The Future of Digital Health with Federated Learning

Hancox Jonny
Hancox Jonny
Albarqouni Shadi
Albarqouni Shadi
Galtier Mathieu N.
Galtier Mathieu N.
Landman Bennett
Landman Bennett
Maier-Hein Klaus
Maier-Hein Klaus
Sheller Micah
Sheller Micah

NPJ DIGITAL MEDICINE, pp. 1192020.

Cited by: 0|Views57
Weibo:
This paper considers key factors contributing to this issue, explores how Federated Learning may provide a solution for the future of digital health and highlights the challenges and considerations that need to be addressed

Abstract:

Data-driven machine learning (ML) has emerged as a promising approach for building accurate and robust statistical models from medical data, which is collected in huge volumes by modern healthcare systems. Existing medical data is not fully exploited by ML primarily because it sits in data silos and privacy concerns restrict access to thi...More

Code:

Data:

0
Full Text
Bibtex
Weibo
Introduction
  • Data-driven Machine Learning has emerged as a promising approach for building accurate and robust statistical models from medical data, which is collected in huge volumes by modern healthcare systems.
  • Applied to digital health this means that FL enables insights to be gained collaboratively across institutions, e.g. in the form of a global or consensus model, without sharing the patient data.
Highlights
  • Data-driven Machine Learning has emerged as a promising approach for building accurate and robust statistical models from medical data, which is collected in huge volumes by modern healthcare systems
  • The aim of this paper is to provide context and detail for the community regarding the benefits and impact of Federated Learning (FL) for medical applications (Section 2) as well as highlighting key considerations and challenges of implementing FL in the context of digital health (Section 3)
  • The medical FL use-case is inherently different from other domains, e.g. in terms of number of participants and data diversity, and while recent surveys investigate the research advances and open questions of FL [14, 11, 15], we focus on what it means for digital health and what is needed to enable it
  • Transfer Learning, for example, is a well-established approach of model-sharing that makes it possible to tackle problems with deep neural networks that have millions of parameters, despite the lack of extensive, local datasets that are required for training from scratch: a model is first trained on a large dataset and further optimised on the actual target data
  • By enabling multiple parties to train collaboratively without the need to exchange or centralise datasets, FL neatly addresses issues related to egress of sensitive medical data
Results
  • Whereas the initial application field mostly comprised mobile devices, participating entities in the case of healthcare could be institutions storing the data, e.g. hospitals, or medical devices itself, e.g. a CT scanner or even low-powered devices that are able to run computations locally.
  • The participating entities have to agree on a collaboration protocol and the high-dimensional medical data, which is predominant in the field of digital health, poses challenges by requiring models with huge numbers of parameters.
  • Note that aggregation strategies do not necessarily require information about the full model update; clients might choose to share only a subset of the model parameters for the sake of reducing communication overhead of redundant information, ensure better privacy preservation [10] or to produce multitask learning algorithms having only part of their parameters learned in a federated manner.
  • Transfer Learning, for example, is a well-established approach of model-sharing that makes it possible to tackle problems with deep neural networks that have millions of parameters, despite the lack of extensive, local datasets that are required for training from scratch: a model is first trained on a large dataset and further optimised on the actual target data.
  • To adopt this approach into a form of collaborative learning in a FL setup with continuous learning from different institutions, the participants can share their model with a peer-to-peer architecture in a ”round-robin” or parallel fashion and train in turn on their local data.
  • This unique characteristic of FL in healthcare brings opportunities as well as challenges such as (1) how to ensure data integrity when communicating; (2) how to design secure encryption methods to take advantage of the computational resources; (3) how to design appropriate node schedulers and make use of the distributed computational devices to reduce idle time.
  • Future efforts to apply artificial intelligence to healthcare tasks may strongly depend on collaborative strategies between multiple institutions rather than large centralised databases belonging to only one hospital or research laboratory.
  • Large-scale initiatives such as the MELLODDY project 2, the HealthChain project 3, the Trustworthy Federated Data Analytics (TFDA) and the German Cancer Consortium’s Joint Imaging Platform (JIP) 4 represent pioneering efforts to set the standards for safe, fair and innovative collaboration in healthcare research.
Conclusion
  • By enabling multiple parties to train collaboratively without the need to exchange or centralise datasets, FL neatly addresses issues related to egress of sensitive medical data.
  • The authors truly believe that its potential impact on precision medicine and improving medical care is very promising
Summary
  • Data-driven Machine Learning has emerged as a promising approach for building accurate and robust statistical models from medical data, which is collected in huge volumes by modern healthcare systems.
  • Applied to digital health this means that FL enables insights to be gained collaboratively across institutions, e.g. in the form of a global or consensus model, without sharing the patient data.
  • Whereas the initial application field mostly comprised mobile devices, participating entities in the case of healthcare could be institutions storing the data, e.g. hospitals, or medical devices itself, e.g. a CT scanner or even low-powered devices that are able to run computations locally.
  • The participating entities have to agree on a collaboration protocol and the high-dimensional medical data, which is predominant in the field of digital health, poses challenges by requiring models with huge numbers of parameters.
  • Note that aggregation strategies do not necessarily require information about the full model update; clients might choose to share only a subset of the model parameters for the sake of reducing communication overhead of redundant information, ensure better privacy preservation [10] or to produce multitask learning algorithms having only part of their parameters learned in a federated manner.
  • Transfer Learning, for example, is a well-established approach of model-sharing that makes it possible to tackle problems with deep neural networks that have millions of parameters, despite the lack of extensive, local datasets that are required for training from scratch: a model is first trained on a large dataset and further optimised on the actual target data.
  • To adopt this approach into a form of collaborative learning in a FL setup with continuous learning from different institutions, the participants can share their model with a peer-to-peer architecture in a ”round-robin” or parallel fashion and train in turn on their local data.
  • This unique characteristic of FL in healthcare brings opportunities as well as challenges such as (1) how to ensure data integrity when communicating; (2) how to design secure encryption methods to take advantage of the computational resources; (3) how to design appropriate node schedulers and make use of the distributed computational devices to reduce idle time.
  • Future efforts to apply artificial intelligence to healthcare tasks may strongly depend on collaborative strategies between multiple institutions rather than large centralised databases belonging to only one hospital or research laboratory.
  • Large-scale initiatives such as the MELLODDY project 2, the HealthChain project 3, the Trustworthy Federated Data Analytics (TFDA) and the German Cancer Consortium’s Joint Imaging Platform (JIP) 4 represent pioneering efforts to set the standards for safe, fair and innovative collaboration in healthcare research.
  • By enabling multiple parties to train collaboratively without the need to exchange or centralise datasets, FL neatly addresses issues related to egress of sensitive medical data.
  • The authors truly believe that its potential impact on precision medicine and improving medical care is very promising
Funding
  • This research was supported by the UK Research and Innovation London Medical Imaging & Artificial Intelligence Centre for Value Based Healthcare, by the Wellcome Trust, by the Intramural Research Program of the National Institutes of Health Clinical Center, as well as by the Helmholtz Initiative and Networking Fund (project ”Trustworthy Federated Data Analytics”)
  • His lab has received research support from Ping An and NVIDIA
  • Author SA is supported by the PRIME programme of the German Academic Exchange Service (DAAD) with funds from the German Federal Ministry of Education and Research (BMBF)
  • Author SB is supported by the National Institutes of Health (NIH)
  • Author MNG is supported by the HealthChain (BPIFrance) and Melloddy (IMI2) projects
Reference
  • Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning,” nature, vol. 521, no. 7553, p. 436, 2015.
    Google ScholarLocate open access versionFindings
  • G. Chartrand, P. M. Cheng, E. Vorontsov, M. Drozdzal, S. Turcotte, C. J. Pal, S. Kadoury, and A. Tang, “Deep learning: a primer for radiologists,” Radiographics, vol. 37, no. 7, pp. 2113–2131, 2017.
    Google ScholarLocate open access versionFindings
  • J. De Fauw, J. R. Ledsam, B. Romera-Paredes, S. Nikolov, N. Tomasev, S. Blackwell, H. Askham, X. Glorot, B. O’Donoghue, D. Visentin et al., “Clinically applicable deep learning for diagnosis and referral in retinal disease,” Nature medicine, vol. 24, no. 9, p. 1342, 2018.
    Google ScholarLocate open access versionFindings
  • C. Sun, A. Shrivastava, S. Singh, and A. Gupta, “Revisiting unreasonable effectiveness of data in deep
    Google ScholarFindings
  • [6] L. Rocher, J. M. Hendrickx, and Y.-A. De Montjoye, “Estimating the success of re-identifications in incomplete datasets using generative models,” Nature communications, vol. 10, no. 1, pp. 1–9, 2019.
    Google ScholarLocate open access versionFindings
  • [7] F.-C. Yeh, J. M. Vettel, A. Singh, B. Poczos, S. T. Grafton, K. I. Erickson, W.-Y. I. Tseng, and T. D. Verstynen, “Quantifying differences and similarities in whole-brain white matter architecture using local connectome fingerprints,” PLoS computational biology, vol. 12, no. 11, p. e1005203, 2016.
    Google ScholarLocate open access versionFindings
  • [8] C. Wachinger, P. Golland, W. Kremen, B. Fischl, M. Reuter, A. D. N. Initiative et al., “Brainprint: A discriminative characterization of brain morphology,” NeuroImage, vol. 109, pp. 232–248, 2015.
    Google ScholarLocate open access versionFindings
  • [9] B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, “Communication-efficient learning of deep networks from decentralized data,” in Artificial Intelligence and Statistics, 2017, pp. 1273–1282.
    Google ScholarLocate open access versionFindings
  • [10] T. Li, A. K. Sahu, A. Talwalkar, and V. Smith, “Federated learning: Challenges, methods, and future directions,” arXiv preprint arXiv:1908.07873, 2019.
    Findings
  • [11] Q. Yang, Y. Liu, T. Chen, and Y. Tong, “Federated machine learning: Concept and applications,” ACM Transactions on Intelligent Systems and Technology (TIST), vol. 10, no. 2, p. 12, 2019.
    Google ScholarLocate open access versionFindings
  • [12] W. Li, F. Milletarı, D. Xu, N. Rieke, J. Hancox, W. Zhu, M. Baust, Y. Cheng, S. Ourselin, M. J. Cardoso et al., “Privacy-preserving federated brain tumour segmentation,” in International Workshop on Machine Learning in Medical Imaging. Springer, 2019, pp. 133–141.
    Google ScholarLocate open access versionFindings
  • [13] M. J. Sheller, G. A. Reina, B. Edwards, J. Martin, and S. Bakas, “Multi-institutional deep learning modeling without sharing patient data: A feasibility study on brain tumor segmentation,” in International MICCAI Brainlesion Workshop. Springer, 2018, pp. 92–104.
    Google ScholarLocate open access versionFindings
  • [14] P. Kairouz, H. B. McMahan, B. Avent, A. Bellet, M. Bennis, A. N. Bhagoji, K. Bonawitz, Z. Charles, G. Cormode, R. Cummings et al., “Advances and open problems in federated learning,” arXiv preprint arXiv:1912.04977, 2019.
    Findings
  • [15] J. Xu and F. Wang, “Federated learning for healthcare informatics,” arXiv preprint arXiv:1911.06270, 2019.
    Findings
  • [16] “Ibm’s merge healthcare acquisition,” https://www.reuters.com/article/us-mergehealthcare-m-a-ibm/ibm-to-buy-merge-healthcarein-1-billion-deal-idUSKCN0QB1ML20150806, 2015 (accessed February 10, 2020).
    Findings
  • [17] “Nhs scotland’s national safe haven,” https://www.gov.scot/publications/charter-safehavens-scotland-handling-unconsented-datanational-health-service-patient-records-supportresearch-statistics/pages/4/, 2015 (accessed February 10, 2020).
    Findings
  • [18] M. Cuggia and S. Combes, “The french health data hub and the german medical informatics initiatives: Two national projects to promote data sharing in healthcare,” Yearbook of medical informatics, vol. 28, no. 01, pp. 195–202, 2019.
    Google ScholarLocate open access versionFindings
  • [19] “Health data https://www.hdruk.ac.uk/, 10, 2020.
    Findings
  • [20] O. Sporns, G. Tononi, and R. Kotter, “The human connectome: a structural description of the human brain,” PLoS computational biology, vol. 1, no. 4, 2005.
    Google ScholarLocate open access versionFindings
  • [21] C. Sudlow, J. Gallacher, N. Allen, V. Beral, P. Burton, J. Danesh, P. Downey, P. Elliott, J. Green, M. Landray et al., “Uk biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age,” PLoS medicine, vol. 12, no. 3, 2015.
    Google ScholarLocate open access versionFindings
  • [22] K. Clark, B. Vendt, K. Smith, J. Freymann, J. Kirby, P. Koppel, S. Moore, S. Phillips, D. Maffitt, M. Pringle et al., “The cancer imaging archive (tcia): maintaining and operating a public information repository,” Journal of digital imaging, vol. 26, no. 6, pp. 1045–1057, 2013.
    Google ScholarLocate open access versionFindings
  • [23] X. Wang, Y. Peng, L. Lu, Z. Lu, M. Bagheri, and R. M. Summers, “Chestx-ray8: Hospital-scale chest x-ray database and benchmarks on weaklysupervised classification and localization of common thorax diseases,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 2097–2106.
    Google ScholarLocate open access versionFindings
  • [24] K. Yan, X. Wang, L. Lu, and R. M. Summers, “Deeplesion: automated mining of large-scale lesion annotations and universal lesion detection with deep learning,” Journal of medical imaging, vol. 5, no. 3, p. 036501, 2018.
    Google ScholarLocate open access versionFindings
  • [25] K. Tomczak, P. Czerwinska, and M. Wiznerowicz, “The cancer genome atlas (tcga): an immeasurable source of knowledge,” Contemporary oncology, vol. 19, no. 1A, p. A68, 2015.
    Google ScholarLocate open access versionFindings
  • [26] C. R. Jack Jr, M. A. Bernstein, N. C. Fox, P. Thompson, G. Alexander, D. Harvey, B. Borowski, P. J. Britson, J. L. Whitwell, C. Ward et al., “The alzheimer’s disease neuroimaging initiative (adni): Mri methods,” Journal of Magnetic Resonance Imaging: An Official Journal of the International Society for Magnetic Resonance in Medicine, vol. 27, no. 4, pp. 685–691, 2008.
    Google ScholarLocate open access versionFindings
  • [27] G. Litjens, P. Bandi, B. Ehteshami Bejnordi, O. Geessink, M. Balkenhol, P. Bult, A. Halilovic, M. Hermsen, R. van de Loo, R. Vogels et al., “1399 h&e-stained sentinel lymph node sections of breast cancer patients: the camelyon dataset,” GigaScience, vol. 7, no. 6, p. giy065, 2018.
    Google ScholarLocate open access versionFindings
  • [28] B. H. Menze, A. Jakab, S. Bauer, J. KalpathyCramer, K. Farahani, J. Kirby, Y. Burren, N. Porz, J. Slotboom, R. Wiest et al., “The multimodal brain tumor image segmentation benchmark (brats),” IEEE transactions on medical imaging, vol. 34, no. 10, pp. 1993–2024, 2014.
    Google ScholarLocate open access versionFindings
  • [29] A. L. Simpson, M. Antonelli, S. Bakas, M. Bilello, K. Farahani, B. Van Ginneken, A. Kopp-Schneider, B. A. Landman, G. Litjens, B. Menze et al., “A large annotated medical image dataset for the development and evaluation of segmentation algorithms,” arXiv preprint arXiv:1902.09063, 2019.
    Findings
  • [30] R. Shokri, M. Stronati, C. Song, and V. Shmatikov, “Membership inference attacks against machine learning models,” in 2017 IEEE Symposium on Security and Privacy (SP). IEEE, 2017, pp. 3–18.
    Google ScholarLocate open access versionFindings
  • [31] A. Sablayrolles, M. Douze, Y. Ollivier, C. Schmid, and H. Jegou, “White-box vs black-box: Bayes optimal strategies for membership inference,” arXiv preprint arXiv:1908.11229, 2019.
    Findings
  • [32] C. Zhang, S. Bengio, M. Hardt, B. Recht, and O. Vinyals, “Understanding deep learning requires rethinking generalization,” arXiv preprint arXiv:1611.03530, 2016.
    Findings
  • [33] N. Carlini, C. Liu, U. Erlingsson, J. Kos, and D. Song, “The secret sharer: Evaluating and testing unintended memorization in neural networks,” in 28th {USENIX} Security Symposium ({USENIX} Security 19), 2019, pp. 267–284.
    Google ScholarLocate open access versionFindings
  • [34] M. Abadi, A. Chu, I. Goodfellow, H. B. McMahan, I. Mironov, K. Talwar, and L. Zhang, “Deep learning with differential privacy,” in Proceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security. ACM, 2016, pp. 308– 318.
    Google ScholarLocate open access versionFindings
  • [35] R. Shokri and V. Shmatikov, “Privacy-preserving deep learning,” in Proceedings of the 22nd ACM SIGSAC conference on computer and communications security. ACM, 2015, pp. 1310–1321.
    Google ScholarLocate open access versionFindings
  • [36] J. Konecny, H. B. McMahan, D. Ramage, and P. Richtarik, “Federated optimization: Distributed machine learning for on-device intelligence,” arXiv preprint arXiv:1610.02527, 2016.
    Findings
  • [37] A. G. Roy, S. Siddiqui, S. Polsterl, N. Navab, and C. Wachinger, “Braintorrent: A peer-to-peer environment for decentralized federated learning,” arXiv preprint arXiv:1905.06731, 2019.
    Findings
  • [38] A. Lalitha, O. C. Kilinc, T. Javidi, and F. Koushanfar, “Peer-to-peer federated learning on graphs,” arXiv preprint arXiv:1901.11173, 2019.
    Findings
  • [39] H. B. McMahan, D. Ramage, K. Talwar, and L. Zhang, “Learning differentially private recurrent language models,” International Conference on Learning Representations (ICLR), 2018.
    Google ScholarLocate open access versionFindings
  • [40] K. Chang, N. Balachandar, C. Lam, D. Yi, J. Brown, A. Beers, B. Rosen, D. L. Rubin, and J. KalpathyCramer, “Distributed deep learning networks among institutions for medical imaging,” Journal of the American Medical Informatics Association, vol. 25, no. 8, pp. 945–954, 2018.
    Google ScholarLocate open access versionFindings
  • [41] H.-C. Shin, H. R. Roth, M. Gao, L. Lu, Z. Xu, I. Nogues, J. Yao, D. Mollura, and R. M. Summers, “Deep convolutional neural networks for computeraided detection: Cnn architectures, dataset characteristics and transfer learning,” IEEE transactions on medical imaging, vol. 35, no. 5, pp. 1285–1298, 2016.
    Google ScholarLocate open access versionFindings
  • [42] N. Tajbakhsh, J. Y. Shin, S. R. Gurudu, R. T. Hurst, C. B. Kendall, M. B. Gotway, and J. Liang, “Convolutional neural networks for medical image analysis: Full training or fine tuning?” IEEE transactions on medical imaging, vol. 35, no. 5, pp. 1299–1312, 2016.
    Google ScholarLocate open access versionFindings
  • [43] M. McCloskey and N. J. Cohen, “Catastrophic interference in connectionist networks: The sequential learning problem,” in Psychology of learning and motivation. Elsevier, 1989, vol. 24, pp. 109–165.
    Google ScholarFindings
  • [44] I. J. Goodfellow, M. Mirza, D. Xiao, A. Courville, and Y. Bengio, “An empirical investigation of catastrophic forgetting in gradient-based neural networks,” arXiv preprint arXiv:1312.6211, 2013.
    Findings
  • [45] Z. Li and D. Hoiem, “Learning without forgetting,” IEEE transactions on pattern analysis and machine intelligence, vol. 40, no. 12, pp. 2935–2947, 2017.
    Google ScholarLocate open access versionFindings
  • [46] N. Shoham, T. Avidor, A. Keren, N. Israel, D. Benditkis, L. Mor-Yosef, and I. Zeitak, “Overcoming forgetting in federated learning on non-iid data,” arXiv preprint arXiv:1910.07796, 2019.
    Findings
  • [47] G. Song and W. Chai, “Collaborative learning for deep neural networks,” in Advances in Neural Information Processing Systems, 2018, pp. 1832–1841.
    Google ScholarLocate open access versionFindings
  • [48] S. Ruder, “An overview of multi-task learning in deep neural networks,” arXiv preprint arXiv:1706.05098, 2017.
    Findings
  • [49] P. H. Jin, Q. Yuan, F. Iandola, and K. Keutzer, “How to scale distributed deep learning?” arXiv preprint arXiv:1611.04581, 2016.
    Findings
  • [50] P. Goyal, P. Dollar, R. Girshick, P. Noordhuis, L. Wesolowski, A. Kyrola, A. Tulloch, Y. Jia, and K. He, “Accurate, large minibatch sgd: Training imagenet in 1 hour,” arXiv preprint arXiv:1706.02677, 2017.
    Findings
  • [51] T. Ben-Nun and T. Hoefler, “Demystifying parallel and distributed deep learning: An in-depth concurrency analysis,” ACM Computing Surveys (CSUR), vol. 52, no. 4, pp. 1–43, 2019.
    Google ScholarLocate open access versionFindings
  • [52] M. Yamazaki, A. Kasagi, A. Tabuchi, T. Honda, M. Miwa, N. Fukumoto, T. Tabaru, A. Ike, and K. Nakashima, “Yet another accelerated sgd: Resnet-50 training on imagenet in 74.7 seconds,” arXiv preprint arXiv:1903.12650, 2019.
    Findings
  • [53] B. Wu, S. Zhao, G. Sun, X. Zhang, Z. Su, C. Zeng, and Z. Liu, “P3sgd: Patient Privacy Preserving SGD for Regularizing Deep CNNs in Pathological Image Classification,” arXiv:1905.12883 [cs], May 2019, arXiv: 1905.12883. [Online]. Available: http://arxiv.org/abs/1905.12883 on Computer Communications. IEEE, 2019, pp.2512–2520.
    Findings
  • [56] B. Hitaj, G. Ateniese, and F. Perez-Cruz, “Deep models under the gan: Information leakage from collaborative deep learning,” in Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security, ser. CCS ’17. New York, NY, USA: Association for Computing Machinery, 2017, p. 603–618. [Online]. Available: https://doi.org/10.1145/3133956.3134012
    Locate open access versionFindings
  • [57] X. Li, Y. Gu, N. Dvornek, L. Staib, P. Ventola, and J. S. Duncan, “Multi-site fmri analysis using privacy-preserving federated learning and domain adaptation: Abide results,” arXiv preprint arXiv:2001.05647, 2020.
    Findings
  • [58] T. Li, A. K. Sahu, M. Zaheer, M. Sanjabi, A. Talwalkar, and V. Smith, “Federated optimization in heterogeneous networks,” 2018.
    Google ScholarFindings
  • [59] H. B. McMahan, E. Moore, D. Ramage, S. Hampson et al., “Communication-efficient learning of deep networks from decentralized data,” arXiv preprint arXiv:1602.05629, 2016.
    Findings
  • [60] Y. Zhao, M. Li, L. Lai, N. Suda, D. Civin, and V. Chandra, “Federated learning with non-iid data,” ArXiv, vol. abs/1806.00582, 2018.
    Findings
  • [61] A. Ghorbani and J. Zou, “Data shapley: Equitable valuation of data for machine learning,” arXiv preprint arXiv:1904.02868, 2019.
    Findings
  • [54] L. Zhu, Z. Liu, and S. Han, “Deep leakage from gradients,” in Advances in Neural Information Processing Systems, 2019, pp. 14 747–14 756.
    Google ScholarLocate open access versionFindings
  • [55] Z. Wang, M. Song, Z. Zhang, Y. Song, Q. Wang, and H. Qi, “Beyond inferring class representatives: User-level privacy leakage from federated learning,” in IEEE INFOCOM 2019-IEEE Conference
    Google ScholarLocate open access versionFindings
Your rating :
0

 

Tags
Comments