[1]江慧敏,方立铭,陶龙.基于分级注意力的多示例口腔癌病理分类[J].中国医学物理学杂志,2024,41(8):946-952.[doi:DOI:10.3969/j.issn.1005-202X.2024.08.004]
　JIANG Huimin,FANG Liming,TAO Long.Pathological classification of oral cancer based on multi-instance network and two-level attention[J].Chinese Journal of Medical Physics,2024,41(8):946-952.[doi:DOI:10.3969/j.issn.1005-202X.2024.08.004]
点击复制

基于分级注意力的多示例口腔癌病理分类()
分享到：

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

卷:
41卷
期数:
2024年第8期
页码:
946-952
栏目:
医学影像物理
出版日期:
2024-08-31

文章信息/Info

Title:
Pathological classification of oral cancer based on multi-instance network and two-level attention
文章编号:
1005-202X（2024）08-0946-07
作者:
江慧敏; 方立铭; 陶龙
皖南医学院医学影像学院， 安徽 芜湖 241000
Author(s):
JIANG Huimin;  FANG Liming;  TAO Long
Department of Medical Imaging, Wannan Medical College, Wuhu 241000, China
关键词:
口腔癌; 多示例学习; 两级注意力; 病理图像
Keywords:
Keywords: oral cancer multi-instance learning two-level attention pathological image
分类号:
R780.2；R318
DOI:
DOI:10.3969/j.issn.1005-202X.2024.08.004
文献标志码:
A
摘要:
针对病理数据超大尺寸、标注成本高昂等问题带来的病理分类准确率较低的问题，基于多示例网络，引入分级注意力模块，设计一种兼顾示例级和图像块级损失的病理分类算法。回顾性收集皖南医学院第一附属医院口腔颌面外科186例口腔癌（126例鳞癌、60例腺癌），其数字病理切片划分为验证集、测试集及训练集。首先对病理图像进行前后背景分割，去掉背景中的噪声部分，然后采用ResNet50对分割后的病理图像提取特征，并将特征输入第一级注意力网络，得到基于图像块的注意力得分和损失，再根据注意力得分对图像块进行排序重置标签输入第二级注意力网络，得到基于示例级别的损失，最后将两级注意力的损失作为模型的总损失，通过训练最终网络，得到口腔癌分类结果。实验结果表明，使用两级注意力的多示例网络准确率为78.95%，AUC为0.843 0，相较于基线模型均有更高表现。
Abstract:
Abstract: To address the problems of the low accuracy of pathological classification caused by the large size of pathological data and the high cost of labeling, a pathological classification algorithm for oral cancer is designed based on multi-instance network and two-level attention module, which takes losses at the instance level and image block level into account. A retrospective analysis is conducted on 186 cases of oral cancer (126 cases of squamous cell carcinoma and 60 cases of adenocarcinoma) in the Department of Oral and Maxillofacial Surgery, the First Affiliated Hospital of Wannan Medical College, and the digital pathological sections are divided into training set, verification set and test set. The foreground and background segmentations are performed on the pathological image, and the noise is removed from the background. ResNet50 is used to extract features from the segmented pathological images, and the features are input into the first-level attention network to obtain the attention score and loss based on image block. Then, the image blocks are sorted according to the attention score, and the reset labels are input into the second-level attention network to obtain the loss based on the instance level. The loss of the two-level attention is taken as the total loss of the model, and the prediction result is obtained by training the final network. The experimental results show that the multi-instance network using two-level attention achieves an accuracy of 78.95% and AUC of 0.843 0, demonstrating superior performance than the baseline models.

相似文献/References:

[1]王润堃,陆汉强,黄秋生.口腔癌调强放疗中靶区优化对患者口腔黏膜反应、唾液腺功能的影响[J].中国医学物理学杂志,2024,41(2):145.[doi:DOI:10.3969/j.issn.1005-202X.2024.02.003]
　WANG Runkun,LU Hanqiang,HUANG Qiusheng.Effects of optimizing target volume on oral mucosal reaction and salivary gland function in oropharyngeal cancer patients undergoing intensity-modulated radiotherapy[J].Chinese Journal of Medical Physics,2024,41(8):145.[doi:DOI:10.3969/j.issn.1005-202X.2024.02.003]

备注/Memo

备注/Memo:
【收稿日期】2024-02-15 【基金项目】国家自然科学基金（11801199）；安徽省自然科学基金（1908085QA30）；皖南医学院中青年重点科研项目（WK2023ZZD04） 【作者简介】江慧敏，讲师，研究方向：医学图像，E-mail: 814891714@qq.com
更新日期/Last Update: 2024-08-31