Utility of gray-matter segmentation of ictal-Interictal perfusion SPECT and interictal 18F-FDG-PET in medically refractory epilepsy

Abstract

PurposeWe present a method of gray-matter segmentation of functional neuroimaging for localization of seizure onset zone (SOZ) in epilepsy surgery. 18F-FDG-PET hypometabolism and ictal SPECT hyperperfusion may correspond to SOZ. We hypothesize that limiting functional images to gray matter improves identification of small, subtle, or obscure cortical volumes of 18F-FDG-PET hypometabolism and eliminates hyperperfused seizure propagation pathways within white matter in ictal perfusion SPECT. MethodsTwenty-five adult and pediatric patients age 2–48 years with epilepsy surgery evaluations consisting of MRI, 18F-FDG-PET, ictal and interictal perfusion SPECT, and intracranial EEG (iEEG) monitoring were selected. MRI gray matter segmentation was used to identify cortical regions in coregistered 18F-FDG-PET and Ictal-Interictal SPECT Analysis by SPM (ISAS) as volumes of interest (VOI). VOIs in 18F-FDG-PET and SPECT perfusion clusters were compared to iEEG localization. The level of VOI concordance between two modalities was recorded as the same subgyrus (highest concordance), gyrus, sublobe, lobe, hemisphere, or no concordance. ResultsWith segmentation, 84% (21/25) of cases had at least one area identified on 18F-FDG-PET scan concordant with iEEG SOZ at sublobar or higher levels, and 72% (18/25) of cases had subgyral concordance with iEEG SOZ. Without segmentation, 60% (15/25) of cases had at least one area in 18F-FDG-PET scan concordant with iEEG SOZ at sublobar or higher levels, and 32% (8/25) with subgyral concordance. 83% (10/12) of seizure free patients had subgyral concordance on segmented 18F-FDG-PET. Both segmented and nonsegmented ictal-interictal SPECT perfusion clusters had 56% (14/25) of cases with at least sublobar concordance. Subgyral concordance was achieved by 28% (7/25) of segmented and 20% (5/25) of nonsegmented SPECTs. DiscussionSegmented 18F-FDG-PET scans frequently result in high correspondence to iEEG onset zones with localizations exactly concordant with iEEG SOZ- more than twice as often as without segmentation. Segmentation allows for the identification of small or subtle areas of hypometabolism that are often unappreciated or are obscured by normally hypometabolic white matter. Segmentation of ictal-interictal SPECT clusters did not significantly increase localization with iEEG SOZ over nonsegmented clusters.