【作者】 沈涛；

【导师】 罗守华；

【作者基本信息】 东南大学 ， 生物医学工程， 2017， 硕士

【摘要】 高分辨CT是指分辨率在10微米以内甚至达到数十纳米分辨率的CT系统。高分辨CT具有无损检测、高空间分辨等优点,为生物医学、材料学等学科提供了非常重要的检测手段,被广泛应用于疾病模型建立、新药测试、骨参数测量和材料分析等领域的研究中。高分辨CT由于其高精度的特性,其硬件系统及成像算法都有其固有的特点。本课题以自研的一套分辨率达到亚微米水平的高分辨CT成像系统为平台,研究了高分辨CT的校准问题和内重建问题。高分辨CT系统成像过程中需要对系统中的探测器、射线源和转台之间的空间几何关系进行精确估计,利用估计到的几何参数对投影数据和重建过程进行几何校正,才能实现高精度三维成像目的。传统的几何校正方法难以满足高分辨CT系统的精度指标要求,为此,本文提出了一种基于单点模体的高精度CT系统几何校正算法。该算法根据CT反投影几何构建了关于几何校正参数的评价函数,该函数能够有效表征体空间内一点的重建精度,通过对该目标函数的最小化来求解最优几何参数。为了克服噪声以及误差对优化过程的影响,选用随机搜索优化算法模拟退火对其进行求解。高分辨CT成像实验结果证明,相较于传统基于空间解析几何的校正算法,该算法具有更高的精度和鲁棒性。自研高分辨CT系统中的多镜头探测器,可切换镜头以实现不同分辨率及视野(FOV)的成像。不同探测器所成图像往往在空间上、灰度上存在着差异,需要实现多探测器下图像的配准。本文首先研究了该装置下的图像配准空间映射模型,实现了基于互信息图像配准优化算法。实验证明,互信息图像配准能够有效满足高分辨CT平台下的配准需求。进一步分析高分辨CT几何校正和图像配准的内在联系,结合CT几何校正算法研究工作,将多探测器几何校正和图像配准融入到同一优化过程中,提出了基于单点模体的几何校正和探测器配准联合优化算法。高分辨CT多探测器实验和显微CT探测器平移实验结果表明,该算法能够有效地对多探测器CT系统进行高精度的几何校正和图像配准,校准流程简单,仅一次优化过程可以完成整个系统的校准工作。在高分辨率CT应用中,待检样本的横向尺寸需要小于成像视野(FOV),否则,重建图像中会产生严重的截断伪影。基于多分辨率混合扫描方式,利用低分辨率全局数据来弥补高分辨数据中的截断部分,是解决内重建问题的有效手段。为了在高分辨CT上实现精确稳定的内重建,本文利用高分辨CT系统的多镜头探测器,采用多分辨率混合扫描方式,实现了一种低分辨率图像约束的内重建算法(LRICR)。该方法通过配准低分辨大视场与高分辨小视场数据,以低分辨全局投影数据的重建结果作为先验,引入到高分辨内重建过程中,充分利用了高低分辨数据,实现高质量内重建。实验结果表明,与传统的基于TV正则化的内重建算法和Scout Reconstruction内重建算法相比较,该方法不但可以非常好地解决高分辨中的内重建问题,还可以有效提高感兴趣区域的图像质量。更多还原

【Abstract】 High resolution CT is one kind of imageing system of which the spatial resolution is higher than 10 micrometers and even can reach to tens of nanometers.High resolution CT plays an important role in the field of disease modeling,drug testing,tumor monitoring,bone parameters measurement and many other fields due to its nature of non-destructive,non-intrusive and high resolution.Because of the high resolution characteristic,the hardware system and imaging algorithm in high resolution CT demands higher precision than those in conventional Micro CT.Our laboratory developed a high resolution CT system with a spatial resolution of sub-micron level.This paper mainly focuses on the study of system calibration and interior reconstruction algorithm in this high resolution CT platform.The high resolution CT system has a high demand for the geometric calibration precision,and the traditional methods can not meet the demand.In this paper,a geometric calibration algorithm for high precision CT system based on single point phantom is proposed.The high-precision calibration method in this dissertation is based on optimization theory with a simple spherical phantom.The involved cost function associates the misalignment parameters with the convergence degree of the lines back-projected from all projections of the phantom.In order to avoid being trapped in the local minimum due to noise,the simplex-simulated annealing algorithm is employed to minimize the cost function for solving the optimal parameters.Results on the simulated data and the real data from high-resolution CT prove that the proposed calibration method can get all geometric parameters with higher precision and stronger robustness.The multi-lens detector conversion device in the high resolution CT platform allows the system to switch lens detectors with different resolution and FOV.In order to effectively perform image registration among the lens detectors,this paper realizes the image registration algorithm based on mutual information optimization.Experiments show that the optimization algorithm can effectively meet the the registration requirements in high resolution CT platform.In this paper,the relationship between geometric calibration and image registration of high resolution CT is further analyzed.A joint optimization algorithm for the geometric calibration and image registration of multi-detector is proposed which the solving process of the two problems are integrated into the same optimization.The results of high resolution CT multi-detector experiment and Micro CT detector translation experiment demonstrate that the algorithm can effectively perform high-precision geometric calibration and image registration of multi-detector CT system.The calibration process is simple and only one optimization process can complete the system calibration work.In high resolution CT applications,the diameter of the field of view(FOV)should cover the specimen to allow a complete reconstruction.Otherwise,projections truncation occurs and results in truncation artifacts in reconstruction.Multi-resolution acquisition based methods can be applied to solve this problem by using a low resolution scan to fill up the incomplete information in truncated high-resolution projections.This study proposes a low resolution image constrained reconstruction algorithm(LRICR)for high resolution CT interior tomography.In the proposed method,data of global and interior scans are collected for different resolutions.The reconstructed low resolution image is incorporated as prior knowledge into the high resolution interior reconstruction process in the LRICR algorithm.Two phantoms and a bamboo sample were used in experiment to evaluate the proposed algorithm.Compared with the traditional TV minimization algorithm and the multi-resolution ’Scout-reconstruction’ algorithm,the proposed method shows significant improvements in reducing truncation artifacts and provide ROI reconstruction in higher quality.更多还原