| In the 1980s, the spatial resolution of general medical CT had been about mllimeter in magnitude, but it could only meet the requirement of macro pathological diagnosis roughly. If the spatial resolution could improve to micrometer in magnitude, then the micro-structure of cells and tissues can be observed and it can reach the golden criteria of clinical diagnosis. So, academe began to develop high-resolution Micro-CT. Micro CT uses cone beam, which not only can acquire truly isotropic volume images and improve spatial resolution and x-ray utilization, but also can obtain a much faster speed than fan-beam CT in acquisition of 3D images. Therefore, cone-beam image reconstruction algorithm became the research issue of Micro CT.Micro CT uses cone beam x-ray and two dimensional detector replacing traditional fan-beam x-ray and one dimensional detector. So it can not only scan multilayer tissues, but also improve the speed of acquiring projection data, utilization of x-ray, and axial resolution of reconstructed image. The algorithm of cone-beam image reconstruction includes analytical and iterative algorithm. The analytical algorithm is divided into approximate and exact algorithm. The analytic algorithm becomes the mainstream of practical CT system, because it is faster than iterative algorithm and occupies less memory resources. Moreover, the approximate algorithm is simple, easy to achieve, and can obtain a good reconstruction effect when the cone angle is small enough. So it has a wide application in practice. The famous Feldkamp algorithm is a typical approximate algorithm dealing with segmental cone-beam projection data. However, its performance will deteriorate obviously when the detector array widen and cone angle largen. But because of the excellent time performance, simple theory and easy realization of Feldkamp algorithms, they still have a great value in engineering application, and become the mainstream algorithm of Micro CT, and also is the content of this research.Because of the large computing capacity of CT reconstruction, especially the back-projection process, reconstruction time can't satisfy practical need in many cases. So, improving the speed of back-projection has been the issue of CT reconstruction. In this field, domestic, foreign scholars and institutions had done a lot of research, and proposed many fast algorithms, but these methods are all based on the nature of sine and cosine, so need many memory to calculate and store values of trigonometric functions. Parallel processing can improve reconstruction speed and shorten reconstruction time, so we will research on parallel cone-beam reconstruction algorithm and a parallel FBP algorithm based on ADRT using multi-core programming and parallel processing. Although parallel processing had been used in 2D CT reconstruction as early as 1990, they were implemented in parallel computers or cluster systems. However, parallel in multi-core CPU can bring more parallel processing capability and higher computing speed.Simulation experiment is the basic approach for researchers to verify the accuracy, effectiveness and evaluate the performance of the algorithm. It is even more conspicuous in the field of biomedical image reconstruction. We use 3D Shepp-Logan as reference model to develop cone-beam reconstruction simulation platform of Micro CT under VC++ 6.0, evaluating the performance of various image reconstruction algorithms. This platform can ensure the realization of our algorithms, and more important, it will be the necessary working environment or platform for Micro CT researchers in the future. |
PDF Full Text Request