Abstracts

Rationale: EEG-correlated functional MRI (EEG-fMRI) detects the distribution of brain regions involved with interictal epileptic discharges (IEDs). The goal of this study was to cross-validate EEG-fMRI correlation patterns with intracranial EEG recordings and outcome of surgery as gold standards. We also assessed the added value of Independent Component Analysis (ICA), a data-driven technique applied to fMRI alone, for those cases where no IEDs are present in the EEG. Methods: EEG and fMRI data were acquired of 21 patients who were later implanted with depth electrodes (n=5) or subdural grids (n=16). EEG was corrected for gradient and pulse artifacts and annotated for the presence of IEDs. These events were used as a predictor in the general linear model framework, yielding a correlation pattern indicating the brain regions that were significantly associated with the IEDs. For the validation of these results, a quantitative approach was developed to reveal the spatiotemporal patterns of the IEDs present in the invasive EEG (Van Houdt PJ, Ossenblok PPW, Colon AJ, Boon PAJM, de Munck JC (2012) A framework to integrate EEG-correlated fMRI and intracerebral recordings. NeuroImage 60: 2042-2053). The EEG-fMRI results were also compared to the location of the seizure onset zone and resection area. The fMRI data of those patients in whom EEG-fMRI overlapped with the resection area and who were seizure free after surgery (Engel score 1, n=9), were also analyzed with ICA (Beckmann CF, Smith SM (2004) Probabilistic independent component analysis for functional magnetic resonance imaging. IEEE Trans Med Imaging 23:137-152) for data selections with and without IEDs. Results: Table 1 summarizes the results of the comparison of EEG-fMRI and invasive EEG patterns. At least one EEG-fMRI cluster overlapped with active invasive electrodes (i.e. electrodes reflecting high correlations) in all data sets, whereas EEG-fMRI overlapped with all active ECoG/SEEG electrodes in 50% of the datasets. In 11 of the 16 ECoG datasets, more than one BOLD region was concordant with active electrodes; some regions were related to the onset of spike activity and others to propagation. The EEG-fMRI clusters included the complete seizure onset zone in 83% of the data sets and the resection area in 90% of the data sets. These values are similar to those obtained by comparing interictal invasive patterns to the seizure onset zone (83%) and resection area (95%). Finally, for the 9 datasets in which ICA was performed, one component was concordant with the seizure onset zone and resembled the EEG-fMRI pattern regardless of whether IEDs were present in the corresponding EEG or not. Conclusions: This study shows that EEG-fMRI has substantial predictive value regarding the seizure onset zone and resection area. In addition with the finding that ICA may reveal epileptic components without the presence of IEDs, this suggests that resting-state fMRI could play an important role in the presurgical planning of invasive recordings.