Abstract
The potential for deep learning to improve medical image analysis is often stymied by the difficulty in acquiring and collecting sufficient data to train models. One major barrier to data acquisition is the private and sensitive nature of the data in question, as concerns about patient privacy, among others, make data sharing between institutions difficult. Distributed learning avoids the need to share data centrally by training models locally. One approach to distributed learning is federated learning, where models are trained in parallel at local institutions and aggregated together into a global model. The 2021 Federated Tumor Segmentation (FeTS) challenge focuses on federated learning for brain tumor segmentation using magnetic resonance imaging scans collected from a real-world federation of collaborating institutions. We developed a federated training algorithm that uses a combination of variable local epochs in each federated round, a decaying learning rate, and an ensemble weight aggregation function. When testing on unseen validation data our model trained with federated learning achieves very similar performance (average DSC score of 0.674) to a central model trained on pooled data (average DSC score 0.685). When our federated learning algorithm was evaluated on unseen training and testing data, it achieved similar performances on the FeTS challenge leaderboards 1 and 2 (average DSC scores of 0.623 and 0.608, respectively). This federated learning algorithm offers an approach to training deep learning learning models without the need to share private and sensitive patient data.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Lo Vercio, L., et al.: Supervised machine learning tools: a tutorial for clinicians. Journal of Neural Engineering 17(6), 062001 (Oct 9 2020). https://doi.org/10.1088/1741-2552/abbff2
Hinton, G.: Deep learning-a technology with the potential to transform health care. JAMA - Journal of the American Medical Association, 320(11), pp. 1101–1102. American Med-ical Association (Sep 18 2018). https://doi.org/10.1001/jama.2018.11100
Kaissis, G.A., Makowski, M.R., Rückert, D., Braren, R.F.: Secure, privacypreserving and federated machine learning in medical imaging. Nat. Mach. Intell. 2(6), 305–311 (Jun 2020). https://doi.org/10.1038/s42256-020-0186-1
MacEachern, S.J., Forkert, N.D.: Machine learning for precision medicine. Genome 64(4), 416–425 (2021). https://doi.org/10.1139/gen-2020-0131. Epub 2020 Oct 22 PMID: 33091314 Apr
Tuladhar, A., Gill, S., Ismail, Z., Forkert, N.D.: Building machine learning models without sharing patient data: A simulation-based analysis of distributed learning by en-sembling. J. Biomed. Inform. 106, 103424 (2020). https://doi.org/10.1016/j.jbi.2020.103424. Jun.
McMahan, H.B., Moore, E., Ramage, D., Hampson, S., Arcas, B.A.: Communication-Efficient Learning of Deep Networks from Decentralized Data. Arxiv (2016)
Yang, Q., Liu, Y., Chen, T., Tong, Y.: Federated machine learning: Concept and applications. ACM Trans. Intell. Syst. Technol. 10(2), 1–19 (2019). https://doi.org/10.1145/3298981. Jan.
Kaissis, G.A., Makowski, M.R., Rückert, D., Braren, R.F.: Secure, privacy-preserving and federated machine learning in medical imaging. Nature Machine Intelligence 2(6), 305–311 (2020). https://doi.org/10.1038/s42256-020-0186-1
Chang, K., et al.: Distributed deep learning networks among institutions for medical imaging. J. Am. Med. Informatics Assoc. 25(8), 945–954 (2018). https://doi.org/10.1093/jamia/ocy017. Aug.
Remedios, S.W., et al.: Distributed deep learning across multisite datasets for generalized CT hemorrhage segmentation. Med. Phys. 47(1), 89–98 (2020). https://doi.org/10.1002/mp.13880. Jan.
Reina, G.A., et al.: OpenFL: An open-source framework for Federated Learning. arXiv preprint arXiv:2105.06413 (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
Bakas, S., et al.: Advancing the cancer genome atlas glioma MRI collections with expert segmentation labels and radiomic features. Nature Scientific Data 4, 170117 (2017). https://doi.org/10.1038/SDATA.2017.117
Bakas, S., et al.: Segmentation labels and radiomic features for the pre-operative scans of the TCGA-GBM collection. The Cancer Imaging Archive (2017). https://doi.org/10.7937/K9/TCIA.2017.KLXWJJ1Q
Bakas, S., et al.: Segmentation labels and radiomic features for the pre-operative scans of the TCGA-LGG collection. The Cancer Imaging Archive (2017). https://doi.org/10.7937/K9/TCIA.2017.GJQ7R0EF
Acknowledgements
This work was supported by the Natural Sciences and Engineering Research Council of Canada Postdoctoral Fellowship (AT) and Discovery Grant (NDF), MITACS Globalink Research Internship (LT), University of Calgary BME Research Scholarship (RS), Canada Research Chairs program (NDF), the River Fund at Calgary Foundation (NDF), and the Canadian Institutes of Health Research (CIHR).
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Ethics declarations
Conflicts of Interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Tuladhar, A., Tyagi, L., Souza, R., Forkert, N.D. (2022). Federated Learning Using Variable Local Training for Brain Tumor Segmentation. In: Crimi, A., Bakas, S. (eds) Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries. BrainLes 2021. Lecture Notes in Computer Science, vol 12963. Springer, Cham. https://doi.org/10.1007/978-3-031-09002-8_35
Download citation
DOI: https://doi.org/10.1007/978-3-031-09002-8_35
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-09001-1
Online ISBN: 978-3-031-09002-8
eBook Packages: Computer ScienceComputer Science (R0)