Please note that the ITEA Office will be closed from 25 December 2024 to 1 January 2025 inclusive.
Code for federated training of segmentation model
- Project
- 20044 ASSIST
- Type
- New library
- Description
Code for federated training of a segmentation model through the federated learning framework FEDn
- Contact
- Mattias Åkesson, Scaleout Systems
- mattias@scaleoutsystems.com
- Research area(s)
- Federated learning, medical imaging, radiotherapy
- Technical features
A 3D U-Net is trained through federated learning using the FEDn framework. An arbitrary number of clients is supported. Model performance can be monitored using Scaleout Studio.
- Integration constraints
The segmentation model expects four MR volumes per patient, and ground truth segmentations for brain tumor and brain stem. The code is written in Python.
- Targeted customer(s)
Anyone working with federated learning or radiotherapy
- Conditions for reuse
See Github repository
- Confidentiality
- Public
- Publication date
- 09-09-2024
- Involved partners
- Linköping University (SWE)
- Scaleoutsystems (SWE)