Application Of Nonlinear Dynamics Analysis Of EEG: Epileptic Seizyre Prediction

Recently the development of nonlinear dynamics accelerates the research of epilepsy. It is reported that the processes of epileptic seizure include interictal state, ictal state and preictal state. The preictal state is the transition of brain activity towards an epileptic seizure, which differs from the interictal state and ictal state. Detecting the preictal state prior to the epileptic seizure and succeeding in the epileptic seizure prediction, the preventive action or therapeutic measures could be taken to minimize risk or injury and greatly improve the quality of life of many people with epilepsy. Aiming at prediction of epilepsy, the whole process of epileptic seizure is studied and analyzed using nonlinear dynamics method. The nonlinear dynamical changes of brain electrical activity during preictal state are analyzed, and the epileptic seizure prediction is studied. We do the following work: The Recurrence Quantification Analysis (RQA), which can measure the complexity of a short and non-stationary signal with noise, is proposed to describe dynamical characteristics of EEG recordings on rat experiments. Calculating RQA variables based on moving-window technique, the hidden dynamics characteristics of epileptic EEG data can be tracked, and the law of complexity change of EEG recordings during the epileptic seizure, which could be applied to predict the epileptic seizure. In order to reconstruct the phase space of EEG recordings, the multiple-autocorrelation and Cao's method are employed to calculate the parameters of embedding dimension and delay time. A Heavyside function in correlation integral is replaced by a Gaussian function on calculation fuzzy similarity index, which eliminates the effect from the crisp boundary of Heavyside function. A lot of EEG recordings of rat experiment are analyzed to compare the prediction performance of fuzzy similarity with those of dynamical similarity.