TY - JOUR T1 - <strong>Image Quality evaluation of analytic reconstruction algorithms for long axial field-of-view PET scanner</strong> JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 240 LP - 240 VL - 60 IS - supplement 1 AU - Yong Zhao AU - Zilin Deng AU - Yang Lv AU - Debin Hu AU - Yu Ding AU - Yun Dong Y1 - 2019/05/01 UR - http://jnm.snmjournals.org/content/60/supplement_1/240.abstract N2 - 240Objectives: Analytic reconstruction algorithm such as Filtered backprojection (FBP) was computationally efficient and considered to be quantitatively accurate for PET imaging. However, the images reconstructed by FBP are very noisy in conventional clinical PET scanner because of its high demand of counts. UEXPLORER, the world first total body PET/CT scanner, is recently manufactured by United Imaging Healthcare. It is able to provide ultra-high sensitivity due to the two-meter long axial FOV. The objective of this study was to investigate the image performance of FORE+FBP when the maximum accept angle changes, and compare it with other reconstruction methods as FOREX+FBP, 3DRP, and 3DOSEM. Methods: UEXPLORER was designed with 8 units, where each unit has axial length of 242mmand transverse diameters of 786mm. The maximum unit difference was 4, corresponding to an accept angle of 56°. It is the largest accept angle has ever achieved for a clinical PET scanner. Because the accuracy of FORE is degraded at large angles, in this study, we changed the accept angle from 17° to 56° to find the optimal accept angle for trade-off between the contrast recovery coefficient and noise level. For comparison, other reconstruction methods like FOREX+FBP, 3DRP and 3D-OSEM were also used with varied accept angles. For analytic algorithms, a rectangular apodizing window was used and the cutoff frequency was set to be 1. For OSEM algorithm, the TOF and PSF techniques were not used. The subset number and the iteration number were set to be 20 and 3 respectively. Two phantoms were scanned on real system for validation: 1) NEMA image quality phantom with four hot spheres and two cold spheres, and a NEMA scatter phantom was put close to the image quality phantom; 2) the 3D Hoffman brain phantom; The comparison metrics include contrast recovery, background variability and the overall image quality. Result: Fig.1 shows the contrast recovery and noise level of six spheres in NEMA image quality phantom for FORE+FBP. We found that the hot sphere contrast deteriorates as the accept angle increasing, and all spheres noise level reduces at the beginning and then increases with the accept angle increasing. Based on this tradeoff between the contrast and noise, we find an optimal accept angle of FORE algorithm for uEXPLORER. Based on this optimal accept angle, we also illustrate the recovery coefficient and noise level for FORE+FBP and 3DOSEM in table 1. And the center slices of NEMA image quality and Hoffman brain phantom are shown in fig 2. Visually comparison of the images did not show significant difference between the two algorithms. Conclusion: Although the FORE+2D FBP could get visually comparable images as 3D-OSEM due to the increased data statistic, this method is not the best choice for uEXPLORER. The degraded accuracy of FORE at large angles makes it cannot fully use all oblique LORs. FOREX, an exact rebinning method, is superior to FORE and 3DRP for its better image quality and less time consumption. In the future, timing of flight information will be used in the analytic reconstruction algorithms to achieve better image quality. ER -