Split Learning for collaborative deep learning in healthcare
Разделённое обучение для совместного глубокого обучения в здравоохранении
2019-12-27
SCID: 54.1/vm85bdtf
Discuss with AI
chest X-rayscollaborative deep learningdistributed machine learningfundus photographsmedical image classificationsplit learning
Figures from the paper
Abstract (AI)
Shortage of labeled data has been holding the surge of deep learning in healthcare back, as sample sizes are often small, patient information cannot be shared openly, and multi-center collaborative studies are a burden to set up. Distributed machine learning methods promise to mitigate these problems. We argue for a split learning based approach and apply this distributed learning method for the first time in the medical field to compare performance against (1) centrally hosted and (2) non collaborative configurations for a range of participants. Two medical deep learning tasks are used to compare split learning to conventional single and multi center approaches: a binary classification problem of a data set of 9000 fundus photos, and multi-label classification problem of a data set of 156,535 chest X-rays. The several distributed learning setups are compared for a range of 1-50 distributed participants. Performance of the split learning configuration remained constant for any number of clients compared to a single center study, showing a marked difference compared to the non collaborative configuration after 2 clients (p < 0.001) for both sets. Our results affirm the benefits of collaborative training of deep neural networks in health care. Our work proves the significant benefit of distributed learning in healthcare, and paves the way for future real-world implementations.
Key Findings
1
Across 1–50 participants, split learning performance remained constant relative to single-center training, despite distributing the learning process.
2
Split learning significantly outperformed non-collaborative configurations after two clients for both medical tasks (p < 0.001).
3
Split learning was evaluated on fundus-photo binary classification and chest-X-ray multi-label classification datasets containing 9,000 and 156,535 images, respectively.
4
The findings support collaborative distributed training as a way to address limited labeled data, privacy constraints, and multi-center collaboration burdens in healthcare.
5
The study introduces split learning as a distributed deep-learning approach for collaborative healthcare research, representing its first medical application.
Research Object
Collaborative deep learning in healthcare using fundus photographs and chest X-rays
Research Subject
The performance and scalability of split-learning-based distributed training compared with centralized and non-collaborative configurations across varying numbers of participants
Publication Details
Publication Date
2019-12-27
Journal
arXiv (Cornell University)
Publisher
Cornell University
ISSN
2331-8422
Cited by
225
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest
References available in scid.ai5
Deep Residual Learning for Image Recognition2016
Adam: A Method for Stochastic Optimization2014
ImageNet: A large-scale hierarchical image database2009
Deep learning for healthcare: review, opportunities and challenges2017
Detailed comparison of communication efficiency of split learning and federated learning2019