Abstracts

RATIONALE: Independent Component Analysis (ICA), a technique for separating mixed, statistically independent spatiotemporal signal components, has been applied to EEG artifact removal [Vigario 1997, Jung et al. 2000], as well as separating spike components [Kobayashi 1999] and decomposing absence spike and waves. We wished to attempt to verify some of these results with high-density 128 channel EEG recordings. METHODS: A patient being evaluated for epilepsy surgery, after finishing standard clinical video-EEG monitoring for four nights, was recorded for an hour with a 128 channel EEG system. During the last few minutes of the recording he had a typical seizure with concomitant movement artifact that obscured the record. We applied ICA [Bell and Sejnowski 1995] to a two minute segment of EEG just before behavioral seizure onset. RESULTS: ICA decomposed the EEG into 128 components. By visual inspection one component was clearly related to eyeblinks and another to spikes and sharp waves. Most of the other components were either associated with artifact or no obvious pattern. CONCLUSIONS: With high-density EEG, ICA was able to separate ocular and other artifacts from epileptiform activity, thus highlighting the latter. Each of the ICA components has an associated spatial pattern which can be explained with a distinct set of neural generators. The additional use of high-density EEG can lead to better localization of each component. Bell, Sejnowski 1995 Neural Comput 7(6): 1129-59 Jung et al. 2000 Psychophysiol 37: 163-178 Kobayashi et al. 1999 Clin Neurophysiol 110: 1755-1763 McKeown et al. (in press) Brain Topog Vigario 1997 Electroenceph CLin Neurophyiol 103: 395-404