Small-voxel reconstructions significantly influence SUVs in PET imaging

Abstract

Dear Sir, We read with great interest the recent article by FuentesOcampo et al. entitled BDigital vs. analog PET/CT: intrasubject comparison of the SUVmax in target lesions and reference regions^ [1]. The authors aimed to assess whether digital photon counting technology in digital PET provides better quantification of SUVmax in lesions and reference regions than analogue PET. They conclude that both improved photon counting technology in digital PET and the effects of delayed increased uptake and retention significantly increase SUVmax, and that this has to be taken into account before the systems can be used interchangeably in follow-up studies. In our opinion, there is an important limitation in their PET image reconstruction method that is not mentioned in their paper. Fuentes-Ocampo et al. compared digital PET using 2mm voxel reconstructions with conventional PETusing 4-mm voxel reconstructions. This difference in reconstruction approach invalidates their comparison, because the introduction of 2-mm voxel reconstructions on a conventional PET system already significantly increases SUVs [2, 3]. FuentesOcampo et al. did not even cite these published studies. For example, in a conventional FDG PET study in 39 patients, we found an increase in SUVmax of 32% across 66 lesions when using small 2-mm voxels instead of 4-mm voxels, and the value increased to as high as 44% for lesions smaller than 11 mm diameter [2]. In comparison, Fuentes-Ocampo et al. reported a mean percentage difference of 35% across 87 lesions. This puts the conclusion of their paper on digital PET in a different light. Perhaps it was not possible to perform 2-mm voxel reconstructions on the conventional PET system used by Fuentes-Ocampo et al. However, at least it should have been mentioned in the Discussion as a main study limitation, since it could have had a larger effect than the effects of two other limitations that were mentioned: the number of iterations and subsets, and the use of point-spread function modelling. We emphasize that the image voxel size has a large effect on SUVmax, even on conventional non-digital scanners, and should therefore not be ignored in PET comparison studies. The use of both photon counting technology and different reconstruction settings, including voxel size, should be taken into account before different systems can be used interchangeably in follow-up studies.