| The pathways by which epileptic seizures propagate through the brain are unknown. Since the speed of electrical activity in the brain is rapid, visual analysis of EEG gives limited information as to how the electrical activity propagates. This thesis gives results using the autoregressive (AR) spectral analysis method to estimate time differences of electrical activity between different sites of the brain and propagation routes of epileptic seizures.; Methodology developed and used in this study includes the EEG data acquisition software, the multichannel AR and the FFT spectral analysis methods, and an automatic method for time delay estimation. The study was done using two types of signal: simulated EEG data, which consisted of various sinusoidal signals plus Gaussian noise, and actual EEG data recorded by intracerebral electrodes from patients undergoing complex partial seizures.; Major findings of this study are: (1) Based on analysis of the simulated EEG data, the AR spectral analysis method is superior to the FFT spectral analysis method in terms of frequency resolution, variance of the estimation and accuracy of time delay estimation. The AR method is more powerful and reliable; its result is easily interpreted; and it does not require any smoothing algorithm. (2) Using the estimated time delays, the propagation routes through which seizure activity spreads were traced. The overall result for propagation routes is consistent with the neurological assessment, although some pathways at variance with present neurological understanding of EEG signal propagation were found. This method may prove useful as a pre-surgical evaluation tool. (3) The signal coherence and propagation routes change considerably during the initiation of the seizure, suggesting that this method could be extended to detect the starting point of seizure activity. |
