《计算机应用》唯一官方网站 ›› 2022, Vol. 42 ›› Issue (1): 302-309.DOI: 10.11772/j.issn.1001-9081.2021020258

• 前沿与综合应用 • 上一篇    

基于改进VGG网络的弱监督细粒度阿尔兹海默症分类方法

邓爽(), 何小海, 卿粼波, 陈洪刚, 滕奇志   

  1. 四川大学 电子信息学院,成都 610065
  • 收稿日期:2021-02-22 修回日期:2021-04-28 接受日期:2021-04-29 发布日期:2021-05-12 出版日期:2022-01-10
  • 通讯作者: 邓爽
  • 作者简介:邓爽 (1995—),女,四川绵阳人,硕士研究生,主要研究方向:图像处理、模式识别、人工智能
    何小海(1964—),四川绵阳人,教授,博士生导师,博士,主要研究方向:图像处理、模式识别、图像通信
    卿粼波(1982—),男,四川简阳人,副教授,博士,主要研究方向:图像处理、模式识别、视频通信
    陈洪刚(1991—),男,四川成都人,助理研究员,博士,主要研究方向:图像/视频理解、复原及压缩编码
    滕奇志(1961—),女,四川成都人,教授,博士,主要研究方向:数字图像处理、模式识别、三维图像重建及分析。
  • 基金资助:
    成都市重大科技应用示范项目(2019-YF09-00120-SN)

Weakly supervised fine-grained classification method of Alzheimer’s disease based on improved visual geometry group network

Shuang DENG(), Xiaohai HE, Linbo QING, Honggang CHEN, Qizhi TENG   

  1. College of Electronics and Information Engineering,Sichuan University,Chengdu Sichuan 610065,China
  • Received:2021-02-22 Revised:2021-04-28 Accepted:2021-04-29 Online:2021-05-12 Published:2022-01-10
  • Contact: Shuang DENG
  • About author:DENG Shuang, born in 1995, M. S. candidate. Her research interests include image processing, pattern recognition, artificial intelligence.
    HE Xiaohai, born in 1964, Ph. D., professor. His research interests include image processing, pattern recognition, image communication.
    QING Linbo, born in 1982, Ph. D., associate professor. His research interests include image processing, pattern recognition, video communication.
    CHEN Honggang, born in 1991, Ph. D., research assistant. His research interests include image/video understanding, restoration and compression coding.
    TENG Qizhi , born in 1961, Ph. D., professor. Her research interests include digital image processing, pattern recognition, 3D image reconstruction and analysis.
  • Supported by:
    Chengdu Major Science and Technology Application Demonstration Project(2019-YF09-00120-SN)

摘要:

针对阿尔兹海默症(AD)患者和正常(NC)人之间核磁共振成像(MRI)图像差别小、分类难度大的问题,提出了基于改进VGG网络的弱监督细粒度AD分类方法。该方法以弱监督数据增强网络(WSDAN)为基本模型,主要由弱监督注意力学习模块、数据增强模块及双线性注意力池化模块等构成。首先,通过弱监督力注意学习模块生成特征图和注意力图,并利用注意力图引导数据增强,将原图和增强后的数据同时作为输入数据进行训练;然后,通过双线性注意力池化算法将特征图和注意力图按元素进行点乘,进而得到特征矩阵;最后,将特征矩阵作为线性分类层的输入。将以VGG19作为特征提取网络的WSDAN基本模型应用到AD的MRI数据上,实验结果表明,仅使用图像增强的模型的准确性、敏感性和特异性分别比WSDAN基本模型提高了1.6个百分点、0.34个百分点和0.12个百分点;仅利用VGG19网络的改进的模型的准确性和特异性相较WSDAN基本模型分别提高了0.7个百分点和2.82个百分点;以上两个方法结合使用的模型与WSDAN基本模型相比,准确性、敏感性和特异性分别提高了2.1个百分点、1.91个百分点和2.19个百分点。

关键词: 改进VGG网络, 弱监督, 细粒度分类, 数据增强, 阿尔兹海默症

Abstract:

In order to solve the problems of small difference of Magnetic Resonance Imaging (MRI) images between Alzheimer’s Disease (AD) patients and Normal Control (NC) people and great difficulty in classification of them, a weakly supervised fine-grained classification method for AD based on improved Visual Geometry Group (VGG) network was proposed. In this method, Weakly Supervised Data Augmentation Network (WSDAN) was took as the basic model, which was mainly composed of weakly supervised attention learning module, data augmentation module and bilinear attention pooling module. Firstly, the feature map and the attention map were generated through weakly supervised attention learning network, and the attention map was used to guide the data augmentation. Both the original image and the augmented data were used as the input data for training. Then, point production between the feature map and the attention map was performed by elements via bilinear attention pooling algorithm to obtain the feature matrix. Finally, the feature matrix was used as the input of the linear classification layer. Experimental results of applying WSDAN basic model with VGG19 as feature extraction network on MRI data of AD show that, compared with the WSDAN basic model, the proposed model only with image enhancement has the accuracy, sensitivity and specificity increased by 1.6 percentage points, 0.34 percentage points and 0.12 percentage points respectively; the model only using the improvement of VGG19 network has the accuracy and specificity improved by 0.7 percentage points and 2.82 percentage points respectively; the model combing the two methods above has the accuracy, sensitivity and specificity improved by 2.1 percentage points, 1.91 percentage points and 2.19 percentage points respectively.

Key words: improved Visual Geometry Group (VGG) network, weakly supervised, fine-grained classification, data augmentation, Alzheimer’s Disease (AD)

中图分类号: