TY - JOUR T1 - Impact of attenuation map generation schemes in lesion quantification in prostate cancer patients with PET/MRI JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 651 LP - 651 VL - 59 IS - supplement 1 AU - Borjana Bogdanovic AU - Jorge Cabello AU - Sylvia Schachoff AU - Matthias Eiber AU - Stephan Nekolla Y1 - 2018/05/01 UR - http://jnm.snmjournals.org/content/59/supplement_1/651.abstract N2 - 651Objectives: The whole-body MR-PET system Siemens Biograph mMR is a well established imaging system for clinical diagnostics as well as research studies. Since its introduction in 2010, several implementations to generate PET attenuation correction were provided by the vendor. The latest variant was a five-compartment model including major bones and extending the MR's field-of-view with HUGE (B0 homogenization using gradient enhancement). Starting with attenuation correction schemes which suffered from truncations in the arms and solely based on a two-point Dixon-VIBE sequence (including only air, fat, water and soft tissue), incremental improvements became available. In this study, we compare the impact of four attenuation correction methods, namely Dixon-VIBE, HUGE with bones included, HUGE without bones, and MLAA, on lesion quantification in prostate cancer patients using F18 labelled PSMA. Methods: Ten prostate cancer patients injected with 340 ± 39 MBq of [18F]-PSMA were scanned for 25-30 minutes (5 min per bed position) in the PET/MR scanner 90 min p.i., Images were reconstructed with the four aforementioned attenuation correction methods using 3 iterations with 21 subsets, filtered with a 4 mm FWHM Gaussian kernel. Visual analysis was performed to assess potential differences in the investigated methods. For a more quantitative assessment, we analyzed SUVmean and SUVmax for each of a total of 14 lesions located in the pelvic region (whereby no lesions were located in the bone), as well as the average SUV in the background using the PERCIST reference VOIs using commercial software (Syngo TrueD). In absence of absolute ground truth data, uptake values were compared using linear regression against the plain DIXON methodology. Finally, we performed the Wilcoxon rank-sum tests to check for significance of the differences between the four reconstruction methods. Results: No visually perceivable differences were noted. The Wilcoxon rank-sum test showed no statistical significance for mean lesion SUV (DIXON 6.9±2.6, HUGE with bones 6.7±2.6, HUGE without bones 6.9±2.6, MLAA: 7.1±2.6 ), maximal lesion SUV (DIXON 9.7±3.8, HUGE with bones 9.9±3.8, HUGE without bones 9.7±3.8, MLAA: 10.1±3.8 ), and background. With Dixon-VIBE taken as reference, the average individual bias in mean SUVs measured with each method was 2.95% (with [MIN, MAX] = [0, 5] %), 1.55% ([-2, 4] %), and 6.63% ([1, 12] %) for HUGE with bones, HUGE without bones, and MLAA, respectively. In the same way, and same order, the average bias in maximum SUVs was 2.5% (with [MIN, MAX] = [0, 8] %), 1.14% ([-2, 5] %), and 5.92% [0, 15] %). The calculated R2 correlation coefficients for both mean SUVs and max SUVs between each of the three methods and Dixon-VIBE were all in the range 0.994—0.999. Even for the background VOIs R2 values in the range 0.607—0.846 with average SUV values of 0.5±0.1 [MIN: 0.27, MAX: 0.83] were found. Conclusions: Based on these results, in the case of lesions found in the soft tissue region in the pelvic area as well as background values, there was no significant difference found between the Dixon-VIBE and each of the three sequentially improved attenuation correction methods under evaluation. We therefore conclude that including bones and compensating for truncations in the arms did not significantly change prostate lesion tracer uptake quantification. This finding is quite reassuring especially for serial studies: MR based whole-body attenuation correction is a stable method independent of the actual level of sophistication. ER -