Research On The Fast Iterative Reconstruction Algorithm Of Cone Beam Computer Tomography

In recent years,the cone beam CT has found an increasingly wide utilization in medical diagnosis and industrial testing.Image reconstruction algorithm is a key technology of cone beam CT.Iterative reconstruction algorithm can get high quality image,has lower requirement of projection data,and is effective to reduce noise.But its drawback is that the reconstruction speed is slow.So there is still plenty of room to improve iterative algorithm speed.This research is based on cone beam CT fast iterative reconstruction method,aims at resolving the problems of iterative reconstruction algorithm such as large amount of data,low efficiency,and slow speed.This paper first analyzed the calculation method of the system matrix in the iterative reconstruction and realized the linear model of system matrix calculation and compressed storage;Analyzed principles of three reconstruction methods ART,EM and ISRA,discussed the results of three types of iterative reconstruction algorithm;On this basis we accelerated the iterative reconstruction algorithm from the following aspects:through ordered subset method to accelerate reconstruction,researched the influence on image quality and the rate of convergence when subset partition is different;The acceleration on the calculation of the system matrix is from two aspects:Compressed reconstruction object model and used the geometric relationship to calculate;Rotated the reconstruction space to accelerate reconstruction speed;Disposed the circular ring artifacts due to image rotation.Results show that with the increase of subset partition number,the convergence speed is accelerated,we can get better image quality by less iteration number;but in the case of adding noise to projection data,the iterative number of times decreased as the number of subsets in the condition of convergence.When the reconstruction object model is compressed from cube to cylinder,the system matrix computation speed increased by 22.94%;For the reconstruction of the traditional tube and detector rotating way,can only use system on x-y plane symmetry relations,the system matrix computing speed can improve 57.61%;Methods of reconstruction for the reconstruction space rotation can only calculated 1/8 of system matrix and stored under an angle,and the rest calculated by geometric symmetry relations,the speed increased by 86.96%,and compressed the storage space.Using the method of reconstruction space rotation the speed increases by 75.85%;We disposed the circular ring artifacts due to the rotation of the image by three methods,the results show that all methods have certain effect on the round ring artifact,and the reconstruction speed still has a lot of advantages than traditional tube and detector rotating way.