摘要
目的/意义探究应用联邦学习开展临床研究,在保护患者隐私数据的同时开展大模型训练,推动医学研究发展.方法/过程介绍联邦学习技术框架,重点分析其在医学影像、疾病预测、个性化治疗和新药研发等领域应用的巨大潜力和可能遇到的问题.结果/结论 联邦学习提供了一种在医学大数据分析中合作而不共享数据的能力,为跨机构的协同合作提供可能.目前联邦学习在医学研究中尚面临数据异质性、通信效率、模型泛化及安全性等问题,有待进一步深入研究.
Abstract
Purpose/Significance To explore the application of federated learning to conduct clinical research,and to carry out large model training while protecting patients'privacy data,so as to promote the development of medical research.Method/Process The pa-per introduces the federated learning technology framework,and analyzes its great potential and possible problems in the fields of medical imaging,disease prediction,personalized therapy,new drug development,etc.Result/Conclusion Federated learning provides the ca-pability to collaborate without sharing data in medical big data analysis,and provides the possibility for cross-institutional collaboration.At present,the problems of federated learning in medical research,such as data heterogeneity,communication efficiency,model general-ization and safety,need to be further studied.
NETL
NSTL
万方数据