Asymmetry index in anatomically symmetrized FDG-PET for improved epileptogenic focus detection in pharmacoresistant epilepsy

Abstract

OBJECTIVE Positron emission tomography (PET) imaging has assumed an essential role in the presurgical evaluation of epileptogenic foci in drug-resistant epilepsy by identifying the hypometabolic cerebral cortex. The authors herein designed a pilot study to test a novel technique of PET asymmetry after anatomical symmetrization coregistered to MRI (PASCOM), utilizing interhemispheric metabolic asymmetry on interictal fluorine 18–labeled fluorodeoxyglucose (FDG)-PET to better localize the epileptogenic zone. METHODS The authors analyzed interictal FDG-PET scans from 23 patients with drug-resistant epilepsy, mean (± SD) age 20.9 ± 13.1 years old, who had an Engel class I postsurgical outcome while followed up for > 12 months. T1-weighted and FLAIR MRI were used to create a patient-specific, structurally symmetrical template. The asymmetry index (AI) image was computed to detect the cerebral region of hypometabolism using different z-score threshold criteria to optimize sensitivity and specificity. The detected regions were compared with the resection cavity on postoperative MRI using predefined anatomical labels. PASCOM was compared with the visual analysis of FDG-PET by a nuclear medicine consultant blinded to other clinical data (VIS) and visual analysis during multidisciplinary team discussion (MDT). The efficacy of each technique was compared based on a performance score (S), sensitivity, specificity, and correct lateralization of epileptogenicity. RESULTS The mean S was maximum (1.30 ± 1.23) for AI images when thresholded at z > 4 and retaining the cluster of more than 100 voxels containing the peak AI value (Z4C) with 73.03% sensitivity and 96.43% specificity. The mean S was minimum for VIS (0.27 ± 0.31). The mean sensitivity was maximum for MDT (85.04%) and minimum for Z5C (AI images thresholded at z > 5 and clustered; 59.47%), whereas the mean specificity was maximum for Z5C (97.77%) and minimum for VIS (64.60%). Z3C (AI images thresholded at z > 3 and clustered) and Z4C were able to correctly identify the side of epileptogenicity in all the patients. CONCLUSIONS The PASCOM technique with a Z4C threshold had a maximum performance score with good sensitivity and specificity in localizing and lateralizing the epileptogenic zone. The described technique outperformed the conventional visual analysis of FDG-PET and hence warrants further prospective verification.