Research On The Key Imaging Technologies For Over FOV Size Object In Cone-beam CT

Cone-beam computed tomography (CBCT) has been widely used in industrial areas ofnon-destructive inspection, reverse engineering and material organization analysis. In traditionalscan modes, the field of view (FOV) of CBCT is often constrained by the size of hardware.While, in many applications, it is necessary to test over FOV size (referred to as over FOV size:(OFS)) objects. So in order to broaden the scope of applicantion of CBCT, researching on how toextend the FOV has important theoretical significance and practical value.In this paper, the theory of CT image reconstruction and the existing scan modes areinvoloved to carry out the research on the key technologies of testing the OFS object in CBCT.We focus on the reconstruction algothrims for OFS object, geometric artifact calibration underthe OFS scan modes and the parallel acceleration of reconstruction algothrims for OFS object.The major achievements include the following aspects:1. A geometric artifact calibration algorithm in helical scan mode in CBCT is proposed. Theproposed algorithm takes the moving axis of turntable as the research object, constructs thenonlinear least squares solution model for the whole moving axis according to the relations ofprojection geometric. Then, the Gauss-Newton algorithm is used to solve this model and thegeometric calibration parameters of any point on the moving axis are got. The experimentalresults show that the geometric parameters of the proposed algorithm based on the lift axis havea significant effect in the eometric artifact calibration of the reconstructed images for the helicalscan mode in the actual CBCT system.2. To examine the long object with long transverse radius, a multiple helical scan mode andcorresponding reconstruction algorithm are proposed. The method is briefly described as follows.In the step of scanning, multiple helical scans can be made performed by translations oftranslating the turntable, and consequently multiple sets of helical projection data are acquired.When reconstructing, first the multiple sets of projection data are rebinned into one set ofoblique parallel projection data which covers the whole object in transverse direction. Then, thefiltered back-projection algorithm is implemented to obtain the reconstruction result. The methodcan also extend the FOV of CBCT in the horizontal and vertical direction and has highcomputational efficiency and good reconstruction image quality. The results of simulationexperiments show that, the proposed algorithm is available, even more it realizes two times ofthe longitudinal expansion and2.73times of the transverse radius expansion of the FOV.3. To improve the reconstruction speed of the cone-beam reconstruction algorithm, a fastalgorithm based on the compute unified device architecture (CUDA) is proposed. The methodacquires higher computa-tional efficiency by more careful optimization techniques, includingreasonable mode of thread assigning, collecting and pre-computing the variables which areirrelevant with the voxel and the decreasing of number of global memory accesses. The results ofsimulation experiments show that, the fully optimized algorithm can run with a400timesspeedup ratio and makes no precision reduction.