计算机工程与应用 ›› 2020, Vol. 56 ›› Issue (4): 122-127.DOI: 10.3778/j.issn.1002-8331.1810-0421

• 模式识别与人工智能 • 上一篇    下一篇

CNN多层特征融合与ELM的乳腺疾病诊断方法

赵京霞,钱育蓉,南方哲,张晗,行艳妮   

  1. 新疆大学 软件学院,乌鲁木齐 830046
  • 出版日期:2020-02-15 发布日期:2020-03-06

Method with CNN Multi-Layer Feature Fusion and ELM Diagnosis for Breast Diseases

ZHAO Jingxia, QIAN Yurong, NAN Fangzhe, ZHANG Han, XING Yanni   

  1. College of Software, Xinjiang University, Urumqi 830046, China
  • Online:2020-02-15 Published:2020-03-06

摘要:

针对传统计算机辅助诊断方法准确率低、耗时长的问题,提出卷积神经网络(Convolutional Neural Networks,CNN)多层特征融合与极限学习机(Extreme Learning Machine,ELM)的乳腺疾病诊断方法。利用CNN从乳腺X光图像中提取多层特征;提出多尺度池化操作将各层提取的特征进行融合;使用极限学习机分类器进行乳腺疾病的快速诊断。实验结果表明,该乳腺疾病检测方法平均准确率高达97.13%,诊断时间是6.43 ms。该方法能有效地提高乳腺疾病诊断的准确率,缩短诊断时间,且具有较好的鲁棒性和泛化能力。

关键词: 多特征融合, 多尺度池化, 极限学习机, 乳腺疾病诊断

Abstract:

Aiming at the problems of low accuracy and long time-consuming in traditional computer-aided diagnosis, a new method of breast disease diagnosis based on Convolutional Neural Networks(CNN) multi-layer feature fusion and Extreme Learning Machine(ELM) is proposed. CNN is used to extract multi-layer features from mammograms, multi-scale pooling is used to fuse the features extracted from each layer, and extreme learning machine classifier is used to diagnose breast diseases quickly. The experimental results show that the average accuracy of the proposed method is 97.13%, and the diagnosis time is 6.43 ms. This method can effectively improve the accuracy of breast disease diagnosis, shorten the diagnosis time, and has good robustness and generalization ability.

Key words: multi-feature?fusion, multi-scale pooling, extreme learning machine, diagnosis of breast diseases