Effects of discrete versus continuous prior image in sparse-view CT

Abstract

Sparse-view CT is a viable option for low-dose CT, and much efforts have been made to develop image reconstruction algorithms for sparse-view CT. Iterative image reconstruction algorithms are choices of reconstruction which discretize a continuous imaging model by voxelizing the image and by approximating the x-ray transform based on the voxels. Prior image has been utilized to further reduce the number of views in sparse-view CT, but the utilization of such a prior image in discrete domain may result in a suboptimal image quality due to the approximation. In this paper, we present a comparison study on the effects of using projections from a continuous prior image versus a discrete prior image. We implemented a total-variation (TV) minimization algorithm that can reconstruct the image from sparse-view data using prior image knowledge. It is shown that higher-quality images can be obtained by use of the projections of a continuous prior image in the sparse-view CT.