MR-based PET Attenuation Correction – Method and Validation

In combined PET/MR, the high soft tissue contrast of Magnetic Resonance Imaging (MRI) and the functional information of Positron Emission Tomography (PET) are combined in one machine. Although this combination still poses many technological challenges, recent progress [1] indicates that it is only a matter of time before PET/MR machines enter the market. On the software side, attenuation correction, which accounts for radiation attenuation properties of the tissue, is mandatory for obtaining PET images that are sufficiently accurate for quantification. Usually the attenuation map is obtained from a transmission scan, which uses a rotating source. In the case of a PET/MR, there is insufficient space for the rotating source and the attenuation map needs to be determined in another way. Ideally, one would want to calculate it from the MR image. This is inherently difficult: The intensities in the MR image are proportional to the proton density; however, the PET attenuation value is not directly related to the proton density. For example bone compacta and air both do not contribute a MR signal whereas their attenuation values are maximally distinct. Thus, standard intensity based segmentation is bound to fail. In order to determine the attenuation value for a position we have to include additional knowledge. We show how such knowledge can be extracted from local image patches and the registration with an atlas.