Localization of epileptogenic foci using artificial neural networks

Current neuropsychological tests based clinical methods are often difficult to interpret for localizing seizure foci in epilepsy patients. The purpose of this study is to predict with a high degree of certainty the location of the epileptogenic foci in epilepsy patients. First, neuropsychological tests data containing information thought to be relevant to the decision making process was extracted from patients' files. Next, the collected data was normalized and based on statistical analysis techniques. A set of best features was selected. These selected features were then analyzed using different classification techniques. The performance of each classifier was compared through the receiver operating characteristic analysis. Results show that the radial basis function classifier yielded the most promising results although other classification techniques produced satisfactory results as well.