Selection of optimal EEG channels for classification of signals correlated with alcohol abusers

Many brain events and disorders can be detected by analyzing electroencephalograms (EEGs). Also the availability of quantitative biological markers that are correlated with qualitative psychiatric phenotypes helps us to utilize automated methods to diagnose and classify these phenotypes. One such a psychiatric phenotype is alcoholism. In this study a method to select an optimal subset of EEG channels for the purpose of practical classification of alcohol abusers from normal subjects is proposed, which is based on combination of model-based spectral analysis and correlation matrices. The EEG signals were recorded when the subjects were represented with single trial visual stimuli. The proposed method proved successful in selecting an optimum number of channels which achieved acceptable average classification accuracy.