Comparison of PiB and FDG in the Differential Diagnosis of Alzheimer’s Disease
Objectives : Both quantitative and visual assessment of Aβ deposition as measured by PiB-PET has shown a robust difference between Alzheimer's disease (AD) and healthy controls
(HC). (Ng et al., 2007) The accuracy for PiB of commonly used anatomic normalization
techniques such as NeuroStat has not yet been determined, nor has it been compared to FDG. This study compared the accuracy of visual assessment, NeuroStat, and semiquantitative region-of-interest analysis for PIB and FDG images in the same subjects.
Methods : Seventy-two participants (27 HC, 38 AD, and 7 frontotemporal dementia –FTD-) were evaluated. PiB and FDG images were assessed separately by two experienced nuclear medicine physicians, blinded to clinical diagnosis. PiB and FDG images were graded aS normal, AD, or FTD. For sensitivity and specificity calculations purposes, AD was compared separately against HC and FTD. Using the cerebellar cortex as reference region, SUVR thresholds of 1.6 and 1.0 were selected to define normal or abnormal values in the semiquantitative analysis of PiB and FDG, respectively.
Results : NeuroStat anatomical normalized images showed the same accuracy (83%) in the classification of FDG images as the semiquantitative approach (84%) and visual reading (81%). PiB was more accurate than FDG, irrespective of the method used: NeuroStat (93%); semiquantitative (98%); or visual assessment (94%).
Conclusions : FDG-PET images appear to be less accurate than PiB-PET images for the diagnosis of AD. Anatomical normalization techniques work well with PiB-PET, providing similar accuracy to experienced readers.