4th MICCAI Workshop on
“Distributed, Collaborative and Federated Learning”
Important Dates
- Location: Meeting Room 11 @ Vancouver Convention Center (East)
- Submission Deadlines: 30th of June 2023 AOE (updated)
- Notification of Acceptance: 16th of July 2023
- Camera Ready: 23rd of July 2023
- Workshop Date: 12th of October 2023
News
[October] Let's meet at Meeting Room 11 @Vancouver Convention Center!
[May] Call for papers and reviewers!
[May] Submission is open on CMT system.
[March] The DeCaF Workshop has been approved for MICCAI 2023
Distributed, Collaborative and Federated Learning
Deep learning, AI's fastest-growing field, empowers enormous advances in applications in both science and real-world scenarios. It has reached a consensus that models could be further improved with a growing amount of data. However, enabling learning on these huge datasets or training huge models in a timely manner requires distributing the learning on several devices. One particularity in the medical imaging setting is that data sharing across different institutions often becomes impractical due to strict privacy regulations as well as data ownership concerns, making the collection of large-scale diverse centralized datasets practically impossible.
Some of the problems, therefore, become: how can we train models in a federated way on several devices? And is it possible to achieve models as strong as those that can be trained on large centralized datasets without sharing data and breaching the restrictions on privacy and property? How can we ensure data privacy and mode generalizability? Federated learning (FL) allows different institutions to contribute to building more powerful models by performing collaborative training without sharing any training data. The trained model can be distributed across various institutions instead of the actual data. We hope that with FL and other forms of distributed and collaborative learning, the objective of training better and more robust models with higher clinical utility while protecting the privacy within the data can be achieved.
Call for Papers
Through the fourth MICCAI Workshop on Distributed, Collaborative and Federated Learning (DeCAF), we aim to provide a discussion forum to compare, evaluate and discuss methodological advancements and ideas around federated, distributed, and collaborative learning schemes that are applicable in the medical domain. We invite full paper (8-page) submissions using the MICCAI 2023 template through CMT (https://cmt3.research.microsoft.com/DeCaF2023/). Topics include but are not limited to:
- Federated, distributed learning, and other forms of collaborative learning
- Server-client and peer-to-peer learning
- Advanced data and model parallelism learning techniques
- Optimization methods for distributed or collaborative learning
- Privacy-preserving technique and security for distributed, federated, and collaborative learning
- Efficient communication and learning (multi-device, multi-node)
- Adversarial, inversion and other forms of attacks on distributed , federated, and collaborative learning
- Dealing with unbalanced (non-IID) data in federated and collaborative learning
- Diverse decentralized medical imaging data analysis
- Security-auditing system for federated learning
- Asynchronous learning
- Software tools and implementations of distributed, federated, and collaborative learning
- Model sharing techniques, sparse/partial learning of models
- Applications of federated/distributed/collaborative learning techniques: multi-task learning, model agnostic learning, meta-learning, etc.
Program
Keynote Session
Aline Talhouk
University of British Columbia, Canada
Dr. Aline Talhouk is an assistant professor in the Department of Obstetrics and Gynecology in the Faculty of Medicine at the University of British Columbia (UBC). She is principal investigator at OVCARE, where she directs a data science and informatics laboratory. Dr. Talhouk holds a PhD from UBC and has expertise in computational statistics and machine learning, specifically in the development and implementation of predictive models in women’s health and oncology. Her research leverages statistical computing, machine learning and artificial intelligence to translate -omics discoveries to clinical applications and bring individualized care to ovarian and endometrial cancer patients. Dr. Talhouk has also developed a nationally funded precision prevention program that uses prediction modeling to identify those at high risk for uterine cancer and direct them to risk-reducing interventions, targeted screening and surveillance. Dr. Talhouk is a Michael Smith Health Research BC Scholar and holds several grants from the Canadian Institutes of Health Research, the Canada Foundation for Innovation and the Canadian Cancer Society.
Topic: Overcoming Barriers to Implementing Federated Machine Learning in Healthcare: Policy Options for Ethical and Secure Data Collaboration
Abstract: In the world of healthcare, leveraging the power of artificial intelligence (AI) algorithms that learn from data can offer tremendous benefits. However, accessing the large volumes of health data required to train AI models has traditionally required centralizing data. This is often hindered by regulatory requirements due to concerns surrounding ethics, privacy, security. Federated machine learning (FL) emerges as a promising technical solution, enabling multiple parties to collaborate on training models without sharing their data, thereby addressing concerns associated with data pooling. Nonetheless, several barriers still impede the practical implementation of FL for health research. We conducted a comprehensive analysis by synthesizing information from four data sources: document and literature reviews, expert interviews, a validation workshop, and a survey of privacy, ethics, and security recommendations pertinent to FL. Through this analysis, we identified twelve technical, ethical, and socio-legal challenges faced by consortia seeking to implement FL in practice. To address these challenges, we proposed policy options that are proportionate to the potential risks and benefits. These policies create a conducive environment for innovation while upholding privacy standards and promoting the common good. By adopting the proposed policy options, we strive to overcome barriers hindering the implementation of FL in healthcare.
Patrick Foley
OpenFL, Intel
Patrick Foley is the lead architect for OpenFL, an open-source deep learning framework for Federated Learning, at Intel. He also serves as the chairman of the OpenFL Technical Steering committee under the LF AI & Data Foundation. During his 12-year tenure at Intel, Patrick has served as technical lead for federated machine learning research projects applied to genomics and medical imaging data. Additionally, he has contributed to the Broad Institute's Genomic Analytics Toolkit and was responsible for bringing Intel accelerator support into Microsoft's OnnxRuntime Inference Framework. Patrick obtained his M.S. from Georgia Tech in computer science with an emphasis in machine learning.
Topic: Scaling Secure Federated Learning
Abstract: Federated Learning has demonstrated enormous potential in the healthcare domain, but the onboarding of hospitals and research institutions is a bespoke process that can take months of preparation and code vetting before private data can be accessed. In this talk, I will address the software design principles that led to successfully onboarding more than 70 healthcare institutions across 6 continents for the FeTS Initiative. I’ll discuss threats posed by malicious actors as the size and trust boundary of federations grow, and share the role of software and hardware techniques in mitigating these risks. Finally, I’ll look to the future with a preview of federated learning standards, how they will mature the field, and their role in breaking down barriers to collaboration.
Meet the Organising Team
Contact: xiaoxiao.li(at)ece.ubc.ca
Shadi Albarqouni
University Hospital Bonn | Helmholtz Munich, Germany
Xiaoxiao Li
University of British Columbia, Canada
Nicola Rieke
NVIDIA, Germany
Daguang Xu
NVIDIA, USA
Meet the Outreach Comittee Team
Contact: chunyinh(at)ece.ubc.ca, manuela.bergau(at)ukbonn.de
Manuela Bergau
University Hospital Bonn, Germany
Chun-Yin Huang
University of British Columbia, Canada
Submission Guidelines
Format: Papers will be submitted electronically following Lecture Notes in Computer Science (LNCS) style of up to 8 + 2 pages (same as MICCAI 2023 format). Submissions exceeding page limit will be rejected without review. Latex style files can be found from Springer, which also contains Word instructions. The file format for submissions is Adobe Portable Document Format (PDF). Other formats will not be accepted.
Double Blind Review: DeCaF reviewing is double blind. Please review the Anonymity guidelines of MICCAI main conference, and confirm that the author field does not break anonymity.
Paper Submission: DeCaF uses the CMT system for online submission.
Supplemental Material: Supplemental material submission is optional, following same deadline as the main paper. Contents of the supplemental material would be referred to appropriately in the paper, while reviewers are not obliged to read them.
Submission Originality: Submissions should be original, no paper of substantially similar content should be under peer review or has been accepted for a publication elsewhere (conference/journal, not including archived work).
Proceedings: The proceedings of DeCaF 2023 will be published as part of the joint MICCAI Workshops proceedings with Springer (LNCS)
