Detecting Epileptic Seizures in EEG Signals with Complementary Ensemble Empirical Mode Decomposition and Extreme...

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• Epilepsy is a prevalent nervous system disorder marked by recurring seizures.
• Electroencephalograms (EEG) are crucial for diagnosing epilepsy by recording brain activity.
• Accurate and automated seizure detection remains a challenge in clinical practice.

Purpose of the Study:

• To propose and evaluate an automated method for epileptic seizure detection using EEG signals.
• To enhance the accuracy and reliability of epilepsy diagnosis through advanced signal processing and machine learning.

Main Methods:

• Complementary Ensemble Empirical Mode Decomposition (CEEMD) was used to decompose EEG signals, mitigating mode mixing and end effects.
• Multi-domain features were extracted from raw and decomposed EEG data and selected based on importance scores.
• Extreme Gradient Boosting (XGBoost) was employed to build a robust epileptic seizure detection model.

Main Results:

• The proposed CEEMD-XGBoost method demonstrated superior performance on benchmark epilepsy EEG datasets (Bonn and CHB-MIT).
• The model achieved significant enhancements in sensitivity, specificity, and overall accuracy compared to existing EEG classification models.
• Feature selection based on importance scores contributed to the model's effectiveness.

Conclusions:

• The CEEMD-XGBoost approach offers a powerful and accurate tool for automated epileptic seizure detection from EEG.
• This method holds promise for improving the clinical diagnosis and management of epilepsy.
• Integrating CEEMD with XGBoost provides a synergistic approach for complex biological signal analysis.