Machine learning and deep learning for breast cancer: A decade systematic review of detection, classification, prognosis, and explainability

Breast cancer remains the most prevalent cancer worldwide, driving demand for accurate and interpretable diagnostic tools. This systematic review synthesizes studies from 2016–2025 (n = 131) using Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) on machine learning (ML) and deep learning (DL) for breast cancer detection, classification, prognosis, and explainability. The review compares algorithmic families, data modalities, hybrid and ensemble approaches, and the role of explainable AI (XAI). Results show that imaging dominates detection and classification, genomics underpins prognosis, and multimodal integration offers the highest robustness. Hybrid and ensemble strategies consistently improve performance, while XAI methods enhance trust but remain limited by shallow explanations. Key challenges include data scarcity, bias, external validation gaps, and clinical integration. Future directions point to federated learning, multimodal and longitudinal modelling, and human-centred XAI. This review provides actionable insights for bridging algorithmic innovation with clinical oncology practice.