作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程

• 图形图像处理 • 上一篇    下一篇

基于多特征联合监督字典学习的乳腺图像分类

刘利卉,徐军,龚磊   

  1. (南京信息工程大学 信息与控制学院 江苏省大数据分析技术重点实验室,南京 210044)
  • 收稿日期:2017-02-28 出版日期:2018-03-15 发布日期:2018-03-15
  • 作者简介:刘利卉(1991—),女,硕士研究生,主研方向为医学图像处理;徐军,教授、博士后;龚磊,硕士。
  • 基金资助:

    国家自然科学基金(61273259);江苏省自然科学基金 (BK20141482)。

Breast Image Classification Based on Multi-feature Joint Supervised Dictionary Learning

LIU Lihui,XU Jun,GONG Lei   

  1. (Jiangsu Key Laboratory of Big Data Analysis Technique,School of Information and Control,Nanjing University of Information Science and Technology,Nanjing 210044,China)
  • Received:2017-02-28 Online:2018-03-15 Published:2018-03-15

摘要:

针对无监督字典学习算法图像分类精度不高的问题,提出一种结合多种图像特征的有监督字典学习分类算法。利用卷积神经网络检测和分割细胞以提取细胞结构形状纹理特征,在细胞对应的病理图像块中提取多种纹理特征后,提取全图的SIFT和SURF特征。为缩小分类误差,对无监督字典学习和二分类函数进行联合训练,将多特征取代图像作为字典学习输入,最终实现乳腺病理图像分类。在2个乳腺病理数据库上的实验结果表明,多特征监督字典学习分类算法的分类准确率达92.15%,优于无监督字典学习算法。

关键词: 多特征, 有监督, 字典学习, 细胞分割, 细胞检测

Abstract:

Aiming at the problem that the unsupervised dictionary learning algorithm has low image classification accuracy,a supervised dictionary learning classification algorithm which combines with multiple image features is proposed.It uses the convolution neural network to detect and divide cells to extract the texture of the cell structure.It extracts a variety of texture signatures for the cells corresponding to the pathological image of the cells,and then extracts the SIFT and SURF characteristics of the whole picture.In order to reduce classification errors,unsupervised dictionary learning and binary classification functions are jointly trained,and images are replaced by multi-feature as dictionary learning input,and breast pathological images are classified.Two breast pathological databases are compared,and experimental results show that multi-feature supervised dictionary learning algorithm classification accuracy is up to 92.15% and classification performance is better than unsupervised dictionary learning algorithm.

Key words: multi-feature, supervised, dictionary learning, cell segmentation, cell detection

中图分类号: