Sparse-View CBCT Reconstruction Using Combined DRUNet and HQS

Sparse-view cone beam computed tomography (CBCT) is commonly used in C arm for clinical diagnosis with reducing radiation dose and low cost advantage. However, due to the existence of circuit noise obeying a Gaussian distribution in the projection data, CBCT images reconstructed by traditional iterative algorithms are not of high quality and still exist in a certain amount of noise. In this paper, we propose a plug-and-play half quadratic splitting algorithm (PnP-HQS), that can be used to solve CBCT image reconstruction problems, which allows the insertion of a pretrained denoiser (DRUNet) into the HQS iterative algorithm to solve subproblems like de-noising. The experimental results based on the three-dimensional Shepp-Logan brain model show that the algorithm has a better noise suppression effect in sparse-view CBCT reconstruction, and is superior to the classical algorithm in quantitative and qualitative evaluation.