Overview of the Hoffman phantom-based approach for estimating spatial resolution in PET images. This approach involves the matching of the physical Hoffman phantom to a certain element of library images constructed from smoothed versions of a digital phantom image.
The aim of the standard approach to estimate resolution in PET images is to bring Hoffman phantom images from different scanner models to a common spatial resolution.
As shown in the diagram, an incremental Full Width at Half Maximum (or FWHM) Gaussian kernel library of smoothed versions of a digital phantom image needs to be created. Then, the resolution of each physical Hoffman phantom is estimated by matching it to each element in that library using the Euclidean distance as a metric of proximity between images.
However, Hoffman phantom images may not always be readily available for all PET imaging sites involved in multi-center clinical trials. In addition, one would need constant monitoring of scanner updates or replacements, including changes to the scanning protocols and image reconstruction parameters.
Finally, the basic underlying assumption that actual human brain PET images inherit the same spatial resolution of Hoffman phantom images scanned on identical scanner models can be difficult to justify in most cases.