Scholarlink

Breast cancer is one of the leading causes of mortality for women globally. Finding an advanced stage makes treatment challenging. Conversely, early identification can significantly enhance the lives of millions of women. To improve breast cancer classification, this study integrated pre-trained models such as MobileNet with EfficientNetB0 and ResNet50 with EfficientNetV2B3, using whole slide images (WSIs) for the histology images to distinguish between IDC and non-IDC tissue. We proposed two hybrid pre-trained models, data augmentation, and downsizing every image from 50 × 50 pixels to 32 × 32 pixels to solve this problem. The test results show that the ResNet50 with EfficientNetV2B3 model has a high level of accuracy (96%), precision (91.1%), recall (86.7%), and F1 score (88.9%). Finally, we will compare the accuracy of the suggested model to that of the MobileNet with the EfficientNetB0 model and other DL models.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Hybrid Deep Learning Approach for Breast Cancer Classification Based on Histology Images

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Sameh Zarif,
  • Hatem Abdulkader,
  • Ibrahim Sayed Elaraby,
  • Marian Wagdy

摘要

Breast cancer is one of the leading causes of mortality for women globally. Finding an advanced stage makes treatment challenging. Conversely, early identification can significantly enhance the lives of millions of women. To improve breast cancer classification, this study integrated pre-trained models such as MobileNet with EfficientNetB0 and ResNet50 with EfficientNetV2B3, using whole slide images (WSIs) for the histology images to distinguish between IDC and non-IDC tissue. We proposed two hybrid pre-trained models, data augmentation, and downsizing every image from 50 × 50 pixels to 32 × 32 pixels to solve this problem. The test results show that the ResNet50 with EfficientNetV2B3 model has a high level of accuracy (96%), precision (91.1%), recall (86.7%), and F1 score (88.9%). Finally, we will compare the accuracy of the suggested model to that of the MobileNet with the EfficientNetB0 model and other DL models.