医学教育管理 ›› 2026, Vol. 12 ›› Issue (5): 655-663.doi: 10.3969/j.issn.2096-045X.2026.05.013

• 临床教学 • 上一篇    下一篇

AI支持下医学影像精准学习教学框架构建与肝脏病例教学原型平台设计

  

  1. 1.上海交通大学医学院附属瑞金医院放射科,上海 200025; 2.上海交通大学医学院医学技术学院医学影像技术系,上海 200025
  • 收稿日期:2026-05-20 修回日期:2026-08-26 出版日期:2026-11-13 发布日期:2026-11-13

Construction of an AI-supported precision learning framework for medical imaging education and prototype design using liver lesion cases

  1. 1. Department of Radiology, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai 200025, China; 2. Department of Medical Imaging Technology, College of Medical Technology, Shanghai Jiao Tong University School of Medicine, Shanghai 200025, China
  • Received:2026-05-20 Revised:2026-08-26 Online:2026-11-13 Published:2026-11-13

摘要:

目的 构建人工智能(artificial intelligence,AI)支持下医学影像精准学习教学框架,并以肝脏占位性病变计算机断层扫描(computed tomography,CT)和磁共振成像(magnetic resonance imaging,MRI)诊断教学为例进行原型平台设计。方法 以胜任力导向医学教育、精准学习、学习分析、认知诊断和知识图谱为理论基础,构建由能力指标、资源与任务、分析与反馈、治理与改进4个模块组成的教学框架;按照最终诊断明确、影像资料完整、具有教学代表性或鉴别诊断价值及完成脱敏处理等标准,遴选30例肝脏占位性病变病例构建AI学习库。并邀请5名放射科带教教师,采用自制评价表,从框架合理性、病例库与任务设计可操作性、AI提示安全性、教师工作量可接受性、后续推广可行性和总体评价6个方面进行Likert 5级评分,并收集开放性意见。结果 形成AI支持下医学影像精准学习四模块框架、肝脏病例库标签方案和分层学习任务。5名放射科带教教师对框架合理性、病例库与任务设计可操作性、AI提示安全性、教师工作量可接受性、后续推广可行性和总体评价的评分分别为(4.80±0.45)分、(4.80±0.45)分、(4.80±0.45)分、(4.00±0.00)分、(4.60±0.55)分和(4.40±0.55)分,总分为(27.40±1.14)分。结论 该框架可依托教师审核的AI病例库,整合能力画像、薄弱环节识别、任务匹配与形成性反馈,为医学影像个体化训练和过程性支持提供参考,其教学效果仍需后续实证研究验证。

Abstract:

Objective To develop an artificial intelligence (AI)-supported precision learning framework for medical imaging education and to design a prototype platform using computed tomography (CT) and magnetic resonance imaging (MRI) diagnostic teaching of focal liver lesions as an example. Methods  Based on competency-based medical education, precision learning, learning analytics, cognitive diagnosis, and knowledge graph theory, a four-module framework was developed, comprising competency indicators, resources and tasks, analysis and feedback, and governance improvement. Thirty focal liver lesion cases were selected to establish a de-identified AI-assisted learning library according to predefined criteria, including a confirmed final diagnosis, complete imaging data, educational representativeness or differential diagnostic value, and appropriate de-identification. Five radiology instructors were then invited to evaluate the framework and prototype design using a self-developed evaluation form. The evaluation covered six aspects: framework rationale, operability of the case library and task design, safety of AI-generated prompts, acceptability of instructor workload, feasibility of further implementation, and overall evaluation. Each item was rated on a 5-point Likert scale, and open-ended comments were also collected.Results  The study established a four-module AI-supported precision learning framework, a liver case-library annotation scheme, and layered learning tasks. The ratings provided by the five radiology instructors for framework rationale, operability of the case library and task design, safety of AI-generated prompts, workload acceptability, implementation feasibility, and overall evaluation were 4.80±0.45, 4.80±0.45, 4.80±0.45, 4.00±0.00, 4.60±0.55, and 4.40±0.55, respectively; the total score was 27.40±1.14 out of 30. Conclusion  Based on an instructor-reviewed AI case library, this framework may support competency profiling, identification of learning gaps, task matching, and formative feedback for individualized medical imaging training. Its educational effectiveness should be further evaluated in empirical studies.