基于类别方差的三维医学图像分割新方法  

A New Method of 3D Medical Images Segmentation Based on Class Variance

作　　者：杜宇慧[1] 桂志国[1] 李晋华[1] 

机构地区：[1]中北大学信息与通信工程学院,山西太原030051

出　　处：《中北大学学报（自然科学版）》2007年第2期166-170,共5页Journal of North University of China(Natural Science Edition)

基　　金：山西省青年科学基金资助项目;医学图像体数据建模技术项目(20051021)

摘　　要：提出一种交互式的三维医学图像分割算法.结合Otsu单阈值图像分割算法,提出了一种基于类别方差的双阈值分割算法.利用该算法对三维医学图像体数据的直方图进行了分析,最终得到的上下阈值使得分割结果具有最大类间方差.该算法采用迭代法实现,简单快速,且可保证分割出的组织包含目标组织.再对此阈值分割结果进行数学形态学的相关操作和三维区域生长,最终得到目标组织的准确分割和它的三维显示.实验证明,分割效果较好,三维重建满足要求.An interactive approach of 3D medical images segmentation was improved in this paper. Firstly, based on the Otsu＇ single threshold segmentation method, a double-threshold segmentation method founded on class variance was proposed. It analyzed the histogram of volume data of 3D medical images, and then got two thresholds automatically that the result of segmentation had max variance between classes. This algorithm applying an iterative procedure was simple and fast, and ensured that the result included the goal tissue. Secondly, the 3D medical images segmented were processed by mathematical morphology operation and 3D region growing. Finally, the exact 3D segmentation and 3D visualization of goal tissue were achieved. The results of experiment show that the algorithm is feasible, and 3D reconstruction meets the goal.

关 键 词：三维医学图像分割 类别方差 阈值 形态学 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]