ESTRO 2024 - Abstract Book

S4821

Physics - Quality assurance and auditing

ESTRO 2024

Conclusion:

We have demonstrated that federated training of segmentation models for radiation therapy, using clinical data from oncology departments, is feasible. The obtained Dice scores are not sufficiently high to be used in a clinical workflow, but our work should be seen as a proof of concept and is ongoing. Future work will also include Umeå as a third Swedish node.

Keywords: federated learning, segmentation model

References:

Matthew Brett, Christopher J. Markiewicz, Michael Hanke, Marc-Alexandre Cˆot e, Ben Cipollini, Paul McCarthy, Dorota Jarecka, Christopher P. Cheng, Yaroslav O. Halchenko, Michiel Cottaar, Eric Larson, Satrajit Ghosh, Demian Wassermann, Stephan Gerhard, Gregory R. Lee, Zvi Baratz, Hao-Ting Wang, Erik Kastman, Jakub Kaczmarzyk, Roberto Guidotti, Jonathan Daniel, Or Duek, Ariel Rokem, Cindee Madison, Dimitri Papadopou- los Orfanos, Anibal S olon, Brendan Moloney, F elix C. Morency, Mathias Goncalves, Ross Markello, Cameron Riddell, Christopher Burns, Jarrod Millman, Alexandre Gramfort, Jaakko Lepp akangas, Jasper J.F. van den Bosch, Robert D. Vincent, Henry Braun, Krish Subramaniam, Andrew Van, Krzysztof J. Gorgolewski, Pradeep Reddy Raamana, Julian Klug, B. Nolan Nichols, Eric M. Baker, Soichi Hayashi, Basile Pinsard, Christian Hasel- grove, Mark Hymers, Oscar Esteban, Serge Koudoro, Fernando P erez-Garc ıa, J erˆome Dock`es, Nikolaas N. Oosterhof, Bago Amirbekian, Horea Christian, Ian Nimmo-Smith, Ly Nguyen, Samir Reddigari, Samuel St-Jean, Egor Panfilov, Eleftherios Garyfallidis, Gael Varoquaux, Jon Haitz Legarreta, Kevin S. Hahn, Lea Waller, Oliver P. Hinds, Ben- net Fauber, Fabian Perez, Jacob Roberts, Jean-Baptiste Poline, Jon Stutters, Kesshi Jor- dan, Matthew Cieslak, Miguel Estevan Moreno, Tom aˇs Hrnˇciar, Valentin Haenel, Yannick Schwartz, Benjamin C Darwin, Bertrand Thirion, Carl Gauthier, Igor Solovey, Ivan Gon- zalez, Jath Palasubramaniam,