Amyloid-specific PET tracers provide quantitative measurements for determining amyloid load in vivo. Many cutpoints have been proposed to define amyloid positivity; those based on post-mortem pathology are highly specific, but may not be sensitive to the initial signs of amyloid deposition. Cutpoints are also highly dependent on many analytical issues including: region of interest (ROI), reference region, partial volume correction, and the statistical criteria for choosing the cutpoint. In the absence of CT scanning, quantifying amyloid load with combined PET/MR scanners requires novel techniques for attenuation correction. We examined these aspects in a large UK sample of individuals born in the same week in 1946, all scanned on the same PET/MR scanner.


      PET and MR data were acquired on 250 participants enrolled in Insight 1946, a sub-study of the MRC National Survey of Health and Development (NSHD). Scanning was performed on a Siemens Biograph PET-MR scanner using 18F-florbetapir. PET images were reconstructed from 50 to 60 minutes post-injection using two different attenuation correction techniques: an ultrashort echo time (UTE) and a pseudo CT (pCT) method. Volumetric T1 MR data were parcellated into ROIs and co-registered to PET to compute standard uptake value ratios (SUVR) against four commonly used reference regions, with and without partial volume correction. Cutpoints for amyloid positivity were created by fitting Gaussian mixture models (with 1 to 3 clusters) and using the 99th percentile of the Gaussian representing the amyloid negative population.


      240 participants with suitable T1 and amyloid PET data were included in the analysis. The mixture modelling for cortical ROIs typically resulted in 2 clusters, confirming the expected bimodal distribution of amyloid deposition. Across the different reference regions, the rate of amyloid positive individuals was consistently 15 – 18% without PVC correction and 19-23% with PVC correction (Table). Precuneus and posterior cingulate SUVRs classified slightly more participants as amyloid positive, while those based on occipital lobe were much lower (Figure). Subcortical ROIs provided inconsistent evidence of bimodal distributions.


      Quantification of SUVR based measures of amyloid load using data acquired on PET/MR produces consistent rates of amyloid positivity across a variety of analysis options.
      Figure thumbnail fx1
      TableMean cortical GM SUVR results, showing the mean and standard deviations of the two groups based on the defined cutpoints (99% of amyloid negative cluster). UTE=Ultrashort echo time method, the vendor-supplied method for attenuation correction. pCT=pseudo CT, the in-house method for attenuation correction that builds a synthetic CT image. PVC=Partial Volume Correction, performed using the Iterative Yang approach with a full-width half-maximum kernel of 6.8 mm and 10 iterations (Erlandsson, Phys. Med. Biol., 57(21): R119-159, 2012)
      Reference ROIMethodAmyloid Negative Mean (SD)Amyloid Positive Mean (SD)Cutpoint% Amyloid Positive
      Whole CerebellumpCT, no PVC0.959 (0.053)1.218 (0.100)1.08017.5%
      UTE, no PVC0.996 (0.054)1.255 (0.106)1.12018.5%
      pCT, no PVC0.759 (0.059)1.129 (0.183)0.89923.3%
      UTE, with PVC0.784 (0.063)1.153 (0.190)0.91724.5%
      Cerebellar Grey MatterpCT, no PVC1.103 (0.077)1.424 (0.113)1.27515.4%
      UTE, no PVC1.156 (0.077)1.485 (0.113)1.32614.7%
      pCT, with PVC1.019 (0.109)1.581 (0.267)1.25320.4%
      UTE, with PVC1.053 (0.106)1.574 (0.269)1.26024.1%
      PonspCT, no PVC0.575 (0.040)0.754 (0.066)0.66415.8%
      UTE, no PVC0.557 (0.037)0.729 (0.058)0.64117.6%
      pCT, with PVC0.462 (0.049)0.729 (0.118)0.57418.8%
      UTE, with PVC0.446 (0.046)0.704 (0.105)0.55720.1%
      Subcortical White MatterpCT, no PVC0.541 (0.036)0.688 (0.053)0.62117.5%
      UTE, no PVC0.535 (0.035)0.677 (0.049)0.61116.8%
      pCT, with PVC0.444 (0.055)0.705 (0.105)0.56517.8%
      UTE, with PVC0.443 (0.050)0.739 (0.150)0.55819.5%