Klunk et al (2015) presented a method for standardization of quantitative amyloid imaging measures by scaling the outcome of each particular analysis method or tracer to the Centiloid scale. Herein we present our work converting florbetapir SUVr to the Centiloid scale.


      Florbetapir and PiB images were acquired 50-60 and 50-70 minutes post injection (respectively) for 46 subjects [13 young cognitively normal (YCN), 6 cognitively normal elder controls, 3 at-risk elderly, 7 MCI, 3 possible AD, and 14 AD] on two separate (18±20) days. SUVr values were calculated by previously published methods (Klunk et al, 2015; Joshi et al, 2015), which in turn were converted to Centiloids according to Klunk et. al. The first step was to validate our image registration pipeline and determine average PiB SUVr anchor points for stereotypical YCN (n=34) and AD (n=45) subjects using data from the Global Alzheimer’s Association Interactive Network (GAAIN). The second step was to employ linear regression to convert florbetapir and PiB SUVr to the Centiloid scale.


      Replicate analysis of GAAIN PiB data resulted in mean SUVr for YCN (n=34) of 1.004±0.048 and AD (n=45) of 2.067±0.204 (R2 0.997, slope 0.996, intercept -0.002). Our independent study showed that SUVr values (n=46) for florbetapir and PiB were correlated (R2 = 0.900). The regression equating florbetapir SUVr by the Joshi et al method to PiB SUVr by the Klunk et al method was y = 0.514x + 0.451 and appeared to support a linear relationship between the two tracers (Figure 1). The average YCN SUVr for florbetapir (0.98±0.07) and PiB (1.01±0.03), were converted to 1.51±12.06 and 0.32±2.61 Centiloids, respectively. 74 YCN florbetapir scans from a previous study (Clark et al, 2011), with average SUVr of 0.96±0.04 were converted to Centiloids using the same regression equations yielding an average of -1.60±7.59 Centiloids. Applied to the same study, a threshold of approximately 22 Centiloids, 3 SD above the young controls, was optimal for distinguishing cases with neuropathologically verified no/sparse vs moderate to frequent plaques.


      These findings provide for conversion of florbetapir SUVr to Centiloids, potentially allowing improved tracer independent amyloid quantitation.
      Figure thumbnail fx1
      Figure 1aSUVr for 46 subjects of mixed amyloid pathology who were scanned with both florbetapir and PiB.
      Figure thumbnail fx2
      Figure 1bCentiloids for 46 subjects obtained by linear regression through Figure 1a.