Abstract
With the ever increasing importance and requirements to ensure data privacy, federated learning emerges as an promising technology for training deep learning models without having hospitals to share the raw data. MICCAI Federated Tumor Segmentation Challenge 2021 is the first international challenge on federated learning to strengthen the understanding of real-world challenges and create practical solutions in the related area. In the challenge of this year, we proposed a series of new aggregation strategies towards improving the learning performance in the context of non-IID and imbalanced data distribution. We also designed a simple collaborator selection scheme to shorten the training time while achieving a good level of model performance for brain tumor segmentation.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
The mean Dice score refers to the metric ‘valid_dice’ computed by the base package, which denotes the average dice across the dice of all the four label 0, 1, 2 and 4.
- 2.
We refer to the records of one patient as one data sample.
- 3.
The convergence score is calculated as the area under the curve of validation Dice where the horizontal axis is the runtime, so a higher convergence score indicates a better convergence performance.
- 4.
We tried 20 rounds once but the test run ended up with termination at 15-th rounds due to simulation time limit, so 10 rounds should be sufficient for comparison among different methods.
- 5.
For example, ten FL rounds over partitioning 2 using FedAvg with all collaborator selected can take roughly 40 h to complete on our DGX workstation.
References
Modern Medicine. https://www.cnbc.com/2018/02/22/medical-errors-third-leading-cause-of-death-in-america.html
Diagnostic Errors. https://psnet.ahrq.gov/primers/primer/12
Tresp, V., Marc Overhage, J., Bundschus, M., Rabizadeh, S., Fasching, P.A., Yu, S.: Going digital: a survey on digitalization and large-scale data analytics in healthcare. Proc. IEEE 104(11), 2180–2206 (2016)
Chen, M., Qian, Y., Chen, J., Hwang, K., Mao, S., Hu, L.: Privacy protection and intrusion avoidance for cloudlet-based medical data sharing. IEEE Trans. Cloud Comput. 8(4), 1274–1283 (2020)
Mcmahan, H.B., Moore, E., Ramage, D., Hampson, S., Arcas, B.A.A.: Communication-efficient learning of deep networks from decentralized data. In: International Conference on Artificial Intelligence and Statistics, vol. 54, pp. 1273–1282 (2017)
Li, X., Huang, K., et al.: On the convergence of FedAvg on non-IID data. In: International Conference on Learning Representations (ICLR) (2020)
FeTS 2022 Challenge. https://www.synapse.org/#!Synapse:syn28546456/wiki/617246
Sheller, M.J., Reina, G.A., Edwards, B., Martin, J., Bakas, S.: Multi-institutional deep learning modeling without sharing patient data: a feasibility study on brain tumor segmentation. In: Crimi, A., Bakas, S., Kuijf, H., Keyvan, F., Reyes, M., van Walsum, T. (eds.) BrainLes 2018. LNCS, vol. 11383, pp. 92–104. Springer, Cham (2019). https://doi.org/10.1007/978-3-030-11723-8_9
Li, W., Milletarì, F., et al.: Privacy-preserving federated brain tumour segmentation. In: Proceedings of the International Workshop on Machine Learning in Medical Imaging, pp. 133–141 (2019)
Yi, L., Zhang, J., Zhang, R., et al.: SU-Net: an efficient encoder-decoder model of federated learning for brain tumor segmentation. In: International Conference on Artificial Neural Networks, pp. 761–773 (2020)
Guo, P., Wang, P., Zhou, J., Jiang, S., Patel, V.M.: Multi-institutional collaborations for improving deep learning-based magnetic resonance image reconstruction using federated learning. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 2423–2432 (2021)
Pati, S., et al.: The federated tumor segmentation (FeTS) challenge. arXiv preprint arXiv:2105.05874 (2021)
Sheller, M.J., et al.: Federated learning in medicine: facilitating multi-institutional collaborations without sharing patient data. Nat. Sci. Rep. 10, 12598 (2020). https://doi.org/10.1038/s41598-020-69250-1
Baid, U., et al.: The RSNA-ASNR-MICCAI BraTS 2021 benchmark on brain tumor segmentation and radiogenomic classification. arXiv preprint arXiv: 2107.02314 (2021)
Reina, G.A., et al.: OpenFL: an open-source framework for federated learning. arXiv preprint arXiv: 2105.06413 (2021)
Karargyris, A., Umeton, R., Sheller, M., Aristizabal, A., George, J., Bala, S.: MedPerf: open benchmarking platform for medical artificial intelligence using federated evaluation. arXiv preprint arXiv: arXiv:2110.01406 (2021)
Hsu, T.M.H., Qi, H., Brown, M.: Measuring the effects of non-identical data distribution for federated visual classification. arXiv preprint arXiv:1909.06335 (2019)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Wang, Y., Kanagavelu, R., Wei, Q., Yang, Y., Liu, Y. (2023). Model Aggregation for Federated Learning Considering Non-IID and Imbalanced Data Distribution. In: Bakas, S., et al. Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries. BrainLes 2022. Lecture Notes in Computer Science, vol 14092. Springer, Cham. https://doi.org/10.1007/978-3-031-44153-0_19
Download citation
DOI: https://doi.org/10.1007/978-3-031-44153-0_19
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-44152-3
Online ISBN: 978-3-031-44153-0
eBook Packages: Computer ScienceComputer Science (R0)
