Two-stage breast cancer histopathological image classification method based on convolutional neural network and Transformer
10.19745/j.1003-8868.2024227
- VernacularTitle:基于CNN和Transformer的两阶段乳腺癌病理图像分类方法研究
- Author:
Kun-cai XU
1
;
Ning ZHANG
;
Yi-long LIAO
;
Xuan LIU
;
You ZHOU
Author Information
1. 贵阳信息科技学院智能工程学院,贵阳 550025
- Publication Type:Journal Article
- Keywords:
convolutional neural network;
Transformer;
breast cancer;
histopathological image;
histopathological image classification
- From:
Chinese Medical Equipment Journal
2024;45(12):1-8
- CountryChina
- Language:Chinese
-
Abstract:
Objective To propose a two-stage breast cancer histopathological image classification method based on CNN and Transformer to improve the accuracy of breast cancer histopathological image classification.Methods Firstly,Macenko normalization and color deconvolution were used to process breast cancer histopathological images to reduce pixel differences.Secondly,a dual-branch feature extraction path based on CNN and Transformer was constructed to extract local and global features of breast cancer histopathological images,respectively.Thirdly,local and global features were effectively fused based on BiFusion mechanism to enhance the expressive power of the features;finally,the BreakHis dataset was applied to validating the method proposed.Results The method achieved AUC values of 0.991,0.982,0.982 and 0.963 for the 40×,100×,200× and 400× magnifications on the BreakHis dataset,respectively,with high overall performance.Conclusion The proposed classi-fication method is of high value for the classification of breast cancer histopathological images.[Chinese Medical Equipment Journal,2024,45(12):1-8]
