图像质量驱动的锥形束CT几何校正方法研究

发布时间：2018-05-16 16:00

本文选题：锥形束CT + 几何校正　；参考：《南方医科大学》2017年硕士论文

【摘要】：X射线计算机断层成像技术(Computed Tomography,CT)是一种利用物体不同角度的射线投影来重建物体内部组织结构的成像技术。其中,锥形束CT(Cone-beamCT,CBCT)相比传统的二维CT(螺旋CT),具有扫描速度快、射线利用率高、空间分辨率高、重建三维图像等优点。CBCT利用平板探测器采集被检测物体在不同角度的X射线投影数据,采用重建算法重构出被检测物体的三维断层图像。在重建算法中,机械几何结构需要满足:第一,射线源中心射线应经过旋转轴并垂直于平板探测器的中心;第二,旋转中心轴与探测器的列方向平行。但是,实际情况中,由于机械设计和安装水平等原因,CBCT的几何结构与重建算法要求的理想几何结构之间存在一定程度的偏移,导致重建图像中产生几何伪影。几何伪影严重影响图像的细节和医生对病人病情的诊断。针对CBCT图像中的几何伪影问题,本文以优化重建图像质量为目标,进行以下三方面的研究:第一,提出一种几何校正效果的评价模体和方法。评价模体由低密度正方体和高密度球体组成。首先通过CT扫描模体获取三维图像并通过Hough变换计算出中心层图像高低密度的边界,然后将高低密度区域CT平均值的差值的绝对值作为评价几何校正效果的指数。通过设置不同几何参数的数字模体和真实模体实验,结果表明提出的评价指数能够客观评价CT图像的几何校正效果,可为评价几何校正效果提供统一的标准。第二,提出一种图像质量驱动的单投影几何校正方法。通过引入上文中的评价指数作为衡量几何伪影严重程度的目标函数,通过迭代修正校正模体上标记点的坐标,最终达到校正几何参数和消除CBCT图像几何伪影的目的。为提高运行效率,采用GPU对CBCT图像重建部分进行加速。通过仿真实验、bench-top CBCT、口腔CT数据实验结果表明,系统几何参数的误差由19.07%降到0.115%,消除了 CBCT图像的几何伪影。该方法还降低了校正模体加工误差对校正精度的影响,大幅度减少校正模体制作成本。第三,提出一种基于简易模体的整体投影几何校正方法。在圆轨道CBCT下,由机械运动和安装调试引起的几何误差,虽与真实值有一定的偏差但每个角度下投影的几何参数比较相近。该方法通过采集简易模体的整体投影,解析计算系统的几何参数,并以评价指数为目标函数,对几何参数进一步迭代修正。实验结果表明,系统几何参数的误差由2.2%降到0.105%,该方法进一步提高了现有解析几何校正方法的校正精度,改善重建图像的质量。本文针对CBCT几何伪影问题,提出了一种评价几何校正效果的标准,并以优化重建图像质量为目标提出了基于单张投影和整体投影的CBCT几何校正方法,效果明显,实用价值突出。
[Abstract]:Computed Tomography (CTT) is an imaging technique for reconstruction of the internal structure of an object by means of ray projection at different angles. The Cone-Beam CTT (CBCTT) of cone-beam has higher scanning speed, higher ray-utilization ratio and higher spatial resolution than the conventional 2D CTT (helical CTT). CBCT uses flat panel detector to collect X-ray projection data of the detected object at different angles, and uses the reconstruction algorithm to reconstruct the 3D tomography image of the detected object. In the reconstruction algorithm, the mechanical geometry should be satisfied: first, the central ray of the ray source should pass through the axis of rotation and perpendicular to the center of the plate detector; second, the axis of the center of rotation should be parallel to the column direction of the detector. However, in practice, due to mechanical design and installation level, there is a certain degree of deviation between the geometric structure of CBCT and the ideal geometric structure required by the reconstruction algorithm, resulting in geometric artifacts in the reconstructed image. Geometric artifacts seriously affect the details of the image and the doctor's diagnosis of the patient's condition. Aiming at the problem of geometric artifacts in CBCT images, this paper aims at optimizing the quality of reconstructed images, and studies the following three aspects: first, a geometric correction evaluation motif and method are proposed. The evaluation motif consists of a low density cube and a high density sphere. Firstly, 3D images are obtained by CT scanning motifs and the boundary of high and low density of center layer images is calculated by Hough transform. Then, the absolute value of the difference of CT mean values in high and low density regions is taken as the index to evaluate the geometric correction effect. Through digital and real motifs experiments with different geometric parameters, the results show that the proposed evaluation index can objectively evaluate the geometric correction effect of CT images and provide a unified standard for evaluating geometric correction effects. Secondly, a single projection geometric correction method driven by image quality is proposed. By introducing the evaluation index mentioned above as the objective function to measure the severity of geometric artifacts, and by iteratively modifying the coordinates of the labeled points on the correction motifs, the purpose of correcting geometric parameters and eliminating geometric artifacts of CBCT images is finally achieved. In order to improve the running efficiency, GPU is used to accelerate the reconstruction of CBCT image. The experimental results of oral CT data show that the error of geometric parameters of the system is reduced from 19.07% to 0.115, and the geometric artifacts of CBCT images are eliminated. The method also reduces the effect of machining error on the correction accuracy and greatly reduces the cost of the calibration mode system. Thirdly, a global projection geometric correction method based on simple motifs is proposed. The geometric error caused by mechanical motion and installation and debugging in circular track CBCT is different from the real value, but the geometric parameters projected at each angle are close to each other. By collecting the global projection of the simple motifs, the geometric parameters of the system are analyzed, and the evaluation index is taken as the objective function to further iterate and modify the geometric parameters. The experimental results show that the error of geometric parameters of the system is reduced from 2.2% to 0.105. This method can further improve the correction accuracy of the existing analytical geometry correction methods and improve the quality of reconstructed images. In this paper, aiming at the problem of CBCT geometric artifact, a standard for evaluating geometric correction effect is put forward. Aiming at optimizing the quality of reconstructed image, a CBCT geometric correction method based on single projection and global projection is proposed. The effect is obvious and the practical value is outstanding.
【学位授予单位】：南方医科大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;O434.1

【参考文献】