Nonlinear time series analysis and its applications to absence seizure EEG

Abstract

An EEG is the recording of electrical activity produced by the firing of neurons within the brain. It gives a view of neural activity and has become one of the most important diagnostic tools in clinical neurophysiology, especially with respect to epilepsy. Since the 1980s, new measures based on discipline of nonlinear dynamical systems (chaos) have been developed for analyzing EEG data, which have become very powerful tools for characterizing hidden dynamic structures within epileptic EEG recordings. However chaos-based approaches must assume that EEG data possesses a non-evolving, low-dimensional attractor and require a long, stationary, and noiseless EEG data to compute the reconstructed attractor’s properties. To overcome the drawbacks of traditional nonlinear methods and meet the requirements of absence seizure EEG analysis, this dissertation applies new methods to characterize EEG changes in different absence-seizure states. First, to investigate whether information extracted from the EEG can provide evidence for the existence of a pre-seizure state in absence epilepsy, dynamic similarity measure and recurrence quantification analysis are used to indicate the dynamic characteristics of EEG in different absence seizure states. The results show that the average similarity measures between EEG segments within the seizure-free state are close to one, suggesting that the EEG segments within the seizure-free state share the same dynamic characteristics. The similarity measures between EEG segments across different seizure states are typically smaller. Furthermore, the determinism measure DET of pre-seizure EEG data are significantly higher than those of seizure-free states but lower than those of seizure states. These results demonstrate that the nonlinear characteristics of pre-seizure EEG are different from those of seizure-free and seizure EEG. Second, in order further to investigate hidden, nonlinear dynamic characteristics in EEG data for differentiating absence seizure states, order time series analysis is applied to analyse absence EEG data. The results show that the order time series analysis can track the dynamic changes of EEG data so as to describe transient dynamics prior to the absence seizures. Our results demonstrate that dissimilarity index and permutation entropy successfully can detect the pre-seizure state in 62 and 60 in 110 seizures, respectively. Compared with the sample entropy, the order time series analysis is more suitable to describe the nonlinear activity of EEG data, or better to extract the pattern of EEG data for the prediction of absence seizure. Finally, multiscale permutation entropy (MPE) is proposed as a tool to evaluate the dynamic characteristics of EEG during the seizure-free and seizure state, respectively. Simulation results show that the MPE method may be able to distinguish between noise and chaos series. In combination with the LDA method, MPE is used to analyse the seizure-free and seizure EEG data. It can be seen that the data separate into well-defined clusters. These results suggest that the MPE method might be a powerful tool to reveal the hidden characteristics of the epileptic EEG signals. Moreover, because neuronal oscillation and synchrony are associated closely with epileptic seizures, a permutation conditional mutual information method, which integrates order time series analysis and conditional mutual information, is applied to estimate a directionality index between two EEG recordings. A coupled mass neural model is used to demonstrate numerically the performance of the method; the results show that this method is superior to the conditional mutual information method for identifying the coupling direction between unidirectional or bidirectional neuronal populations.

Date of Award	15 Jul 2010
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Chuangyin DANG (Supervisor)