多维度动态增强核磁共振乳房区域全自动分割  

Multi-dimensional Automatic Segmentation of the Breast Area in DCE-MRI

作　　者：黄丽娟[1] 厉力华[1] 范明[1] 

机构地区：[1]杭州电子科技大学生命信息与仪器工程学院,浙江杭州310018

出　　处：《杭州电子科技大学学报（自然科学版）》2015年第4期88-92,共5页Journal of Hangzhou Dianzi University：Natural Sciences

基　　金：国家自然科学基金资助项目(61271063);国家重点基础研究发展计划资助项目(2013CB329502)

摘　　要：提出了一种基于乳房水平面与乳房矢状面相结合的多维度DCE-MRI乳房图像全自动分割方法。方法分为3部分,即基于乳房水平面的分割,基于乳房矢状面的分割,以及水平面与矢状面相结合的分割方法。首先,基于水平面的分割方法通过阈值确定乳房外边缘,经过梯度算法后,根据乳房与胸大肌分界面的特点设定约束条件,得到分界面的分割曲线。其次,基于乳房矢状面的分割方法使用双边滤波、边缘提取法对图像预处理,分区计算分割曲线。最后,矢状面的分割结果根据图像三维大小按比例映射到水平面上,将两者的分割结果结合,然后根据相邻图像之间的相关性,从而进行优化,输出分割结果。通过对24例DCE-MRI病例测试,与手动分割得到的结果对比,平均重叠率为93.33%,平均差异度为8.14%。In this paper a multi-dimensional segmentation method based on the combination of breasthorizontal and breast-sagittal was proposed. This method is composed of three parts, which are the segmentation of breast-horizontal, the segmentation of breast-sagittal and the combination of those two segmentations. First,we obtained outer edge of breast by using a certain threshold. The interface of breastpectoral is calculated based on its characteristics by implementing gradient algorithm. Second,we calculated segmentation curve by using bilateral filtering and edge extraction for pre-processing of image. Third,the segmentation results for sagittal are mapped to horizontal in proportion to the three-dimensional size. This method thus obtained optimized results by combining the results of these two levels according to the correlation between the adjacent images. Our segmentation method has been tested on 24 DCE-MRI studies. The results showed the mean percentages of overlay and volume difference compared with manual segmentation are93. 33% and 8. 14%,respectively.

关 键 词：多维度 动态增强核磁共振 图像分割 

分 类 号：TN911.73[电子电信—通信与信息系统]