摘要
As the integration of medical big data and artificial intelligence advances,the secure sharing of medical data has become a key driving force for advancing disease research and clinical diagnosis.Federated learning,a distributed approach enabling collaborative data processing without sharing raw data,offers promising solutions to challenges in multi-center medical data sharing.This review summarizes the progress of federated learning in multi-center medical data processing,analyzed from four perspectives:system architectures,data distribution strategies,clinical tasks,and algorithmic models.At the same time,this paper explores the challenges in practical applications,such as data heterogeneity,communication overhead,and privacy concerns.It proposes driving future research development by optimizing algorithms,strengthening privacy protection mechanisms,and enhancing computational efficiency.
基金
supported and funded by the National Natural Science Foundation of China(82101079)
the Key R&D Program of Jiangsu Province(BE2023836)
the National Key Research and Development Program of China(SQ2023YFC2400025).
