FedZaCt: Federated Learning with Z Average and Cross-Teaching on Image Segmentation
In Federated Learning (FL), data communication among clients is denied. However, it is difficult to learn from the decentralized client data, which is under-sampled, especially for segmentation tasks that need to extract enough contextual semantic information. Existing FL studies always average clie...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-10-01
|
| Series: | Electronics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2079-9292/11/20/3262 |
| _version_ | 1797473679201796096 |
|---|---|
| author | Tingyang Yang Jingshuang Xu Mengxiao Zhu Shan An Ming Gong Haogang Zhu |
| author_facet | Tingyang Yang Jingshuang Xu Mengxiao Zhu Shan An Ming Gong Haogang Zhu |
| author_sort | Tingyang Yang |
| collection | DOAJ |
| description | In Federated Learning (FL), data communication among clients is denied. However, it is difficult to learn from the decentralized client data, which is under-sampled, especially for segmentation tasks that need to extract enough contextual semantic information. Existing FL studies always average client models to one global model in segmentation tasks while neglecting the diverse knowledge extracted by the models. To maintain and utilize the diverse knowledge, we propose a novel training paradigm called Federated Learning with Z-average and Cross-teaching (FedZaCt) to deal with segmentation tasks. From the model parameters’ aspect, the Z-average method constructs individual client models, which maintain diverse knowledge from multiple client data. From the model distillation aspect, the Cross-teaching method transfers the other client models’ knowledge to supervise the local client model. In particular, FedZaCt does not have the global model during the training process. After training, all client models are aggregated into the global model by averaging all client model parameters. The proposed methods are applied to two medical image segmentation datasets including our private aortic dataset and a public HAM10000 dataset. Experimental results demonstrate that our methods can achieve higher Intersection over Union values and Dice scores. |
| first_indexed | 2024-03-09T20:19:51Z |
| format | Article |
| id | doaj.art-d97ca2eb1d084d44b44d33767246da0d |
| institution | Directory Open Access Journal |
| issn | 2079-9292 |
| language | English |
| last_indexed | 2024-03-09T20:19:51Z |
| publishDate | 2022-10-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Electronics |
| spelling | doaj.art-d97ca2eb1d084d44b44d33767246da0d2023-11-23T23:52:20ZengMDPI AGElectronics2079-92922022-10-011120326210.3390/electronics11203262FedZaCt: Federated Learning with Z Average and Cross-Teaching on Image SegmentationTingyang Yang0Jingshuang Xu1Mengxiao Zhu2Shan An3Ming Gong4Haogang Zhu5State Key Laboratory of Software Development Environment, Beihang University, Beijing 100191, ChinaState Key Laboratory of Software Development Environment, Beihang University, Beijing 100191, ChinaSchool of Information Science and Technology, North China University of Technology, Beijing 100144, ChinaJD Health International Inc., Beijing 100176, ChinaBeijing Laboratory for Cardiovascular Precision Medicine, Department of Cardiac Surgery, Beijing Anzhen Hospital, Capital Medical University, Beijing 100029, ChinaState Key Laboratory of Software Development Environment, Beihang University, Beijing 100191, ChinaIn Federated Learning (FL), data communication among clients is denied. However, it is difficult to learn from the decentralized client data, which is under-sampled, especially for segmentation tasks that need to extract enough contextual semantic information. Existing FL studies always average client models to one global model in segmentation tasks while neglecting the diverse knowledge extracted by the models. To maintain and utilize the diverse knowledge, we propose a novel training paradigm called Federated Learning with Z-average and Cross-teaching (FedZaCt) to deal with segmentation tasks. From the model parameters’ aspect, the Z-average method constructs individual client models, which maintain diverse knowledge from multiple client data. From the model distillation aspect, the Cross-teaching method transfers the other client models’ knowledge to supervise the local client model. In particular, FedZaCt does not have the global model during the training process. After training, all client models are aggregated into the global model by averaging all client model parameters. The proposed methods are applied to two medical image segmentation datasets including our private aortic dataset and a public HAM10000 dataset. Experimental results demonstrate that our methods can achieve higher Intersection over Union values and Dice scores.https://www.mdpi.com/2079-9292/11/20/3262Federated LearningsegmentationZ-averageCross-teaching |
| spellingShingle | Tingyang Yang Jingshuang Xu Mengxiao Zhu Shan An Ming Gong Haogang Zhu FedZaCt: Federated Learning with Z Average and Cross-Teaching on Image Segmentation Electronics Federated Learning segmentation Z-average Cross-teaching |
| title | FedZaCt: Federated Learning with Z Average and Cross-Teaching on Image Segmentation |
| title_full | FedZaCt: Federated Learning with Z Average and Cross-Teaching on Image Segmentation |
| title_fullStr | FedZaCt: Federated Learning with Z Average and Cross-Teaching on Image Segmentation |
| title_full_unstemmed | FedZaCt: Federated Learning with Z Average and Cross-Teaching on Image Segmentation |
| title_short | FedZaCt: Federated Learning with Z Average and Cross-Teaching on Image Segmentation |
| title_sort | fedzact federated learning with z average and cross teaching on image segmentation |
| topic | Federated Learning segmentation Z-average Cross-teaching |
| url | https://www.mdpi.com/2079-9292/11/20/3262 |
| work_keys_str_mv | AT tingyangyang fedzactfederatedlearningwithzaverageandcrossteachingonimagesegmentation AT jingshuangxu fedzactfederatedlearningwithzaverageandcrossteachingonimagesegmentation AT mengxiaozhu fedzactfederatedlearningwithzaverageandcrossteachingonimagesegmentation AT shanan fedzactfederatedlearningwithzaverageandcrossteachingonimagesegmentation AT minggong fedzactfederatedlearningwithzaverageandcrossteachingonimagesegmentation AT haogangzhu fedzactfederatedlearningwithzaverageandcrossteachingonimagesegmentation |