The effect of the scatter correction obtained using single scatter simulations with CT- and MR-based attenuation maps for F-18-FDG brain PET

Abstract

Attenuation correction (AC) and scatter correction (SC) are essential steps for accurate PET quantification in positron emission tomography (PET)/computed tomography (CT) and PET/magnetic resonance (MR) imaging. SC with a single scatter simulation (SSS) algorithm is usually performed by employing an attenuation map using CT and MR data for AC. The purpose of this study was to evaluate the effect of SC using a SSS algorithm with CT- and MR-based attenuation maps in phantom and volunteer studies for F-18-FDG brain PET. We investigated the effect of the SC on two MR-based attenuation maps, which included a four-segmentation Dixon (soft tissue, fat, lung, and air) and a five-segmentation Model, which is Dixon plus bone, and we compared those maps with a standard CT-based attenuation map. In the phantom study, the difference in (%) recovery coefficients before and after SC for a CT-based attenuation correction (CTAC) and a MR-based attenuation correction (MRAC) were 20.36 and 21.01, respectively. The difference in the scatter fractions on the sinogram was about 2% between the CTAC and the MRAC (Dixon). In the volunteer study, the scatter fractions were 24.08 for the CT, 23.09 for the Dixon, and 23.11 for the Model. The bias in the scatter ratio image, after calculating the before and after SCs from the PET images, between the CTAC and the two MRACs was less than 1% (% MAE = 0.43 for CT-Dixon and 0.39 for CT-Model). A voxel-wise analysis of the scatter ratio before and after the SC from PET images showed no significant differences between the CTAC and the two MRACs (p > 0.05). We found that despite the different attenuation maps, the difference in the effects of the SC between the two MR and the CT attenuation maps was minimal. This implies that the SC obtained using the SSS algorithm was rarely affected by the attenuation map.