NONLINEAR ANALYSIS OF THE DYNAMICAL CHARACTERISTICS OF EEG SIGNALS IN EPILEPTIC BRAINS

Abstract

Background and Purpose

 Brain seizures generally occur as a disorder of the nervous system that can cause fatal issues for patients due to their abrupt and unpredictable occurrence. Nonlinear analysis of electroencephalogram (EEG) signals from these patients can reveal substantial information about the underlying mechanism of the brain during the interictal, preictal and ictal stages.

 Methods

This paper analyses the nonlinear characteristics of 150 EEG signals collected from 10 epileptic patients to investigate their dynamical differences as the brain progresses through the interictal, preictal and ictal stages of a seizure. These characteristics were evaluated using time series analysis, phase space, power spectral density, and recurrence plots with corresponding quantification measures.

Results

The results illustrate that the ictal stage has a reduced amount of complexity and enhanced predictability through the nonlinear measures of Higuchi Fractal Dimension and Hurst exponent respectively, compared to the interictal stage. Furthermore, from the Recurrence plot, we have identified the long diagonal structures that characterize the periodic behavior of the ictal EEG signals. The phase portrait and the spectral analysis also justify these observations.

Conclusions

Compared to the interictal and preictal stages, the ictal stage has high values for average diagonal length but low values for Recurrence Time Entropy, confirming the highly recurrent and less complex nature of seizure activity. Interestingly, the DET measure is greater than 0.94 for all three seizure stages, indicating the deterministic nature of the epileptic brain. Our study concludes that the underlying dynamical processes of the epileptic brains during the three seizure stages are highly deterministic.

Keywords: Epilepsy, Nonlinear time series analysis, Recurrence Quantification Analysis, Higuchi Fractal dimension, Hurst exponent