| The EEG, like many other biological phenomena, is quite likely governed by non-linear dynamics. Certain characteristics of the underlying dynamics have recently been quantified by computing the correlation dimensions (D{dollar}sb2{dollar}) of EEG time series data. In this dissertation, D{dollar}sb2{dollar} of the unbiased autocovariance (UAD{dollar}sb2{dollar}) function of the scalp EEG data was used to detect electrographic seizure activity and seizure focus localization. Digital EEC data were acquired at a sampling rate of 200 Hz per channel and organized in continuous frames (duration 2.56 sec, 512 data points). To increase the reliability of D{dollar}sb2{dollar} computations with short duration data, raw EEG data were initially simplified using unbiased autocovariance analysis to highlight the periodic activity that is present during seizures. The D{dollar}sb2{dollar} computation was then performed from the unbiased autocovariance function of each channel using the Grassberger-Procaccia method with Theiler's box-assisted correlation algorithm. Even with short duration data, this preprocessing proved to be computationally robust and displayed no significant sensitivity to implementation details such as the choices of embedding dimension and box size. As a result of clinical experiments, the system achieved a sensitivity of 86% and specificity of 95%, overall system performance was 90%. |
