Motion-compensated image reconstruction for cardiac CT with sinogram-based motion estimation

Abstract

Motion-compensated image reconstruction (MCIR) has been widely studied to reduce artifacts induced by cardiac motion. In this paper, we propose a new approach to estimating the motion of the coronary arteries. We use a simple linear model for the motion of each coronary artery during an axial scan and we estimate the model parameters in the sinogram domain. The purpose of the method is not to estimate the motion of the coronary arteries precisely, but rather to provide good initial estimates for subsequent joint image reconstruction and motion estimation.. The idea was evaluated with a cardiac CT simulation. Simulation results illustrate that our proposed method can provide reasonably good motion estimates that may be useful for initializing the joint estimation of both motion parameters and reconstructed images.