CenTauR Project


 

Tau positron emission tomography (PET) imaging is the most recent addition to the arsenal of tools for the in vivo assessment of neurodegenerative proteinopathies.

Differences between tracers, quantitative pipelines and regions of interest lead to disparities in PET-derived standardized uptake value ratios (SUVR) measurements which decrease reproducibility and pose a challenge when trying to compare tau PET outcomes across cohorts or in therapeutic trials that use different tau tracers.

The aim of the CenTauR project is to standardize tau PET results by expressing them in a universal standard scale, the units of which are termed CenTauRs. This project includes both the CenTauRZ approach, which summarizes the location and amount of tau tangles in the brain (Villemagne et al., Alzheimer's Dement 2023 ), and the Joint Propagation Model (JPM), a mixed-effects nonlinear model that provides a set of mapping equations converting SUVR values to CenTauRs (Leuzy et al., Alzheimer's Dement 2024 ).

The distinction between CenTauRZ and the JPM lies in the anchor points they use. CenTauRZ is anchored solely on cognitively unimpaired (CU) individuals, whereas CenTauRs from the JPM, like the Centiloid, are anchored on both CU individuals and younger (age < 75) mild Alzheimer's disease (AD) cases.

In simple terms, CenTauRZ measures tau abnormality, while JPM-based CenTauRs reflect the amount of tau. Both approaches can be implemented at the voxel level or using regions of interest (ROIs). For the ROI-based approach, users can either use the five CenTauR masks (universal, mesial temporal, meta temporal, temporo-parietal, and frontal) or by applying their own mask(s) to the parametric CenTauR image.

DATA SETS

CenTauRZ

The CenTauR ROIs developed as part of the CenTauRZ approach were intended to be used with SPM8. When the approach was published, there was no public calibration datasets or software were available. Subsequently, the Aging and Dementia Imaging Research (ADIR) Lab at Mayo Clinic normalized the SPM8-based ROIs to higher resolution MNI152-nonlinear and Mayo Clinic Adult Lifespan Template (MCALT) voxel spaces using SPM12. They then implemented a modernized pipeline using SPM12 that measures SUVR in MRI space with tissue-class masking and population-matched MCALT templates and priors (Schwarz et al. Human Amyloid Imaging Conference, 2024). The dataset, ROIs and software are available below.


Calibration Dataset

This dataset contains T1-weighted MRI and Flortaucipir tau PET images from 100 participants in the Mayo Clinic Study of Aging and Mayo Clinic Alzheimer’s Disease Research Center. These participants came from two groups: a) 50 (25M+25F) adults age 65+ with a clinical diagnosis of probable Alzheimer’s disease with an amnestic phenotype, a positive PiB amyloid PET scan, and no additional neurologic diagnoses and b) 50 (25M+25F) cognitively unimpaired older adults age- and sex-matched to group A, with a negative amyloid PET scan. 65 of these individuals have included an additional follow-up scan to enable analyses of longitudinal stability. The intended use of this dataset is for a user to measure these images using both their own pipeline and the CenTauR standard (aka Centiloid standard) SPM8 pipeline, and perform linear regression to calibrate their pipeline’s units to the standard pipeline, and to CenTauR units.


CenTauR ROIs

The five CenTauR ROIs (universal, mesial temporal, meta temporal, temporo-parietal, and frontal) can be downloaded below. These can be used whether using the CenTauRZ approach or the JPM (see below).

Software

These files, provided by the ADIR Lab at Mayo Clinic, include the CenTauR standardized VOIs transformed to the space of several popular neuroimaging templates, and automated software that calculates SUVR and CenTauRZ units using these VOIs for input pairs of summed tau PET images and T1-weighted MRI. Two versions are included: one implements the SPM8 “Centiloid standard” pipeline of Klunk et al. 2015. The other is a SPM12-based pipeline that measures SUVR and CenTauRZ in MRI space with tissue-class masking and population-matched MCALT templates and priors. Both are pre-calibrated to output CenTauRZ units for all supported tau PET tracers.


Joint Propagation Model

The steps describing how to convert SUVR values to CenTauRs using the JPM are described in Leuzy et al., Alzheimer's Dement 2024b . In brief, once a user has validated their specific image processing pipeline using a dataset specific to their tau PET tracer, the user is free to implement the JPM conversion equations, either at the voxel (i.e., convert tau PET images to parametric CenTauR images and sample with a mask of their choice) or ROI-level (i.e., using one or several of the CenTauR ROIs and region specific conversion equations).

Tau PET tracer specific datasets for pipeline validation are available for download below. Each folder contains 50 tau PET images with corresponding T1-weighted MRIs in native space, with the exception of flortaucipir, which contains the 100 cases from the above described CenTauRZ calibration dataset.