Objective: Alzheimer’ disease (AD) is characterized by increase of cortical amyloid deposition
in prodromal stage and subsequent decrease of cerebral glucose metabolism as disease progresses.
The present study introduces the voxel-wise metabolism to amyloid deposits ratio (MAR) image and to
evaluate its reliability for the diagnosis of AD.
Methods: Consecutive one-hundred and forty-three subjects with AD and 181 normal subjects who underwent
both 18F-FDG PET and 18F-florbetapir (AV-45) PET at baseline were included to this study
from the database of Alzheimer's disease neuroimaging initiative (ADNI). After normalizing to a standard
stereotactic space, the MAR image was created by dividing each FDG-PET image by corresponding
AV-45 PET image using with voxel-wise inter-image computation. We examined voxel wise comparison
in the MAR images between AD subjects and normal subjects and compared the diagnostic performances
between the MAR image and FDG-PET and AV-45 image.
Results: In the voxel wise comparison, the MAR images of AD subjects exhibited severe and extensive
decrease compared with normal subjects in the affected region in both FDG-PET and AV-45, especially
in the precuneus /posterior cingulate. The highest t-value was equivalent to FDG-PET and the voxel extent
was much greater than the other images. In the ROI analysis, the diagnostic accuracies were 82.6%
(sensitivity: 86.7%, specificity: 79.5%), 80.7% (sensitivity: 77%, specificity: 83.4%), and 78.8% (sensitivity:
75.2%, specificity: 81.5%) for the MAR image, FDG-PET, and AV-45, respectively. AUC for
the MAR image was 0.904 (95%CI: 0.867-0.942), and was larger than those for FDG-PET (AUC:
0.884, 95%CI: 0.843-0.926), and AV-45 (AUC: 0.847, 95%CI: 0.798-0.897).
Conclusion: MAR image reflected not only amyloid deposition but the cerebral glucose metabolisms
and successfully classified the subjects with AD. These data suggest that the MAR image might be a
more proper appropriate diagnostic marker for AD reflecting cerebral metabolisms and amyloid deposition.