东北大学学报（自然科学版）

东北大学学报(自然科学版) ›› 2013, Vol. 34 ›› Issue (6): 792-796.DOI: -

• 信息与控制 • 上一篇    下一篇

基于特征选择ELM的乳腺肿块检测算法

王之琼1,2，康雁1，于戈2，赵英杰3   

  1. (1东北大学中荷生物医学与信息工程学院，辽宁沈阳110819；2东北大学信息科学与工程学院，辽宁沈阳110819；3辽宁省肿瘤医院影像科，辽宁沈阳110042)
  • 收稿日期:2012-12-01 修回日期:2012-12-01 出版日期:2013-06-15 发布日期:2013-12-31
  • 通讯作者: 王之琼
  • 作者简介:王之琼(1980-)，女，黑龙江哈尔滨人，东北大学讲师，博士研究生；康雁(1964-)，男，辽宁沈阳人，东北大学教授，博士生导师；于戈(1962-)，男，辽宁大连人，东北大学教授，博士生导师；赵英杰(1959-)，女，辽宁沈阳人，辽宁省肿瘤医院主任医师.
  • 基金资助:
    国家自然科学基金资助项目(61100022)；辽宁省科技计划项目(20120323).

Breast Tumor Detection Algorithm Based on Feature Selection ELM

WANG Zhiqiong1,2， KANG Yan1， YU Ge2， ZHAO Yingjie3   

  1. 1. SinoDutch Biomedical and Information Engineering School， Northeastern University， Shenyang 110819， China; 2. School of Information Science & Engineering， Northeastern University， Shenyang 110819， China; 3. Medical imaging Department， Tumor Hospital of Liaoning Province， Shenyang 110042， China.
  • Received:2012-12-01 Revised:2012-12-01 Online:2013-06-15 Published:2013-12-31
  • Contact: KANG Yan
  • About author:-
  • Supported by:
    -

RichHTML

0

PDF (PC)

885

摘要： 乳腺肿块检测是防治乳腺癌的有效途径，基于乳腺X射线图像特征模型的极限学习机(ELM)分类算法已被应用于计算机辅助检测乳腺肿块中.针对由于特征间的依赖性导致的ELM学习效率和检测准确度低的问题，提出了基于特征选择ELM的乳腺肿块检测算法.利用影响值选择、序列前向选择和遗传选择等方法进行特征选择，进而利用该结果提高ELM的性能.通过490例来自辽宁省肿瘤医院的乳腺X射线图像的实验表明，基于特征选择ELM的乳腺肿块检测算法能有效提升乳腺肿块检测的效果，其中以遗传选择对ELM性能提升最明显.

关键词: 极限学习机, 遗传选择, 影响值选择, 序列前向选择

Abstract: Breast tumor detection is an effective way for preventing breast cancer， and the classification algorithm of extreme learning machine(ELM) that based on XRay image feature model of breast had been used in computer aided detection of breast tumor. Due to the low learning efficiency and detection accuracy of ELM caused by the dependence between features， a breast tumor detction algorithm was proposed in this paper based on features selection ELM. The methods of impact value selection， sequential forward selection and genetic algorithm were used to improve the performance of ELM. The 490 XRay images used in the experiment came from Tumor Hospital of Liaoning Province， and the results showed that the precision of breast tumor detection could be improved with the proposed method especially for genetic selection algorithm.

Key words: extreme learning machine, genetic selection, impact value selection, sequential forward selection

中图分类号: