摘要
本研究构建一种医疗AI风险分层的动态监管框架,以应对伦理、隐私与技术挑战,推动标准化与可持续发展。从医疗AI的应用场景和遇到的挑战2个方面,设计了一个7级监管体系(0~6级),涵盖无AI干预的场景到完全自主决策。提出建议包括标准和指南的构建、医疗专业人员的培训和监管机构的作用,并结合IBM沃森健康、Butterfly iQ、Zebra Medical Vision与清华AI医院4个案例验证可行性。分层框架从任务自动化、决策自主、患者安全3个维度提供了多层次的监督方案,提出了动态校准、国际标准化、抑制幻觉的技术路径。医疗AI风险分级监管框架,既是业界共同遵守的标准准则,也为监管机构提供方法,未来需深化算法透明度、数据治理与跨学科协作等内容,以实现安全高效、普及性的全球医疗AI发展。
This study builds a dynamic regulatory framework for risk stratification in healthcare AI to address ethical,privacy,and technical challenges and drive standardization and sustainability.From the two aspects of application scenarios and challenges encountered by medical AI,a 7-level regulatory system(0-6levels)is designed,covering scenarios without AI intervention to fully autonomous decision-making.The recommendations include the construction of standards and guidelines,the training of Medical professionals and the role of regulators,and the feasibility is verified by combining four cases of IBM Watson Health,Butterfly iQ,Zebra Medical Vision and Tsinghua AI Hospital.The hierarchical framework provides a multi-level supervision scheme from three dimensions of task automation,decision autonomy and patient safety,and proposes a technical path of dynamic calibration,international standardization and suppression of hallucinations.The medical AI risk classification regulatory framework is not only a standard for the industry to comply with,but also provides a method for regulators to deepen algorithmic transparency,data governance and interdisciplinary collaboration in the future to achieve safe,efficient and universal global medical AI development.
作者
刘英
尹婉宜
袁晴
柳琪林
LIU Ying;YIN Wanyi;YUAN Qing;LIU Qilin(The First Hospital of Hebei Medical University,Shijiazhuang,Hebei 050023,China;School of Public Health,North China University of Science and Technology,Tangshan,Hebei 063210,China;Department of Standardized Residency Training,Chinese Medical Doctor Association,Beijing 100027,China)
出处
《社区医学杂志》
2025年第5期141-145,共5页
Journal Of Community Medicine
关键词
医疗人工智能
分级体系
风险管理
智慧医疗
medical artificial intelligence
classification system
risk management
smart medical care
