Artificial intelligence in radiology education: from AI-supported learning to clinical innovation-a narrative review.
Authors
Affiliations (4)
Affiliations (4)
- Cancer Institute, Cellular Therapeutics School of Medicine, Xuzhou Medical University, Xuzhou, China.
- Department of Radiology, The Second Affiliated Hospital of Xuzhou Medical University, Xuzhou, China.
- Department of Radiology, Taizhou Hospital of Traditional Chinese Medicine, Taizhou, China.
- Department of Radiology, The 71st Group Army Hospital of the People's Liberation Army of China, Xuzhou, China.
Abstract
Artificial intelligence (AI) is changing diagnostic radiology, but education and training have not kept pace. This narrative review brings together current evidence on AI-supported learning in radiology education and how such learning can move into clinical innovation. Surveys from different countries point to the same gap: in a 25-country international survey of radiographers, radiologists, and students, only 10.3% had received formal AI training; in one national survey of Turkish residents, 84.5% had received none. The review maps competency frameworks and curriculum models that define AI literacy and support staged integration into residency. It also assesses AI as a teaching tool, including AI-assisted case-based learning, computer-aided diagnosis training, precision-education knowledge recommenders, and virtual, augmented, and extended reality. Controlled and randomized studies show measurable gains: computer-aided diagnosis training raised resident scores above self-study controls (mean score 52 vs. 38 on a 100-point scale), while an AI-based knowledge recommender shortened reporting time by 162 s per case and relative improvement in graded reporting accuracy scores of 14% (95% CI 8.1-19.8%). Large language models are promising but limited: GPT-4 approaches board-style examination performance (80.6%) yet remains poor at image-based reasoning (47.8%). The review closes by considering how education can feed clinical innovation, along with algorithmic and automation bias, ethical governance, and future research needs. AI education should therefore span the full continuum-from undergraduate training to continuing professional development-if radiology is to innovate safely, effectively, and equitably.