System Matrix Modeling Methods In CT Iterative Image Reconstruction Algorithm And Their Comparison In TV

For medical Computed Tomography ?,excessive scanning doses are harmful to patients.An effective method of low-dose scanning is to get the projection data via less sampling angles.However,sparse projection data based on analytic reconstruction will produce severe strip artifacts that can affect the diagnosis.In 2006,CS theory was proposed.Based on this theory,the optimization algorithm can reconstruct high-quality images via sparse projection data.TV minimization algorithm is an effective CS algorithm,which has been studied deeply both in theory and practice;it also has made a series of progress.The TV algorithm is an essentially iterative method,and its imaging problem can be modeled as a linear system of equations.The coefficient matrix of the linear equations can be called as system matrix,which shows how to get the projection data.System matrix modeling method is a projection method.The selection of the system matrix modeling method has some effect on the performance of the TV algorithm.In view of the fact that the problem has not been studied in depth,this paper focus on this problem in order to reveal the influence of system matrix modeling method on TV reconstruction on the basis of comparing various projection methods.This paper has finished the following research:(1)Design and implement four system matrix modeling methods: pixel drive,distance drive,improved distance drive and ray drive.(2)Compare their modeling speed and accuracy.(3)Compare the influence of four modeling methods on the convergence performance in ASD-POCS algorithm.(4)Compare their sensitivity degree to noise.(5)Combining the unmatched theory with ASD-POCS algorithm,the performance of different unmatched pairs in ASD-POCS algorithm is experimentally verified.The results show that the distance-driven modeling method is the fastest to generate system matrix,and the precision of the improved distance is the highest.The convergence speed of the ray drive in the ASD-POCS algorithm is the fastest.The matched projector and backprojector can always make the ASD-POCS convergence but some unmatched PBP can lead to diverge,especially when dealing with noisy projection data,the unmatched PBP is more prone to divergence.In all unmatched PBP pairs,DD-RD has a good convergence performance that can be a recommended unmatched PBP pair.These studies reveal the influence of the selection of the projection algorithm on the image reconstruction based on the TV algorithm,and provide the system matrix selection theory for the medical image reconstruction.