TY - JOUR T1 - Patlak parametric images are less sensitive to attenuation bias than SUV images JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 1754 LP - 1754 VL - 59 IS - supplement 1 AU - Baixuan Xu AU - Shulin Yao AU - YouJun Sun AU - Yun Dong AU - Jun Bao AU - Hongdi Li AU - Wentao Zhu Y1 - 2018/05/01 UR - http://jnm.snmjournals.org/content/59/supplement_1/1754.abstract N2 - 1754Objectives: Low-dose brain PET/CT imaging is not optimal with associated CT radiation. However, CT is usually performed anyway because attenuation correction is critical to the quantitative accuracy of SUV. Dynamic PET imaging, on the other hand, may be less sensitive to AC"s inaccuracy. The purpose of this study is to compare the error of Patlak and SUV images due to AC"s bias, and to demonstrate that Patlak is a more robust metric in quantitation. Methods: Ten patients were included in the study. The dynamic PET protocol included 80min PET scan post 18F-FDG injection, and was coupled with a CT scan serving for attenuation correction. SUV images corresponding to 40, 60, 80 min post injection were computed. Voxelwise Patlak images were also computed, with image-based blood input function computed from the reconstructed frames. For comparison, the known attenuation maps were replaced with coarse approximations by filling uniform water-AC within the head boundary. The biased SUV and Patlak images could then be computed again with these erroneous umaps. Bias of ROIs within the brain was computed with the paired SUV and Patlak images. Statistical analysis was performed to evaluate the robustness of each imaging metric against the error in attenuation maps. Results: > the mean relative bias of SUV images at 40min for ROIs within white matter, striatum, and cerebellum were 4.88%, 3.76%, and 5.68% respectively. For SUV at 60 and 80min, the mean relative bias became 4.93%, 3.75%, 5.78% and 4.80%, 3.70%, 5.72% respectively. On the other hand, Patlak parametric images were less sensitive than SUV images to the error in attenuation maps. The mean relative bias of Patlak slope images for same ROIs was 1,60%, 2.55%, and 2.56% respectively, much less than that of SUV images. The image contrast between striatum and cerebellum changed 5.6% for SUV at 60min. However, only a 3.3% change was observed for Patlak slope image, again much less. Conclusions: the results indicated that Patlak images were more robust than SUV images against the error of attenuation maps. For low dose applications, PET/MR studies, or other studies when the attenuation map is subject to error, it is still possible to obtain adequately accurate Patlak images while SUV"s quantitation is impaired. ER -