Abstracts

The introduction of dynamic nonlinear systems tools into the field of EEG analysis has shown great promise in the areas of seizure prediction, seizure detection and in constructing large scale models of nonlinear interactions amongst removal populations. While both surface and intracranial recordings have been used for these purposes there are few direct comparisons of the two data sets. Generally speaking intracranial data has proven to be a more powerful predictor and detector of seizure activity than surface recorded data. This raises the question of the degree to which intracranial activity is accurately reflected in extracranial recordings. The purpose of this study was to determine the degree to which intracranial nonlinear dynamic changes are accurately reflected on the surface of the scalp and to determine the surface and intracranial spatial distribution of those dynamic changes. Ten pediatric aged patients were included in this University of Chicago IRB approved protocol. Three seizures were selected from each patient. eigenvalues, correlation dimensions (least squares and maximum likelihood), Kolmogorov entropy and Z (a measure of global non-linearity) were calculated for each of thirty consecutive thirty second epochs, separated by one second (moving window approach). The software package RRCHAOS was used to calculate the above named measures. These moving windows were calculated for involved and uninvolved intracranial leads (12 to 34 electrodes) and cross correlated to eight surface electrodes (four on each side). The spatial distributions of these correlations were then analyzed. Delays between surface and intracranial changes were also calculated, as were delays attributable to propagation across the intracranial grid. Values of cross correlation varied considerably as a function of intracranial electrode location, extracranial electrode and the dependent variable. The highest correlations were observed for eigenvalues, where the values for the highest correlation usually exceeded 0.90. Kolmogorov entropy also demonstrated relatively high (0.75 to 0.95), but more variable results. The lowest cross correlation were found for the least squares cross correlations and the Z score was intermediate in value. As expected correlations were generally lowest for distant and contralateral electrodes, although there were some notable, but consistent exceptions. In general the waveform delays between surface and intracranial data were minimal as expected. The surface recorded waveform of the temporal evolution of nonlinear dynamic changes accurately reflects the behavior of intracranial resources. The degree to which surface recorded activity is distorted or diminished by the intervening tissues depends upon the dependent variable and the geometric relationship of the recording electrode to the source. (Supported by Falk Medical Trust Foundation)