Compensation of patient motion in PET using a rotator and list-mode reconstruction

PET images are degraded by a number of physical factors present in the imaging process. One of these factors that can potentially have a large impact on PET image quality is patient motion. Due to the relatively large scanning time involved in a whole-body PET study, it is common for patients to move during the imaging study. Furthermore, physiological motion such as respiratory and cardiac motion can also degrade image quality. There are two basic approaches for developing PET iterative reconstruction methods that compensate for patient motion. The first approach is to model patient motion in the reconstruction (image) space. The second approach is to model the relative patient motion by moving the detector, and correcting the measured lines of response (LORs) in event space. In this study, an image-space based, list-mode ordered-subsets, expectation maximization (OSEM) algorithm is evaluated and tested. This method forms frames of list-mode data for each motion state, and uses a fast 3D Gaussian rotator to model motion in the projection and backprojection operations. In addition, a simpler method involving post-reconstruction rotation of each motion state is also evaluated. To study these approaches, a GATE Monte Carlo based computer simulation study using the NCAT anthropomorphic phantom was used to evaluate and compare the motion compensation methods. Quantitative analysis indicated that the two motion correction methods provided good fidelity to the reconstruction obtained from simulated measurements with no patient motion.