Use of Generative Artificial Intelligence in Nuclear Medicine Research, Education, and Clinical Practice: Results from a 2025 Society-Wide Survey

Abstract

Purpose Generative artificial intelligence (AI) is rapidly expanding within medical practice, yet its use among nuclear medicine professionals has not been systematically evaluated. This study provides the first society-wide assessment of generative AI adoption among members of the Korean Society of Nuclear Medicine (KSNM). Methods An anonymous online survey was distributed to approximately 670 KSNM members. The questionnaire assessed AI use in two domains (research/education and clinical practice), including platforms, frequency, purposes, and perceived usefulness and trust. Descriptive statistics and chi-square tests were used. Results A total of 122 members responded (estimated response rate, 18.2%); 105 (86.1% of respondents) reported using generative AI. Among users, 51.4% used AI daily and 51.4% used paid services; paid subscriptions were more frequent among daily than non-daily users (58.9% vs. 34.4%, p = 0.021). ChatGPT was the most commonly used platform (89.5%), followed by Gemini (55.2%) and Perplexity (26.7%). Perceived usefulness was higher in research/education (n = 96) than in clinical practice (n = 60): 85.4% vs. 41.7% rated AI as helpful/very helpful (p < 0.001). In research/education, translation/ language editing showed high prevalence and high utility, whereas code generation had lower prevalence but high perceived utility. Conclusion Generative AI showed a high adoption rate among the survey respondents, with higher perceived usefulness in research/education than in clinical practice. Given the widespread real-world use of third-party generative AI tools, our findings underscore the need for institutional guidance and privacy-aware governance to ensure safe clinical adoption.