Abstracts

Rationale: Children with intractable focal epilepsy and normal MRI are a challenging group for epilepsy surgery. 2-deoxy-2[F-18]fluoro-D-glucose positron emission tomography (FDG PET) has been used to assist presurgical localization of the seizure focus in these patients. Objective, accurate detection of potentially epileptogenic regions by PET is technically difficult in children. Our previous study achieved up to 75% accuracy by using an adult control group (Kumar et al., J Nucl Med, 2010;51:1901-7); however, this approach has not been validated below 6 years of age. In the present study, we evaluated the performance of an optimized analytic approach utilizing a large pediatric pseudo-control group (a concept originally used by Chugani et al., Ann Neurol, 1987;22:487-97 and recently applied in voxel-by-voxel comparison by Archambaud et al. EJNMMI Res, 2013;3:2) for an objective analysis of FDG PET scans of epileptic children above 1 year of age. Methods: Eighteen children (mean age: 9.1 years; range: 1.2-15 years) with medically refractory epilepsy of neocortical origin and non-localizing MRI underwent two-stage epilepsy surgery with chronic subdural EEG monitoring at the Children's Hospital of Michigan (Detroit). FDG PET scans of 64 children (pseudo-control group; mean age: 9.0 years; age range: 1-18 years) with non-lesional focal epilepsy and normal MRI and FDG PET (on visual analysis) were processed to create 5 pediatric PET templates for different age groups. Statistical Parametric Mapping (SPM) was used to compare the PET scans of individual surgical subjects with those of the appropriate age-matched control subgroup. After an optimization procedure using control subgroups via grid-search algorithm, a statistical threshold of uncorrected p=0.001 was applied to each patient's scan for detection of hypo- and hypermetabolic clusters (with a 50-voxel minimal extent). SPM-detected locations of hypo- and hypermetabolic clusters were compared to lobar locations of surgical resection as well as surgical outcome as the gold standard. Localization accuracy was determined for the whole group and for subgroups with and without frequent spiking on EEG detected during PET. Results: In the 18 patients, a total of 40 lobar regions were surgically resected due to epileptogenicity determined by intracranial EEG recordings. Presurgical FDG-PET with optimized SPM analysis detected hypometabolic clusters in 30 (23 true-positive for resection site as the gold standard) and hypermetabolic clusters in 18 (16 true-positive) cerebral lobes ipsilateral to the resection. Hypo- and hypermetabolic clusters occurred with a closely equal distribution in patients with and without frequent interictal spiking on PET-EEG, and their lobar location was largely complementary (occurring in the same lobe in only 22% of the cases). The combined group of hypo+hypermetabolic clusters detected the epileptogenic (resected) lobes with 80% sensitivity and 75% specificity (78% accuracy). The performance of SPM-defined metabolic abnormalities was similar in post-operatively seizure-free patients. Detection accuracy was even higher in those patients (n=11) where PET-EEG showed no frequent spiking (88% sensitivity, 78% specificity, 82% accuracy). Conclusions: The applied optimized approach for objective PET analysis, employing age-matched pediatric pseudocontrol groups and including both hypo- and hypermetabolic clusters, provides a superior accuracy (around 80%) for detecting epileptogenic lobar regions in children above 1 year of age. This approach may enhance presurgical localization of epileptic foci in children with normal MRI across a wide age range. Funding: NIH grants NS089659, NS064033