Distributed contrastive learning for medical image segmentation

计算机科学 上传 人工智能 特征(语言学) 匹配(统计) 分割 机器学习 深度学习 监督学习 半监督学习 数据挖掘 人工神经网络 统计 操作系统 哲学 语言学 数学
作者
Yawen Wu,Dewen Zeng,Zhepeng Wang,Yiyu Shi,Jingtong Hu
出处
期刊:Medical Image Analysis [Elsevier]
卷期号:81: 102564-102564 被引量:28
标识
DOI:10.1016/j.media.2022.102564
摘要

Supervised deep learning needs a large amount of labeled data to achieve high performance. However, in medical imaging analysis, each site may only have a limited amount of data and labels, which makes learning ineffective. Federated learning (FL) can learn a shared model from decentralized data. But traditional FL requires fully-labeled data for training, which is very expensive to obtain. Self-supervised contrastive learning (CL) can learn from unlabeled data for pre-training, followed by fine-tuning with limited annotations. However, when adopting CL in FL, the limited data diversity on each site makes federated contrastive learning (FCL) ineffective. In this work, we propose two federated self-supervised learning frameworks for volumetric medical image segmentation with limited annotations. The first one features high accuracy and fits high-performance servers with high-speed connections. The second one features lower communication costs, suitable for mobile devices. In the first framework, features are exchanged during FCL to provide diverse contrastive data to each site for effective local CL while keeping raw data private. Global structural matching aligns local and remote features for a unified feature space among different sites. In the second framework, to reduce the communication cost for feature exchanging, we propose an optimized method FCLOpt that does not rely on negative samples. To reduce the communications of model download, we propose the predictive target network update (PTNU) that predicts the parameters of the target network. Based on PTNU, we propose the distance prediction (DP) to remove most of the uploads of the target network. Experiments on a cardiac MRI dataset show the proposed two frameworks substantially improve the segmentation and generalization performance compared with state-of-the-art techniques.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
更新
大幅提高文件上传限制,最高150M (2024-4-1)

科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
刚刚
下一秒发布了新的文献求助10
3秒前
英俊的铭应助松帅采纳,获得10
4秒前
CipherSage应助鱼鱼鱼采纳,获得10
5秒前
kdqiu完成签到,获得积分10
6秒前
丰富的大地完成签到,获得积分10
6秒前
xiao123789发布了新的文献求助10
7秒前
zxk完成签到,获得积分10
8秒前
w婷完成签到 ,获得积分10
8秒前
Slemon完成签到,获得积分10
8秒前
Photon完成签到,获得积分10
9秒前
10秒前
顾矜应助tian采纳,获得10
10秒前
科研通AI2S应助ZLY采纳,获得10
13秒前
无敌小天天完成签到 ,获得积分10
13秒前
炸鸡完成签到 ,获得积分10
14秒前
强强强强完成签到,获得积分10
15秒前
16秒前
phase完成签到,获得积分10
21秒前
22秒前
22秒前
100完成签到,获得积分10
24秒前
晓晨完成签到 ,获得积分10
25秒前
佳期发布了新的文献求助10
27秒前
科研通AI2S应助科研通管家采纳,获得10
27秒前
脑洞疼应助科研通管家采纳,获得10
28秒前
28秒前
完美世界应助科研通管家采纳,获得10
28秒前
隐形曼青应助科研通管家采纳,获得10
28秒前
情怀应助科研通管家采纳,获得200
28秒前
英俊的铭应助科研通管家采纳,获得10
28秒前
搜集达人应助科研通管家采纳,获得10
28秒前
领导范儿应助科研通管家采纳,获得10
28秒前
华仔应助科研通管家采纳,获得10
28秒前
Joy应助科研通管家采纳,获得70
29秒前
薰硝壤应助科研通管家采纳,获得10
29秒前
赘婿应助科研通管家采纳,获得10
29秒前
哎嘿应助科研通管家采纳,获得10
29秒前
乐乐应助科研通管家采纳,获得10
29秒前
情怀应助房天川采纳,获得10
29秒前
高分求助中
中国国际图书贸易总公司40周年纪念文集: 史论集 2500
Sustainability in Tides Chemistry 2000
Дружба 友好报 (1957-1958) 1000
The Data Economy: Tools and Applications 1000
How to mix methods: A guide to sequential, convergent, and experimental research designs 700
Mantiden - Faszinierende Lauerjäger – Buch gebraucht kaufen 600
PraxisRatgeber Mantiden., faszinierende Lauerjäger. – Buch gebraucht kaufe 600
热门求助领域 (近24小时)
化学 医学 生物 材料科学 工程类 有机化学 生物化学 物理 内科学 纳米技术 计算机科学 化学工程 复合材料 基因 遗传学 催化作用 物理化学 免疫学 量子力学 细胞生物学
热门帖子
关注 科研通微信公众号,转发送积分 3111664
求助须知:如何正确求助?哪些是违规求助? 2761878
关于积分的说明 7667857
捐赠科研通 2416960
什么是DOI,文献DOI怎么找? 1282924
科研通“疑难数据库(出版商)”最低求助积分说明 619212
版权声明 599512