Research And Optimization Of PET 3D Image Reconstruction Algorithm Based On OSEM

PET is a kind of medical imaging method,which has accepted clinical application value for disease diagnosis and treatment monitoring.How to reconstruct high quality PET image quickly and accurately has been a hot topic in the current research field.The PET image reconstruction algorithm includes analytical method and iterative method.FBP algorithm is a classical analytical method,which is simple and fast to calculate,but the quality of the reconstructed image is not high.The statistical iterative reconstruction algorithm in the iterative method is widely used.The quality of the reconstructed image is improved by introducing the physical model into the image reconstruction.At present,the statistical iterative reconstruction algorithm widely used in clinic is the OSEM algorithm.This paper researched the related work of PET 3D OSEM image reconstruction.Investigated the PET imaging principles,data collection methods and processing methods,image reconstruction algorithms of PET system,and achieved basic PET 3D OSEM image reconstruction.In order to improve the quality of PET image,we applied 3D mean-median filter to process the projection data before reconstruction,and the reconstructed image was visually and quantitatively evaluated.In short,the research content of this topic is mainly divided into three parts:1.Building an experimental platform: Installed the software for tomographic image reconstruction.Configured the STIR software operating environment to run STIR.Installed PET analysis simulator ASIM to simulate 3D projection data.2.Data acquisition and processing: The acquired data was sorted according to michelogram.Converted the sorted data to the Interfile format supported in the STIR.Attenuation correction,normalized correction and scattered correction were applied to the data.3.Implementation and optimization of statistical iterative image reconstruction algorithm: The basic 3D image reconstruction of PET was realized by using the OSEM algorithm and MRP algorithm of STIR to reconstruct the corrected data.In order to further improve the quality of PET image,we applied 3D mean-median filter to smooth nosie in the sonogram.OSEM algorithm of STIR was used to reconstruct the projection data before and after filtering.And the influence of filter parameter on OSEM reconstruction images was evaluated qualitatively and quantitatively.The experimental results show that the noise level and edge preservation effect of the reconstructed image are very sensitive to the selection of filter parameters.We could select the range of filtering parameter according to the distribution of gradient histogram.Then the appropriate parameter is chosen by combining with the gradient distribution ratio,so as to remove noise and retain edge characteristic.