A radial basis function neural network model for classification of epilepsy using EEG signals

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Epilepsy is a significant neurological disorder impacting cortical excitability.
• Accurate epilepsy syndrome diagnosis is crucial for effective treatment and prognosis.
• Current diagnostic methods can be complex, necessitating advanced analytical tools.

Purpose of the Study:

• To evaluate epileptic patients and classify epilepsy groups using two distinct neural network models.
• To compare the classification performance of Radial Basis Function Neural Network (RBFNN) and Multilayer Perceptron Neural Network (MLPNN).
• To assess the potential of these AI models as decision support tools in clinical epilepsy diagnosis.

Main Methods:

• Utilized data from 418 epilepsy patients diagnosed per International League Against Epilepsy (ILAE 1981) criteria.
• Trained RBFNN and MLPNN models using patient EEG signals and clinical characteristics.
• Expert neurologists provided initial classifications for training and validation.

Main Results:

• Both RBFNN and MLPNN models demonstrated high accuracy on learning datasets.
• RBFNN achieved a total classification accuracy of 95.2%.
• MLPNN achieved a total classification accuracy of 89.2%, performing less successfully than RBFNN.

Conclusions:

• RBFNN significantly outperformed MLPNN in classifying epilepsy groups.
• The RBFNN model shows promise as a clinical decision support tool for confirming epilepsy classifications.
• Further development could integrate RBFNN into routine clinical practice for improved epilepsy diagnosis.