CenTauR_z: A standardized quantification of tau PET scans

Background: Over the past decade, several PET tracers were developed to visualise and quantify tau pathology in vivo. However, all these tracers have distinct off-target binding, different dynamic ranges and likely different levels of non-specific binding resulting in large variability in semiquantification. We propose to standardise the sampling and the quantification across all available tau tracers. Method: 549 participants underwent tau scans with either 18F-FTP (Cognitively Unimpaired (CU)=54/AD=14), 18F-MK6240 (CU=186/AD=89), 18F-PI2620 (CU=17/AD=21), 18F-PM-PBB3 (CU=30/AD=28), 18F-GTP1 (CU=7/AD=38) or 18F-RO948 (CU=35/AD=30). All CU individuals were Aβ- and all AD were Aβ+. The tau scans were spatially normalized using CapAIBL and the cerebellar cortex was used as reference region. We constructed a “universal” tau mask from the intersection of all the specific tau tracer masks, after subtracting AD from CU. All tau PET studies were sampled with a Mesial Temporal (MTL) and a Meta Temporal (MetaT) composites constrained by the universal mask. For each tracer and in composite, the mean and standard deviation of the Aβ- CU SUVR for each tau tracer were used to generate z-scores (CenTauR_z). Result: Using a threshold of 2 CenTauR_z in the MetaT regions, all tracers highly discriminated Aβ+ AD from Aβ- CU (ACC=[0.94-1], sens=[0.84-1], spec=[0.96-1]) with mean CenTauR_z for the different AD cohorts ranging from 8 to 14. Lower accuracy was observed in the MTL (ACC=[0.78-1]) due to lower sensitivity in some cohorts [0.65-1] however, the specificity was similar to that in the MetaT composite (spec=[0.94,1]). Conclusion: All tracers exhibited comparably high discriminative power to separate Aβ+ AD from Aβ- CU, where AD Aβ+ displayed a consistent range of CenTauR_z across tracers. However, there were some differences between cohorts. For example, different PET scanners, with different sensitivities were used. For some cohorts, scans were selected as extreme representative cases, while for others the scans were more representative of clinical settings, with AD patients at early stages (with low or negative tau scans) or with suspected hippocampal sparing subtype that likely explains the lower accuracy in the MTL for some cohorts. Further studies with larger cohorts to validate the universal mask and CenTauR_z scale are ongoing.