Our SPITFIRE™ approach was applied to each of the n=311 human amyloid brain PET images and n=-15 Hoffman phantom images from a cohort of subjects participating in a Phase 2, multi-center, Alzheimer's disease clinical trial sponsored by Cognito Therapeutics.

For presentation purposes, the images were arranged according to their corresponding site ID, manufacturer, scanner model, and reconstruction parameters. The mean and standard deviation values of both axial and in-plane FWHMs for human brain PET data and phantoms are shown in the figure.

In most cases, the FWHM estimation of the human brain data is less than the voxel size of the matching phantom FWHM estimation.

Exceptions are the C and E site IDs where the mean in-plane resolution estimation is a bit smaller than one voxel size of the mean in-plane estimation of the matching phantoms. However, even in these cases, the corresponding confidence interval (i.e. accounting for the standard deviation as well) overlap, meaning that those mean values could be considered statistically close each other.