Abstracts

EEG source localization has been utilized to identify epileptogenic generators non-invasively. As propagation of electrical activities in ictal events is evident, the task for ictal source analysis is to disentangle the neural sources responsible for seizure onset from the sources due to seizure propagation. The aim of the present study is to develop a novel ictal source analysis method for localizing epileptogenic foci from scalp EEG by distinguishing seizure onset sources from propagation sources., The new ictal source analysis method includes three steps. First, a spatio-temporal source localization, i.e. FINE (Ding & He, IEEE BME, in press), is used to locate the multiple sources underlying ictal events. Secondly, an autoregressive model is employed to characterize the dynamic behaviors of the sources. Finally, the causal interaction patterns among the sources are estimated by Granger causality based on the autoregressive model. These patterns are used to distinguish the seizure onset sources from the propagation sources. Twenty seizures from five patients (2-7 per subject) with medically intractable partial epilepsy and MRI lesion were analyzed using the new ictal source analysis method. The patient[apos]s EEGs were recorded using 31 electrodes positioned according to the modified 10/20 system. The results were then validated by comparing to the lesions seen on MRI., Most seizure onset sources (19/20) were localized at the vicinity of MRI lesions. The propagation sources could either appear close to the lesional areas or far away from the lesions. Fig. 1 shows an example from one patient. Two sources were identified: one within the area of the visible lesion (top); and another just outside of the visible lesion (bottom). The waveforms for each of these sources are shown in green. The bivariate Granger causality estimation demonstrates a strong directional causal interaction from the source within the lesion, defined as the seizure onset source, to the other located just outside of the lesion, defined as the propagation source., We have developed a novel approach to characterize ictal events in space, time, and frequency domains. The causal interaction patterns were used to distinguish seizure onset sources from propagation sources. The present pilot study demonstrates that the seizure onset sources, estimated using the present method, are consistent with the MRI diagnosis for lesional patients. The present method may be applicable to non-lesional patients.[figure1], (Supported by NIH EB00178 and NSF BES-0411898. LD was supported by a DDF fellowship from the Graduate School of the UMN.)