计算机应用 ›› 2020, Vol. 40 ›› Issue (7): 2104-2109.DOI: 10.11772/j.issn.1001-9081.2019122233

• 虚拟现实与多媒体计算 • 上一篇    下一篇

基于同一特征空间的多模态脑肿瘤分割方法

陈浩, 秦志光, 丁熠   

  1. 电子科技大学 信息与软件工程学院, 成都 610054
  • 收稿日期:2020-01-07 修回日期:2020-02-18 出版日期:2020-07-10 发布日期:2020-04-23
  • 通讯作者: 秦志光
  • 作者简介:陈浩(1991-),男,山东济南人,博士研究生,主要研究方向:深度神经网络、医学图像分割;秦志光(1956-),男,四川成都人,教授,博士生导师,博士,主要研究方向:网络安全、数据挖掘、深度学习、医学图像处理;丁熠(1985-),男,四川成都人,副教授,博士,主要研究方向:深度学习、医学图像处理。
  • 基金资助:
    国家自然科学基金广东联合基金资助项目(U1401257)。

Multi-modal brain tumor segmentation method under same feature space

CHEN Hao, QIN Zhiguang, DING Yi   

  1. School of Information and Software Engineering, University of Electronic Science and Technology of China, Chengdu Sichuan 610054, China
  • Received:2020-01-07 Revised:2020-02-18 Online:2020-07-10 Published:2020-04-23
  • Supported by:
    This work is partially supported by the Guangdong Joint Fund of National Natural Science Foundation of China (U1401257).

摘要: 脑胶质瘤的分割依赖多种模态的核磁共振成像(MRI)的影像。基于卷积神经网络(CNN)的分割算法往往是在固定的多种模态影像上进行训练和测试,这忽略了模态数据缺失或增加问题。针对这个问题,提出了将不同模态的图像通过CNN映射到同一特征空间下并利用同一特征空间下的特征来分割肿瘤的方法。首先,不同模态的数据经过同一深度CNN提取特征;然后,将不同模态的特征连接起来,经过全连接层实现特征融合;最后,利用融合的特征实现脑肿瘤分割。模型采用BRATS2015数据集进行训练和测试,并使用Dice系数对模型进行验证。实验结果表明了所提模型能有效缓解数据缺失问题。同时,该模型较多模态联合的方法更加灵活,能够应对模态数据增加问题。

关键词: 多模态, 脑肿瘤分割, 同一特征空间, 卷积神经网络, 核磁共振成像, 数据缺失

Abstract: Glioma segmentation depends on multi-modal Magnetic Resonance Imaging (MRI) images. Convolutional Neural Network (CNN)-based segmentation algorithms are often trained and tested on fixed multi-modal images, which ignores the problem of missing or increasing of modal images. To solve this problem, a method mapping images of different modalities to the same feature space by CNN and using the features in the same feature space to segment tumors was proposed. Firstly, the features of different modalities were extracted through the same deep CNN. Then, the features of different modal images were concatenated, and passed through the fully connected layer to realize the feature fusion. Finally, the fused features were used to segment the brain tumor. The proposed model was trained and tested on the BRATS2015 dataset, and verified with the Dice coefficient. The experimental results show that, the proposed model can effectively alleviate the problem of data missing. At the same time, compared with multi-modal joint method, this model is more flexible, and can deal with the problem of modal data increasing.

Key words: multi-modal, brain tumor segmentation, same feature space, convolutional neural network, Magnetic Resonance Imaging (MRI), data missing

中图分类号: