• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学

• 论文 • 上一篇    下一篇

基于标签化医学图像的多阈值三维重建算法

肖洪旭,杨志永,姜杉,黄哲,赵胜丽   

  1. (天津大学机械工程学院,天津 300350)
  • 收稿日期:2015-06-08 修回日期:2016-08-20 出版日期:2017-10-25 发布日期:2017-10-25
  • 基金资助:
    国家自然科学基金(51775368);天津市重大科技专项(14ZCDZGX00490);广东省省级科技计划
    (2017B020210004)

A multi-threshold 3D reconstruction algorithm
based on labeled medical images

XIAO Hong-xu,YANG Zhi-yong,JIANG Shan,HUANG Zhe,ZHAO Sheng-li   

  1. (School of Mechanical Engineering,Tianjin University,Tianjin 300350,China)
     
  • Received:2015-06-08 Revised:2016-08-20 Online:2017-10-25 Published:2017-10-25

摘要:

针对传统面绘制重建方法MC无法一次性提取多个阈值器官的问题,提出一种将MRI图像中靶区及其周边组织进行标签化分割的
多阈值三维重建算法。该算法通过将二维图像中的多阈值器官标签化为简单的整数,降低提取等值面时的数据存储量,进而
提高等值面的绘制速度。同时,定义了多阈值三维重建时体素顶点索引方式和等值面相交形态,减少了传统MC算法提取多个
阈值器官时存在三角面片与顶点复用的情况,且仅需一次性扫描即可重建多个器官等值面。实验结果表明,本文算法较传统
MC算法,重建器官数量越多,器官彼此结构越紧密,三角面片与顶点复用情况减少越明显,在保证重建效果的同时,绘制速
度可提高30%。
 
 

关键词: 三维重建, 分割, 标签, 移动立方体算法

Abstract:

As a traditional surface reconstruction algorithm the marching cubes (MC) is unable to extract multi-
threshold organs at one time. We propose a multi-threshold reconstruction algorithm to divide the target area
and surrounding tissues in the MRI image into several label values. Changing the lables of multi-threshold
organs into brief numbers can decrease the memory space and accelerate the speed of drawing contour surfaces.
We also define the index way of voxel vertexes and isosurface intersecting morphology of multi-threshold
three-dimensional reconstruction, which can avoid the situation of triangle facets and vertexes reuse of the
MC. Furthermore, the contour surface reconstruction of the multi-organ requires only one time scanning of the
MRI image. Experimental results show that the number of facets and vertices is decreased more significantly
when the number of organs is larger and their relationship is closer in comparison with the traditional MC
algorithm, and the reconstruction speed is improved by 30% while the reconstruction effect is guaranteed.

Key words: 3D reconstruction, segmentation, label, marching cubes algorithm