Research On Algebraic Iterative CT Image Reconstruction Algorithm

In recent decades, computer tomography CT has a rapid development in clinical medicine, industrial detection, aviation and other fields. CT check has become an indispensable part of radiological diagnosis field. However, CT radiation will be harm to human body, excessive X ray radiation can cause cancer. Modern medical technology hope that when scanning some organs, as far as possible with the least amount of X-ray doses to get the best reconstruction image quality.Under the condition of limited hardware level, in order to get high precision and convergence speed of image, the CT image reconstruction algorithm becomes crucial. After the years of development, researchers have proposed many CT image reconstruction algorithm, mainly divided into two categories, parsing algorithm and iteration algorithm. Parsing algorithm is the traditional reconstruction algorithm, the main representative algorithm is Filtered Back Projection FBP, which is a direct inversion of projection data refactoring. In complete projection data set can get good quality reconstructed image, this algorithm is simple and fast convergence speed, but in the case of incomplete projection data easily affected by noise lead to reconstruction image has serious strips artifacts. Iterative algorithm including algebraic iterative reconstruction algorithm and statistical iterative reconstruction algorithm, and Algebraic Reconstruction Techniques ART algorithm is the most classical algebraic iterative reconstruction algorithm, it is suitable for different data acquisition mode. It can reconstruct high quality image under the condition of the incomplete projection data and uneven projection angle, but large calculation amount occupied space memory lead to reconstruction speed is slow, but with the rapid development of computer hardware level, the slow convergence speed of iterative reconstruction algorithm can also be solved. To improve the accuracy and speed up the image, in this paper studied three kind improved algorithm based on ART algorithm.This thesis mainly proposed ART algorithm based on non-uniform sampling scanning orbit can improve the image quality precision. Under the same number of sampling means that the same radiation doses, to Sheep-Logan phantom do simulation experiment base on the method we proposed and uniform circular sampling scanning track ART algorithm, and analysis of simulation results. Then put forward minimum related projection data access method based on ART algorithm, which makes the correlation that between rays under the continuous access projection angles is minimum, to speed up the convergence speed. Finally, we make a simulation comparison with the method we proposed and sequential projection and random projection. This paper also studied the Total Variation-Projection Onto Convex Set TV-POCS algorithm based on the theory of compression sensing under the sparse view sampling, and when make a simulation experimental for a two-dimensional Shepp-Logan phantom, the quality of reconstruction used by TV-POCS)algorithm is obviously better than ART algorithm.This thesis mainly content is divided into three parts:1. This paper introduces the CT imaging principle, research background and significance and the development of image reconstruction algorithm, both at home and abroad present situation and future trends.2. The mathematical and physical basis of CT reconstruction algorithm and the principle of several kinds of common image reconstruction algorithm and evaluate the quality of reconstruction image.3. Put forward non-uniform circular sampling orbits ART algorithm under the same number of projection angle with uniform circular sampling orbits of ART algorithm, and used the two methods makes simulation analysis for Sheep-Logan phantom.4. Put forward the minimum correlation projection data access way based on ART algorithm, and comparing with the order projection and random projection in the reconstruction quality and rate of convergence.5. This paper introduces the current situation of the development of compressed sensing theory and TV-POCS algorithm based on compressed sensing theory of sparse sampling. Using two-dimensional Shepp-Logan phantom makes a simulation experiment by TV-POCS algorithm and ART algorithm based on order projection、 random projection and least related projection. Then analyzing the result of simulation experiment.