Truncation artifacts in whole-body PET/MR - Compensation and evaluation

Objectives In whole-body PET/MR examinations the reduced transaxial FOV of the MR (40cm), in comparison to that of the PET scanner (60cm), leads to truncation artifacts (TAs) in the MR-based attenuation map (MRMap). We investigated the quantitative influence of TAs on the reconstructed PET image volumes and developed a new method for truncation compensation (TC).

Methods 20 patient investigations with arms down (FDG, BMI 16-38) from a Philips Ingenuity PET/MR were reconstructed using different techniques for TC: (a) no TC (PET-noTC), (b) vendor-provided TC (PET-TC1), (c) newly developed TC (PET-TC2) based on automatic 3D segmentation of PET-noTC. SUVmax/mean deviations of tracer accumulations at different locations (e.g. heart, bladder, kidneys, arms etc.) were analysed. Moreover, a correlation analysis between all voxels in the trunk and truncated arms was performed.

Results Residual artifacts in the truncation compensated MRMaps were found in 11 (TC1) and 3 (TC2) out of 20 cases. Average SUVmax differences between PET-noTC and PET-TC2 were: -3.3% (bladder), -3.6% (kidney), -3.3% (heart), -3.4% (abdomen), -38.9% (arms). The correlation analysis showed average SUV underestimations of -10% to -1% and -68% to -12% depending on patients' BMI for all voxels in the trunk and truncated arms, respectively.

Conclusions In whole-body PET/MR imaging with arms down, TAs in the MRMap lead to an average SUV underestimation inside the trunk of ca. -3% . In the truncated areas and their immediate vicinity substantial SUV underestimation (-10 to -70%) can be expected. Compensation of TAs based on PET-noTC is straightforward, improves quantification accuracy distinctly and has turned out to be well suited for clinical routine.