AI-Driven Breast Cancer Diagnosis: A Systematic Review of Imaging Modalities, Deep Learning, and Explainability
Margo Sabry1, Hossam Magdy Balaha2, Khadiga M Ali3
1Information Systems Department, Assiut University, Assiut 71515, Egypt.
Cancers
|May 4, 2026
Summary
Artificial intelligence (AI) significantly enhances breast cancer (BC) diagnosis using deep learning across imaging. While AI shows high accuracy, challenges in data, generalizability, and clinical integration require further research for optimal outcomes.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Overview of artificial intelligence (AI) and deep learning in breast cancer (BC) diagnosis.
- Focus on advancements across diverse imaging modalities.
Purpose of the Study:
- To systematically review recent AI and deep learning applications in BC diagnosis.
- To analyze diagnostic performance, architectural developments, and clinical integration.
Main Methods:
- Systematic review adhering to PRISMA guidelines.
- Comparative analysis of 65 peer-reviewed studies (2018-2024).
Main Results:
- Convolutional Neural Networks (CNNs) achieve up to 98.5% accuracy in mammography; Vision Transformers reach 96% in histopathology.
- Leading research domains include diagnostic accuracy, risk prediction, and personalized screening.
- Explainable AI (XAI) methods (SHAP, LIME, Grad-CAM) enhance transparency and trust.
Conclusions:
- AI is crucial for early BC detection and treatment optimization.
- Challenges include data heterogeneity, model generalizability, and clinical workflow integration.
- Future research should focus on multi-dataset validation and standardized implementation frameworks.
Keywords:
breast cancer (BC)computer-aided diagnosis (CAD)deep learning (DL)eXplainable artificial intelligence (XAI)machine learning (ML)
