人工智能医疗器械测评数据库评估方法研究.pdfVIP

专题策划——人工智能医疗器械监管研究 Special Planning 人工智能医疗器械测评数据库评估方法研究 张晨光 彭亮 张宇晶 刘枭寅 王雅文 贺伟罡 【摘要 】目的：以临床准入和应用评估衔接为导向，研究用于人工智能医疗器械确认测试的第三方数据库即人 工智能医疗器械测评数据库评估方法。方法：根据人工智能医疗器械产品安全有效评价方法、测评数据库的数 据分布特性和产品确认方法，结合产品测评方式和测评环境进行分析。结果：给出测评数据库基本要素，数据 库建设、更新、维护的通用要求和权威性、科学性、规范性、多样性、封闭性、动态性等专用要求。结论：人 工智能医疗器械测评数据库需符合数据库建设的通用要求以及权威性、科学性、规范性、多样性、封闭性、动 态性等专用要求。 【关键词 】人工智能医疗器械；测评数据库；评估方法；非临床研究 Doi:10.3969/j.issn.1673-7571.2023.05.006 【中图分类号 】R197.39 ；TP392 Research on evaluation method of AI medical device evaluation database ZHANG Chenguang, PENG Liang, ZHANG Yujing, LIU Xiaoyin, WANG Yawen, HE Weigang. Center for Medical Device Evaluation, National Medical Products Administration, Beijing 100081, China Corresponding author: PENG Liang, Email: pengliang@ Abstract Objective To study the evaluation method of the third-party database for the validation test of AI 【 】 medical devices, guided by the connection between clinical access and application evaluation. Methods According to the effective and safety evaluation methods of AI medical device, the data distribution characteristics of the evaluation database and the method of products validation were evaluated and analyzed considering the evaluation method and evaluation environment. Results The basic elements of the evaluation database were given, as well as the general requirements for the construction, data update and maintenance of the database and the special requirements of authority, scientificity, normativeness, diversity, closure and dynamics. Conclusion The AI medical device evaluation database should meet the general requirements for database construction and special requirements such as authority, scientificity, normativeness, diversity, closure and dynamics. Keywords AI medical device; Evaluation database; Assessment methods;