PDF(467 KB)
PDF(467 KB)
PDF(467 KB)
人工智能独立医用软件监管研究
Supervision System of AI-based Software as a Medical Device
科学的监管体系可以促进新兴事物的蓬勃和规范化发展。人工智能（AI）独立医用软件是人工智能时代赋能医疗健康领域的重要产物。随着人工智能的深入发展，人工智能独有的黑盒算法及自主学习能力引起了巨大的监管挑战。AI 独立医用软件的监管需要与时俱进，为最大程度降低人工智能医疗软件不良事件发生率和风险影响，我们亟待寻求更为科学合理的监管应对方案。对此，本文从AI 技术特征监管应对出发，针对算法变更管理、质量控制、安全追溯等监管上存在的痛点和难点问题梳理了国内外AI 独立医用软件相关监管制度和支撑资源的现状。分析总结了我国AI 独立医用软件监管在制度层面、支撑资源层面仍面临的一些问题与挑战，并针对上市后监管短板提出了措施建议：系统完善AI 独立医用软件监管制度，深化AI 独立医用软件监管支撑体系。以期能够为进一步完善我国AI 独立医用软件科学监管提供参考。
A scientific supervision system enhances vigorous and standardized development of emerging things. Artificial intelligencebased software as a medical device (AI-Based SaMD) is an important product in the health field enabled by artificial intelligence (AI). As AI develops further, its unique black box algorithm and independent learning ability have posed major supervision challenges. AIBased SaMD supervision needs to keep pace with the times and a more scientific supervision strategy is urgently required to minimize the adverse events of AI-Based SaMD and their risk impact. This article summarizes the current status of supervision systems and supporting resources of AI-Based SaMD in China and abroad considering the difficulties in algorithm change management, quality control, and safety traceability. In addition, we explore the problems and challenges of AI-Based SaMD in China. Furthermore, we suggest that the AI-Based SaMD supervision and support systems should be improved in China to overcome the disadvantages of postmarket supervision.
software as a medical device (SaMD) / artificial intelligence (AI) / supervision science
| [1] |
IMDRF SaMD Working Group. Software as a Medical Device (SaMD): Key definitions [C].Washington D.C.: IMDRF, 2013.
|
| [2] |
Matheny M E, Whicher D, Israni S T. Artificial intelligence in health care: A report from the National Academy of Medicine [J]. The Journal of the American Medical Association, 2020, 323:509– 510.
|
| [3] |
Benjamens S, Dhunnoo P，Meskó B. The state of artificial intelligence-based FDA-approved medical devices and algorithms: An online database [J]. npj Digital Medicine, 2020, 3(1): 118.
|
| [4] |
European Commission. ATI - Artificial Intelligence-based software as a medical device [R]. Brussels: European Commission, 2020.
|
| [5] |
FDA. Proposed regulatory framework for modifications to Artificial Intelligence/ Machine Learning (AI/ML)—Based Software as a Medical Device (SaMD) [R]. Silver Spring: FDA, 2019.
|
| [6] |
IMDRF. Software as a Medical Device: Possible framework for risk categorization and corresponding considerations [C]. Washington D.C.: IMDRF, 2014.
|
| [7] |
Abràmoff M D, Lavin P T, Birch M, et al. Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy in primary care offices [J]. npj Digital Medicine, 2018, 1: 39.
|
| [8] |
严舒, 徐东紫, 欧阳昭连. 美国人工智能医疗器械监管与应用分 析 [J]. 中国医疗设备 2021, 36(2): 117–122. Yan S, Xu D Z, Ouyang Z L. Analysis on the regulation and Application of United States Intelligence medical device [J]. China Medical Devices, 2021, 36(2): 117–122.
|
| [9] |
FDA. How is the FDA considering regulation of Artificial Intelligence and machine learning medical devices?[R/OL]. (2021-09-21)[2021-11-10]. https://www.fda.gov/medical-devices/ software-medical-device-samd/artificial-intelligence-andmachine-learning-software-medical-device#regulation.
|
| [10] |
国家药品监督管理局. 人工智能医用软件产品分类界定指导原 则 [R]. 北京：国家药品监督管理局, 2021. National Medical Products Administration. Guiding principles for the classification and definition of Artificial Intelligence medical software products [R]. Beijing: National Medical Products Administration, 2021.
|
| [11] |
国家药品监督管理局. 监管科学研究稳步推进 重点项目攻关成 果丰硕 [N]. 中国医药报, 2021-04-15(002). National Medical Products Administration. Scientific regulatory research has steadily made fruitful results in key projects [N]. China Medical News, 2021-04-15(002).
|
| [12] |
赵飞, 高明献, 刘璞, 等. 136例国际医疗器械严重不良事件的统 计与分析 [J].中国医疗器械杂志, 2020, 44(2):166–171. Zhao F, Gao M X, Liu P, et al. Statistics and analysis of 136 cases of adverse events of international medical devices [J]. Chinese Journal of Medical Instrumentation, 2020, 44 (2): 166–171.
|
| [13] |
蔡小舒, 吕晖, 于广军. 美国FDA医疗人工智能软件审核指南研 究 [J].中国数字医学, 2019, 14(11): 34–37, 33. Cai X S, Lyu H,Yu G J. Study on audit guidelines of US FDA medical Artificial Intelligence software [J].China Digital Medicine, 2019, 14(11): 34–37, 33.
|
| [14] |
Ministry of Health, Labour and Welfare (MHLW). Update on medical device and IVD regulation in Japan [R/OL].(2021-03-20) [2021-11-10]. https://www.pmda.go.jp/files/000234056.pdf.
|
| [15] |
FDA. Artificial Intelligence/Machine Learning(AI/ML)-Based Software as a Medical Device(SaMD) action plan[R/OL]. (2021-01-15)[2021-11-10].https://www.fda.gov/media/145022/ download.
|
| [16] |
ITU.Whitepaper for the ITU/WHO focus group on Artificial Intelligence for health [EB/OL]. (2020-10-13)[2021-11-10]. https:// www.itu.int/go/fgai4h.
|
| [17] |
AI4EU.Verifiable and explainable risk forecasting Artificial Intelligence framework [EB/OL].（2021-11-26）[2021-11-28]. https://www.ai4europe.eu/business-and-industry/healthcare.
|
| [18] |
National Institutes for Food and Drug Control. The kick-off meeting of the “Development and application demonstration of the full life cycle testing platform for medical Artificial Intelligence products” was held online[EB/OL].（2020-08-03）[2021-11-10]. https:// www.nifdc. org.cn/nifdc/gzdt/ywdt/20200803095421924.html.
|
| [19] |
Wang Z F, Huang X L, Yang J，et al. Universal adversarial perturbation generated by attacking layer-wise relevance propagation [C]. New Yok: IEEE, 2020: 431–436.
|
| [20] |
王士泉，李言生，苏明亮，等. 医疗人工智能产品应用效果的评 估框架与流程研究 [J]. 医疗卫生装备, 2020, 41(1): 62–65. Wang S Q, Li Y S, Su M L, et al. Medical Artificial Intelligence product application evaluation framework and process research [J]. Chinese Medical Equipment Journal, 2020, 41(1): 62–65.
|
/
| 〈 |
|
〉 |
