Basis-image reconstruction directly from sparse-view data in spectral CT

In the work, we investigate the feasibility of using sparse data from spectral CT systems with multiple data sets. We develop an optimization-based reconstruction method that may avoid the limitations of standard data-based decomposition methods in spectral CT, such as overlapping rays and dense angular sampling. We generated data from two distinct diagnostic range kVp spectra based on two non-overlapping cone-beam CT scanning geometries. Realistic noise level is added to the data. In specific, we employ the constrained total-variation minimization reconstruction program with beam-hardening-corrected adaptives-teepest-descent-projection-onto-convex-sets algorithm and apply to the full and sparse data for directly reconstructing basis images. Results indicate that images reconstructed from sparse data can be visually comparable to those from full data, with both free of beam-hardening artifacts, demonstrating the feasibility of the sparse data acquisition in spectral CT with multiple data sets for potentially reducing radiation dose.