Detection of seizures in EEG using subband nonlinear parameters and genetic algorithm

Area of Science:

• Biomedical Engineering
• Neuroscience
• Signal Processing

Background:

• Electroencephalogram (EEG) based seizure detection is crucial for epilepsy management but faces challenges from myogenic artifacts and time constraints.
• Accurate discrimination between interictal EEG in epileptic and normal subjects is clinically significant yet difficult to achieve.

Purpose of the Study:

• To develop and evaluate an automated method for seizure detection in EEG signals.
• To enhance the accuracy and efficiency of EEG seizure detection using nonlinear parameters and a genetic algorithm.
• To assess the method's capability in differentiating interictal EEG from normal EEG.

Main Methods:

• EEG signals were decomposed into subband components using discrete wavelet transform.
• Nonlinear parameters were extracted as features for Support Vector Machine (SVM) classifiers (linear and RBF kernels).
• A genetic algorithm (GA) was employed for feature selection and weight adjustment to optimize classifier performance.

Main Results:

• The proposed method achieved high seizure detection sensitivities: 90.8% for SVML and 94.0% for SVMRBF with GA feature selection.
• Further optimization using GA for weight adjustment increased the SVMRBF sensitivity to 95.8%.
• The method demonstrated effectiveness in discriminating interictal EEG of epileptic subjects from normal EEG.

Conclusions:

• The combination of subband nonlinear parameters and genetic algorithms offers a robust and accurate approach for automatic EEG seizure detection.
• The developed method shows promise for clinical application, particularly in distinguishing epileptic from normal EEG patterns.
• This automated system can potentially reduce the time and effort required for manual EEG analysis.