无网格激发荧光断层成像方法研究

发布时间：2018-03-02 03:28

本文关键词： 分子影像激发荧光断层成像三维重建算法无网格方法分段常数水平集方法　出处：《北京交通大学》2017年博士论文　论文类型：学位论文

【摘要】：分子影像(Molecular Imaging)是一门近年来不断发展的新颖的影像学科,结合了生物化学、多模态成像技术、生物数学、生物信息学、细胞分子生理学、生物物理学、药理学等多种影像学和生物相关学科技术。它为疾病的早期诊断和量化分析、治疗监测和评估以及综合生理学的发展提供了新的技术平台。激发荧光断层成像(Fluorescence Molecular Tomography,FMT)是MI中光学影像模态的一种,它通过靶向生物组织内部特定分子的荧光物质产生的荧光来获取荧光在生物组织内分布的三维成像,从而获取靶向的生理或病理组织的三维量化信息。相比于其他光学分子影像方式,激发荧光断层成像具有灵敏度高、成本低和安全可靠等特点。近年来发展迅速,成为光学分子影像技术的研究前沿和研究热点。本文针对如何提高激发荧光断层成像(FMT)的成像精度、速度和鲁棒性进行了研究,提出了基于无网格方法的FMT前向问题和逆向问题的求解方法,并提出了基于分段常数水平集的FMT重建方法,主要研究内容包含以下几点:1.提出了无网格FMT前向问题求解方法。该方法基于紧支径向基函数(Compactly Supported Radius Basis Function,CSRBF),首先采用一系列彼此独立的离散节点及形状函数来精确表征离散成像区域,然后采用CSRBF求解FMT光子传输模型。由于成像空间离散节点彼此独立,无需进行网格剖分等人工操作,且CSRBF具有彼此独立、非线性表征离散区域的特性,重建效率和精确度显著提高。数值仿真及生物模型实验结果均说明了无网格方法可有效提高FMT的成像精度和效率。2.提出了无网格FMT逆向问题求解方法。我们将CSRBF引入FMT逆向问题求解当中,将经典的有限元网格求解问题转化为无网格求解问题,并结合多种最优化算法,如Tikhonov正则化、L1范数迭代收缩法以及稀疏自适应匹配追踪法等方法来求解FMT逆向问题。数值仿真以及在体实验表明无网格FMT逆向问题求解较经典有限元法可以有效提高FMT逆向问题求解精度和计算效率。3.提出了 FMT逆向问题区域求解方法。该方法基于分段常数水平集理论(Piece Wise Constant Level Set,PCLS),结合光源稀疏性和高对比度的先验知识,在传统的L1范数正则化方法的基础上建立了分段常数水平集重建框架。在具体求解过程中,采用增益拉格朗日交替迭代法求解FMT逆向问题,并结合共轭梯度法和其他加速算法来快速精确求解PCLS水平集函数及成像区域光强分布。数值仿真和生物模型实验证明该方法较传统Tikhonov正则化方法和L1-IS方法能够更快速精确的刻画出成像区域内部荧光光源区域。
[Abstract]:Molecular Imaging is a novel imaging subject that has been developed in recent years. It combines biochemistry, multimodal imaging technology, biomathematics, bioinformatics, cellular and molecular physiology, biophysics. Pharmacology and many other imaging and biorelated disciplines. It provides early diagnosis and quantitative analysis of diseases. Therapeutic monitoring and evaluation, as well as the development of comprehensive physiology, provide a new technical platform. Excitation fluorescence Molecular Tomography (FMTT) is a kind of optical imaging mode in MI. It gets three dimensional imaging of the distribution of fluorescence in biological tissues by targeting fluorescence produced by specific molecules within biological tissues. Compared with other optical molecular imaging methods, excitation fluorescence tomography has the characteristics of high sensitivity, low cost, safety and reliability, and has developed rapidly in recent years. This paper focuses on how to improve the imaging accuracy, speed and robustness of the excited fluorescence tomography (FMTT). In this paper, a method for solving FMT forward problem and reverse problem based on meshless method is proposed, and a FMT reconstruction method based on piecewise constant level set is proposed. The main contents of this paper are as follows: 1. A meshless FMT forward problem solution method is proposed. This method is based on compact Supported Radius Basis function. Firstly, a series of independent discrete nodes and shape functions are used to accurately solve the problem. Characterization of discrete imaging regions, Then CSRBF is used to solve the FMT photon transmission model. Because the discrete nodes in the imaging space are independent of each other, there is no need for manual operations such as mesh generation, and the CSRBF is independent of each other, and nonlinear characterizes the discrete region. The results of numerical simulation and biological model experiments show that the meshless method can effectively improve the imaging accuracy and efficiency of FMT. 2. A meshless FMT inverse problem solution method is proposed. We use CSRBF to solve the inverse problem. In the process of solving FMT reverse problem, The classical finite element mesh solving problem is transformed into a meshless solution problem and combined with various optimization algorithms. For example, Tikhonov regularized L 1 norm iterative contraction method and sparse adaptive matching tracing method are used to solve FMT inverse problem. Numerical simulation and in vivo experiments show that meshless FMT inverse problem solution is more effective than classical finite element method. To improve the accuracy and computational efficiency of FMT reverse problem, a method for solving FMT reverse problem region is proposed, which is based on piecewise constant level set theory, and combines the priori knowledge of light source sparsity and high contrast, based on piecewise constant level set theory. Based on the traditional L1 norm regularization method, a piecewise constant level set reconstruction framework is established. In the concrete solution, the gain Lagrangian alternating iteration method is used to solve the FMT inverse problem. The conjugate gradient method and other accelerated algorithms are used to solve the PCLS level set function and the intensity distribution of the imaging region quickly and accurately. Numerical simulation and biological model experiments show that the proposed method is more effective than the traditional Tikhonov regularization method and L1-IS method. Fast and accurate characterization of the imaging region of the internal fluorescent light source region.
【学位授予单位】：北京交通大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】