Optimization-based image reconstruction from low-dose patient breast CT Data

Current dedicated breast-CT prototypes use analytic-based algorithms such as FDK for image reconstruction, which require a large number of densely sampled views. Because the total imaging dose delivered to a patient in a breast-CT scan is kept about the same as that in a typical two-view mammography exam, the use of a large number of views thus can lead to projection data of low SNR and images with high noise, which makes reconstruction improvement challenging. Recently, there exists increased interest in development and evaluation of optimization-based (i.e. iterative) image reconstruction algorithms for low-dose cone-beam CT (CBCT). In the work, we focus on investigation of optimization-based image reconstruction for low-dose breast CT by tailoring a TV-minimization-based algorithm, adaptive-steep-descent (ASD)-projection-onto-convex-set (POCS) algorithm, for image reconstruction from low-SNR patient data. We performed inverse-crime studies for verifying if the algorithm is solving the designed optimization program, and studied the effect of optimization program parameter, ε, on the reconstruction images. We also studied the change of image power spectra with ε and iteration numbers. The results indicate that optimization-based algorithms may improve image quality over analytic-based algorithm for low-dose dedicated breast CT.