胜任力驱动的智能交互式虚拟病人实训架构OA
Competency-driven Responsive Intelligence for Simulated Patient Architecture
针对临床胜任力培养中理论向实践转化的"断层"问题,提出一种临床胜任力驱动的智能交互式虚拟病人实训架构(competency-driven responsive intelligence for simulated patient architecture,CRISPA).将临床能力评价指标解构成病史采集、临床推理、医患沟通及临床操作 4个维度,并建立其与技术实现机制的映射关系.在技术层面,通过语义重构实现结构化病理知识的口语化表达;基于检索增强生成与医学知识图谱构建知识约束机制,以限制生成边界并提升医学一致性;通过多模态情感计算实现人文沟通的量化评估;结合有限状态机驱动病情动态演化、教学策略适配与学习路径动态适配.系统形式化为受控交互五元组模型,采用"知识支撑-逻辑推理-多模态交互-评价反馈"的协同架构.以内窥镜检查场景为例,阐述CRISPA 在动态教学交互中的运行机制,并构建涵盖系统性能、医学逻辑依从性、能力评估准确性及教学适配性的多维评价体系.该研究为智能化医学模拟教学提供了一种高保真、可量化且具备可解释性的建模框架.
To bridge the"gap"between theoretical knowledge and clinical practice in competency-based medical training,this study proposes the Competency-driven Responsive Intelligence for Simulated Patient Architecture(CRISPA)model.The model categorizes clinical competency assessment indicators into four dimensions—medical history collection,clinical reasoning,doctor-patient communication and clinical operations—and establishes their mapping relationships with the corresponding technical implementation mechanisms.At the technical level,colloquial expression of structured pathological knowledge is achieved through semantic reconstruction.A knowledge constraint mechanism based on retrieval-augmented generation(RAG)and a medical knowledge graph restricts the generation scope and enhances medical accuracy.Multimodal affective computing is used to evaluate humanistic communication skills quantitatively.Additionally,a finite-state machine drives dynamic disease evolution,enabling adaptive teaching strategies and dynamic learning path adjustment.The system is formalized as a controlled,interactive,five-tuple model adopting a collaborative architecture comprising"knowledge support—logical reasoning—multimodal interaction—evaluation feedback".Using an endoscopy examination scenario as an example,this paper explains how CRISPA operates in dynamic teaching interactions and describes how it constructs a multi-dimensional evaluation system covering system performance,compliance with medical logic,accuracy of competency assessment and teaching adaptability.Overall,this study provides an interpretable,quantifiable and high-fidelity modelling framework for intelligent medical simulation teaching.
宋祎宁;蒋君铭
徐州医科大学 医学人工智能学院,江苏 徐州 221004徐州医科大学 医学人工智能学院,江苏 徐州 221004
社会科学
医学教学智能模拟病人大语言模型检索增强生成知识图谱
medical educationintelligent simulated patientslarge language modelsretrieval-augmented generationknowledge graphs
《徐州工程学院学报(自然科学版)》 2026 (3)
71-79,9
在徐高校服务"343"产业发展项目(gx2024017)
评论