Radiotherapy has raised increasing interest based on artificial intelligence algorithms such as ma- chine learning or deep learning. Federated learning is a machine learning approach that enables multiple parties to collaboratively train a model without sharing their data.

The overall goal of this project is to implement a federated learning framework that can be used to train models that translate medical images from one modality to another, without the need to cen- tralize all the data in one location. The key advantage of adopting federated learning, especially in the context of image translation within the domain of radiotherapy, lies in its ability to foster collab- orative training while upholding the utmost privacy of confidential medical information.
Unlike conventional methods that require consolidating data in a centralized repository, federated learning allows disparate healthcare institutions to collectively enhance model performance with- out compromising the security of individual datasets. Moreover, the versatility of this federated learning framework extends to the broader medical domain. By training models on local datasets from various hospitals, the resulting models can be finely tuned to accommodate the unique char- acteristics of each institution's patient population. This tailored approach not only enhances the model's accuracy but also ensures its adaptability to new data emanating from diverse healthcare settings.