Examining nonlinearity using complexity and entropy

Abstract

A method based on complexity and Shannon entropy along with surrogate data testing is described to detect nonlinearity in biosignals. The importance of denoising is illustrated in the detection of nonlinearity. The procedure is tested on synthetic linear and Lorenz data and on a large set of surface and intracranial electroencephalographic (EEG) signals. This method provides a measure of the complexity and entropy associated with nonlinearity. The results indicate that EEG signals measured during a seizure and from intracranial recordings show more nonlinearity when compared with surface EEG data and eyes open more than eyes closed signals.