Global Trends and Thematic Evolution of Machine Learning in Breast Cancer Research

Breast cancer continues to be a significant public health challenge worldwide and demands new diagnostic and prognostic technologies. Although artificial intelligence (AI) has made notable contributions to oncology, few computational approach have integrated bibliometric mapping with experimental validation, enabling global visualization of AI breast cancer research during 2020-2025. We report a systematic bibliometric analysis of 5,032 Scopus-indexed articles, integrative Python-based analytics using VOSviewer, and computational experiments on the GSE45827 gene expression dataset. The PRISMA-based literature screening identified an exponential growth (25.6 % CAGR), which 389 papers were published in 2020, and we estimated 1,141 publications in 2024; this trend was statistically significant (R² = 0.98, p<0.001) and accounted for a total of 94,295 citations (H-index:119). China and India led in publication productivity, while the United States experienced dominance in citation impact and collaborative centrality. The development trends have shifted from imaging-focused AI to the multimodal integration of radiomics, gene expression, and explainable AI. Importantly, our experimental verification showed the Support Vector Machine (Accuracy: 89.1 %, AUC: 0.991) to be the best-performing model, and SHAP-based explanation identified pivotal features (POSTN, COL1A1, CDKN3), thus empirically validating our bibliometric trends towards biologically interpretable findings. The field has developed into an integrated research discipline, but it still faces challenges in clinical validation and interpretability. Ongoing priorities should focus on clinical translation to benefit patients, fair model architecture, federated learning infrastructure for large-scale health data sharing, ethical principles and regulations for transparent deployment, and integration of multimodal omics to connect algorithmic innovation to equitable precision oncology practice.