Scholarlink

The integration of deep learning models into breast cancer diagnosis has revolutionized medical imaging, offering enhanced accuracy and automation. Yet, adoption remains limited due to the “black-box” nature of these systems, raising concerns over transparency, trust, and ethics. This paper reviews the role of Explainable Artificial Intelligence (XAI) in enhancing the interpretability and reliability of AI-based diagnostic tools. Following a systematic review, 110 studies were screened and 40 selected for final analysis. These involved mammography, ultrasound, and histopathology, applying interpretability methods such as Grad-CAM, LIME, and SHAP. Findings show XAI improves clinician trust, supports decision-making, reduces workload, and enhances patient outcomes. The review also addresses ethical challenges, current limitations, and the need for patient-centered trust measures. Finally, future research directions are proposed to close the gap between algorithmic transparency and clinical applicability, promoting responsible and transparent AI use in oncology.

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

Toward Transparent and Trustworthy AI for Breast Cancer Diagnosis: Deep Learning, XAI, and Ethical Perspectives

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Nabanita Choudhury,
  • Cinu C. Kiliroor

摘要

The integration of deep learning models into breast cancer diagnosis has revolutionized medical imaging, offering enhanced accuracy and automation. Yet, adoption remains limited due to the “black-box” nature of these systems, raising concerns over transparency, trust, and ethics. This paper reviews the role of Explainable Artificial Intelligence (XAI) in enhancing the interpretability and reliability of AI-based diagnostic tools. Following a systematic review, 110 studies were screened and 40 selected for final analysis. These involved mammography, ultrasound, and histopathology, applying interpretability methods such as Grad-CAM, LIME, and SHAP. Findings show XAI improves clinician trust, supports decision-making, reduces workload, and enhances patient outcomes. The review also addresses ethical challenges, current limitations, and the need for patient-centered trust measures. Finally, future research directions are proposed to close the gap between algorithmic transparency and clinical applicability, promoting responsible and transparent AI use in oncology.