Article révisé par les pairs
Résumé : In this work we use the NEC transformation method of Nuyts et al. in the OSEM reconstruction technique as a procedure to partially restore the Poisson characteristics of 3D whole-body PET pre-corrected data. We compare the use of FORE+NEC-OSEM with a fully 3D implementation of NEC-OSEM, in terms of lesion detectability, using a 3D version of the non-prewhitening matched filter numerical observer. In FORE+NEC-OSEM, only a multiplicative weight was applied (NEC-scaling) that accounts for the effects of both the quantitative corrections and FORE itself. For the fully 3D reconstruction, the generalized NEC-transformation was used (NECG-OSEM3D), with both multiplicative (NEC-scaling) and additive (NEC-shifting) factors to account for the effects of the corrections. In addition to NEC-OSEM, FORE+AW-OSEM (attenuation weighted) and ANW-OSEM3D (attenuation and normalization weighted) were also tested. We used simulated data of 3D whole-body PET exams with 1 cm diameter inserted lesions to perform the study. From the results we conclude: for the existence of an optimum post-smoothing filter with FWHM of 5.5 mm; for an overall better performance of both ANW-OSEM3D and FORE+NEC-OSEM over FORE+AW-OSEM. The comparison between NECG-OSEM3D and the other algorithms depends very much on the number of iterations. The NPWMF numerical observer seems not to be the best figure-of-merit to determine the optimum number of iterations the NPWMF.