Five-dimensional respiratory and cardiac motion compensation for simultaneous PET/MR

We propose a new method for 5D motion-compensated PET reconstruction, which is based on radial MR data and which compensates for both respiratory and cardiac patient motion. The new method is evaluated for simultaneous ¹⁸F-FDG PET/MR acquisitions of four patients with a total acquisition time of 5min per bed position. PET/MR data are sorted into 20 overlapping respiratory phases and 12 overlapping cardiac phases. Respiratory and cardiac motion vector fields are estimated sequentially from MR data employing a novel algorithm, which alternates between MR image reconstruction and motion estimation. For 5D motion-compensated PET reconstruction, the resulting motion vector fields are incorporated into the system matrix. 5D motion-compensated PET images reveal less motion blurring than 3D motion average PET and considerably lower noise levels than 5D double-gated PET reconstructions. Evaluation of the myocardium of the four patients shows a quantitative improvement in terms of SUV_mean and contrast when compared to the motion average.