Preliminary study on reduction of cartoon artifact in the iteratively reconstructed images from sparse projection views

Abstract

In this study, we developed an approach that can help reducing cartoon artifact in the reconstructed images from sparse projection views. Sparse-view CT has been actively investigated as a viable option for low-dose CT, and compressive-sensing-inspired image reconstruction frameworks have been developed accordingly. One of the problems in such algorithms is the over-smoothing in the reconstructed images often called the cartoon artifact. Sparsity of the transformed image is maximized during the iterative image reconstruction steps from the insufficiently sampled data, which helps preserve the edge information but may cause such an over-smoothing. The cartoon artifact not only degrades accuracy of the reconstructed images but also may confuse the clinician’s interpretation. In this study, we analyzed the noise statistics of the reconstructed images and performed a preliminary study to reduce cartoon artifact.