Scholarlink

Breast cancer is a very commonly diagnosed disease and is a leading cause of cancer deaths in women. Many of the early diagnosis systems have been developed, but the performance of these systems is quite low. Deep learning is an emerging and promising area for medical imaging and other applications, in which classification of breast cancer is a very challenging application. In this work, the cross-level attention (CLA) module is designed to improve the feature gradient in the feature maps to assure the performance improvement of the suggested system. The fusion of all CLA features is concatenated and finally classified by the fully connected neural network. The VGG19 network is considered the base network because of its less complex structure, easy implementation, and faster training. The proposed model gives 98.04% accuracy, which is 1.07% better in comparison with the other existing methods for breast cancer diagnosis form histopathology images.

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

Cross-Level Attention Feature Fusion-Based Deep Learning Process for Breast Cancer Diagnosis

  • 章节

  • 发表日期：

    2025年

查看全文
  • Vivek Patel,
  • Vijayshri Chaurasia,
  • Rajesh Mahadeva,
  • Vaibhav Gupta,
  • Ebtsam Ahmad Siddiqui,
  • Mamta Patankar,
  • Vinay Gupta

摘要

Breast cancer is a very commonly diagnosed disease and is a leading cause of cancer deaths in women. Many of the early diagnosis systems have been developed, but the performance of these systems is quite low. Deep learning is an emerging and promising area for medical imaging and other applications, in which classification of breast cancer is a very challenging application. In this work, the cross-level attention (CLA) module is designed to improve the feature gradient in the feature maps to assure the performance improvement of the suggested system. The fusion of all CLA features is concatenated and finally classified by the fully connected neural network. The VGG19 network is considered the base network because of its less complex structure, easy implementation, and faster training. The proposed model gives 98.04% accuracy, which is 1.07% better in comparison with the other existing methods for breast cancer diagnosis form histopathology images.