• 中国核心期刊数据库收录期刊
  • 中文科技期刊数据库收录期刊
  • 中国期刊全文数据库收录期刊
  • 中国学术期刊综合评价数据库统计源期刊等

快速检索引用检索图表检索高级检索

监管科学

《医疗器械分类目录》优化的实践探索与AI辅助框架的思考

  • 王越 ,
  • 崔乐 ,
  • 江潇 ,
  • 朱炯 ,
  • 张春青 ,
  • 郑佳
展开
  • 1.中国药品检定研究院,北京 100050;

    2.广东省医疗器械质量监督检验所,广东 广州 510663

王越,女,副主任药师,研究方向:医疗器械安全性评估

收稿日期: 2026-03-17

  修回日期: 2026-06-29

  录用日期: 2026-07-22

  网络出版日期: 2026-10-09

Optimizing the Medical Device Classification Catalog: Practices and Pathways with Artificial Intelligence

Expand
  • 1.National Institutes for Drug Control, Beijing 100050, China;
    2.Guangdong Medical Devices Quality Surveillance and Test Institute, Guangdong Guangzhou 510663, China

Received date: 2026-03-17

  Revised date: 2026-06-29

  Accepted date: 2026-07-22

  Online published: 2026-10-09

摘要

目的:分析医疗器械注册备案中的不规范问题及其规律,发现《医疗器械分类目录》实际运行中的堵点,并在此基础上探索人工智能(AI)辅助的分类目录优化路径。方法:基于2024—2025年度2225条医疗器械注册备案产品信息研判结论中的“不规范/不准确”数据,系统梳理问题类型、高发领域和具体理由。结果:医疗器械注册备案产品不规范问题的5种类型中,管理类别判定错误(34.7%)和分类编码识别错误(24.8%)是两类最集中的问题,合计占比59.5%。“专科目录与通用目录混淆”模式占分类编码判定错误近40.0%,集中于“09物理治疗器械”目录向“16眼科器械”“18妇产科器械”“20中医器械”等专科目录的误转移;管理类别识别错误高度集中于肿瘤标志物试剂、高频手术电极、无创呼吸机等依赖“关键限定词”判定的产品,其中大多与关键信息缺失直接相关;药械组合产品属性界定争议聚焦于薄荷脑、苯扎氯铵等《中华人民共和国药典》收载成分;关键信息缺失集中于“是否在内窥镜下使用”“探头使用部位”等决定类别升降级的字段。结论:医疗器械注册备案产品分类界定现存问题说明《医疗器械分类目录》在规则传导、信息颗粒度、组合界定、系统衔接等方面存在优化空间。本研究所构建的“规则+AI”辅助分类研判框架,通过智能导航、动态问答、风险预判、一致性校验4个功能模块分别回应上述问题,尝试推动企业申报从“事后纠错”向“事前提示预警、事中辅导”转变,为医疗器械分类管理的智能化辅助提供参考。

 

本文引用格式

王越 , 崔乐 , 江潇 , 朱炯 , 张春青 , 郑佳 .

《医疗器械分类目录》优化的实践探索与AI辅助框架的思考

[J]. 中国医药导刊, 2026 , 28(7) : 789 -789-796 . DOI: 10.1009-0959.2026.080008

Abstract

Objective: To analyze the non-compliant issues and their patterns in medical device registration and filing, identify the bottlenecks in the operation of the Medical Device Classification Catalog, and explore AI-assisted pathways for improving the catalog.Methods: Based on 2225 records identified as “non-compliant/inaccurate” from medical device registration and filing data between 2024 and 2025, this study systematically reviewed and categorized the types of problems, high-risk areas, and specific reasons for non-compliance.Results: In the 5 types of non-compliant issues of medical device registration and filing, incorrect classification of management category (34.7%) and classification code errors (24.8%) were the two predominant problem types, together accounted for 59.5% of all non-compliant cases.The “general-to-specialty misclassification” pattern constituted nearly 40.0% of classification code errors, primarily involving products misclassified into sub-catalog 09 (physical therapy devices) that should have been assigned to specialty sub-catalogs such as 16 (ophthalmology devices), 18 (obstetrics and gynecology devices), and 20 (traditional Chinese medicine devices). Product classification errors were highly concentrated in products whose classification depends on “key qualifiers” in their intended use—including tumor marker reagents, high-frequency surgical electrodes, and non-invasive ventilators—with most of such errors directly attributable to missing critical information. Combination product disputes predominantly involved ingredients listed in the Chinese Pharmacopoeia, such as menthol and benzalkonium chloride. Missing critical information was concentrated in fields determining class upgrades or downgrades, including “use under endoscopy” and “probe application site.” Conclusion: The identified registration and filing issues reveal areas for optimization in the Medical Device Classification Catalog, including rule transmission, information granularity, combination product definition, and system integration. This study proposes a classification assistance framework based on a “rule + AI” hybrid architecture, with four functional modules—smart navigation, dynamic Q&A,risk pre-assessment, and consistency verification—addressing the identified problems respectively. The framework represents a conceptual shift from “post-hoc correction” to “real-time guidance” in classification management, offering a reference for intelligent assistance in medical device classification.


参考文献

        [1] 原国家食品药品监督管理总局.总局关于发布医疗器械分类目录的公告(2017年第104号)[EB/OL].(2017-08-31)[2026-03-02].https://www.nmpa.gov.cn/ylqx/ylqxggtg/20170904150301537.html.

         [2]  吴欣然,刘洋.大语言模型在药品监管典型场景应用分析[J].中国食品药品监管,2025,(9):30-39.

         [3]  国家药品监督管理局.国家药监局关于印发《药品监管网络安全与信息化建设“十四五”规划》的通知[EB/OL].(2022-04-24)[2026-03-02].https://www.nmpa.gov.cn/xxgk/fgwj/gzwj/gzwjzh/20220511110329171.html.

         [4]  李瑶.药品审评改革持续深化,CDE筹划AI辅助审评[N].医药经济报,2025-09-15(001).

         [5]  国家药品监督管理局.国家药监局关于调整《医疗器械分类目录》部分内容的公告(2025年第132号)[EB/OL].(2025-12-30)[2026-03-02].https://www.nmpa.gov.cn/xxgk/ggtg/ylqxggtg/ylqxqtggtg/20260104173409116.html.

         [6]  国家药品监督管理局.国家药监局关于发布第一类医疗器械产品目录的公告[EB/OL].(2021-12-31)[2026-03-02].https://www.beijing.gov.cn/zhengce/zhengcefagui/qtwj/202204/t20220407_2668581.html.

         [7]  季冬梅,郝天祎.全生命周期框架下人工智能医疗器械的智慧监管应对路径[J].中国市场监管研究,2025,(5):44-50.

         [8]  原国家食品药品监督管理总局.医疗器械分类规则(国家食品药品监督管理总局令第15号)[EB/OL].(2015-07-14)[2026-01-02].https://www.samr.gov.cn/zw/zfxxgk/fdzdgknr/bgt/art/2023/art_24dbff6e15494c9cb112ea15ed158001.html.

         [9]  国家药品监督管理局.医疗器械软件注册审查指导原则(2022年修订版)[EB/OL].(2022-03-07)[2026-03-02].https://www.cmde.org.cn//xwdt/shpgzgg/gztg/20220309091454609.html.

         [10] 国家药品监督管理局.人工智能医疗器械注册审查指导原则[EB/OL].(2022-03-09)[2026-03-02].https://www.cmde.org.cn//flfg/zdyz/zdyzwbk/20220309091014461.html.

         [11] 国家市场监督管理总局.医疗器械注册与备案管理办法[EB/OL].(2021-08-26)[2026-03-02].https://www.gov.cn/gongbao/content/2021/content_5654783.htm.

         [12] 李屹玲,赵琪,魏旗鹏,等.优化人工智能医疗器械全生命周期监管:国际经验与中国路径[J].中国食品药品监管,2026,(3):42-51.

         [13] 张建楠,李莹莹,周佳卉,等.人工智能独立医用软件监管研究[J].中国工程科学,2022,24(1):198-204.

         [14] 王翔宇,侯艳,杨蕾,等.人工智能在药品监管应用中的法律挑战探讨[J].中国食品药品监管,2026,(4):4-15.

         [15] 施颂华.人工智能在提升临床试验质量管理体系效能中的应用[J].中国食品药品监管,2026,(1):52-65.

         [16] Han Yu, Ceross A, Bergmann J. Regulatory frameworks for AI-enabled medical device software in China:comparative analysis and review of implications for global manufacturer[J].JMIR AI, 2024,3:e46871.

         [17] 邱英鹏,姚涵,任平,等.我国医学人工智能监管沙盒模式构建思路[J].中国卫生信息管理杂志,2025,22(2):262-268.

         [18] 潘威,刘峰,武宵宵,等.数智药监一体化平台在新疆药品智慧监管中的探索和应用[J].中国医药导刊,2026,28(4):497-503.

         [19] 石镇维,刘再毅.重视医学影像人工智能数据库的标准化建设[J].协和医学杂志,2021,12(5):599-601.

         [20] 国家药品监督管理局.国家药监局关于发布优化全生命周期监管支持高端医疗器械创新发展有关举措的公告(2025年第63号)[EB/OL].(2025-06-07)[2026-03-02].https://www.nmpa.gov.cn/yaowen/ypjgyw/ylqxyw/20250703163951182.html.

         [21] 国家药品监督管理局.国家药监局综合司关于印发药品监管人工智能典型应用场景清单的通知[EB/OL].(2024-06-13)[2026-03-02].https://www.gov.cn/zhengce/zhengceku/202406/content_6958094.htm.

         [22] 刘涛,张钰,李钻,等.人工智能在药品审评中的应用展望[J].中国医药导刊,2024,26(5):504-510.

         [23] 朱文博.人工智能在药品审评审批中的应用[J].中国科技信息,2025,(12):148-150.

         [24] 张龄文,秦宇,董志宏,等.公立医院医疗设备管理体系构建的研究进展[J].中国医疗设备,2026,41(5):171-180.

         [25] 肖敏,陈超,姚金成.人工智能赋能药品全流程监管效能提升的路径研究[J].中国战略新兴产业,2026,(15):76-78.

文章导航

/

〈 〉