Scholarlink

<p>Breast cancer (BC) continues to pose a substantial worldwide health concern, with early and precise identification being essential for improving patient survival rates. This systematic review examines generative adversarial networks (GANs) applications across five critical domains: model performance enhancement, image segmentation, data augmentation, super-resolution imaging, and imbalanced dataset management. Through comprehensive analysis of 33 studies (2021–2025), using databases including ScienceDirect, Web of Science, Scopus, PubMed, Springer, IEEE, and Google Scholar. Our analysis reveals data augmentation as the most clinically viable application, achieving 89.71–100% accuracy (mean: 96.2%), followed by super-resolution applications improving diagnostic accuracy from 95.82 to 99.11% while enhancing image quality metrics (SSIM: 0.7260 to 0.8005). Segmentation applications demonstrated robust performance with 85–94.27% Dice coefficients. We introduce a novel architectural comparison matrix providing practical guidance for selecting appropriate GAN architectures based on computational constraints and clinical requirements. Key findings establish a clinical translation hierarchy: immediate deployment of data augmentation GANs, followed by super-resolution for mammographic enhancement, then segmentation requiring multi-center validation. Critical challenges include dataset heterogeneity and computational demands. This framework provides the first comprehensive roadmap for clinical integration, potentially transforming diagnostic accuracy and accessibility in BC detection.</p>

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

Applications of GAN Models in Breast Cancer Detection: A Comprehensive Review

  • 期刊论文

  • 发表日期：

    2025年8月7日

查看全文
  • Jafar Majidpour,
  • Hunar A. Ahmed,
  • Mohammed H. Ahmed,
  • Shayan I. Jalal,
  • Hossein Arabi

摘要

Breast cancer (BC) continues to pose a substantial worldwide health concern, with early and precise identification being essential for improving patient survival rates. This systematic review examines generative adversarial networks (GANs) applications across five critical domains: model performance enhancement, image segmentation, data augmentation, super-resolution imaging, and imbalanced dataset management. Through comprehensive analysis of 33 studies (2021–2025), using databases including ScienceDirect, Web of Science, Scopus, PubMed, Springer, IEEE, and Google Scholar. Our analysis reveals data augmentation as the most clinically viable application, achieving 89.71–100% accuracy (mean: 96.2%), followed by super-resolution applications improving diagnostic accuracy from 95.82 to 99.11% while enhancing image quality metrics (SSIM: 0.7260 to 0.8005). Segmentation applications demonstrated robust performance with 85–94.27% Dice coefficients. We introduce a novel architectural comparison matrix providing practical guidance for selecting appropriate GAN architectures based on computational constraints and clinical requirements. Key findings establish a clinical translation hierarchy: immediate deployment of data augmentation GANs, followed by super-resolution for mammographic enhancement, then segmentation requiring multi-center validation. Critical challenges include dataset heterogeneity and computational demands. This framework provides the first comprehensive roadmap for clinical integration, potentially transforming diagnostic accuracy and accessibility in BC detection.