Harmonizing standardized uptake value recovery between two PET/CT systems from different manufacturers when using resolution modelling and time-of-flight

PET iterative reconstruction algorithms with resolution modelling (RM) can be used to improve spatial resolution in the images. However, RM has a significant impact on quantification, which raises issues for harmonization across multicentre networks or collaborations. This investigation compared quantification from two modern time-of-flight (TOF) PET/CT systems from different manufacturers with RM with the intention to harmonize recovery. Images of a National Electrical Manufacturers Association image quality phantom with a sphere-to-background concentration ratio of 4: 1 were acquired on a GE Discovery 710 and a Siemens Biograph mCT and reconstructed with RM and TOF. Voxel dimensions and image noise (background coefficient of variation) were matched. One to five iterations were used with 2 and 4 mm Gaussian filters. Mean and maximum contrast recovery (CR) were measured for the 10, 13, 17 and 22 mm hot phantom spheres. Notable differences in CR for images reconstructed with matched reconstruction parameters were observed between the scanners. A set of parameters was found that reduced differences in CR between scanners. Using these parameters, relative differences for the Biograph compared with the Discovery were -8.1, -3.7, +7 and +0.7% for mean CR and -23.1, -6.9, +9.1 and +0.9% for maximum CR in the 10, 13, 17 and 22 mm spheres, respectively. This study has used a technique of harmonizing standardized uptake value recovery on PET/CT systems from different vendors with advanced reconstructions including TOF and RM using phantom data. Considerable quantitative differences may occur in images, which highlights the need to apply methods such as those used in this work for multicentre studies.