Enhancement of four-dimensional cone-beam computed tomography by compressed sensing with Bregman iteration

In four-dimensional (4D) cone-beam computed tomography (CBCT), there is a spatio-temporal tradeoff that currently limits the accuracy. The aim of this study is to develop a Bregman iteration based formalism for high quality 4D CBCT image reconstruction from a limited number of low-dose projections. The 4D CBCT problem is first divided into multiple 3D CBCT subproblems by grouping the projection images corresponding to the phases. To maximally utilize the information from the under-sampled projection data, a compressed sensing (CS) method with Bregman iterations is employed for solving each subproblem. We formulate an unconstrained optimization problem based on least-square criterion regularized by total-variation. The least-square criterion reflects the inconsistency between the measured and the estimated line integrals. Furthermore, the unconstrained problem is updated and solved repeatedly by Bregman iterations. The performance of the proposed algorithm is demonstrated through a series of simulation studies and phantom experiments, and the results are compared to those of previously implemented compressed sensing technique using other gradient-based methods as well as conventional filtered back-projection (FBP) results. The simulation and experimental studies have shown that artifact suppressed images can be obtained with as small as 41 projections per phase, which is adequate for clinical 4D CBCT reconstruction. With such small number of projections, the conventional FDK failed to yield meaningful 4D CBCT images, and CS technique using conjugate gradient was not able to recover sharp edges. The proposed method significantly reduces the radiation dose and scanning time to achieve the high quality images compared to the 4D CBCT imaging based on the conventional FDK technique and the existing CS techniques.