[1]田恒屹,肖洪兵,计亚荣,等.基于三维UNet与混合焦点损失函数的脑肿瘤全自动分割算法[J].中国医学物理学杂志,2023,40(9):1114-1120.[doi:DOI:10.3969/j.issn.1005-202X.2023.09.009]
　TIAN Hengyi,XIAO Hongbing,JI Yarong,et al.Fully automatic segmentation algorithm of brain tumor based on three-dimensional UNet and hybrid focal loss function[J].Chinese Journal of Medical Physics,2023,40(9):1114-1120.[doi:DOI:10.3969/j.issn.1005-202X.2023.09.009]
点击复制

基于三维UNet与混合焦点损失函数的脑肿瘤全自动分割算法()
分享到：

《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]

卷:
40卷
期数:
2023年第9期
页码:
1114-1120
栏目:
医学影像物理
出版日期:
2023-09-26

文章信息/Info

Title:
Fully automatic segmentation algorithm of brain tumor based on three-dimensional UNet and hybrid focal loss function
文章编号:
1005-202X（2023）09-1114-07
作者:
田恒屹; 肖洪兵; 计亚荣; Rahman Md Mostafizur
北京工商大学人工智能学院， 北京 100048
Author(s):
TIAN Hengyi;  XIAO Hongbing;  JI Yarong;  Rahman Md Mostafizur
School of Artificial Intelligence, Beijing Technology and Business University, Beijing 100048, China
关键词:
脑肿瘤; 深度学习; 3D UNet; 混合焦点损失函数
Keywords:
Keywords: brain tumor deep learning 3D UNet hybrid focus loss function
分类号:
R318；TP317.4
DOI:
DOI:10.3969/j.issn.1005-202X.2023.09.009
文献标志码:
A
摘要:
针对脑肿瘤分割方法中由于正常脑组织、脑肿瘤等不同类别的数据量严重不平衡，导致分割精度受到极大影响的问题，提出一种结合混合焦点损失函数与三维UNet（3D UNet）的全自动脑肿瘤分割算法。在3D UNet模型框架中，使用包含焦点损失与改进的焦点Tversky损失的混合损失函数， 两种损失函数可以优势互补，分别缓解输入与输出数据类不平衡带来的不利影响，使分割模型聚焦在难以分类和学习的样本上。利用公开的脑肿瘤数据集进行相关实验，提出的混合焦点损失函数分割模型在完整肿瘤区域、核心肿瘤区域（TC）和增强肿瘤区域（ET）的Dice均值分别可达89.01%、88.67%与83.74%，豪斯多夫距离均值分别为14.29、5.01与3.84 mm，实验结果表明，基于混合损失函数的深度学习分割模型可以显著提升由于数据类不平衡导致的难以分类区域（TC和ET）的分割效果。
Abstract:
Abstract: A fully automatic brain tumor segmentation algorithm combining hybrid focus loss function and 3D UNet is proposed to address the problem of low segmentation accuracy caused by the serious data imbalance of different categories such as normal brain tissue and brain tumor. In 3D UNet framework, the hybrid focus loss function including focus loss and improved focus Tversky loss which possess complementary advantages can effectively alleviate the adverse effects caused by the class imbalance of input and output data, and make the segmentation model focus on the samples which are difficult to be classified and learned. The experiments on public brain tumor dataset show that the proposed segmentation model with hybrid focal loss function can achieve a mean Dice coefficient of 89.01%, 88.67%, and 83.74% in whole tumor, tumor core and enhanced tumor, respectively, and that the mean Hausdorff distances were 14.29, 5.01 and 3.84 mm, respectively. The segmentation model based on hybrid focal loss function can significantly improve the segmentation accuracy of tumor core and enhanced tumor which are difficult to be classified due to class imbalance.

相似文献/References:

[1]王晓飞,聂生东,王远军,等.基于MRI的脑肿瘤计算机辅助检测技术研究进展[J].中国医学物理学杂志,2014,31(01):4635.[doi:10.3969/j.issn.1005-202X.2014.01.010]
[2]万俊,聂生东,王远军,等.基于MRI的脑肿瘤分割技术研究进展[J].中国医学物理学杂志,2013,30(04):4266.[doi:10.3969/j.issn.1005-202X.2013.04.013]
[3]赖胜圣,刘虔铖,张刚平.基于模糊C均值自动随机游走算法在脑肿瘤分割中的应用[J].中国医学物理学杂志,2015,32(05):707.[doi:doi:10.3969/j.issn.1005-202X.2015.05.021]
[4]谷珊珊,张怀文,王运来.基于水平集稳健特征统计算法的脑肿瘤自动分割研究[J].中国医学物理学杂志,2016,33(1):63.[doi:10.3969/j.issn.1005-202X.2016.01.014]
　[J].Chinese Journal of Medical Physics,2016,33(9):63.[doi:10.3969/j.issn.1005-202X.2016.01.014]
[5]徐艳红,张晋建,黎静,等.影像增强剂对脑瘤螺旋断层放疗剂量计算的影响[J].中国医学物理学杂志,2016,33(8):757.[doi:10.3969/j.issn.1005-202X.2016.08.001]
　[J].Chinese Journal of Medical Physics,2016,33(9):757.[doi:10.3969/j.issn.1005-202X.2016.08.001]
[6]宋婷婷,盛晓芳,汪洋,等.Compass剂量验证系统在VMAT脑肿瘤瘤床同步追量中的应用[J].中国医学物理学杂志,2016,33(10):1012.[doi:10.3969/j.issn.1005-202X.2016.10.008]
　[J].Chinese Journal of Medical Physics,2016,33(9):1012.[doi:10.3969/j.issn.1005-202X.2016.10.008]
[7]陶源,王佳飞,杜俊龙,等.基于卷积神经网络的细胞识别[J].中国医学物理学杂志,2017,34(1):53.[doi:10.3969/j.issn.1005-202X.2017.01.011]
　[J].Chinese Journal of Medical Physics,2017,34(9):53.[doi:10.3969/j.issn.1005-202X.2017.01.011]
[8]门阔,戴建荣. 利用深度反卷积神经网络自动勾画放疗危及器官[J].中国医学物理学杂志,2018,35(3):256.[doi:DOI:10.3969/j.issn.1005-202X.2018.03.002]
　MEN Kuo,DAI Jianrong. Automatic segmentation of organs at risk in radiotherapy using deep deconvolutional neural network[J].Chinese Journal of Medical Physics,2018,35(9):256.[doi:DOI:10.3969/j.issn.1005-202X.2018.03.002]
[9]邓金城,彭应林,刘常春,等. 深度卷积神经网络在放射治疗计划图像分割中的应用[J].中国医学物理学杂志,2018,35(6):621.[doi:DOI:10.3969/j.issn.1005-202X.2018.06.001]
　DENG Jincheng,PENG Yinglin,LIU Changchun,et al. Application of deep convolution neural network in radiotherapy planning image segmentation[J].Chinese Journal of Medical Physics,2018,35(9):621.[doi:DOI:10.3969/j.issn.1005-202X.2018.06.001]
[10]查雪帆,杨丰,吴俣南,等. 结合迁移学习与深度卷积网络的心电分类研究[J].中国医学物理学杂志,2018,35(11):1307.[doi:DOI:10.3969/j.issn.1005-202X.2018.11.013]
　ZHA Xuefan,YANG Feng,WU Yunan,et al. ECG classification based on transfer learning and deep convolution neural network[J].Chinese Journal of Medical Physics,2018,35(9):1307.[doi:DOI:10.3969/j.issn.1005-202X.2018.11.013]
[11]方新林,方艳红,王迪.基于多模态特征融合的脑瘤图像分割方法[J].中国医学物理学杂志,2022,39(6):682.[doi:DOI:10.3969/j.issn.1005-202X.2022.06.005]
　FANG Xinlin,FANG Yanhong,WANG Di.Brain tumor image segmentation method based on multi-modal feature fusion[J].Chinese Journal of Medical Physics,2022,39(9):682.[doi:DOI:10.3969/j.issn.1005-202X.2022.06.005]

备注/Memo

备注/Memo:
【收稿日期】2023-02-24 【基金项目】北京市自然科学基金-北京市教育委员会科技计划重点项目（KZ202110011015） 【作者简介】田恒屹，硕士，研究方向：图像处理与模式识别，E-mail: thy8562@sina.com 【通信作者】肖洪兵，博士，副教授，研究生导师，研究方向：图像处理与模式识别，E-mail: x.hb@163.com
更新日期/Last Update: 2023-09-26