• Nonlinear analysis methods are crucial for extracting diagnostic information from these signals.
• Entropy estimation is a key nonlinear technique, with estimates serving as features for signal classification.

Purpose of the Study:

• To compare the performance of four entropy estimation methods: Distribution entropy (DistEn), Shannon entropy (ShanEn), Renyi entropy (RenEn), and Lempel-Ziv complexity (LempelZiv).
• To evaluate these methods as classification features for detecting epileptic seizures (ES) from surface Electroencephalography (sEEG) signals.

Main Methods:

• Experiments were performed on sEEG data from 23 subjects, sourced from the CHB-MIT database.
• The study analyzed and compared the efficacy of DistEn, ShanEn, RenEn, and LempelZiv for ES detection.
• Feature extraction involved calculating entropy estimates from sEEG signals.

Main Results:

• Shannon entropy (ShanEn), Renyi entropy (RenEn), and Lempel-Ziv complexity demonstrated significant potential as classification features.
• These methods proved effective for accurate and consistent detection of ES.
• The findings were consistent across multiple EEG channels and subjects.

Conclusions:

• ShanEn, RenEn, and LempelZiv are promising features for automated epileptic seizure detection using sEEG.
• Nonlinear entropy-based analysis provides a robust approach for diagnosing neurological conditions from EEG.
• Further research can leverage these findings for improved clinical diagnostic tools.