Chaotic Time Series Analysis Using Short and Noisy Data Sets: Application to a Clinical Epilepsy Seizure

Abstract

In recent years the methods of chaotic time series analysis have been applied to data sets from a number of experimental systems. In this note we report on work in progress on the usefulness of chaotic time series analysis as a potential diagnostic tool in the classification of epileptic seizure activity [1]. Seizure episodes are classified by correlation dimension and estimated largest Lyapunov exponent. The exponent is found using a modified version of Wolf’s method [2]. We begin with a brief discussion of some of these modifications before proceeding to a discussion of their application to an epilepsy data set.