Evaluation of Attenuation Correction Using Dixon based Bone Model for 18F FDG Brain PET/MR: A Volunteer Study

The present study aims to evaluate PET quantification for the fluoro deoxy glucose (FDG) brain PET images using two MR attenuation maps by comparing them with images using CT attenuation map in a healthy volunteer. The two MR attenuation maps include four segments of the Dixon sequence (soft tissue, fat, lung, and air), and five segmentations of the Dixon based bone model (soft tissue, fat, lung, air, and bone). Quantifications of the standardized uptake values (SlUVs) for 133 regions of FDG brain PET images using voxel wise analyses revealed significant differences between CT and Dixon AC (mean difference = 0.93 0.25). However, applying the Dixon based bone model AC reduced the estimation error compared with CTAC (mean difference = 0.66 0.21). An attenuation map including bone information using the Dixon based bone model can reduce underestimation of SUV compared to Dixon from brain PET/MR imaging. In particular, errors were decreased more in the near skull regions than in central regions.