CHANGES IN PERSISTENT LONG-RANGE TEMPORAL CORRELATIONS OF EEG DURING A SEIZURE

Rationale: Although changes in EEG activity during a seizure can be observed, statistical metrics allowing for automated seizure detection have proven difficult to obtain. Here we present results from studying EEG and ECoG using detrended fluctuation analysis (DFA).
Methods: DFA is a tool for examining the presence of persistent long range-temporal correlations in time series data such as EEG and ECoG. The raw EEG signal is squared to produce the energy of the signal. Next, the signal is divided into n blocks. For each block, the signal is demeaned and integrated. This signal can then be used to calculate the rms fluctuation, F (n), of the signal as a function of the number of blocks. A pot in log-log coordinates of the rms-fluctuation as a function of the number of blocks is linear. The slope of the linear trend is the DFA exponent.
Results: This method was applied to scalp and subduralictal and interictal ecorded data (as determined by the clinical neurophysiologist MHK) in two patients. The method indicates a statistically significant increase in the mean DFA exponent during the ictal state in both patients. There is no statistically significant difference in DFA-exponent between scalp and subdural recordings within a given patient and seizure state as has been suggested by similar DFA-exponents reported across separate studies using either scalp or subdural recordings [1,2].
Conclusions: These results indicate that persistent long range temporal correlations increase during the ictal phase. In addition, the linearity in log coordinates of the EEG fluctuations is reminiscent of many other complex systems, where local interacting subunits self-organize into a state displaying long range spatiotemporal correlations; a mechanism which has now come to be called self-organized criticality.
Acknowledgments: This work is supported by the Falk Foundation.