Whereas Monte Carlo (MC) simulation is widely utilized in estimation of the scatter component, a simulation model which can calculate the scatter fraction (SF) of each patient is needed for making an accurate image quality assessment for clinical PET images based on the noise equivalent count. In this study, an MC simulation model was constructed which can calculate the SF for various phantoms. We utilized the Geant4 toolkit based on MC simulation to make a model of a PET scanner with a scatter phantom, and SFs calculated with this model were compared with the SF (SFconstant: 44%) measured with use of an actual PET scanner. Additionally, the SF values for an anthropomorphic phantom were calculated from its voxel phantom. Furthermore, we evaluated the impact on the SF due to the difference in the source distribution inside the phantom. The SF calculated from the scatter phantom in the MC simulation was 44%, the same as the SFconstant value. The average SF for the anthropomorphic phantom was 41%, but there was a maximum of 14 percentage points difference between each scan range, and the maximum difference in the SF was 8 percentage points for the difference in the source distribution. We constructed an MC simulation model which can calculate SFs for various phantoms. The SF was confirmed to be affected significantly by the source distribution. We judged that the actually measured SFconstant obtained from the PET scanner with the scatter phantom was not suitable for the assessment of the quality of all patient images.
Keywords: 18F-FDG PET/CT; Geant4; Monte Carlo simulation; Noise equivalent count rate; Scatter fraction; Voxel phantom.