Federico L, Fusaro D D, Coppola G C, Gregori M, Durante S
Department of Radiology, IRCCS Azienda Ospedaliero-Universitaria di Bologna, 40138, Bologna, Italy; Alma Mater Studiorum, University of Bologna, 40138, Bologna, Italy; Italian Association of Interventional Radiographers (AITRI), Via S. Gregorio, 53, Milan 20124, Italy.
Alma Mater Studiorum, University of Bologna, 40138, Bologna, Italy.
Radiography (Lond). 2025 Jul;31(4):102972. doi: 10.1016/j.radi.2025.102972. Epub 2025 May 20.
Conversational Artificial Intelligence (CAI) is transforming healthcare by introducing innovative tools for decision support and training. ChatGPT 4.0, an advanced generative language model, represents a promising resource for radiographers, particularly in complex areas such as Interventional Radiology (IR). This study evaluates the utility of ChatGPT 4.0 in supporting radiographers with varying levels of experience in radiological dose management.
A questionnaire consisting of 12 questions was developed, divided into three levels of difficulty (basic, intermediate, and advanced) relating to dose management in IR. The responses generated by ChatGPT 4.0 were evaluated by six radiographers divided into two groups: experts (more than five years of experience) and non-experienced (with experience in general radiology but no direct involvement in IR), according to three criteria: clarity, accuracy, and reliability, using a five-point Likert scale. Data were analyzed using the U test, Wilcoxon test, and ANOVA to identify differences between the groups.
The Mann-Whitney U test showed general agreement between the groups, with only one question approaching statistical significance (p = 0.059). The Wilcoxon test showed no significant differences for individual questions (p > 0.05). ANOVA showed significant differences in clarity (p < 0.05) and accuracy (p < 0.05), whereas differences in reliability were less pronounced.
ChatGPT 4.0 proves to be an effective tool for supporting non-experienced radiographers, facilitating learning, and reducing professional anxiety. However, greater technical sophistication is required to meet the needs of experts, that expressed higher expectations for clarity and accuracy.
Integrating CAI into radiography can enhance continuous education and optimize clinical practices, supporting both the professional development of radiographers and operational safety. Future adaptations and updates could make such tools more useful for advanced tasks.
对话式人工智能（CAI）正在通过引入用于决策支持和培训的创新工具来改变医疗保健。ChatGPT 4.0是一种先进的生成式语言模型，对放射技师来说是一种很有前景的资源，尤其是在介入放射学（IR）等复杂领域。本研究评估了ChatGPT 4.0在支持不同经验水平的放射技师进行放射剂量管理方面的效用。
编制了一份包含12个问题的问卷，分为与IR剂量管理相关的三个难度级别（基础、中级和高级）。ChatGPT 4.0生成的回答由六名放射技师进行评估，他们被分为两组：专家（超过五年经验）和非专家（有普通放射学经验但未直接参与IR），根据清晰度、准确性和可靠性三个标准，使用五点李克特量表进行评估。使用U检验、威尔科克森检验和方差分析来分析数据，以确定组间差异。
曼-惠特尼U检验显示两组之间总体一致，只有一个问题接近统计学显著性（p = 0.059）。威尔科克森检验显示单个问题无显著差异（p > 0.05）。方差分析显示在清晰度（p < 0.05）和准确性（p < 0.05）方面存在显著差异，而可靠性方面的差异不太明显。
ChatGPT 4.0被证明是支持非专家放射技师、促进学习和减轻职业焦虑的有效工具。然而，需要更高的技术复杂性来满足专家的需求，专家对清晰度和准确性的期望更高。
将CAI整合到放射学中可以加强继续教育并优化临床实践，支持放射技师的专业发展和操作安全。未来的调整和更新可以使这些工具对高级任务更有用。
