Neural network recognition of otoneurological vertigo diseases with comparison of some other classification methods

Abstract

We have studied computer-aided diagnosis of otoneurological diseases which are difficult, even for experienced specialists, to determine and separate from each other. Since neural networks require plenty of training data, we restricted our research to the commonest otoneurological diseases in our database and to the very most essential parameters used in their diagnostics. According to our results, neural networks can be efficient in the recognition of these diseases provided that we shall be able to add our available cases concerning those diseases which are rare in our database. We compared the results yielded by neural networks to those given by discriminant analysis, genetic algorithms and decision trees.