计算机与现代化 ›› 2021, Vol. 0 ›› Issue (02): 1-6.

• 图像处理 •    下一篇

基于纹理特征的超声图像乳腺肿块识别

  

  1. (1.河海大学计算机与信息学院,江苏南京211100;2.河海大学无线通信与智能系统研究所,江苏南京211100)
  • 出版日期:2021-03-01 发布日期:2021-03-01
  • 作者简介:李梓龙(1997—),男,江西上饶人,硕士研究生,研究方向:医学图像处理,神经网络,E-mail: 2471763183@qq.com; 吕勇(1979—),男,江苏南京人,副教授,博士,研究方向:多媒体信号处理,音频信号处理,E-mail: yonglu@hhu.edu.cn; 谭国平(1975—),男,教授,博士,研究方向:无线多媒体通信,随机网络优化与控制,E-mail: gptan@hhu.edu.cn; 严勤(1977—),女,教授,博士,研究方向:视音频网络及编码,语音信号处理,E-mail: yanqin@ieee.org。
  • 基金资助:
    国家自然科学基金重点项目(61832005)

Breast Mass Recognition Based on Texture Features in Ultrasound Images

  1. (1. College of Computer and Information, Hohai University, Nanjing 211100, China;
    2. Institute of Wireless Communications and Intelligent Systems, Hohai University, Nanjing 211100, China)
  • Online:2021-03-01 Published:2021-03-01

摘要: 针对乳腺超声图像,提出一种基于图像纹理特征提取的乳腺肿块识别方法,从而有助于使用计算机辅助鉴别的方法判断乳腺肿块是否发生癌变,辅助放射科医生对影像的性质作出预判。首先对乳腺超声图像进行最大响应滤波处理,在保证一定边缘组织结构完整的同时去除主要的噪声干扰。在此基础上,提取乳腺图像的一阶和二阶纹理特征,然后用人工神经网络对特征进行识别分类。在从医院拿到的真实数据集上验证本文方法的准确性,并分别从预处理、特征提取和分类方法3个方面与其他方法进行对比分析,结果表明,本文方法在降低算法复杂度的基础上提升了乳腺肿块的识别率。

关键词: 乳腺肿块, 超声图像, 图像纹理特征, 滤波处理, 人工神经网络, 计算机辅助检测

Abstract: Aiming at the breast ultrasound image, a method of breast mass recognition based on  texture feature extraction is proposed, which is helpful to use computer-aided identification method to judge whether breast mass is cancerous or not, and assist radiologists to predict the nature of images. Firstly, the max-response filter is applied to the breast ultrasound image to remove the main noise interference while ensuring the integrity of a certain edge tissue structure. On this basis, the first-order and second-order texture features of breast images are extracted, and then the features are identified and classified by artificial neural network. The accuracy of the method is verified on the real data set obtained from a hospital and the proposed method is compared with other methods from three aspects: preprocessing, feature extraction and classification. The results show that the proposed method improves the recognition rate of breast masses on the basis of reducing the complexity of the algorithm.

Key words: breast mass, ultrasound image, image texture features, filter processing, artificial neural network, computer aided detection