A New Pattern Discriminant Method for Evoked Potentials

Abstract

A statistical approach to evaluate evoked potentials (EPs), consisting of three stages in procedure, was proposed with an application to study on postictal changes of visually evoked potentials (VEPs).

At the first step, time series data of EPs were reduced by the means of using an autoregressive (AR) model. In this way, EPs were described in terms of AR coefficient vectors. Then statistical distances which are well known as the Mahalanobis' distances were determined as a scale for dissimilarity between patterns of EPs. At the third stage, a retrospective classification through a numerical taxonomy based on similarity of patterns was demonstrated. This procedure will tell us some standards for future experiments as well as an integrative insight of results from present experiment.

By the above method, it was suggested that there may be two different processes in postictal recovery of kindled cats.