The Research And Optimization Of Pet Image Reconstruction Algorithm

Computer science and technology develop so rapidly that it affects all aspects of our life and work. In clinical medicine, computer-aided applications are more and more commonly used. PET is so an important clinical diagnosis technology that it get more and more widely used. The main principle of PET is:patient's body is injected dose of radioisotope pharmaceutical, the radioactive particles released positive electron in the decay process. Annihilation occurs, when the negative electron encounters the positron, and the reverse movement of two photons was resulted. PET equipment detects the photons value which is launched by the body regions, and then according to the measured data, the grayscale image of the organism-specific organs was reconstruction. The images show the prosperity of the organ's metabolism level, which can identify whether the organ has lesions or even cancer.Three-dimensional (3-D) image reconstruction process is the key point of PET imaging. PET image reconstruction algorithm can be divided into analytical method and iterative method. The representative of analytical method is filtered back projection algorithm (FBP).The most classical algorithm of statistical iterative algorithm is MLEM algorithm, which is proposed by Shepp and Vardi in 1982. MLEM algorithm is pixel-based iterative algorithm, the value of each voxel is to be estimated as parameters. By continuous iterative updating, so that the likelihood function get maximum approximation and thus gain the last value of the parameter.The ideas is used in a variety of follow-up improved algorithm, and on this basis, OSEM and MOSEM algorithm were developed.Classical MLEM algorithm costs large amount of computation, and the speed of reconstruction is very slow, so it is very difficult to be directly applied to practical engineering. Therefore, this paper focuses on the process of traditional MLEM algorithm. by double improvement of MLEM algorithm's calculus process,it greatly reduce the iteration time of PET reconstruction.The characteristic of MLEM algorithms is slow convergence. The improvement of MLEM algorithm is OSEM algorithm. In OSEM algorithm, by initializing voxel value with FBP results, it can reduce the iteration computation time and the whole number of convergence iterations. So that the entire process of image reconstruction is speeded up. And the time of the reconstruction is smaller.