基于分层多模板匹配的乳腺肿块自动检测  

Automatic Breast Mass Detection Based on Hierarchical Multi-Template Matching

作　　者：徐胜舟[1] 胡怀飞[2] 

机构地区：[1]中南民族大学计算机科学学院,武汉430074 [2]中南民族大学生物医学工程学院,武汉430074

出　　处：《中南民族大学学报（自然科学版）》2013年第2期92-96,共5页Journal of South-Central University for Nationalities：Natural Science Edition

基　　金：国家自然科学基金资助项目(61075010);中南民族大学自然科学基金资助项目(CZQ11027)

摘　　要：为提高乳腺癌检测的精准度和效率,提出了一种基于多模板匹配的乳腺X线摄片肿块自动检测方法.该方法设计了一种基于灰度直方图先验信息的阈值分割法来提取乳房区域;构造了不同尺度的模板对肿块进行分层匹配;对匹配结果提取了灰度、形态学及层间特征以便尽可能地去掉假阳性区域.与DDSM数据库中的已确诊的306幅乳腺X线射片进行了比较,实验结果表明:该算法在检测各种不同类型和尺寸的肿块时,具有高敏感度和低假阳性率.In order to provide a valuable opinion for improving accuracy and efficiency of detecting breast cancer in the clinical environment, an automatic detection method based on hierarchical multi-template matching has been proposed in this paper to detect breast mass on mammograms. First, a histogram thresholding method based on priori information has been proposed to extract the breast region. Second, templates with different size have been designed to match the mass. At last, hierarchical features, gray-scale and morphologie features have been extracted to discriminate the true positive and false positive. The results based on 306 mammograms from DDSM validate that the proposed algorithm can achieve good results on a variety of different types, different sizes of masses, with higher sensitivity and lower false positive rate.

关 键 词：肿块 检测 分割 多模板匹配 乳腺X线摄片 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]