目的 研究健康档案数据集的质量控制策略，旨在促进该领域的人工智能医疗器械专用标准规范。方法 根据健康档 案数据集的技术特点和开发场景，对此类数据集的特殊性进行分析，梳理质控要点丨采用不同方法，对具体的数据模态开 展抽样检验实验，比较和评估资源消耗。结果 与生产验收抽验方法相比，质量监督抽验方法需要的样本量更低，更具有经 济性和可行性。结论 本文研究的思路和方法有助于促进健康档案数据集的质量控制，加速智能化快速部署医院等特殊情景 的数据集建设。

Objective To study the strategy for the quality control of health record datasets with the aim of promoting specific standard specifications for artificial intelligence medical devices in this field. Methods The particularity of such datasets was analyzed and the key points of quality control were analyzed according to technical features and development scenarios of health record datasets. Different methods were used to carry out sampling inspection experiments on specific data modes to compare and evaluate the consumption of resources. Results Comparison with the production acceptance sampling method, the sampling method of quality supervision required lower sample size, which was more feasibility in practice. Conclusion The ideas and methods studied in thispaper are helpful to promote the quality control of health record datasets, and accelerate the development of corresponding datasets for special scenarios such as intelligent rapid deployable hospital.