TY - JOUR T1 - Development and Evaluation of Penalized Image Reconstruction for the Total-Body EXPLORER JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 1773 LP - 1773 VL - 59 IS - supplement 1 AU - Xuezhu Zhang AU - Ramsey Badawi AU - Simon Cherry AU - Jinyi Qi Y1 - 2018/05/01 UR - http://jnm.snmjournals.org/content/59/supplement_1/1773.abstract N2 - 1773Objectives: The EXPLORER consortium is building a 194-cm axial field of view total-body PET scanner to create new opportunities for biomedical research and clinical applications [1]. This first EXPLORER prototype is currently under construction in collaboration with United-Imaging Healthcare. The scanner consists of 564,480 crystals and has ~100 billion lines-of-response (LORs), which presents a daunting challenge for data corrections and image reconstruction. In this work, we develop the penalized likelihood (PL) image reconstruction for the total-body scanner and evaluate the performance for image quantification and lesion detection. Methods: We conducted simulation studies using the SimSET Monte-Carlo toolkit for the EXPLORER scanner. The scanner consists of 8 axial rings, each formed by 24 modules of 5x14 detector blocks with an array of 6x7 LYSO scintillator crystals of size of 2.76 x 2.76 x 18.1 mm3 [2]. The detector modules have a measured energy resolution of 11.7 ± 1.5% and timing resolution of 409 ± 39 ps. We employed the XCAT2.0 phantom and modeled a clinical 18F-FDG scan with an injected activity of 25 MBq and 20-minute scan duration (1-hour post injection). A total of 1 billion true coincidences were detected. Randoms and scatters were not included in this study. The data were reconstructed using a 3D TOF list-mode OS-EM algorithm and PL algorithm with a Fair penalty. For fast reconstruction, we used the Siddon single-ray tracing projector in combination with a shift-variant image-based point spread function (PSF) model. Images were reconstructed into a 475x475x1355 matrix with 1.425-mm cubic voxels to cover a 68 cm transaxial and 194 cm axial FOV. The OSEM reconstruction was initialized with a uniform image and the PL reconstruction was initialized with one iteration OSEM reconstruction. Both algorithms used 20 subsets. We compared the image quality by evaluating the contrast recovery coefficient (CRC) of a 10-mm spherical lesion in the liver vs. the background noise in the liver. The CRC and background noise curves were generated by varying the iteration number of the OSEM algorithm and by varying the regularization parameter β value of the PL method. Results: Images reconstructed by both OSEM and PL algorithms show good image quality with low noise, thanks to the extremely high sensitivity of the EXPLORER scanner. The image quality was further improved by using the PL method. At a CRC level between 0.6 and 0.8, the PL method reduced the background standard deviation by nearly a factor of 2. When the regularization strength is reduced to the minimum, the PL method approaches the same performance of the OSEM algorithm as expected. Conclusion: This study demonstrated that the EXPLORER with penalized likelihood reconstruction can achieve superior noise performance compared to OSEM for total-body imaging. Research Support: Support for this work includes NIH grants R01 CA206187 and R01 CA170874. We acknowledge Drs. Hongdi Li and Weiping Liu from United Imaging Healthcare for sharing the configuration of the UIH EXPLORER scanner. ER -