To be held virtually on January 20th 2023
The Artificial Intelligence and Machine-Learning Subgroup (AIMS) of the North American Imaging in MS Cooperative (NAIMS) will be hosting a virtual workshop focused on federated learning. Federated learning is a technique allowing for training of AI models across multiple servers/centers without directly sharing the underlying data. Rather than sharing the data, the model is shared between institutions instead, thereby circumventing the intrinsic challenges associated with data privacy and governance.
The purpose of the workshop is to gather a critical mass of people to advance understanding of federated learning applications in MS imaging and to brainstorm on suitable solutions for MS lesion segmentation and other practical AI problems in the context of a multi-site study. There will be a series of talks given by experts, with hands-on experience in federated learning for medical image segmentation, followed by a discussion session.
Please mark it on your calendars! All levels of experience and backgrounds are welcome.
The workshop is free 🙂
👉 Please register here so we have an idea of the number of attendees.
🎥 Zoom link
🗒 Miro board. This board is used as a brainstorming platform. You will need to create a login to be able to view/edit this board.
This is a single day workshop. We hope that times will accomodate people on the west and east coast of North America.
Times are Eastern Time Zone (UTC-05:00):
| Time | Topic |
|---|---|
| 10:00 | Brief intro (what is FL, purpose of the workshop) |
| 10:15 | Introduce Miro whiteboard: It will allow people having ideas/suggestions during the workshop to enter them in the whiteboard |
| 10:30 | Alexandros Karargyris (IHU Strasbourg): https://www.medperf.org/ 👉 Slides |
| 10:50 | Marco Lorenzi (INRIA): https://fedbiomed.gitlabpages.inria.fr/ 👉 Slides |
| 11:10 | Sarthak Pati (University of Pennsylvania): The Federated Tumor Segmentation initiative (https://www.med.upenn.edu/cbica/fets/) |
| 11:30 | Julien Cohen-Adad (Polytechnique Montreal / Mila): CODA (https://github.com/coda-platform). 👉 Slides |
| 12:00 | Break (zoom link stays open) |
| 13:00 | Brennan Nichyporuk (McGill University / Mila): Generalization across cohorts/sources (and accounting for biases) for MS lesion (https://arxiv.org/pdf/2108.00713.pdf) |
| 13:20 | Farhad Imam (Gates Ventures): GRIP Architecture (Gates Ventures): GRIP Architecture |
| 13:40 | Peter Chang (University of California, Irvine): TensorFlow framework for medical FL |
| 14:00 | Lisa Nicole Schneider (Charité Universitätsmedizin, Berlin): Hands-on experience with NVIDIA Flare |
| 14:30 | Discussion, next steps, assign tasks |
| 17:00 | Adjourn |