Research On Geometric Artifacts Correction For Cone-beam CT

Cone-beam computed tomography (CBCT) has been widely used in various applicationssuch as medical diagnosis and industrial nondestructive testing (NDT) for its advantages ofisotropy and fast nondestructive inspection. However, due to the discrepancies between idealmathematical model of reconstruction algorithm and actual CT system, the reconstructed imagein CBCT contains a lot of artifacts which severely degrade the image quality. So research ongeometric calibration is of great importance for improving the system performance and is one ofthe hotspots and difficulties in CT field.Diverse scanning trajectory are adopted in CBCT for specific applications,while onescanning trajectory has one imaging geometric model. Therefore, different calibration methodsneed to be studied according to different characteristics of the geometric model. Here we focuson the research of calibration method of CBCT in RT trajectory, spiral trajectory and circulartrajectory. The major achievements are as follows:1. An algorithm of geometry calibration is proposed for RT trajectory in CBCT, aiming atbroadening the horizontal FOV. Based on machine vision theory, in according with trajectoryproperty, system parameters are separated into intrinsic parameters and extrinsic parameters, inorder to reduce the correlation between them. System parameters are directly related to orbitparameters, through the projection procedure of the calibration phantom. Projection parametersof two different positions are used to build the numerical model, which greatly enhances thestability of the algorithm. The experimental results show that the proposed algorithm has highsolving accuracy and proved to be effective for the correction of geometric artifacts inrecontructed images.2. An algorithm of geometry calibration is put forward for spiral trajectory in CBCT, aimingat enlarging the longitudinal FOV. The orbit of the calibration phantom is predicted according toimaging geometric model, and a least square model is established based on the error between thepredicted orbit and the idle orbit. Geometric parameters are solved by L-M algorithm.Parameters are reduced by half according to the correlation between the direction vector ofascending axis and the focal spot, and the dimension of the model is reduced effectively. Theglobal convergence performance of the model is also enhanced. Experimental results show thatthe presented approach has the advantage of fast convergence rate and high precision, and cansignificantly calibrate geometric artifact in reconstructed image.3. As for conventional imaging application, a self-calibration method based on spatial highfrequency energy for single orbit trajectory is proposed. An optimization model is constructed based on spatial high frequency energy of the reconstructed image. Part of the parameters aredirectly extracted from the projection image to diminish the search scope, and the optimalsolution of geometry parameters is achieved by NM-simplex method that makes the spatial highfrequency energy of CT image maximize. To improve the speed of the algorithm, GPU is used toaccelerate the process of image reconstruction to reduce the reconstruction time. Withoutreprocessing the reconstructed image, the proposed method has less computing complexitycompared with the existing ones. The experimental results show that the presented method has asignificant effect in geometric artifact calibration of the reconstructed images with high accuracy.It can notably reduce iterations without precision reduction, and improve the efficiency.