Attenuation correction in PET/MR: A performance assessment of UTE-based and Dixon-based methods for PET attenuation correction

Hybrid positron emission tomography/magnetic resonance imaging (PET/MR) scanners are one of the latest tools available in the field of medical imaging, and are expected to outperform the well-established PET/X-ray computed tomography (CT) scanners in a large range of fields. The perhaps largest challenge that has to be overcome before this can be achieved, is that of attenuation correction (AC) of the acquired PET images, as there is no direct relation between the MR image intensity of a tissue and its attenuating properties, as is the case in CT.This study investigated the performance of two PET AC methods provided with the biograph mMR PET/MR scanner installed at St. Olavs Hospital (Trondheim, Norway); one for head imaging based on an ultra-short echo-time (UTE) sequence, and one for whole-body imaging based on a Dixon sequence. These AC methods were compared to the `gold standard' of CT-based AC, based on activity concentrations in PET images from mMR examinations of lymphoma and lung cancer patients, corrected with the different AC methods (UTE, Dixon and CT).The results of the study show that the UTE-based AC method leads to an underestimation of PET activity in the brain of up to 9 \% in the investigated regions of interest. This is caused by underestimation of bone in the cranial region. The exclusion of bone in the Dixon-based AC method leads to underestimation of PET activity in the thorax/abdomen, indicated by an underestimation of 4 \% in the liver. The two MR-based AC methods are thus not sufficiently accurate to be utilised for quantification in PET imaging