文献知网节
  • 记笔记

Brain Tumors Segmentation Based on 3D Dilated-U-Net

Yong Xu1Xinyu Zhao1Xiaoyu He1Jinxin Liu2Xiangfeng Chang2Bin Xie1

1. School of Automation,Central South University2. Mobile Health Ministry of Education-China Mobile Joint Laboratory

摘要:Magnetic resonance imaging(MRI) is a prevailing method to the clinical diagnosis of brain tumors.However,manual brain tumors segmentation from multimodal MRI images is a challenging and time-consuming task.In this paper,we proposed a deep convolutional neural network,namely 3 D Dilated-U-Net,which make automatic segmentation of brain tumors perform well.The network uses dilated convolution to reduce accuracy loss caused by pooling process.The dilated convolution expands the receptive field of the convolution layer to reduce the number of down-sampling layers.To improve the performance on the category imbalance problem,we proposes adaptive Dice Loss(A-Dice Loss) based on "Dice Loss".ADice Loss achieves better results when segmenting small areas.The Dice’s coefficient(DSC) of each region is higher than the traditional U-Net method on Brats 2017 dataset.
会议名称:

2019计算智能、工程与信息技术世界大会

会议时间:

2019-06-29

会议地点:

中国上海

  • 专辑:

    医药卫生科技; 信息科技

  • 专题:

    神经病学; 肿瘤学; 计算机软件及计算机应用; 自动化技术

  • DOI:

    10.26914/c.cnkihy.2019.048729

  • 分类号:

    R739.41;TP391.41;TP183

  • 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

  • CAJ下载
  • PDF下载

下载手机APP用APP扫此码同步阅读该篇文章

下载:4 页码:531-537 页数:7 大小:456k

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