Investigation On The Fully Three Dimensional Reconstruction Algorithms Of Cone Beam Computed Tomography

Fully three dimensional cone beam CT is the evolution of the original CT on the three dimensional space, which acquires volume projection data by using two dimensional detectors and reconstructs isotropic low dose volume images by using cone beam reconstruction algorithms. It has become the focus of the theoretical study and application with its great spatial resolution and low X-ray dose. But its projection data incompleteness caused by big cone angle scan and low reconstruction efficiency caused by computing huge data prevent the development of Fully three dimensional cone beam CT. So Investigation on the Fully Three Dimensional Reconstruction Algorithms of Cone Beam Computed Tomography was done to improve the reconstructed image quality and efficiency of cone beam reconstruction algorithms.Firstly, this paper reviewed the history of the fully three dimensional CT and its reconstruction algorithms’evolving process and indicated its problems and challenges by analyzing the characters of different scan trajectories, which led the aims and directions of our research in this paper.Secondly, the angle mismatch problem between the off-center plane and center plane and the Radon shadow zone problem in circular cone beam scan was indicated in this paper by analyzing the reasons of big cone angle intensity drop artifacts. Basing on introducing the weighting function to the FDK cone beam reconstruction algorithms, we proposed a new heuristic weighting function of FDK algorithms, which can restrain the affect of angle mismatch, reduce the big cone angle intensity drop artifacts and expand the application range of FDK algorithms. The formation reasons and geometry of the Radon shadow zone of circular cone beam scan were also investigated in this paper. By this method, we proposed a distance-weighted interpolation padding method of the shadow zone, which had a better padding effect than constant padding methods and handled the shadow zone problem in circular cone beam scan.Thirdly, we indicated the importance of PI-line segments selecting method on reconstruction results and unsuitable selecting method may affect image quality and reconstruction efficiency by analyzing the characters and geometry of helical trajectory. So we proposed a PI-line segments selecting method based on FOV, which can insure the uniform intersection between PI-line segments and FOV and reduce computing complexity and improve the reconstructed image quality of BPF algorithms.Fourthly, by analyzing and investigating the problems in the algorithms of super short scan and the relationship between VOI and scan trajectory, a super short scan method based on the M parameter was proposed in this paper, which was suitable for the cone beam reconstructions, avoided inaccurate image out of VOI and Robust to noise. By inversely solving the formula of PI line segments and the three dimensional space, a novel PI-line segments selecting method was proposed, which can low X-ray dose by reducing scan time.Finally, by introducing the geometric relevant and symmetry properties between off-center planes and the middle plane in the back projection process of FDK reconstruction, we proposed an accelerate FDK algorithm based on the geometric relevant. Based on this approach, different off-center planes can be reconstructed synchronously, which reduces the computations in the reconstructions. By optimizing BPF reconstruction algorithms using texture mapping technology, we proposed a new BPF reconstruction algorithms based on GPU. In this approach, the back projection process was moved form CPU to GPU and the reconstruction efficiency was improved, which fulfilled the real time reconstruction requirements.