Abstracts

Rationale: Evidence for the role of interictal EEG-fMRI as a localizing investigational tool in the presurgical evaluation of refractory focal epilepsy is growing. A prerequisite for the clinical implementation of this technique is a straightforward EEG-fMRI processing. Accurate manual identification of interictal spikes recorded on EEG during fMRI can be challenging and time-consuming. We therefore propose a new semi-automatic method using the morphological characteristics of patient-specific spikes and compare the results with the conventional procedure based on manual spike identification.Methods: Our morphology-based method starts from a patient specific spike-template, built after averaging interictal spikes recorded during EEG-telemetry outside the scanner. The template was created as a time-amplitude map per channel starting from the spike onset to the peak of the slow wave. After standard artifact removal of the EEG measured during fMRI, we calculated for each channel the temporal coherence between the template and the preprocessed EEG using a moving time window of the same length as the template. The time-series of this coherence were averaged over all channels, squared, convolved with the canonical hemodynamic response function and resampled to the frequency of the fMRI recording. This function was used as regressor of interest in the SPM analysis. To validate our new method, we applied it to 8 adult refractory focal epilepsy patients with a well-defined ictal onset zone (IOZ) and with the presence of unifocal interictal spikes during simultaneous EEG-fMRI. We delineated the IOZ as the region of subtraction ictal SPECT co-registered to MRI (SISCOM) hyperperfusion (Z 1.5) within the resection zone in patients who underwent successful epilepsy surgery (3/8) or within the presumed epileptogenic zone based on all available concordant investigations after a full presurgical work-up (5/8). Blood oxygen level dependent (BOLD) responses were determined from the SPM analysis using a statistical threshold of uncorrected p < 0.001 at the voxel level and corrected p < 0.05 at the cluster level. An overlap between suprathreshold voxels and the IOZ indicated that the method was able to find BOLD activations in the IOZ. We compared these results with the conventional analysis.Results: Three patients had temporal lobe epilepsy, while the IOZ was localized extratemporal in the other 5 patients. BOLD activations in the IOZ were seen in 7/8 (87.5%) patients using our morphology based semi-automatic method. Using the conventional method, we found 6/8 (75%) patients all of whom were also found with our method.Conclusions: We propose a new semi-automatic morphology-based analysis to determine the regressor of interest in EEG-fMRI. This approach shows similar performance for detecting BOLD changes in the ictal onset zone as the conventional method. This is an important step towards a more standardized, less reader-dependent and time-consuming analysis of EEG-fMRI, which is crucial for the translation of the technique into clinical practice. Funding: IWT-TBM grant No.080658