Amyloid pathology is believed to play an initiating role in Alzheimer’s disease (AD). PET imaging can reveal the amount and localization of these abnormal proteins in the brain and three different tracers have been approved for clinical use, but their uptake cannot be directly compared. To overcome this limitation a standardization method has been developed and named ‘Centiloid’ (see https://doi.org/10.1016%2Fj.jalz.2014.07.003). This method is based on performing a linear mapping of the uptake values of each tracer so, in analogy with the Centigrade scale, a value of 0 (zero) corresponds to young healthy controls and 100 (one hundred) to the amyloid level in a reference group of typical AD patients.
Centiloid values are being currently used to describe the natural evolution of AD and as an outcome to test novel drugs to remove amyloid from the brain. To these goals, a better understanding of the uncertainty associated to Centiloid values.
To this end, we have collected amyloid PET scans with the 3 approved tracers from more than 500 individuals in the European project Amypad (https://amypad.eu/) and have designed 32 different quantification pipelines, all rendering Centiloid values. We want to understand how uncertainty in the amyloid PET scans are propagated through the various transformations in order to quantify the uncertainty across the full range of variation depending on the tracer, quantification pipeline, and other relevant factors. The project involves generating virtual samples of Centiloid values and bootstrapping them to render 95% confidence intervals for all combinations of factors.

This project is in collaboration with Fundació Pasqual Maragall.