基于迭代重建的PET图像空间分辨率优化研究OA

Objective Through experimental research on the self-made miniature Derenzo phantom,to explore the influence of different imaging conditions(the number of iterations,the number of subsets)in the iterative reconstruction algorithm on the image quality of positron emission tomography(PET).Methods The spatial resolution measurement experiment of the preclinical MadicLab Ultra PET/CT equipment was conducted using the miniature Derenzo phantom(list-mode mode,with a acquisition time of 20 min).Image reconstruction was carried out using the iterative reconstruction algorithm.The iterative range was 1 to 40,the subset was 5 to 30 with an interval of 5,the voxel size was 0.314 mm,the matrix size was 257×257,and the full width at half maximum(FWHM)of the Gaussian post-filter was 1.57 mm.Through visual evaluation,signal-to-noise ratio(SNR),contrast,coefficient of variation(CV),and contrast-to-noise ratio(CNR)were used to evaluate the image quality of the 0.8 mm hot rod in the 8 slices at the center of the reconstructed phantom PET image.Finally,through an experiment on healthy mouse,the influence of the optimal reconstruction parameters was empirically studied.Results In the study of the miniature Derenzo phantom,an image resolution of 0.6 mm could be clearly identified.When the 3D-maximum likelihood expectation maximized(3D-MLEM)algorithm was adopted,the SNR of the image reached the peak in 15 iterations and tended to converge after 25 iterations.When the 3D-ordered subset expectation maximized(3D-OSEM)algorithm(with 5 subsets)was adopted,the differences in image changes from 8 to 12 iterations were relatively small.The SNR tended to converge after approximately 7 iterations,and the contrast,CNR and CV values all showed an upward trend.In animal experiments,3D-MLEM(30 iterations)and 3D-OSEM(5 subsets,6 iterations)were used for image reconstruction,and the local tissue structures such as the brain,heart and kidneys of mice could be clearly observed.Conclusion The optimal reconstruction parameters are crucial for obtaining high-resolution images and quantitative accuracy.By using 30 to 40 iterative update times(subset×iteration),0.314 mm reconstructed voxels and 1.57 mm Gaussian post-filter FWHM for image reconstruction,high SNR and high-resolution PET image quality can be obtained.In animal empirical studies,this reconstruction parameter is suitable for high-precision identification and imaging of tiny lesions.