ANN for electrophysiological analysis of neurological disease

The aim of this study was to develop a discriminant analysis based both on classical linear methods, as Fisher's Linear Discriminant (FLD) and Likelihood Ratio Method (LRM), and non-linear Artificial Neural Network (ANN) classifier in order to distinguish between patients affected by Huntington's disease (HD) and normal subjects. R.O.C. curve analysis revealed ANN to be the best classifier. Moreover the network classified gene-carrier relatives as normal thus suggesting the EEG to be a marker of the evolution of the HD.