RESUMO
UNLABELLED: The aim of this study was to compare optimized voxel-based morphometry (VBM) and (18)F-FDG PET for discrimination between patients with Alzheimer's disease (AD) and healthy subjects in relation to age. METHODS: The study population consisted of 2 groups; the first group (27 AD patients and 40 control subjects) was used to determine the locations of significant abnormalities for both PET and VBM using statistical parametric mapping, and the second group (34 AD patients and 50 control subjects) was used to compare the diagnostic performance of PET and VBM. In the second group, a z-score map for PET or VBM of each subject was obtained by comparison with the mean and SD of PET or gray-matter MR images of the control subjects. Receiver-operating-characteristic (ROC) curve analysis was then performed to compare the diagnostic performance between PET and VBM. Furthermore, group 2 was divided into the early- and late-onset subgroups, and ROC analysis was performed for each subgroup. RESULTS: In the first group, VBM revealed a significant decrease in gray-matter concentration in the hippocampus complex in AD, whereas PET showed a significant reduction in (18)F-FDG uptake in the posterior cingulate and parietotemporal lobe. The diagnostic performance of PET (0.988 +/- 0.008; mean +/- SE), as measured by the area under the ROC curve, was higher than that of VBM with (0.782 +/- 0.059) or without (0.832 +/- 0.049) modulation. PET yielded a sensitivity, specificity, and overall accuracy of 100%, 92%, and 95%, respectively, whereas for VBM the sensitivity, specificity, and accuracy were 74%, 92%, and 85%. Modulation for the VBM did not improve these values (56%, 94%, and 79%, respectively). When the early- and late-onset subjects were analyzed separately, the superiority of (18)F-FDG PET was significant only in the early-onset subgroup. CONCLUSION: The present study indicates that the detection of metabolic alteration by (18)F-FDG PET yields a better diagnostic performance for the discrimination between AD patients and healthy control subjects than does the morphologic approach by VBM, particularly in the early-onset subjects.