Abstracts

Rationale: Automated seizure detection often miscategorizes artifact or sleep patterns as spikes or seizures. This results in numerous files without clinical significance. Dynamic measures are useful in assessing complex systems analysis of biological phenomena, like those findings of EEG. The purpose of this study was to determine the sensitivity of dynamic systems analysis to these nuisance variables. Methods: Files of 13 children with routine EEG s with spikes, stage II sleep, or artifact were used for the analysis. Segments of 30-60 seconds of the EEG were selected. The particular EEG channel which showed the finding best was used. The segment of the EEG was converted to a CVS file for use in Matlab macros that prepared the data for the statistical program rrChaos. Calculations include principle Eigen value, average Mann-Whitney, Correlation Dimension and Correlation Dimension with least square fit, Kolomolov Entropy in bits/second and Kolmogrov in cycles/second and Maximum Possible Kolmologrov Entropy in bits/second, Average absolute deviation in units of time series, average cycle frequency, absolute noise level as dimensionless units, absolute noise level as units of time series, and an average of b. All data points were plotted with time on the x-axis and the data values on the y-axis for the selected measures. Average sores of the calculations ere made and averaged over time.Results: There were 6 EEG s with spikes, 8 with artifact, 7 with sleep, and 2 with seizures. All subjects were children between the ages of 3 and 16 years. Comparisons over the three conditions found no changes in Eigen values or z-scores, both of which dramatically change with seizure onset. Sleep and spikes on EEG showed no differences in z-scores, Entropy values, or correlation dimension. With both spikes and seizures, however, entropy calculations were elevated above baseline by 20 to 50 times or more for all subjects. The results of the calculations remained stable over the course of the EEG time period. This is unlike seizures where there is clear biphasic changes in entropy values with the onset of a seizure. Absolute noise values are unchanged in the three conditions and between subjects.Conclusions: 1. Eigen values do not change with spikes, artifact, or sleep but do drop with seizure onset consistently for greater that 90% of the time. 2. Entropy could distinguish between artifact and spikes or sleep, but not between spikes and sleep, but Entropy changes with seizures onset, which is different than both. 3. Dynamic system measures of seizures have a low likelihood of being confused with sleep spikes, or artifact.