Study Of Limited-View CBCT Reconstruction

Computed Tomography (CT) has been widely used in medical diagnosis andindustrial testing since its advent. In the last decades, CT technology has made greatprogress and breakthrough, majorly on scanning speed and reconstruction quality. CTimaging is composed of two stages: projection data acquisition by rotating X-raytube around the object; CT image reconstruction via captured data. The commonlyused CT technologies are parallel-beam, fan-beam and come-beam. Among them,cone beam CT(CBCT) can capture2D data in one shot, and is more helpful toimprove the ray utilization rate and scanning speed. Due to its complicatedstructure,the complexity of the CBCT reconstruction algorithm increases. Moreover,we may only gather limited angle range of projection data in clinic practices. In suchcases, strong artifacts will be observed on images by traditional reconstructionmethods.In this paper, we studied limited view CBCT reconstruction algorithm and themain work is organized as follows:1. In order to improve the accuracy and effectiveness of the limited anglecone-beam CT reconstruction algorithm, we describe its model design and computersimulation of system matrix in detail which uses3D standard head phantom forexample. First, we list the parameters for different internal structure of the model;then, combined with illustrations and code, we introduce the operation steps andmethods to calculate system matrix.2. As ASD-POCS method utilizes a number of control factors for step-sizecontrol and these factors have to be empirically determined which restrict thepopularization of the method. In this paper, we proposed a new step-size controlalgorithm for the steepest descent in the TV optimization stage. Different with theASD-POCS, our algorithm adjust the step-size according to changes of the updatedprojection data, or the consistency term at each iteration. The consistency controlledstep-size descent (CCSD) algorithm reduce the number of empirical control factors.Besides the step-size control, another contribution of this work is a new stoppingcriterion for the TV steepest descent iterations. Instead of fixed number of loops, the number of iterations is adaptively adjusted according to changes of the reconstructedimages.. Simulated experimental results demonstrate the flexibility and feasibility ofour proposed method.3. Based on the CCSD-POCS algorithm, we add prior knowledge to theconstraint. Due to the fact that the same organ of different people are similar, we usePCA method to process the large number of data by dimension reduce and buildfeature image space. Then we reduce the dimension of the reconstructed image andproject it to the space. Through incorporating priori information, we can furtherreduce the ray dose and the experiment results demonstrate that we can get goodreconstruction results in the condition of few iterations.