【作者】 王振天；

【导师】 康克军； 张丽；

【作者基本信息】 清华大学 ， 核科学与技术， 2010， 博士

【摘要】 常规X光源光栅成像技术是当前X射线成像领域里的最新前沿技术和研究热点之一,它在一次成像过程可以同时获得反映物质内部结构的衰减图像、相衬图像和暗场图像,形成“三位一体”的成像模式,即保持了传统X射线衰减成像的优点,又拥有相衬成像和暗场成像两者的优势,因此具有非常重要的科研价值和广阔的应用前景。本论文依托清华大学工程物理系建立的国内第一套基于经典光学的光栅成像实验平台,深入开展了常规X光源光栅成像技术的建模方法、系统性能与优化设计,信息提取算法与CT重建算法等相关方法和技术研究,为常规X光源光栅成像技术的临床应用奠定坚实的理论基础和技术积累,使我国在该领域的前沿研究中占有一席之地。本论文的研究成果可概述如下:第一,研究了基于经典光学的X射线光栅成像的建模方法和系统性能优化设计问题。首先,通过在Geant4中实现X射线折射效应来解决X射线相衬成像蒙特卡洛模拟问题,建立了X射线光栅成像蒙特卡洛模拟平台;然后,从多信息提取算法的误差公式、系统敏感度、成像几何条件约束三个方面分析系统性能和优化条件,确立X射线光栅成像系统优化设计方案。第二,提出了X射线光栅暗场信息定量提取算法和基于三张图的多信息提取算法。首先,暗场信息定量提取算法建立了针对一般高斯散射体的位移曲线对比度退化与散射角分布二阶矩之间的定量关系,将光栅暗场成像中的CT重建问题纳入到传统衰减CT重建的框架中。其次,基于三张图的多信息提取算法解决了光栅成像中快速信息提取问题,在保证图像质量前提下尽可能减少辐射剂量和采集时间。第三,提出了针对一阶相移信息的少量投影数据DART-CS迭代重建算法。首先,根据一阶相移信息的投影过程的特殊性推导了线性偏导数矩阵,并以此提出了基于ART形式的DART迭代重建算法;然后,以DART为基础,结合Compressed Sensing理论提出了少量投影数据DART-CS迭代重建算法,实现了低剂量条件下从少量折射角投影数据迭代精确重建物体折射率三维空间分布。更多还原

【Abstract】 Grating-based X-ray imaging with conventional X-ray tube is one of the cutting-edge and hot research fields in X-ray imaging. It can retrieve three different kinds of information of the sample: absorption, refraction and dark-field, from one single scanning. It maintains the advantages of conventional X-ray absorption imaging, and also has the advantages of phase contrast imaging and dark-field imaging, thus it has important research value and a promsing future for practical applications.Based on the classical-optical grating-based imaging system at the department of Engineering Physics, Tsinghua University, this dissertation investigates several key problems such as the system modeling,analysis and optimization, information retrieving algorithms and imaging reconstruction algorithms. The research in this dissertation makes solid theoretical foundation and technical accumulation for the clinical applications of garting-based imaging and takes our country to the forefront of this area. A series of research results were achieved:Firstly, the system modeling, performance analysis and design method of the classical-optical grating-based imaging method are investigated. A Monte Carlo simulation platform is built by implementing X-ray refraction process in Geant4. And from the three aspects of the error formula of information retrieving algorithm, the system sensitivity and the system geometric constraints, the optimized system design method is given.Secondly, a quantitative dark-field information retrieving algorithm and a multiple information retrieving algorithm based on three images are proposed. In the quantitative dark-field information retrieving algorithm, the relationship between the visibility degradation in grating-based dark-field imaging and the general scattering parameter of the sample is deduced. An important conclusion is that, the reconstruction problem of grating-based dark-field computed tomography (CT) can be solved by conventional reconstruction algorithm in absorption CT. The multiple information retrieving algorithm based on three images can reduce the imaging time consuming and the dose delivered to patients in grating-based imaging, while maintain the image quality.Thirdly, an iterative reconstruction algorithm (DART-CS) which benefits from the Compressed Sensing (CS) theory is proposed for differential phase contrast imaging (DPCI). By discretizing the projection process of DPCI into a linear partial derivative matrix, a differential algebraic reconstruction algorithm (DART) is proposed which can reconstruct the refractive index distribution of the sample from the refraction angle projections directly. Based on DART algorithm, the DART-CS algorithm is proposed by introducing the CS theory into the reconstruction of DPCI. It can accurately reconstruct the refractive index with incomplete projection data in low dose situation.更多还原