李子睿

目的：探讨医学影像全流程智能体的技术内涵、临床应用现状及未来发展趋势，为影像科的智能化转型提供理论参考。方法：从医技工作流视角出发，系统阐述全流程智能体的概念定义与技术特征，分析其在CT、DR、MRI三大检查设备上的具体应用场景，并就自动化偏见、技能退化、人际沟通等挑战展开讨论。结果：全流程智能体是以多模态信息感知为基础、以任务自主编排执行为核心、贯穿检查全周期的智能化系统，具备将临床意图分解为操作序列、持续监测检查过程并动态响应异常的能力。临床应用方面，CT智能体侧重“一扫多查”与心血管后处理自动化，联影智能“一扫多查”可基于单次胸部CT平扫自动检出73种胸部异常；DR智能体聚焦摆位标准化与测量智能化，全幅DR搭载智能测量系统将传统人工测量时间从数分钟压缩至秒级，Cobb角自动测量效率显著提升；MRI智能体力破解操作复杂与患者配合度难题，全球首个“一键式”AI驱动心脏磁共振方案将传统30-50分钟检查压缩至约10分钟，且实现无需屏气的膈肌导航延迟增强成像。三者共同推动工作流从经验驱动转向系统驱动的标准化协作。与此同时，自动化偏见可能诱发决策惰性，人际接触减少可能削弱医患信任温度，基础操作全面自动化可能引发年轻技师技能退化，需在效率与审慎、自动化与人文关怀、系统辅助与技能传承之间寻求平衡。结论：全流程智能体标志着医学影像AI从“单任务工具”向“全流程助手”的范式跃迁，未来将沿多模态深度融合、群体智能协同、自进化能力闭环及基层普惠覆盖四条主线加速演进。医技人员应主动理解其运行逻辑，在技术赋能中坚守专业判断的核心地位。

医学影像；全流程智能体；工作流；范式转移；大模型；多模态融合

[1] Rajpurkar P,Chen E,Banerjee O,et al.AI in health and medicine[J].Nature Medicine,2022,28(1):31-38. [2] Khosravi B,Rouzrokh P,Akinci D'Antonoli T,et al.Agentic AI in radiology:evolution from large language models to future clinical integration[J].Radiology:Artificial Intelligence,2026,8:e250651. [3] Salehi S,et al.Agentic AI and large language models in radiology:opportunities and hallucination challenges[J].Bioengineering,2025,12(12):1303. [4] Yi W,Chen Y,Feng X,et al.Agentic artificial intelligence in radiology workflow:from image interpretation to report quality control[J]. Frontiers in Medicine,2026,13:1927284. [5] Tzanis E,et al.Agentic systems in radiology:principles,opportunities,privacy risks,regulation,and sustainability concerns[J].Diagnostic and Interventional Imaging,2026,107(1):7-16. [6] Raza M,Salem S,Habib A,et al.MedXpert-CAD:a multimodal multi-agentic system for clinical imaging analysis via model context protocol LLM-driven agentic workflows[C]//MICCAI 2025 Workshops.2025:55-64. [7] Kang X,Vorberg L,Maier A,et al.Scan-do attitude:towards autonomous CT protocol management using a large language model agent[C]//AI for Clinical Applications:First International Workshops,Agentic AI 2025,CREATE 2025,and Clinical MLLMs 2025,Held in Conjunction with MICCAI 2025.Cham:Springer,2025:46-54. [8] Melazzini L,Bortolotto C,Brizzi L,et al.AI for image quality and patient safety in CT and MRI[J].European Radiology Experimental,2025,9(1). [9] Fink A,Rau S,Kästingschäfer K,et al.From referral to reporting:the potential of large language models in the radiological workflow[J]. RöFo,2026,198(1):55-63.

点击下载PDF