【作者】 余新明；

【导师】 陈作炳； 艾春庭；

【作者基本信息】 武汉理工大学 ， 机械电子工程， 2003， 硕士

【摘要】 基于医学图像的人工关节三维重建系统研究是一个典型的多学科交叉领域，它涉及到医学图像处理、三维重建、虚拟仿生制造等相关技术。这种匹配式的人工关节对减少病人置换手术后的各种并发症极有帮助，同时医学图像三维重建技术在医学诊断、手术规划等临床医学中也有很重要的应用，因此本研究具有重要的学术意义和使用价值。 针对当前医学CT机都配有DICOM接口，其数据可以输出为符合DICIOM协议的数据文件现状，本文研究了DICOM协议中文件格式部分内容，实现了DICOM文件解码功能，可以直接采用CT机产生的原始数据文件为数据来源。 对骨组织轮廓的分割和提取是保证重建模型精度的前提。对不同的医学图像，采用不同的滤波算法，可以去除医学图像中的部分干扰噪声。针对CT文件的特点，首先进行窗宽窗位处理，增强对象的特征，然后再进行聚类分割，把对象分割为不同的区域，利用边缘跟踪算法得到对象边缘轮廓的离散表示，实现了所提出以CT值为聚类特征的K—均值聚类分割轮廓提取方法。由于骨组织边缘轮廓形状的复杂性，采用弧长和曲率结合的原则进行特征控制点的选取，兼顾了特征控制点均匀性和形状准确性的统一。介绍了三次均匀B样条及其反求方法，以选择的特征点为控制点，反求得到骨组织截面轮廓的样条曲线，实现轮廓的矢量化。根据从CT文件中得到的截面间距，对所有的截面数据进行重新排列，形成规则的三维数据场。 研究了三维重建的关键技术，探讨了两种重建算法，详细分析了Marching Cubes算法的基本原理和实现方法，采用边界跟踪算法确定边界立方体，并且用立方体棱边中点代替三次插值计算作为三角片顶点，可以减少处理立方体的数量和计算时间，解决了数据量大和显示速度慢的不足。 本文最后介绍了基于医学图像的人工关节三维重建系统的相关细节。系统采用以项目管理为中心的管理模式，各功能并行处理的系统结构框架。采用层次结构来组织数据，应用面向对象思想分析并建立了系统中的所有数据，全部封装为类。更多还原

【Abstract】 Artificial Joint Reconstruction System based Medical Image is a typical multi-subjects field. It involves medical image process, three-dimensional reconstruction and virtual bionic manufacturing, etc. This kind of matching artificial joints are very helpful for reducing all kinds of syndromes on patients caused by transpositional operations. At the same time, successful three-dimensional reconstruction technology on medical image is widely used in clinic medical diagnosis and operation layout. So this research is very important in both academic and practical significance.Since current CT machines are equipped with DICOM interface and their data can be output as data files, we may directly take the original CT data files produced by CT machines as the source of data. This article investigates the part of file format in DICOM protocol and realizes OICOM files’ function of decoding.The subdivision on bone organization’s image is the premise of accuracy of model reconstruction. This article discusses some classic algorithm of smoothing. We may get rid of part of noise disturbance by some smoothing process on image data. Aiming at the characteristics of CT files, we rise up K-mean clustering subdivision to obtain profile, which takes CT value as character of clustering. First, we carry out the process of window frame and window level to strengthen the objects’ characteristics. And then we carry out treatment of clustering subdivision, the images was divided into different areas. The profile of objection, made up by discrete points, could be obtained by applying algorithm of profile tracking. Because of the complexities of the figure of bone organization’s border, we adopt both the length of arc and curvature to choose character control points, so that we can take into account of both character control points’ equality and their accuracy. After comparing several kinds of spline curve, we discuss the reverse method of three-degree Uniform B-spline in detail. We take the chosen character points as the control points to reverse and realize profile vectorization. According to the space between sections obtained from CT files, we rearrange all the sections’ data to form a regular three-dimensional data field.Besides, we investigate the key technologies in three-dimensional reconstruction, discuss several kinds of main reconstructing algorithm, and analyze the basic principle and algorithm of Marching Cubes in detail. Because of the large quantity of data and low speed of display, we adopt the algorithm of profile tracking to find cubes which at edge, thus reducing the number of cubes. And then, in order to reduce computation, we take themidpoint of cube’s edge; instead of the point by three-degree interpolating calculate, as theacme of triangle.At last, this article introduces some relevant details of Artificial Joint Reconstruction System based Medical Image. The system adopts the management mode which takes project management as the key point and all the functions can parallel work in the system,. Based the OPP, analyzing all the data in the system and adopts structure of hierarchy to organize die data. It also establishes corresponding data structure and encapsulates the data into classes.更多还原