Abstract
PURPOSE: After a decade of PET/MR, the case of attenuation correction (AC) remains open. The initial four-compartment (air, water, fat, soft tissue) Dixon-based AC scheme has since been expanded with several features, the latest being MR field-of-view extension and a bone atlas. As this potentially changes quantification, we evaluated the impact of these features in PET AC in prostate cancer patients. METHODS: Two hundred prostate cancer patients were examined with either (18)F- or (68)Ga-prostate-specific membrane antigen (PSMA) PET/MR. Qualitative and quantitative analysis (SUV(mean), SUV(max), correlation, and statistical significance) was performed on images reconstructed using different AC schemes: Dixon, Dixon+MLAA, Dixon+HUGE, and Dixon+HUGE+bones for (18)F-PSMA data; Dixon and Dixon+bones for (68)Ga-PSMA data. Uptakes were compared using linear regression against standard Dixon. RESULTS: High correlation and no visually perceivable differences between all evaluated methods (r > 0.996) were found. The mean relative difference in lesion uptake of (18)F-PSMA and (68)Ga-PSMA remained, respectively, within 4% and 3% in soft tissue, and within 10% and 9% in bones for all evaluated methods. Bone registration errors were detected, causing mean uptake change of 5% in affected lesions. CONCLUSIONS: Based on these results and the encountered bone atlas registration inaccuracy, we deduce that including bones and extending the MR field-of-view did not introduce clinically significant differences in PSMA diagnostic accuracy and tracer uptake quantification in prostate cancer pelvic lesions, facilitating the analysis of serial studies respectively. However, in the absence of ground truth data, we advise against atlas-based methods when comparing serial scans for bone lesions.