Whole brain radiomics for clustered federated personalization in brain tumor segmentation

Matthis Manthe, Stefan Duffner, Carole Lartizien
Medical Imaging with Deep Learning, PMLR 227:957-977, 2024.

Abstract

Federated learning and its application to medical image segmentation have recently become a popular research topic. This training paradigm suffers from statistical heterogeneity between participating institutions’ local datasets, incurring convergence slowdown as well as potential accuracy loss compared to classical training. To mitigate this effect, federated personalization emerged as the federated optimization of one model per distribution. We propose a novel personalization algorithm tailored to the feature shift induced by the usage of different scanners and acquisition parameters by different institutions. This method is the first to account for both inter and intra-institution feature shifts (multiple scanners used in a single institution). It is based on the computation, within each centre, of a series of radiomic features capturing the global texture of each 3D image volume, followed by a clustering analysis pooling all feature vectors transferred from the local institutions to the central server. Each computed clustered decentralized dataset (potentially including data from different institutions) then serves to finetune a global model obtained through classical federated learning. We validate our approach on the Federated Brain Tumor Segmentation 2022 Challenge dataset (FeTS2022).

Cite this Paper

BibTeX
@InProceedings{pmlr-v227-manthe24a, title = {Whole brain radiomics for clustered federated personalization in brain tumor segmentation}, author = {Manthe, Matthis and Duffner, Stefan and Lartizien, Carole}, booktitle = {Medical Imaging with Deep Learning}, pages = {957--977}, year = {2024}, editor = {Oguz, Ipek and Noble, Jack and Li, Xiaoxiao and Styner, Martin and Baumgartner, Christian and Rusu, Mirabela and Heinmann, Tobias and Kontos, Despina and Landman, Bennett and Dawant, Benoit}, volume = {227}, series = {Proceedings of Machine Learning Research}, month = {10--12 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v227/manthe24a/manthe24a.pdf}, url = {https://proceedings.mlr.press/v227/manthe24a.html}, abstract = {Federated learning and its application to medical image segmentation have recently become a popular research topic. This training paradigm suffers from statistical heterogeneity between participating institutions’ local datasets, incurring convergence slowdown as well as potential accuracy loss compared to classical training. To mitigate this effect, federated personalization emerged as the federated optimization of one model per distribution. We propose a novel personalization algorithm tailored to the feature shift induced by the usage of different scanners and acquisition parameters by different institutions. This method is the first to account for both inter and intra-institution feature shifts (multiple scanners used in a single institution). It is based on the computation, within each centre, of a series of radiomic features capturing the global texture of each 3D image volume, followed by a clustering analysis pooling all feature vectors transferred from the local institutions to the central server. Each computed clustered decentralized dataset (potentially including data from different institutions) then serves to finetune a global model obtained through classical federated learning. We validate our approach on the Federated Brain Tumor Segmentation 2022 Challenge dataset (FeTS2022).} }
Endnote
%0 Conference Paper %T Whole brain radiomics for clustered federated personalization in brain tumor segmentation %A Matthis Manthe %A Stefan Duffner %A Carole Lartizien %B Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2024 %E Ipek Oguz %E Jack Noble %E Xiaoxiao Li %E Martin Styner %E Christian Baumgartner %E Mirabela Rusu %E Tobias Heinmann %E Despina Kontos %E Bennett Landman %E Benoit Dawant %F pmlr-v227-manthe24a %I PMLR %P 957--977 %U https://proceedings.mlr.press/v227/manthe24a.html %V 227 %X Federated learning and its application to medical image segmentation have recently become a popular research topic. This training paradigm suffers from statistical heterogeneity between participating institutions’ local datasets, incurring convergence slowdown as well as potential accuracy loss compared to classical training. To mitigate this effect, federated personalization emerged as the federated optimization of one model per distribution. We propose a novel personalization algorithm tailored to the feature shift induced by the usage of different scanners and acquisition parameters by different institutions. This method is the first to account for both inter and intra-institution feature shifts (multiple scanners used in a single institution). It is based on the computation, within each centre, of a series of radiomic features capturing the global texture of each 3D image volume, followed by a clustering analysis pooling all feature vectors transferred from the local institutions to the central server. Each computed clustered decentralized dataset (potentially including data from different institutions) then serves to finetune a global model obtained through classical federated learning. We validate our approach on the Federated Brain Tumor Segmentation 2022 Challenge dataset (FeTS2022).
APA
Manthe, M., Duffner, S. & Lartizien, C.. (2024). Whole brain radiomics for clustered federated personalization in brain tumor segmentation. Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 227:957-977 Available from https://proceedings.mlr.press/v227/manthe24a.html.

Related Material