Study Of PET Image Reconstruction Algorithm Based On Co-evolution Genetic Algorithm

Positron emission tomography(PET) technology is the most advanced medical imaging technique in the field of nuclear medicine, and is the bridge between molecular biology and clinical medicine. The information of PET images is rich. It can reflect the anatomical structure and metabolism of living tissue, as well as display organism physiological and biochemical processes at the molecular level. In the present, PET has become a hot research in modern nuclear medicine.Image reconstruction algorithm is the key technology of a PET system, with an important influence in the quality of image. We can inverse the radionuclide concentration within the organization and then estimate physiologic parameters of tissue with PET images reconstructed by collecting a large number of coincidence events. Because of the incompleteness of the collected data, the reconstructing algorithm is ill- posed problem in mathematics.The principles of PET imaging and data acquisition is described in this paper at first, and then the two main reconstruction algorithm, including analytical methods and iterative methods, are discussed. The essence of iterative algorithms are correcting the estimated image repeatedly with a variety of prior knowledge and constraints in the process of iteration, according to certain criteria. Iterative reconstruction algorithms have better performance than the analytical methods, Their reconstructed image has better quality and higher accuracy.Co-evolutionary genetic algorithms are the new algorithms in computer intelligence, the use of these algorithms is just started. With strong adaptive searching capabilities and progressive learning abilities, these algorithms can effectively overcome the premature and slow convergence problems of the traditional genetic algorithms. In the paper, we try to use the co-evolutionary genetic algorithm for reconstructing the PET image. Based on the co-evolutionary principles, we build the algorithm model, and evaluate the individual fitness by the marginal contribution. The image has been initially reconstructed. The results show that the algorithm is effective in PET image reconstruction. Since the algorithm does not require systemtransmission matrix and complex mathematical calculations, computer resources can be efficiently used efficiently. We believe that it is a promising reconstruction algorithm in PET image reconstruction.The experimental data is produced by GATE simulation. GATE combines the advantages of the GEANT4 simulation toolkit including well-validated physics models,and has become the professional simulation platform for PET and SPECT.Taking the cylindrical PET system model as the prototype, we setup a single ring PET scanner with 64 detection modules, and simulate the process of PET imaging. The format of output data is ASII. The data is used in image reconstruction and validate the feasibility of co-evolutionary genetic algorithms.