ABSTRACT
OBJECTIVE: Our aim is to validate the process steps implemented by the French CATI platform to assess amyloid status, obtained from 18F-Florbetapir PET scans, in a cohort of 318 cognitively normal subjects participating in the INSIGHT-preAD study. Our objective was to develop a method with partial volume effect correction (PVEC) on untransformed PET images, using an automated pipeline ("RACHEL") adapted to large series of patients and including quality checks of results. METHODS: We compared RACHEL using different options (with and without PVEC, different sets of regions of interest), to two other methods validated in the literature, referred as the "AVID" and "CAEN" methods. A standard uptake value ratio (SUVR) was obtained with the different methods for participants to another French study, IMAP, including 26 normal elderly controls (NEC), 11 patients with mild cognitive impairment (MCI) and 16 patients with Alzheimer's disease (AD). We determined two cutoffs for RACHEL method by linear correlation with the other methods and applied them to the INSIGHT-preAD subjects. RESULTS: RACHEL including PVEC and a combination of the whole cerebellum and the pons as a reference region allowed the best discrimination between NEC and AD participants. A strong linear correlation was found between RACHEL and the other two methods and yielded the two cutoffs of 0.79 and 0.88. According to the more conservative threshold, 19.8% of the INSIGHT-preAD subjects would be considered amyloid positive, and 27.7% according to the more liberal threshold. CONCLUSIONS: With our method, we clearly discriminated between NEC with negative amyloid status and patients with clinical AD. Using a linear correlation with other validated cutoffs, we could infer our own positivity thresholds and apply them to an independent population. This method might be useful to the community, especially when the optimal cutoff could not be obtained from a population of healthy young adults or from correlation with post-mortem results.
