Computer Engineering and Applications ›› 2015, Vol. 51 ›› Issue (1): 189-194.

Previous Articles     Next Articles

Clustered calcification detection in mammography based on instance selection

LI Yaolin1, FENG Jun1, WANG Xiaodong1, BU Qirong1, CHEN Baoying2   

  1. 1.School of Information Science and Technology, Northwest University, Xi’an 710127, China
    2.Department of Radiology, Tangdu Hospital, The Fourth Military Medical University, Xi’an 710038, China
  • Online:2015-01-01 Published:2015-01-06


李耀琳1,冯  筠1,王小东1,卜起荣1,陈宝莹2   

  1. 1.西北大学 信息科学与技术学院,西安 710127
    2.第四军医大学 唐都医院 放射诊断科,西安 710038

Abstract: A novel approach based on instance selection for calcification cluster in X-ray mammography images is proposed in order to reduce the time and energy in labeling Regions of Interest(ROI) of Computer Aided Detection(CAD) system. The images are partitioned into blocks, then the texture models are established for all instances of negative packages. Afterwards, distances from the instances of unknown packages to the negative average model are calculated, and the instance having the greatest distance is selected  as the suspicious region. The experimental results show that this method not only has the ability to extract the ROI automatically, but also can  reduce the computation time substantially while keeping the detection performance simultaneously.

Key words: mammography, region of suspicious, instance selection, statistical modeling

摘要: 为了减轻传统计算机辅助检测系统中感兴趣区域标定的时间和精力,提出针对钼靶X光乳腺钙化簇检测的示例选择算法。以分块形式对图像打包,对所有负包示例进行纹理建模,并计算每个未知包中的示例与负包平均模型的距离,选取最大距离的示例作为可疑区域。实验结果表明,该算法在不需要人工标注感兴趣区域,不降低钙化簇检测性能的前提下,大幅度减少了运算时间和空间。

关键词: 乳腺, 可疑区域, 示例选择, 统计建模