Image reconstruction of motion corrected sinograms

Abstract

We discuss a motion correction scheme for rigid body motion in PET if the movement is known. Several groups have previously proposed a 2 stage process for motion correction: acquiring the data in list-mode and realigning the events according to the known motion before binning them into a sinogram, followed by reconstruction of the sinogram. However, motion correction of the sinogram data can result in parts of the sinogram not being filled for the full duration of the time frame. This results in image artefacts. We suggest ways to remove these artefacts by either reprojecting data or by determining scale factors for the partially measured parts of the sinogram, or a combination of both. We test this on simulations, phantom data and one patient study.