CUDA-based Simulation Platform For Multi-energy CT Forward Projection

X-ray Computed tomography(CT) is an important medical imaging and non-destructive industrial test technique. It is assumed that X-ray has single spectrum in conventional CT reconstruction algorithms, whereas in practice the x-ray emitted by x-ray source has continuous mixing spectrum. Recently, with the emergence of energy sensitive x-ray detectors, dual-energy CT and multi-energy CT has become a hot spot in CT field.Both the high price of energy-sensitive X-ray detectors and lack of software platform for multi-energy CT projection simulation limited the acquisition of multi-energy CT data and the application of multi-energy CT in the CT-related area. To solve this problem, this thesis developed a multi-energy CT simulation platform based on CUDA acceleration, which could get multi-energy CT projection data faster and more efficiently. Based on original projection theory and multi-energy CT theory, primary works have been done as followings:First, I researched classical projection algorithms, and on which I implemented CUDA acceleration. The classical projection algorithms including the Joseph ray-driven algorithm and the distance driven algorithm, etc. executed slowly, especially when the data volume is large. So I adopted CUDA acceleration in these two above classical projection algorithms, which is the basis of the development of the simulation platform.Second, the multi-energy CT simulation platform was carried out. Based on the multi-energy CT theory, multi-energy CT projection simulation was performed by projection algorithms used CUDA acceleration in different ways. And in this process, forward projection function library was formed and the performances of different ways were compared.Third, I proposed block projection strategy to solve the large-volume data problem in multi-CT simulation. This strategy got different block projections and then rearranged them into a whole projection, which achieved simulation of projection data with large size on the basis of reasonable partition.Then, based on the multi-energy forward projection library, a user-friendly multi-energy forward CT projection simulation platform was developed, which was easy to use and operate for others. And this simulation platform could load in the original picture, set the parameters, save and show the projections and so on.Finally, detailed performance of the simulation platform was tested. This thesis compared CUDA acceleration, projection performance of different methods, the results of the forward projection reconstruction and multi-energy CT simulation results, etc. Experimental results showed that the platform could simulate multi-energy forward projection of data with large size conveniently and fast.