Author Login Editor-in-Chief Peer Review Editor Work Office Work

Computer Engineering ›› 2021, Vol. 47 ›› Issue (8): 308-314. doi: 10.19678/j.issn.1000-3428.0058866

• Development Research and Engineering Application • Previous Articles     Next Articles

Liver Cirrhosis Recognition Based on Spatial Channel Squeeze Excitation Module

WANG Qian1, ZHAO Ximei1,2   

  1. 1. College of Computer Science and Technology, Qingdao University, Qingdao, Shandong 266071, China;
    2. Shandong Provincial Key Laboratory of Digital Medicine and Computer Aided Surgery, Qingdao, Shandong 266071, China
  • Received:2020-07-08 Revised:2020-08-10 Published:2020-08-13

基于空间通道挤压激励模块的肝硬化识别

王倩1, 赵希梅1,2   

  1. 1. 青岛大学 计算机科学技术学院, 山东 青岛 266071;
    2. 山东省数字医学与计算机辅助手术重点实验室, 山东 青岛 266071
  • 作者简介:王倩(1994-),女,硕士研究生,主研方向为医学图像处理;赵希梅,副教授、博士。
  • 基金资助:
    国家自然科学基金(61303079)。

Abstract: The existing Convolutional Neural Networks(CNN) fail to fully learn the feature information, and provides low accuracy in recognition as well as classification. To address the problem, a scSE_MVGG network using spatial channel squeeze excitation module is proposed and applied to liver cirrhosis recognition. The liver cirrhosis images are enhanced to avoid over fitting phenomenon in deep learning training, and the VGG network is improved to adapt to experimental samples of different sizes. At the same time, the scSE module is fused with the improved MVGG network to assist the CNN in the learning of channel information and spatial feature information. So the directivity of extracted network features is enhanced, and thus the performance of CNN in liver cirrhosis recognition is improved. Experimental results show that the accuracy of the proposed network in liver cirrhosis image recognition reaches 98.78%, higher than that of scSE_VGG, scSE_AlexNet and other networks.

Key words: liver cirrhosis recognition, spatial channel squeeze excitation module, Convolutional Neural Network(CNN), VGG network, spatial channel relation

摘要: 针对卷积神经网络对特征信息学习不全面、识别准确率和分类精度不高的问题,提出一种采用空间通道挤压激励模块的scSE_MVGG网络,将其应用于肝硬化识别。对肝硬化图像进行数据增强,以避免深度学习训练出现过拟合现象,改进VGG网络使其适应不同实验样本尺寸,同时将scSE模块与改进的MVGG网络相融合,通过提高网络提取特征的指向性增强肝硬化识别效果。实验结果表明,该网络对肝硬化图像的识别率达到98.78%,较scSE_VGG、scSE_AlexNet等网络识别效果更优。

关键词: 肝硬化识别, 空间通道挤压激励模块, 卷积神经网络, VGG网络, 空间通道关系

CLC Number: