Study On Iterative Reconstruction Algorithm Based On 3D Cone-beam CT

Computed tomography (CT) is widely used in medical diagnosis, spaceflight and industry. The principle of industrial computed tomography (ICT) is similar to medical computed tomography,both are base on Radon transform. With the development of ICT for many years, the scan mode of CT has changed from 2D parallel-beam to 3D cone-beam. Compared with 2D CT, 3D cone-beam CT has much shorter scan time and higher axis resolution because it can make use of the rays more efficiently. The image reconstruction algorithm is an important part of CT technology. There are two kinds of basic reconstruction algorithms: analytic algorithm and iterative algorithm. The analytic algorithm is generally adopted in CT, because it can quickly reconstruct the image with high quality.But the analytic algorithm can not suit the incomplete and noise projections situation,while the interative algorithm can reconstruct better image for this situation.The main disadvantage of iterative algorithm is its slow speed .The time cost of image reconstruction is related to the number of projections,the size of reconstruction image and iteration number. But with the development of computer, the disadvantage of iterative algorithm is become subordinate contradiction. In this paper, iterative reconstruction algorithm is researched for three dimensional cone beam CT.In the ICT application, sometimes can only attain to the incomplete projections, the iterative algorithm has to be applied so that better reconstructed image can be gotten. This paper studies the different iterative algorithms in limited projection situation. In addition, because of the fact that the 3D cone-beam CT using high-density array detector, the cross-talk between detector units caused by X-ray's scatter , there are lots of noise in projections . So we make comparisons between iterative algorithms to the noise projection data. At the same situation, ART(Algebraic Reconstruction Techniques) algorithm is easily influenced by the noise, so we study how to select the relaxation parameter,projection access order and initial value to improve its restrain noise.In algebraic iterative reconstruction algorithm, SART (Simultaneous Algebraic Reconstruction Techniqueis) is an important iterative reconstruction method. The main idea of SART is that, we are through all the rays in a projection to compute the correction between the correct projection and calculational projection, and then add the correction to the image space, the correction will become less and less after each iteration.While there is no restrict to image space, so the image space's noise will increase after the definite number of iteration. In order to solve this problem ,for three dimensional cone-beam CT, a penalty SART is researched.Computer simulation experiment results show that, in cone-beam CT image reconstruction,this means can improve the convergence speed of image reconstruction and reduce the effect coming from the noise.In statistical iterative reconstruction algorithm, OSEM (Ordered Subsets Expectation Maximization) has good reconstructed image quality and costs less computation time. When there has the noise in projection data, the image quality and convergence are influenced by the different subsets size of projective orientation. To three dimensional cone beam CT, OSEM reconstruction algorithm with varied subset size based on image block is researched. The new means partition the image space into equal parts, then for different block image with varied projection subset size to reconstruct image after each iteration. Computer simulation experiment results show that, in cone-beam CT image reconstruction, this means can improve the convergence speed of image reconstruction and reduce the effect coming from the noise.