Study On X-ray In-line Phase-contrast Computed Tomography Reconstruction Algorithms Using Incomplete Projections

X-ray in-line phase-contrast imaging(IL-PCI)is a novel imaging technique based on phase shift mechanism,which can produce high contrast between different materials with similar attenuation properties.By extending IL-PCI to computed tomography(IL-PCCT),IL-PCCT can be used for high-contrast and high-resolution three-dimensional(3D)visualization of the weakly absorbing structures inside biological samples(e.g.soft tissues),and this has great application value for biomedical imaging and preclinical imaging research.Filtered back-projection(FBP)is a standard CT reconstruction algorithm for IL-PCCT.According to the“Nyquist-Shannon”sampling theorem,FBP typically requires a long scan time to reconstruct high-quality CT images.However,the long scan time will subject biological samples to higher radiation doses,and also,it may cause motion artifacts.Considering the superior characteristics of synchrotron radiation,such as high brightness and high coherence,IL-PCCT is currently mainly conducted on this light source.To alleviate its long scan issue,the thesis is aimed at solving this issue by designing new IL-PCCT reconstruction algorithms for incomplete projections.By analyzing the imaging characteristics of biomedical tissues commonly applied in IL-PCCT,as well as considering the physical process and imaging advantages of IL-PCCT,three iterative CT reconstruction algorithms for IL-PCCT are proposed,and they are as follows:(1)Multi-material samples are commonly applied in IL-PCCT.Considering the multi-material artifacts issue of multi-material samples and the imaging characteristics of their IL-PCCT images where there are many fine texture details and edge information,an 8-neighborhood forward and backward diffusion filtering(FAB8)method was designed,and then it was combined with a simultaneous algebraic reconstruction technique(SART),and thus SART-FAB8 algorithm was proposed.In SART-FAB8,SART is used as a fidelity term for CT reconstruction,while FAB8 is used to suppress artifacts(including sparse reconstruction artifacts and multi-material artifacts)and protect edge and texture details in the reconstructed images.Considering that the SART-FAB8 algorithm does not need to rely on a prior knowledge of the samples,this algorithm has a good potential to be applicable for many different samples.The proposed algorithm aims to promote the application and development of IL-PCCT for multi-material samples.(2)Morphological study of the liver fibrous tissue is of great significance for revealing the evolutionary mechanism of liver fibrosis diseases,and this has an important application in IL-PCCT.Based on the piecewise-smooth characteristic of the liver fibrous tissue and the anisotropic feature of its edges,an adaptive-weighted anisotropic Tp V(0<p<1)regularization(Awa Tp V)model is designed.The Aw Tp V model includes not only horizontal and vertical image gradient constraints,but also diagonal image gradient constraints which can be used to enhance the directional continuity in the gradient domain.By combining Awa Tp V regularization model with projection onto convex sets(POCS),the Awa Tp V-POCS algorithm is proposed.Considering that the Awa Tp V-POCS algorithm model is non-convex,a“two-step”strategy is used to ensure the convergence of this algorithm,that is,the“POCS step”and“Awa Tp V step”.In“POCS step”,SART is used as a fidelity term for CT reconstruction;In“Awa Tp V step”,the splitting Bregman method is used to solve the Awa Tp V regularization model,and this procedure can suppress artifacts and noise in CT images,and also,it can improve image quality.The above two processes are performed alternately,and finally an ideal CT reconstruction image can be obtained.Considering that IL-PCCT has the advantages of high contrast and high resolution in fiber-like structures,the Awa Tp V-POCS algorithm aims to promote the imaging application of IL-PCCT on biological samples containing fiber-like structures.(3)The qualitative and quantitative analyses of microvascular networks contribute to full understanding,effective diagnosis and accurate evaluation of the therapy or prognosis of diseases involving vasculopathy,and this has a wide application in IL-PCCT.Based on the homogeneity prior of microvascular samples,a new iterative algorithm is presented.In this algorithm,considering the huge computation cost of traditional iterative methods,FBP is thus used to replace the back-projection operator A~T in Landweber framework.Moreover,the k-means clustering image segmentation method is adopted to extract the prior image of microvascular sample,and then this image is utilized as the guide image in weighted guided image filtering(WGIF)algorithm to suppress artifacts on the intermediate reconstructed image and protect the edges of the microvessels.Considering the huge advantages and the great application of IL-PCCT in microvascular imaging,the proposed method aims to promote the imaging application and development of IL-PCCT in microvascular samples.The effectiveness and performance of three algorithms proposed in this thesis are tested and evaluated via numerical model simulation and synchrotron radiation-based IL-PCCT experiments.The experimental content includes not only the reconstruction results of the proposed algorithms under incomplete projections,but also their reconstruction results from the incomplete projections in presence of statistical noise.The experimental results show that the proposed algorithms can significantly reduce the artifacts introduced by the undersampling projection data and can also effectively protect the texture details and edge information of the reconstruction images,and additionally,they have a good robustness in presence of statistical noise.In summary,three algorithms proposed in this thesis can be used as effective tools to reduce the scan time of IL-PCCT,and they are expected to promote the application and development of IL-PCCT in biomedical imaging and preclinical imaging research.