[1]计亚荣,王瑜,肖洪兵,等.基于TensorMixup的脑胶质瘤全自动分割[J].中国医学物理学杂志,2022,39(12):1502-1508.[doi:DOI:10.3969/j.issn.1005-202X.2022.12.008]
　JI Yarong,WANG Yu,XIAO Hongbing,et al.Fully automated glioma segmentation based on TensorMixup[J].Chinese Journal of Medical Physics,2022,39(12):1502-1508.[doi:DOI:10.3969/j.issn.1005-202X.2022.12.008]
点击复制

基于TensorMixup的脑胶质瘤全自动分割()
分享到：

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

卷:
39卷
期数:
2022年第12期
页码:
1502-1508
栏目:
医学影像物理
出版日期:
2022-12-25

文章信息/Info

Title:
Fully automated glioma segmentation based on TensorMixup
文章编号:
1005-202X（2022）12-1502-08
作者:
计亚荣; 王瑜; 肖洪兵; 邢素霞
北京工商大学人工智能学院， 北京 100048
Author(s):
JI Yarong;  WANG Yu;  XIAO Hongbing;  XING Suxia
School of Artificial Intelligence, Beijing Technology and Business University, Beijing 100048, China
关键词:
脑胶质瘤; TensorMixup; Mixup; 数据增强; 深度学习
Keywords:
Keywords: glioma TensorMixup Mixup data augmentation deep learning
分类号:
R318；TP391
DOI:
DOI:10.3969/j.issn.1005-202X.2022.12.008
文献标志码:
A
摘要:
脑胶质瘤及其子区域的全自动分割对临床脑胶质瘤患者的诊断、治疗与病情监控具有重要意义。本文改进传统Mixup方法，提出TensorMixup模型，并将其应用于三维U-Net脑胶质瘤分割任务。算法核心思想包括，首先从两位患者相同模态的核磁共振脑影像中分别获取肿瘤区域所在边界框的图像序列，并从获取的图像序列中选取尺寸为128×128×128体素的图像块，然后使用一个所有元素均独立采样于贝塔分布的张量，混合图像块的信息，接着将上述张量映射为矩阵，用于混合图像块的独热编码标签序列，从而合成新图像及其标注数据，最后使用合成数据训练模型，以提高模型的分割精度。在BraTs2019数据集的测试结果显示，本文算法在完整肿瘤、肿瘤核心与增强肿瘤区域的平均Dice值依次可达91.32%、85.67%与82.20%，证明使用TensorMixup进行脑胶质瘤分割，具有可行性与有效性。
Abstract:
Abstract: The automated segmentation of glioma and its subregions is of great significance for the diagnosis, treatment and monitoring of brain cancer. A data augmentation method called TensorMixup is proposed based on the improvement of traditional Mixup, and it is applied to the three-dimensional U-Net for glioma segmentation. The image patches with the size of 128×128×128 voxels are selected according to glioma information of ground truth labels from the magnetic resonance imaging data of any two patients with the same modality. A tensor in which all elements are independently sampled from Beta distribution is used to mix the image patches, and then the tensor is mapped to a matrix which is used to mix the one-hot encoded labels of the above image patches, so as to synthesize a new image and obtain its encoded labels. The model is trained using the synthetic data for improving the segmentation accuracy. The experiment on the proposed algorithm on BraTs2019 data set shows that the mean Dice coefficients are 91.32%, 85.67%, and 82.20% in the whole tumor, tumor core, and enhancing tumor region, respectively, which indicates that TensorMixup is feasible and effective for glioma segmentation.

相似文献/References:

[1]周 钢,田 野,陆雪官,等.同步加量技术应用于脑胶质瘤术后调强放疗的剂量学研究[J].中国医学物理学杂志,2014,31(02):4727.[doi:10.3969/j.issn.1005-202X.2014.02.002]
[2]戴红娅,黄江华,陈露,等. RapidArc和IMRT在脑胶质瘤术后放疗中保护海马的剂量学比较[J].中国医学物理学杂志,2018,35(12):1404.[doi:DOI:10.3969/j.issn.1005-202X.2018.12.007]
　DAI Hongya,HUANG Jianghua,CHEN Lu,et al. Dosimetric comparison of RapidArc and IMRT in hippocampus sparing during postoperative radiotherapy for glioma[J].Chinese Journal of Medical Physics,2018,35(12):1404.[doi:DOI:10.3969/j.issn.1005-202X.2018.12.007]
[3]郭飞宝.高分级脑胶质瘤4种同步加量调强放射治疗的剂量学分析[J].中国医学物理学杂志,2019,36(12):1396.[doi:DOI:10.3969/j.issn.1005-202X.2019.12.006]
　GUO Feibao.Dosimetric analysis of 4 types of simultaneous integrated boost radiotherapy for high-grade glioma[J].Chinese Journal of Medical Physics,2019,36(12):1396.[doi:DOI:10.3969/j.issn.1005-202X.2019.12.006]
[4]段欢欢,李书舟,曹瑛,等.基于深度学习方法预测IMRT计划射野的γ通过率[J].中国医学物理学杂志,2021,38(6):677.[doi:DOI:10.3969/j.issn.1005-202X.2021.06.004]
　. School of Nuclear Science and Technology,University of South China,Hengyang 00,et al.Predicting gamma passing rates for intensity-modulated radiotherapy fields based on deep learning method[J].Chinese Journal of Medical Physics,2021,38(12):677.[doi:DOI:10.3969/j.issn.1005-202X.2021.06.004]
[5]黄唯,徐中标,邓官华,等.MRI拉莫尔频率范围内人体脑胶质瘤组织的介电特性[J].中国医学物理学杂志,2021,38(12):1538.[doi:DOI:10.3969/j.issn.1005-202X.2021.12.015]
　HUANG Wei,XU Zhongbiao,DENG Guanhua,et al.Dielectric properties of human glioma tissue at Larmor frequencies in MRI[J].Chinese Journal of Medical Physics,2021,38(12):1538.[doi:DOI:10.3969/j.issn.1005-202X.2021.12.015]
[6]袁红杰,杨艳,张东,等.基于多监督注意力机制神经网络的脑胶质瘤循环肿瘤细胞分割算法[J].中国医学物理学杂志,2022,39(7):828.[doi:DOI:10.3969/j.issn.1005-202X.2022.07.007]
　YUAN Hongjie,YANG Yan,ZHANG Dong,et al.Neural network-based multi-level supervision and attention mechanism algorithm for brain glioma CTC segmentation[J].Chinese Journal of Medical Physics,2022,39(12):828.[doi:DOI:10.3969/j.issn.1005-202X.2022.07.007]
[7]王瑞,刘志强,齐崇,等.基于3D深度残差网络和多模态MRI的脑胶质瘤自动分级[J].中国医学物理学杂志,2022,39(10):1236.[doi:DOI:10.3969/j.issn.1005-202X.2022.10.010]
　WANG Rui,LIU Zhiqiang,QI Chong,et al.Automated glioma grading based on 3D deep residual network and multimodal MRI[J].Chinese Journal of Medical Physics,2022,39(12):1236.[doi:DOI:10.3969/j.issn.1005-202X.2022.10.010]
[8]刘一蓉,张恺,吴剑.脑胶质瘤切除术中辅助检测与精准操作技术概况[J].中国医学物理学杂志,2023,40(2):163.[doi:DOI:10.3969/j.issn.1005-202X.2023.02.006]
　LIU Yirong,ZHANG Kai,et al.Introperative auxiliary detection and precise operation techniques in glioma resection[J].Chinese Journal of Medical Physics,2023,40(12):163.[doi:DOI:10.3969/j.issn.1005-202X.2023.02.006]
[9]游慧霞,张怀岺.深度学习和影像组学在脑胶质瘤诊疗中的研究进展[J].中国医学物理学杂志,2023,40(12):1502.[doi:DOI:10.3969/j.issn.1005-202X.2023.12.008]
　YOU Huixia,ZHANG Huailing.Deep learning and radiomics in diagnosis and treatment of glioma: a review[J].Chinese Journal of Medical Physics,2023,40(12):1502.[doi:DOI:10.3969/j.issn.1005-202X.2023.12.008]

备注/Memo

备注/Memo:
【收稿日期】2022-06-20 【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】计亚荣，硕士研究生，主要从事图像处理与机器学习的研究，E-mail: 2470214219@qq.com 【通信作者】王瑜，博士，教授，主要从事图像处理与模式识别的研究，E-mail: wangyu@btbu.edu.cn
更新日期/Last Update: 2022-12-23