The Modelling Of Multichannel Eeg And Synchronization Analysis

Synchronization is an important characteristic for establishing communications between different areas in the brain. Synchronization of neural signals is recognized as an important potential mechanism in a normal or pathological brain. Multi-channel EEG modelling and analysis can help to discover the information integrates, coordinates and propagates between different areas in the brain and understand the brain disorder mechanism, which is significant for diagnosing, preventing and treating brain diseases. The dissertation proposes a new multi-channel coupled neural model and a new multi-channel synchronization analysis method, which are applied to analyze epileptic EEG signals.Firstly, the study develops a new neural mass model, which can enable us to analyze the synchronization between different areas and understand the mechanism of epileptic seizures and propagation from the point of view of computing neural modeling. By arranging the subpopulations with different kinetics in parallel, a multi-kinetics neural mass model is constructed. By setting up coupling, which is the nonlinear outputs to another area with propagation, a multi-channel coupled neural mass model is constructed. The simulation results show that the new model can generate the EEG time series with varing frequencies ranging from delta to gamma wave (1Hz-50Hz); with the increasing of the coupling strength the simulated signals contain bi-modal and uni-modal at the frequency domain; by changing the weighting and coupling coefficients, the model can simulate epileptic EEG series. The simulations confirm that the abnormal synchronization of the massive neural populations could induce an epileptic seizure.Secondly, a new multi-channel EEG signals analysis method is developed to study the synchronization between different areas in the brain. The existing multivariate synchronization analysis methods are analyzed with the multi-channel coupled neural mass model, including S-estimator, Omega complexity, phase synchronization cluster analysis, Granger causality, frequency flow analysis and complex networks analysis. Some multivariate synchronization methods are based on the bivariate analysis, so the common bivariate synchronization methods are also evaluated, including cross correlation, coherence, phase synchronization, non-linear interdependency, corr-entropy coefficient and mutual information. Then, a new multi-channel EEG signals analysis method is proposed, denoted as correlation matrix analysis based on bivariate synchronization methods and surrogate data method. By using the multi-channel coupled neural mass model to generate the simulated EEG time series with different frequencies, channel numbers, coupling strengths, signal to noise ratios and time window widths, the performance of the correlation matrix analysis method is evaluated. The new method can reduce the effects of random synchronization due to the finite signal length and different frequency, track the change of the global synchronization strength at different epileptic states and frequency bands and give the number of synchronization clusters and the contribution of each channel to the synchronization clusters. Anyway this new method can contain both global and detailed information.Finally, using the correlation matrix analysis method, different epileptic ictal data are analyzed, including the in vitro epileptic data of rats, human local field potential and scalp EEG data. The synchronization change of the different areas at the epileptic interical, ictal and post-ictal stage can help us to understand the mechanism of epileptic seizures and propagation and to preliminarily determin the epileptic focus, the rate of accuracy is about 60%.