Artificial intelligence for medical imaging education: bibliometric and visual analysis

While bibliometric analyses have examined the adoption of Artificial Intelligence (AI) in general medical education contexts, the specific knowledge structures, collaborative networks, and evolutionary pathways of AI for medical imaging education remain poorly characterized quantitatively, particularly regarding the transition from technical feasibility to systematic pedagogical integration and cross-institutional partnerships.

A structured literature search of English-language original articles and reviews published between 2006 and 2025 was performed in the Web of Science Core Collection (WoS) and Scopus, with the former serving as the primary database and the latter for external validation. Bibliometric analyses covering citation trends, collaboration networks, keyword bursts, and journal distributions were conducted using CiteSpace, VOSviewer, and Bibliometrix.

This search retrieved 577 articles from WoS, revealing exponential growth since 2019, with 226 articles published in 2025. These publications were contributed by 3,739 authors from 1,479 institutions across 80 countries/regions. Cross-database validation supported the consistency of these findings (Spearman r = 0.97, p < 0.001), with the United States (166 articles, 28.8%) and China (115 articles, 19.9%) as the primary contributors. BMC Medical Education, Insights into Imaging, and Academic Radiology were the most productive journals. Keyword analysis revealed diversification from foundational algorithms toward interactive, generative AI applications in medical imaging education, documenting an expanded scope integrating technical development with pedagogical innovation and ethical governance.

This study characterizes a dual-polar production structure with the United States and China as dominant contributors, and a journal distribution spanning medical education and imaging technology venues. The post-2022 emergence of generative AI keywords documented a new thematic cluster alongside foundational machine learning research, indicating an evolving methodological repertoire. These findings provide a quantitative baseline for evidence synthesis in this interdisciplinary domain, with implications for curriculum development, algorithmic performance benchmarks, and ethical governance, subject to standard bibliometric data coverage and currency constraints.