Statistical pattern classification of clinical brainstem auditory evoked potentials

The brainstem auditory evoked potentials (BAEPs) recorded in the neurological clinic were classified using the Bayes classifier (BC) and Fisher's linear discriminant function (FLD). The latencies of initial five peaks, interpeak intervals were examined for optimum features to develop classifiers. The accuracy of classification was 85.3% when absolute latencies of peaks III, IV and V were used as features. The BC gave better performance than FLD, indicating that second order statistics of BAEPs for normal and pathological classes are different. The results of this study indicate that latencies alone give enough information for recognizing normal from pathological BAEPs, using physician's evaluation of BAEPs as the reference.