[1]徐华畅,许倩,赵钰琳,等.基于AEViT与先验知识的胶质瘤IDH1突变状态预测[J].智能系统学报,2024,19(4):952-960.[doi:10.11992/tis.202209055]
XU Huachang,XU Qian,ZHAO Yulin,et al.Prediction of glioma IDH1 mutation status based on AEViT and prior knowledge[J].CAAI Transactions on Intelligent Systems,2024,19(4):952-960.[doi:10.11992/tis.202209055]
点击复制
XU Huachang,XU Qian,ZHAO Yulin,et al.Prediction of glioma IDH1 mutation status based on AEViT and prior knowledge[J].CAAI Transactions on Intelligent Systems,2024,19(4):952-960.[doi:10.11992/tis.202209055]
基于AEViT与先验知识的胶质瘤IDH1突变状态预测
《智能系统学报》[ISSN 1673-4785/CN 23-1538/TP] 卷:
19
期数:
2024年第4期
页码:
952-960
栏目:
学术论文—智能系统
出版日期:
2024-07-05
- Title:
- Prediction of glioma IDH1 mutation status based on AEViT and prior knowledge
- 关键词:
- 胶质瘤; 异柠檬酸脱氢酶1; K-Means聚类算法; 伪标签; Auto-Encoder; vision Transformer; 果蝇优化算法; 先验知识
- Keywords:
- glioma; IDH1; K-Means clustering algorithm; pseudo-label; Auto-Encoder; vision Transformer; fruit fly optimization algorithm; priori knowledge
- 分类号:
- TP18
- 摘要:
- 针对目前预测胶质瘤异柠檬酸脱氢酶1（isocitrate dehydrogenase1, IDH1）突变状态存在的数据不足、准确率较低等问题，提出一种基于AEViT（auto-encoder vision Transformer）与先验知识的胶质瘤IDH1突变状态预测方法。首先使用改进的K-Means聚类算法为无IDH1突变状态标签的胶质瘤磁共振成像（magnetic resonance imaging，MRI）数据标注伪标签，并采用ViT（vision Transformer）网络对伪标签进行修正，得到最终的胶质瘤IDH1突变状态。为避免不准确的伪标签数据影响模型精度，采用果蝇优化算法为伪标签数据赋予最优权重；然后提出基于Auto-Encoder和ViT的分类模型AEViT，利用Auto-Encoder提取胶质瘤MRI特征；再将特征输入ViT中对胶质瘤IDH1突变状态进行分类；最后将基于胶质瘤位置信息的先验知识加入模型，达到99.01%的预测准确率。结果表明该方法优于其他现有模型，能够实现胶质瘤数据扩增和术前无创、准确地预测胶质瘤IDH1突变状态，从而辅助诊疗过程。
- Abstract:
- Aiming at the problems of insufficient data and low accuracy in predicting the mutation status of brain glioma isocitrate dehydrogenase1 (IDH1), a prediction method for IDH1 mutation status of brain glioma is proposed based on auto-encoder vision Transformer (AEViT) and priori knowledge. Firstly, an improved K-Means clustering algorithm was used to label the pseudo-labels for MRI data of glioma without IDH1 status labels, and vision Transformer (ViT) network was used to modify the pseudo labels to obtain the final glioma IDH1 mutation status. In order to avoid inaccurate pseudo-label data that affect accuracy of the model, the fruit fly optimization algorithm was used to assign the optimal weight to the pseudo-label data. Secondly, a classification model AEViT based on Auto-Encoder and Vision Transformer was proposed, and Auto-Encoder was used to extract MRI features of glioma, and then the features were input into ViT to classify the IDH1 mutation status of glioma. Finally, the prior knowledge based on glioma location information was added to the model, which achieved a prediction accuracy of 99.01%. The experimental results show that this method is superior to other existing models, and can realize glioma data augmentation and non-invasive and accurate preoperative prediction of glioma IDH1 mutation status, thereby assisting the diagnosis and treatment process.
- 参考文献/References:
-
[1] FU Jie, SINGHRAO K, ZHONG Xinran, et al. An automatic deep learning-based workflow for glioblastoma survival prediction using preoperative multimodal MR images: a feasibility study[J]. Advances in radiation oncology, 2021, 6(5): 100746.
[2] ZHANG Maolei, HUANG Nunu, YANG Xuesong, et al. A novel protein encoded by the circular form of the SHPRH gene suppresses glioma tumorigenesis[J]. Oncogene, 2018, 37(13): 1805–1814.
[3] LOUIS D N, PERRY A, REIFENBERGER G, et al. The 2016 world health organization classification of tumors of the central nervous system: a summary[J]. Acta neuropathologica, 2016, 131(6): 803–820.
[4] CHEN Xi, YAN Yuanliang, ZHOU Jianhua, et al. Clinical prognostic value of isocitrate dehydrogenase mutation, O-6-methylguanine-DNA methyltransferase promoter methylation, and 1p19q co-deletion in glioma patients[J]. Annals of translational medicine, 2019, 7(20): 541.
[5] LOUIS D N, ARIE P, PIETER W, et al. The 2021 WHO classification of tumors of the central nervous system: a summary[J]. Neuro-oncology, 2021, 23(8): 1231–1251.
[6] MILLER J J, SHIH H A, ANDRONESI O C, et al. Isocitrate dehydrogenase-mutant glioma: evolving clinical and therapeutic implications[J]. Cancer, 2017, 123(23): 4535–4546.
[7] 李锐, 马林. IDH基因突变与胶质瘤相关性的研究进展[J]. 中国医学影像学杂志, 2020, 28(2): 142–145
LI Rui, MA Lin. Research progress on the correlation between IDH gene mutation and glioma[J]. China journal of medical imaging, 2020, 28(2): 142–145
[8] WANG Shuo, SHI Jingyun, YE Zhaoxiang, et al. Predicting EGFR mutation status in lung adenocarcinoma on computed tomography image using deep learning[J]. European respiratory journal, 2019, 53(3): 1800986.
[9] CHOI K S, CHOI S H, JEONG B. Prediction of IDH genotype in gliomas with dynamic susceptibility contrast perfusion MR imaging using an explainable recurrent neural network[J]. Neuro-oncology, 2019, 21(9): 1197–1209.
[10] BANGALORE Y C, SHAH B R, VEJDANI-JAHROMI M, et al. A novel fully automated MRI-based deep-learning method for classification of IDH mutation status in brain gliomas[J]. Neuro-oncology, 2020, 22(3): 402–411.
[11] CHOI Y S, BAE S, CHANG J H, et al. Fully automated hybrid approach to predict the IDH mutation status of gliomas via deep learning and radiomics[J]. Neuro-oncology, 2021, 23(2): 304–313.
[12] SAFDAR M F, ALKOBAISI S S, ZAHRA F T. A comparative analysis of data augmentation approaches for magnetic resonance imaging (MRI) scan images of brain tumor[J]. Acta informatica medica, 2020, 28(1): 29–36.
[13] LIU Sidong, SHAH Z, SAV A, et al. Isocitrate dehydrogenase (IDH) status prediction in histopathology images of gliomas using deep learning[J]. Scientific reports, 2020, 10(1): 7733.
[14] LI Qingyun, YU Zhibin, WANG Yubo, et al. TumorGAN: a multi-modal data augmentation framework for brain tumor segmentation[J]. Sensors, 2020, 20(15): 4203.
[15] LEE D. Pseudo-Label : the simple and efficient Semi-Supervised learning method for deep neural networks[C]//ICML 2013 Workshop : Challenges in Representation Learning. Atlanta: ACM, 2013, 3(2): 896.
[16] XIE Qizhe, LUONG M T, HOVY E, et al. Self-training with noisy student improves imageNet classification[C]// 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 10684-10695.
[17] CHEN Dongdong, YUAN Lu, LIAO Jing, et al. Explicit filterbank learning for neural image style transfer and image processing[J]. IEEE transactions on pattern analysis and machine intelligence, 2021, 43(7): 2373–2387.
[18] VASWANI Ashish, SHAZEER Noam, PARMAR Niki, et al. Attention is all you need[EB/OL]. (2017-06-12)[2022-07-21]. https://arxiv.org/abs/1706.03762.
[19] DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al. An image is worth 16x16 words: transformers for image recognition at scale[EB/OL]. (2020-10-22)[2022-07-21]. https://arxiv.org/abs/2010.11929.
[20] 徐华畅, 许倩, 赵钰琳, 等. 基于改进EfficientNetV2的脑胶质瘤IDH1突变状态预测方法[J]. 山东大学学报(理学版), 2023, 58(7): 60–66
XU Huachang, XU Qian, ZHAO Yulin, et al. Prediction method of IDH1 mutation status of glioma based on improved EfficientNetV2[J]. Journal of Shandong University (science edition), 2023, 58(7): 60–66
[21] ZHANG Lifei, FRIED D V, FAVE X J, et al. Ibex: an open infrastructure software platform to facilitate collaborative work in radiomics[J]. Medical physics, 2015, 42(3): 1341–1353.
[22] 刘光宇, 黄懿, 曹禹, 等. 基于灰度共生矩阵的图像纹理特征提取研究[J]. 科技风, 2021(12): 61–64
LIU Guangyu, HUANG Yi, CAO Yu, et al. Research on image texture feature extraction based on grayscale cooccurrence matrix[J]. Technology wind, 2021(12): 61–64
[23] 王铭, 田为中, 张继, 等. 基于TIRM序列的游程矩阵纹理特征联合ADC值预测乳腺癌ki-67表达水平[J]. 放射学实践, 2021, 36(12): 1520–1525
WANG Ming, TIAN Weizhong, ZHANG Ji, et al. Prediction of ki-67 expression level in breast cancer based on run-length matrix texture feature combined with ADC value based on TIRM sequence[J]. Radiology practice, 2021, 36(12): 1520–1525
[24] PAN Wen-tsao. A new fruit fly optimization algorithm: taking the financial distress model as an example[J]. Knowledge-based systems, 2012, 26: 69–74.
[25] 陈璟, 虞继敏. 基于果蝇—广义回归神经网络优化的wsn节点定位算法[J]. 南京师大学报(自然科学版), 2017, 40(2): 31–38
CHEN Jing, YU Jimin. A wsn node localization algorithm based on drosophila-generalized regression neural network optimization[J]. Journal of Nanjing Normal University (natural science edition), 2017, 40(2): 31–38
[26] 彭启伟, 罗旺, 冯敏, 等. 改进二维otsu法和果蝇算法结合的图像分割方法[J]. 计算机应用, 2017, 37(S2): 193–197
PENG Qiwei, LUO Wang, FENG Min, et al. Improved image segmentation method by combining two-dimensional otsu method and fruit fly algorithm[J]. Computer applications, 2017, 37(S2): 193–197
[27] ZHU Hong, HE Hanzhi, XU Jinhui, et al. Medical image segmentation using fruit fly optimization and density peaks clustering[J]. Computational and mathematical methods in medicine, 2018, 2018: 1–11.
[28] WIJNENGA M, VOORT S, FRENCH P J, et al. Differences in spatial distribution between WHO 2016 low-grade glioma molecular subgroups[J]. Neuro-oncology advances, 2019, 1(1): z1.
[29] SONODA Y, SHIBAHARA I, KAWAGUCHI T, et al. Association between molecular alterations and tumor location and MRI characteristics in anaplastic gliomas[J]. Brain tumor pathology, 2015, 32(2): 99–104.
[30] 徐芊芊, 许倩, 徐华畅, 等. 基于CnViT的胶质瘤IDH1突变状态智能预测方法[J]. 山东大学学报(工学版), 2023, 53(2): 127–134
XU Qianqian, XU Qian, XU Huachang, et al. Intelligent prediction of glioma IDH1 mutation status based on CnViT[J]. Journal of Shandong University (engineering edition), 2023, 53(2): 127–134
备注/Memo
收稿日期:2022-09-27。
基金项目:江苏省卫生健康委医学科研项目（Z2020032）；徐州市重点研发计划项目（KC22117）；徐州市卫生健康委员会青年医学科技创新项目（XWKYHT20210586）.
作者简介:徐华畅，硕士，主要研究方向为人工智能、深度学习、智能医学图像处理。E-mail：xuhuachang@xzhmu.edu.cn;许倩，副主任医师，博士，江苏省医学会脑卒中分会第一届青年委员会委员，主要研究方向为中枢神经系统疾病的影像诊断。E-mail：xuqianxz@126.com;朱红，教授，博士，主要研究方向为人工智能、深度学习、模式识别、智能医学图像处理。主持或参与教学、科研课题10余项，发表学术论文40余篇。E-mail：zhuhong@xzhmu.edu.cn
通讯作者:朱红. E-mail：zhuhong@xzhmu.edu.cn
更新日期/Last Update:
1900-01-01