Research And Optimization Of PET Analytical Reconstruction Algorithm Based On 3D Images

Positron emission tomography(PET)is the representative of advanced nuclear medical imaging equipment at present,and it is an effective means to guide the diagnosis and curative effect of tumors,coronary heart disease and brain diseases.PET can obtain physiological and biochemical information with high sensitivity and structural soft tissue information with high signal-to-noise ratio,and realize the image reconstruction of two-dimensional(2D)data and three-dimensional(3D)data.This paper took the PET projection data as the research object,aimed to optimize the analytical reconstruction algorithm of PET images,and explored the filtering method suitable for PET image reconstruction by studying the artifact noise of the image.The main contents are as follows:(1)PET scanning mode,data acquisition mode and data correction method were discussed.In order to solve the problem of data scattering,the scattering correction experiment was carried out by using patient data.The number of scattering events was estimated by single scatter simulation(SSS)algorithm,then the scattering projection data was subtracted from the original projection data to obtain the real ones.(2)Through the comparison and analysis of filter back-projection(FBP),three-dimensional reprojection(3DRP)and single slice rebinning(SSRB)algorithm,the applicable scope of analytical reconstruction algorithm with their advantages and disadvantages of each algorithm was disscussed.(3)Built the analytical simulator(ASIM)simulation platform and the experimental platform of the software for tomographic image reconstruction(STIR).Configured the files required by the ASIM simulation platform and described the operation flow of the STIR experimental platform.The patient data and simulation data were applied to the STIR platform for image reconstruction to verify the effectiveness of the analytical algorithm,obtained unfiltered reconstructed images.(4)In order to reduce the artifact noise,2D and 3D images were filtered respectively.The median filter was applied to the 2D image for reconstruction and compared with the unfiltered 2D images.3D mean-median filtering method was applied to 3D images for denoising,and the filter control parameter K was adjusted to evaluate and analyze the quality of the reconstructed images.Combined with visual and quantitative evaluation,the influence of 3D mean-median filtering parameter K on the quality of reconstructed images were investigated.The results of this study showed that if the K value was too large,the noise suppression effect was obvious,and the edge information of the reconstructed images were reduced;if the K value was too small,the noise suppression effect was poor,but the edge information of the reconstructed images were better preserved.It was concluded that the filtering control parameter K was closely related to the image quality.According to the gradient distribution histogram of projection data,it was calculated that when the proportion of gradient distribution achieved 96.20%,an appropriate filtering parameter K can be determined from the range of filtering parameters in order to achieve the purpose of better preserving image edge information and eliminating artifact noise.