权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Study of Theory and Algorithms for Model Based General Purpose CAD Systems of Three-dimenisional Images in Medicine

基于模型的医学三维图像通用CAD系统理论与算法研究

基本信息

批准号：
13480093
负责人：
TORIWAKI Junichiro
金额：
$ 9.66万
依托单位：
Nagoya University
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (B)
财政年份：
2001
资助国家：
日本
起止时间：
2001 至 2002
项目状态：
已结题

项目摘要

(1) Procedures for model-based extraction of organs We derived basic procedures for recognizing organs from three-dimensional human body. The method consists of the voxel-wise region growing using CT values and model matching based on shape models stored in the system beforehand. Models were described in terms of (1) center line expression by spline curves, (2) Beje curves expression of the surface of organs, (3) sets of shape features, and (4) structure on key-slices. These were applied to recognition of (1) large vessels in mediastinum, (2) lung areas, (3) bronchus, and (4) discriminating aorta and veins in lung, respectively and their effectiveness was shown experimentally. By several tens ofModel matching was performed by minimizing the energy function defined on 3D images using hill cribming algorithm.(2) Survey States of arts in the relating fields was surveyed in detail.(3) Three-dimensional image processing algorithm Past work of this research groups on three-dimensional digital geometry was systematized and published as a monograph and textbook. The vision expert system IMPRESS was extended to three dimensional image processing.(4) CAD concept CAD concept was clarified by dividing three types : general purpose CAD, speical purpose CAD, and daily use CAD.(5) Virtualized human body and MR medicine We proposed the MR (mixed reality) medicine as the unification of the virtualized human body and the real human body for diagnosis and treatment. As its element technology, virtualized endoscope system and navigation-based endoscope system were developed.

(1)基于模型的器官提取流程我们推导了三维人体器官识别的基本流程。该方法包括基于CT值的体素区域生长和基于预先存储在系统中的形状模型的模型匹配。模型采用样条曲线表达中心线，Beje曲线表达器官表面，形状特征集合，键片结构。将这些方法分别应用于(1)纵隔大血管的识别，(2)肺区域的识别，(3)支气管的识别，(4)肺主动脉和静脉的识别，并通过实验证明了它们的有效性。通过几十次的模型匹配，利用hill cricrim算法对三维图像上定义的能量函数进行最小化。(2)对相关领域的艺术现状进行了详细的调查。(3)三维图像处理算法本课题组在三维数字几何方面的研究成果已被系统化，并以专著和教科书的形式出版。将视觉专家系统IMPRESS扩展到三维图像处理。(4) CAD概念将CAD概念划分为通用CAD、专用CAD和日常使用CAD三种类型，明确了CAD的概念。(5)虚拟人体与MR医学我们提出了MR（混合现实）医学作为虚拟人体与真实人体的统一进行诊断和治疗。作为其核心技术，开发了虚拟内窥镜系统和基于导航的内窥镜系统。