| Biological images can be divided into traditional images and molecular images.Traditional imaging by using some imaging medium through the tissue gets the structure of the object.And the molecular images show the biological processes of specific molecules at the tissue,cell or subcellular level.Biological imaging has become an indispensable means in modern medical and biological research.As the most basic medical imaging technology,X-ray CT scanning is of great importance.Because of various factors,intensive scanning is often needed.The amount of projected data is very large,and the reconstruction in the later stage takes a long time.The speed of reconstruction can not keep up with the speed of scanning.On the one hand,it greatly reduces the efficiency of the equipment;on the other hand,it prolongs the waiting time of patients.With the rapid development of computer software and hardware,especially the emergence of GPU parallel computing,brings hope for the problem.It is of great significance to optimize the original algorithm by GPU parallelizing to improve the reconstruction speed.With the improvement of people’s health level,the problem of health damage caused by X-ray radiation has gradually emerged.Reducing the radiation dose of X-ray has become the focus of attention.It is imperative to reduce the radiation dose.Traditional optical projection tomography reconstruction algorithms need dense projection data.Acquisition of intensive projection requires long scanning time,which will lead to a series of problems,such as prolonged anesthesia time,phototoxicity,photobleaching effect,etc.There are many ways to reduce the dose of X-ray radiation,among which reducing the number of scans times is the most direct and effective means.How to use these sparse scanned data to reconstruct better images has become a hot research topic in the field of image reconstruction in recent years.Using these sparse projection data to reconstruct a good quality image is of great significance.The main objectives of this study are as follows:1.GPU is used for parallel computing to improve the reconstruction speed and realize real-time reconstruction.2.Optimizing reconstruction for sparse data to reduce scanning time and radiation dose.The main contents of this study are as follows:1.As the research basis of other algorithms,the existing two-dimensional and three-dimensional reconstruction algorithms are analyzed and discussed in depth.2.According to the characteristics of GPU architecture,the existing FDK algorithm is improved for parallel computing.3.Iterative reconstruction algorithms are optimized,and sparse data is used for reconstruction to reduce scanning time and radiation dose.4.An improved compressed sensing algorithm is proposed to further improve the quality of reconstruction and reduce the scanning time and radiation dose.5.In order to verify the algorithm,a CBCT experimental platform is designed and implemented,and a perfect software system is developed for it.6.As a comprehensive application of this optimization algorithm research,a dual-mode optical projection tomography experimental platform is designed and implemented.The platform has both transmission and emission imaging modes,and can realize simultaneous and in-situ detection.An improved statistical algorithm is proposed to reconstruct dual-mode images,which achieves good image quality with very few projection angles,and effectively improves the scanning speed and reduces the irradiation dose.7.Finally,the proposed optimization algorithms are analyzed and compared,and the conclusions and evaluation are given. |
