Abstracts

Rationale: Medically refractory epilepsy is often amenable to surgery. One technique used to identify the epileptogenic zone is the subtraction ictal SPECT coregistered to MRI (SISCOM). However, SISCOM does have several limitations. One limitation is the threshold to use when comparing areas of ictal hyperperfusion compared to the interictal SPECT study. Traditionally, a z score of 2 is used as the threshold. Some may vary the threshold for each individual, but no study has been performed to systematically examine the specificity and sensitivity of different z scores. Furthermore, several perfusion patterns may emerge at each threshold. These patterns may reflect both the seizure onset and propagation which may provide important localizing information. Most previous studies did not take into account the observed patterns of hyperperfusion and only take the most hyperperfused region to represent the seizure onset zone. In this study we investigated the location and patterns of hyperperfused regions using different z thresholds. We propose that for each patient the z threshold should be based on the observed patterns of hyperperfusion to achieve the most accurate result. Methods: A database of over 300 patients who underwent surgical resection for refractory epilepsy was reviewed. The surgical sites were both temporal and extratemporal. Approximately 70 patients who had an ictal and interictal SPECT and both a pre- and post-operative MRI were identified. From these patients 28 (mean age 33 years old, range 2-65) have remained seizure free for at least 6 months. Post-surgical MRI was co-registered with pre-MRI using MINPAV and SPM (Matlab). SISCOM was performed to both pre- and post-operative MRI. The location and patterns of hyperperfused regions at varying z thresholds were compared using MRIcro. The epileptogenic zone was defined as the site of surgical resection and seizure freedom. Results: We found that the optimal threshold for SPECT-SISCOM localization of the epileptogenic zone was between 1.5 to 2 standard deviations (1.5 SD n= 15, 54%; 2.0 SD n= 12, 43%). Notably, in patients who were either not localized or incorrectly localized according to our standard SPECT-SISCOM localization criteria (n=12), we identified a focus of hyperperfusion that co-localized with the site of resection in 7 patients (58%) when SPECT-SISCOM was thresholded at 1.5 standard deviation. Figure 1 shows representative SPECT images of one patient at varying z thresholds overlaid on the postoperative MRI. The threshold of 1.5 showed optimal sensitivity, with an hour-glass pattern of hyperperfusion overlapped with the resected area. Conclusions: SISCOM is a useful modality in evaluation of patients for epilepsy surgery. Although a z score of 2 is traditionally used, this threshold may not be optimal for every patient. Our preliminary studies suggest a z score of 1.5 may increase the sensitivity in some patients. We propose that the determination of z score threshold should be guided by the observed patterns of hyperperfusion, which reflect different degrees of seizure propagation.