Noise suppression for spectral CT imaging

Abstract

Dual-Energy CT (DECT) expands clinical applications of CT imaging by its capability to decompose CT images into different material images. However, decomposition leads to large noise amplification and limits the quantitative use of DECT. Additionally, current dual-energy CT scanner designs induce additional image noise problems. Because of these two factors, noise suppression methods have been actively pursued since the inception of DECT. The focus of this research is improvement of a previously developed noise suppression algorithm via penalized weighted least-squares with edge preserving regularization. The method performance is enhanced by pixel similarity-based regularization, which substantially enhances the quality of decomposed images by retaining a more uniformly distributed noise power spectrum.