Global research trends and knowledge mapping of artificial intelligence in triple negative breast cancer: a ten-year bibliometric analysis

This study aims to systematically delineate the global research landscape, developmental trajectory, and emerging hotspots of artificial intelligence (AI) in triple negative breast cancer (TNBC) between 2015 and 2025, thereby providing a scientific basis and strategic pathways for the advancement of precision oncology. Relevant literature on artificial intelligence in the field of TNBC from 2015 to 2025 was retrieved from the Web of Science Core Collection (WoSCC). VOSviewer, CiteSpace, and Tableau were used for bibliometric and visualization analysis. The contribution characteristics of countries, institutions, and authors, as well as the co-occurrence patterns of keywords, were systematically analyzed. The analysis of highly cited references is used to evaluate important achievements accumulated in the research field over the medium to long term. The study included 320 articles. Publication volume reached an inflection point of upward growth in 2022, peaking at 104 articles in 2025, accompanied by a simultaneous acceleration in annual citations. China and the United States serve as the primary global contributors to research output in this field, with Fudan University and the MD Anderson Cancer Center identified as the core high-output institutions. The research focus in this domain has shifted from auxiliary diagnosis and molecular subtyping (such as Lehmann typing) toward prediction of pathological complete response (pCR) following neoadjuvant chemotherapy (NAC). Current research trends are characterized by the integration of AI with single-cell sequencing, aiming to further advance the standards of precision oncology diagnosis and treatment through the analysis of tumor microenvironment (TME) heterogeneity. Research on AI in TNBC has evolved from early-stage applications in diagnostic assistance and molecular subtyping toward an integrated system for precision diagnosis and treatment. This study suggests a potential disconnect between algorithm development and clinical implementation, pointing out multi-center validation and global data sharing as critical prerequisites for achieving the practical translation of this technology.