Electroencephalogram signals processing for topographic brain mapping and epilepsies classification.

Area of Science:

• Neuroscience
• Computational Biology
• Signal Processing

Background:

• Epilepsy diagnosis relies on electroencephalography (EEG) analysis, which can be complex.
• Distinguishing between epilepsy types (grand mal, petit mal) and normal EEGs requires sophisticated methods.
• Artifacts in EEG signals can hinder accurate classification.

Purpose of the Study:

• To develop and evaluate a novel method for classifying epilepsy types using topographic brain mapping and wavelet transform-neural networks.
• To improve the accuracy and efficiency of EEG-based epilepsy diagnosis.
• To differentiate between healthy, ictal, and interictal states in epilepsy patients.

Main Methods:

• Preprocessing of EEG signals using Discrete Wavelet Transformation (DWT) to remove artifacts and notch filtering to eliminate power line interference.
• Application of Counties Wavelet Transform (CWT) to capture transient EEG features for classifier input.
• Implementation of a two-stage Learning Vector Quantization (LVQ) neural network classifier operating in time and frequency domains.
• Utilizing topographic brain mapping to identify epilepsy's location within the brain.

Main Results:

• Preprocessing significantly enhanced the speed and accuracy of the wavelet transform and neural network stages.
• The proposed two-stage classifier achieved approximately 80% accuracy in classifying experimental clinical EEG data.
• The method successfully categorized EEG signals into normal, petit mal, and clonic epilepsy types.

Conclusions:

• The combined approach of topographic brain mapping and wavelet transform-neural network analysis offers a promising method for epilepsy classification.
• The developed preprocessing techniques effectively reduce noise and artifacts in EEG signals.
• This study demonstrates the potential of advanced computational methods for improving epilepsy diagnosis and understanding.