人工智能医疗器械用数据集管理与评价方法研究

Strategy for Management and Quality Control of Datasets in Artificial Intelligence Medical Device

数据集是人工智能(Artificial Intelligence,AI)医疗器械研发、训练、验证、优化的重要资源,是产业链上游影响产品质量的重要因素。在人工智能医疗器械产业起步的关键时期,如何规范和引导企业、医院、第三方机构等社会力量建设与管理数据集,关系到行业的长远健康发展与持久繁荣,关系到人工智能医疗器械产品的有效性、安全性和风险性,也关系到我国海量医学数据红利能否释放。国内外对于人工智能医疗器械用数据集的管理与评价尚未建立专业标准,缺乏专用的可操作的方法体系。本文结合数据集建设实践经验,对人工智能医疗器械用数据集管理与评价策略提出具体的建议,从医疗器械质量评价与生产质量管理体系的视角明确具体的工作流程与方法要求,旨在帮助行业加强对数据集质量管理的重视,通过对其质量管理体系的科学管理,确保和提升人工智能医疗器械产品质量。

Datasets are important resources for the development,training,verification and optimization of artificial intelligence medical device,which have important impact on the quality of AI products.In the beginning period of market translation,it is important to regulate and guide companies,hospitals,academia and third party institutes to build and manage datasets for AI correctly,which determines the safety,effectiveness and long-term success of AI medical device,and further impacts the utilization of medical big data in China.However,there are no specific standards and clear methods for the management and assessment of datasets for AI medical devices.With reference to practical experience on dataset development,we proposed detailed advice on the quality management and evaluation of AI datasets,clarified working procedure and requirement,which might help industry recognize the importance of dataset quality management and promote scientific quality management system,further ensure product quality.