文献知网节
摘要:The main manifestations of pneumonia are ground glass shadows and ground glass shadows on the X-rays of the lungs. Therefore, the research on pneumonia recognition is basically focused on the recognition of glass shadows.In this paper, according to the characteristics of pneumonia image in the existing data set, the author also does image registration, enhancement, filtering and denoising, data amplification, binarization, normalization, and other preprocessing on the data set. Based on the VGG16 network, each convolutional layer and a fully connected layer of the supplementary network are used before dating a batch normalization layer. A deep convolutional neural network BNnet with a more complex network level is obtained. And a feature extraction of pneumonia images is associated with it. When training the model, use the migration learning method to date VGG16, combined with the addition of a network layer to train the Kaggle Data Company’s data set, and input the extracted features into a classifier composed of a fully connected layer. Finally, these images are divided into two categories: normal images and pneumonia images. The accuracy rate can reach0.98, which effectively solves the over-fitting problem caused by the imbalance of sample data. At the same time, the mainstream neural network model is dispersed,and its accuracy, sensitivity, and specificity have been improved, and it has better robustness sexuality and generalization.
会议名称:

The 11th International Conference on Computer Engineering and Networks(CENet2021)

会议时间:

2021-10-21

会议地点:

中国广西河池

  • 专辑:

    医药卫生科技; 信息科技

  • 专题:

    呼吸系统疾病; 计算机软件及计算机应用; 自动化技术

  • DOI:

    10.26914/c.cnkihy.2021.045051

  • 分类号:

    R563.1;TP391.41;TP18

  • 手机阅读
    即刻使用手机阅读
    第一步

    扫描二维码下载

    "移动知网-全球学术快报"客户端

    第二步

    打开“全球学术快报”

    点击首页左上角的扫描图标

    第三步

    扫描二维码

    手机同步阅读本篇文献

  • HTML阅读
  • CAJ下载
  • PDF下载

下载手机APP用APP扫此码同步阅读该篇文章

温馨提示:阅读CAJ格式原文,请使用CAJ浏览器

下载:17 页码:504-513 页数:10 大小:1045k

引文网络
  • 参考文献
  • 引证文献
  • 共引文献
  • 同被引文献
  • 二级参考文献
  • 二级引证文献
  • 批量下载
相关推荐
  • 相似文献
  • 读者推荐
  • 相关基金文献
  • 相关法规
  • 关联作者
  • 相关视频