AMYLOID LOAD: A NOVEL BIOMARKER WITH INCREASED SENSITIVITY FOR β-AMYLOID

Abstract

The cortical standardised uptake value ratio (SUVr) derived from PET amyloid tracers has previously been the outcome measure of choice to quantify the β-amyloid (Aβ) burden in the brain. We present a novel outcome measure called amyloid load, AβL along with a quantitative method to calculate it, as a more sensitive biomarker for molecular imaging studies of Aβ. The method to calculate AβL uses two previously derived template images1 in MNI152 space for the tracer non-specific binding (NS) and the carrying capacity (K) (Figure 1). Based on previous work which shows that a logistic growth model can describe the accumulation of in AD1, we hypothesise that any SUVr image volume can be described as a linear combination of these two template images; SUVrfit = αNS + AβLK. A fully automated algorithm is used to calculate (Figure 2) for any individual scan as follows; an individual [18F]-AV45 scan is first converted to an SUVr image by intensity normalisation to the grey matter cerebellum reference region, before being spatially normalised with DARTEL2 into MNI152 space and a voxel-wise regression is then performed to calculate α and AβL. We applied to algorithm to 769 [18F]-AV45 PET scans from the ADNI database (HC=211, EMCI=223, LMCI=203 and AD=132). We evaluated the fit of the SUVrfit images by calculating the percentage of voxels in the brain where |SUVr – SUVrfit|>0.3 for each subject. Finally, we calculated the effect sizes (hedges’ g) between the different clinical groups provided by ADNI using and compared those to effect sizes when using SUVr. AβL was successfully calculated for all 769 subjects. 80% of 769 subjects have less than 10% voxels where |SUVr – SUVrfit|>0.3. Example fitted and difference images are shown in Figure 3. There was an increase in effect size between all groups when compared to cortical SUVr (Figure 4). The mean Increase in effect size between groups was substantial at 56%.