摘要
随着医学技术的发展,基于计算机视觉的医疗图像分割技术逐渐被应用于医学实践中.针对当前医疗图像分割算法在实际应用时性能不足、泛化能力弱的问题,将联邦学习与主动学习结合,既能够保证训练数据的隐私性,又能够主动挑选有利于训练的样本.实验结果表明,所提出的主动联邦方法,在医疗图像分割任务上达到了较高的准确率,为该技术在医疗实践中的进一步发展做出了贡献.
With the development of medical technology,medical image segmentation technology based on computer vision has been gradually applied in medical practice.Aiming at the problems of insufficient performance and weak generalization ability of current medical image segmentation algorithm in practical application,this paper proposes to combine federated learning with active learning,which can guarantee the privacy of training data and actively select samples beneficial to training.Experimental results show that the proposed active federation method achieves high accuracy in medical image segmentation,which contributes to the further development of this technology in medical practice.
作者
刘靖宇
杨诚奕
邓云迪
LIU Jingyu;YANG Chengyi;DENG Yundi(Medicine School of University of Electronic Science and Technology of China,Chengdu Sichuan 611731;Information and Communication Engineering School of University of Electronic Science and Technology of China,Chengdu Sichuan 611731)
出处
《四川文理学院学报》
2022年第5期88-93,共6页
Sichuan University of Arts and Science Journal
基金
四川省科技厅项目(2021YFG0018)。
关键词
联邦学习
主动学习
医疗图像分割
匹配网络
全卷积网络
federal learning
active learning
medical image segmentation
matching network
fully convolutional network
