Abstracts

Rationale: Spike correlated functional Magnetic Resonance Imaging (S-MRI) analysis is a non-invasive method for localizing epileptogenic tissue. However, since S-MRI localizes the source of interictal spikes, it is not clear whether the tissue localized by S-MRI is the same as the ictal (or seizure) onset. This study addresses this issue by comparing the S-MRI localization with an imaging method that is based on ictal events, namely the difference Single Photon Emission Computed Tomography (D-SPECT) analysis.Methods: Data Acquisition: The same set of six patients with intractable partial epilepsy participated in the S-MRI and D-SPECT analyses after informed consents. For the S-MRI analysis, patients stayed still in a 3T MRI scanner during six runs of ten-minute simultaneous EEG-fMRI acquisitions. After the magnet noise was removed from the EEG signal offline, EEG was examined to identify interictal spikes, and the timing of these spikes was used to localize the sources of interictal spikes using a generalized linear model. For the D-SPECT analysis, patients were injected with a radioisotope agent during seizure episodes and were scanned in a SPECT scanner after the seizures ended. Baseline SPECT images were then subtracted from these ictal SPECT images to yield the D-SPECT images (see http://spect.yale.edu). Data Analysis: First, the S-MRI and the D-SPECT images were independently examined by experts for epileptogenic tissue localization. Next, both the S-MRI and the D-SPECT images were spatially normalized to a standard template and negative voxel values were set to zero. The resultant images were thresholded at various levels (varying from 0.3 to 15 multiplied by the standard deviations), and the union of suprathreshold voxels in S-MRI and D-SPECT images were subject to image correlation analyses. Results: The S-MRI and the D-SPECT analyses localized to concordant loci in four out of six patients. Three of these patients (subject #1,2,3 in the figure) had high (r>0.6) image correlations at focal spatial scales (< 1000 voxels), whereas one subject (#4) had a mild (r>0.4) correlation only at broad spatial scales (> 5000 voxels). Note that the number of voxels within a normalized brain was about 100000. Two remaining patients had non-localizing analyses in either S-MRI (#6) or D-SPECT (#5) and had low correlations (r<0.15) at all scales.Conclusions: The result of the image correlation analysis between the S-MRI and the D-SPECT images suggests that there are similarities in the spatial patterns of brain metabolism during interictal spikes and seizures in some patients. A larger number of studies are needed for assessing the concordance between the interictal and ictal spike onsets across a range of pathologies. This research was partly supported by NIH R01-NS47605.