SIA OpenIR  > 工业信息学研究室  > 先进制造技术研究室
面向快速成型的医学图像三维重建关键技术
Alternative TitleKey Techniques of 3D Reconstruction Based on Medical Images for Rapid Prototyping
夏仁波1,2
Department先进制造技术研究室
Thesis Advisor王越超 ; 刘伟军
ClassificationTP391.4
Keyword医学图像处理 三维重建 网格简化 网格光顺
Call NumberTP391.4/X26/2005
Pages140页
Degree Discipline机械电子工程
Degree Name博士
2005-01-09
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract生物医学工程是21世纪最重要的科学研究领域之一。快速成型(Rapid Prototyping,RP)技术以分层离散和逐点的材料叠加为特点,为复杂形状的零件制造提供了可行的解决方法。结合医学影像、计算机图形学和RP技术,可以制造出的人体组织器官的代替物,在诊断医学、手术规划及模拟仿真、整形及假肢外科、放射治疗规划、解剖教学等方面都有重要应用。医学图像三维重建不仅是计算机可视化技术的重要研究内容,也是RP技术应用于医学领域的技术支撑。同时,医学图像三维重建方法可用于基于工业CT(ICT)图像的CAD模型反求。因此,研究基于医学图像的三维重建技术,具有重要的学术意义和应用价值。本文以此背景,对医学图像的三维重建以及相关技术进行了深入系统的研究,主要内容包括医学图像的滤波与分割,组织或器官的复杂三维几何模型重建;重建模型的网格简化与光顺等。本文完成了以下工作: 1. 介绍了基于医学图像的三维曲面重建关键技术的研究背景,总结归纳出现有重建方法的基本原理以及它们的优缺点,结合RP技术在医学中的应用所面临的问题,给出了学位论文的选题意义和主要研究内容。 2. 医学图像的滤波和分割是保证重建模型准确表达组织器官的前提。为了有效地去除噪声和保护边缘,设计出25种掩模用于计算中心像素与邻域像素之间的相似度,提出了一种基于相似度的医学图像滤波方法。由于医学图像的模糊性和复杂性,医学图像的分割仍然是图像处理的难点和热点。提出了一种基于Mumford-Shah泛函的具有全局优化的水平集图像分割方法。该方法对于模糊的、灰度渐进的图像具有较强的分割能力,并且分割质量不受初始轮廓线位置和形状的影响。另外,提出了一种基于矢量距离变换的符号距离函数计算方法,用于水平集函数的初始化,这种方法不但速度快,精度高,而且稳定性好。 3. 对基于体数据场的Marching Cube(MC)曲面重建方法进行了研究。介绍了规则体数据场的构造方法以及等值面的定义,讨论了MC算法抽取等值面的基本原理及实现方法,分析了MC算法中连接方式产生二义性的原因及消除二义性的方法。针对现有曲面重建算法存在拓扑难以保持和精度较差等缺陷,首先对移动立方体内三线性等值面的关键点进行分类,并对其几何属性进行了分析,提出了一种自适应的保持拓扑正确的高精度曲面重构算法。不同于以往的移动立方体算法,新的三维重建算法无需事先生成多边形的基本构造形式列表,也不必进行旋转和互补等操作。 4. 研究了重建后网格模型的简化和光顺等方法。采用MC算法构建的三维几何模型包含了大量的三角面片,不但难以实现实时绘制,也不利于存储、传输和计算等操作。以Garland的二次误差测度简化算法为基础,提出采用顶点平均曲率比值的高斯函数值作为加权系数来重新定义测度矩阵,使得尖锐特征边的折叠代价进一步加大,改变边折叠顺序,从而达到保持网格尖锐特征的目的。通过引入不合法折叠边的判断准则,保证了简化网格与原始网格的具有一致的拓扑结构。提出一种基于单步顶点预测单步位置移动的网格光顺算法。通过分析三角网格的顶点及其邻域之间的几何关系,应用带速度抑制函数的双边滤波器,把原始顶点单步移动到预测的新坐标,再利用基于三角形尖角度的顶点移动算法,得到最终的顶点坐标。该算法在光顺噪声的同时,减轻了过光顺的程度,有效地保持了原始三角网格的特征。 5. 在算法研究的基础上,开发了一个面向RP的医学图像三维重建软件原型系统。简要介绍了原型系统的功能模块、结构和使用方法。为RP技术应用于医学领域提供了强有力的数据前处理工具。
Other AbstractBiomedical engineering is one of the most important scientific research fields for 21st century. Rapid prototyping (RP) technology possesses the capability to fabricate parts with complex structure as the characteristic of additive material drop by drop and layer by layer. Using the medical imaging (CT or MRI), computer graphics and RP technology, a biological substitute for damaged or diseased tissue and organ can be produced and is widely used in diagnostic, surgery planning, radiotherapy planning, and teaching in anatomy. 3D surface reconstruction from medical slice-images is not only a main topic of visualization of 3D volume data but also a premise of applying the RP technology into medical domain. So, study on 3D reconstruction from medical images has important significance on science and worthiness in practical application. The main contributions and work are summarized as follows: 1. The study background of key techniques of 3D surface reconstruction from medical images is analyzed firstly. The fundamental theories, the disadvantages and advantages of different surface reconstruction methods are summarized. Then, the problem that applying the RP technology into medical domain encounters is presented. Finally, the significance of this research is pointed out and the main contents of this dissertation are summarized. 2. Filtering and segmenting medical image are premises of accurately representing the 3D reconstruction models of tissues or organs. To filter out noise and protect features of edges in image, a correlation-degree-based filtering method for medical image is proposed. For blurry and complex medical image, image segmentation is difficult and a hot topic in the image processing field. A global optimal image segmentation method based on Mumford-Shah functional and level set was proposed. In order to further stabilize and fasten the level set evolution procedures, an improved approach to construction of the signed distance function using vector distance transformation is addressed. 3. The Marching Cube method to reconstruct surfaces from medical volume data is studied. The ambiguity of MC method is investigated and the corresponding solution is analyzed to avoid the ambiguity. To obtain topology-preserving and high resolution triangular mesh, a robust triangulation algorithm for constructing a manifold triangular mesh approximation to isosurface is presented. The new algorithm presented produces a robust representation of the surface in the interior of each grid cell without using look-up-table and complementary and rotation operations. 4. The methods for mesh simplification and smoothing are studied. A simplification method preserving the important features of mesh is presented by modifying the coefficient of quadric error metrics to strengthen the effect of sharp vertex on the position of new vertex and change the order of edge collapses. This paper systematically discusses the different topological error and gives the corresponding preservation approaches of manifold topology. A vertex-estimation-based, feature-preserving smoothing technique for meshes is proposed based on the geometric analysis of vertex’s relation with the specific local neighbors. 5.A rapid-prototyping-oriented surface reconstruction system is developed to offer a data preprocess and establish a good foundation for carrying out the research of applying the RP technology into medical domain.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/9563
Collection工业信息学研究室_先进制造技术研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院研究生院
Recommended Citation
GB/T 7714
夏仁波. 面向快速成型的医学图像三维重建关键技术[D]. 沈阳. 中国科学院沈阳自动化研究所,2005.
Files in This Item:
File Name/Size DocType Version Access License
面向快速成型的医学图像三维重建关键技术.(3873KB) 开放获取LicenseApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[夏仁波]'s Articles
Baidu academic
Similar articles in Baidu academic
[夏仁波]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[夏仁波]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.