Component-based normalization for the HRRT for sinogram-mode reconstruction

THE Siemens HRRT scanner is a complex PET tomograph, using 8 LSO/LYSO panel detectors and Depth Of Interaction(DOI) with a total of 119908 detector crystals. The Direct Normalization (DN) is the standard method used for sinogram-mode reconstruction. This paper investigates the use of an improved component-based normalization method for HRRT sinogram reconstruction. This method is originally used for MOLAR LOR listmode reconstruction.Three effects have been identified in the implementation of the original method that caused artifacts and quantification bias in the reconstructed uniformity image. The new CBN method has been implemented that corrects for these effects and produces a uniformity image with the same quality as the DN method. The main effect was the sensitivity drop for LORs with small angles of incidence, which affects all scanners by design.