计算机工程 ›› 2019, Vol. 45 ›› Issue (7): 258-263.doi: 10.19678/j.issn.1000-3428.0052132

• 图形图像处理 • 上一篇    下一篇

基于RV-FCN的CT肝脏影像自动分割算法

张杰妹, 杨词慧   

  1. 南昌航空大学 信息工程学院, 南昌 330063
  • 收稿日期:2018-07-16 修回日期:2018-08-27 出版日期:2019-07-15 发布日期:2019-07-23
  • 作者简介:张杰妹(1992-),女,硕士研究生,主研方向为图像处理、深度学习;杨词慧(通信作者),副教授、博士。
  • 基金项目:
    国家自然科学基金青年项目(61402218);江西省教育厅科学基金(GJJ180516);江西省图像处理与模式识别重点实验室开放基金(TX201304002)。

Automatic Segmentation Algorithm of CT Liver Image Based on RV-FCN

ZHANG Jiemei, YANG Cihui   

  1. School of Information Engineering, Nanchang Hangkong University, Nanchang 330063, China
  • Received:2018-07-16 Revised:2018-08-27 Online:2019-07-15 Published:2019-07-23

摘要: 由于肝脏的大小、形状因人而异,且CT影像中肝脏与其毗邻器官的灰度对比值较低,难以精准地判断肝脏影像的边界信息。为此,提出一种基于全卷积神经网络(FCN)的改进算法,在FCN的基础上引入残差和VGG-16网络,得到肝脏影像的初始分割结果。引入批归一化和PReLU激活函数,提高网络的泛化能力和收敛速度。采用条件随机场方法,进一步优化分割结果,提高分割准确率。通过VTK和ITK系统对二维肝脏影像进行三维重建。在3DIRCADb数据集上的实验结果验证了该算法的有效性和高效性。

关键词: 肝脏分割, 全卷积神经网络, 残差网络, 批归一化, 条件随机场

Abstract: Since the size and shape of the liver vary from person to person,and the grayscale contrast value of the liver and its adjacent organs in the CT image is low,it is difficult to accurately determine the boundary information of the liver image.Aiming at these problems,this paper proposes an improved algorithm based on Fully Convolutional neural Network (FCN).Based on the FCN,the residual and VGG-16 networks are introduced to obtain the initial segmentation result of the liver.The Batch Normalization (BN) and PReLU activation functions are introduced to improve the generalization ability and convergence speed of the network.Conditional Random Field (CRF) method is used to further optimize the segmentation result and improve the segmentation accuracy.The 2-dimensional liver segmentation result is reconstructed into a 3-dimensional structure by the system of VTK and ITK.The effectiveness and efficiency of the algorithm are verified by the experimental results on the 3DIRCADb date set.

Key words: liver segmentation, Fully Convolutional neural Network(FCN), residual network, Batch Normalization(BN), Conditional Random Field(CRF)

中图分类号: