一种双尺度模板的双视图乳腺肿块检测匹配  

Research of Two-view Mammographic Masses Detection and Matching Based on Double-Template

作　　者：宋立新[1] 李东红 裴恒 曾皓[4] 牛滨[4] 

机构地区：[1]哈尔滨理工大学电气与电子工程学院,黑龙江哈尔滨150080 [2]聚束科技(北京)有限公司,北京100176 [3]中国兵器工业导航与控制技术研究所,北京100089 [4]哈尔滨理工大学测控技术与通信工程学院,黑龙江哈尔滨150080

出　　处：《哈尔滨理工大学学报》2017年第2期129-134,共6页Journal of Harbin University of Science and Technology

基　　金：黑龙江省自然科学基金(F200912);哈尔滨创新人才基金(2010RFXXS026)

摘　　要：为减少乳腺肿块检测到假阳性区域,进而提高可疑病灶区域的匹配率,提出了一种基于双尺度模板检测乳腺肿块可疑病灶区域的方法。该方法首先依据CC视图中可疑病灶区域,在MLO视图中构建条形匹配区域带;然后,基于双尺度Sech模板对肿块图像进行检测后,再做归一化互相关计算,检测出相关性高的区域为肿块的可疑病灶区域,依据基于形状、面积特征的规则删除假阳性区域;最后,根据基于互信息的相似性度量方法实现双视图的可疑病灶区域的匹配。实验结果显示:对DDSM数据库中已确诊的100幅肿块图像进行实验对比,有90幅图像能实现双视图肿块的匹配,匹配率达到90%,与基于灰度分层的乳腺肿块的双视图匹配相比,匹配率得到提高。In order to reduce the detection of false positive masses and improve the matching rate of the double view lump, a method of double templates matching to detect the suspicious lesions area has been proposed. This paper uses the suspicious lesions in the CC-view area to identify the matching bar area in the MLO-view firstly. And then, it uses double Sech-template to detect lumps, the areas of high correlation coefficient will be suspicious lesions area. After deleting the false positive regions, it fuse the different size template results based on rules of shape and area. Finally, it is realized that the mass matching by measuring the similarity based on the mutual information. This paper selects 100 pairs of images which have been construed to carry out the experiments of suspicious lesions match. The experimental results show that ninety pairs of images have been achieved double-view mass matching in comparing with the gray-scale layering algorithm.

关 键 词：双视图 模板匹配 双尺度Sech模板 互信息 匹配率 

分 类 号：TN911.73[电子电信—通信与信息系统]