摘要
目的:设计医学人工智能产品检测平台,以实现多模态的人工智能产品的全流程质量评价。方法:该平台采用SpringBoot、SpringCloud、Python Django等工具进行开发,由数据入库、数据标注、数据配置、产品送检、算法测试、数据集质量评价等模块组成。通过鉴权流程设计保证权限控制与访问安全,通过与云平台专线连接的方式进行本地化部署,实现多中心/医院数据的安全存储。结果:该平台支持开展算法部署、数据上传、专家标注、性能验证、算法调优、再评价等全流程服务,可为用户提供人工智能产品研发阶段、上市前、上市后的性能评价。结论:该平台能够按照用户需求提供医学人工智能产品检测评价服务,为医学人工智能产品建立了全生命周期的检测服务体系。
Objective To design a medical artificial intelligence(AI) product testing platform to enable whole-process quality evaluation of multimodal AI products. Methods The platform was developed using SpringBoot, SpringCloud and Python Django, which consisted of several modules for data loading, data annotation, data configuration, product submission, algorithm testing and dataset quality evaluation. Permission control and access security were ensured through authentication process design, and localized deployment was carried out through a dedicated line connection with the cloud platform to achieve secure storage of multi center/hospital data. Results The platform developed facilitated the whole-process service involving in algorithm deployment, data upload, expert labeling, performance verification, algorithm tuning and re-evaluation, and could be used for the performance evaluation of medical AI products at R & D stage and before and after launch. Conclusion The platform is able to provide testing and evaluation services according to user needs, establishing a full life-cycle testing service system for medical AI products.
作者
孟祥峰
王浩
张超
李佳戈
李静莉
MENG Xiang-feng;WANG Hao;ZHANG Chao;LI Jia-ge;LI Jing-li(Institute for Medical Devices Control,National Institutes for Food and Drug Control,Beijing 102629,China)
出处
《医疗卫生装备》
CAS
2023年第1期1-7,共7页
Chinese Medical Equipment Journal
基金
国家重点研发计划项目(2019YFB1404805)。
关键词
医学人工智能
人工智能产品
人工智能产品检测
检测平台
medical artificial intelligence
artificial intelligence product
testing of artificial intelligence product
testing platform
