Computer-assisted lateralization of unilateral temporal lobe epilepsy using Z-score parametric F-18 FDG PET images.

Abstract

Background: To evaluate the use of unbiased computer-assisted lateralization of temporal lobe epilepsy (TLE) by z-score parametric PET imaging (ZPET).

Methods: 38 patients with histologically proven unilateral TLE due to pure hippocampal sclerosis, referred for pre-surgical PET evaluation of intractable seizure over a 5-year period, were included. The F-18 FDG images were oriented along temporal long axis and then transformed into ZPET images on a voxel by voxel basis. Multiple regions of interests (21 in total) were placed on cortical, subcortical and cerebellar structures on twenty-eight out of 38 patients with totally seizure-free (class I) outcome. Paired t-tests with Bonferroni correction were used to determine the location of the most asymmetric regions as variables for subsequent discriminant analysis of the entire group of the patients.

Results: The computer program identified the anterior half of the temporal lobe (p < 0.0005) and thalami (p = 0.021) as the most asymmetric regions in TLE patients with Class I outcome. Discriminant analysis using z-scores from a total of 8 ROIs (in 4 pairs) on these structures correctly lateralized thirty-seven out of 38 (97%) patients (sensitivity = 94%; specificity = 100%). The only false localization came from a patient with equivocal z-scores on the temporal lobes and this patient turned out to have poor outcome.

Conclusion: The computer-assisted lateralization of TLE using ZPET provides an accurate, fast and objective way of seizure evaluation.