Unsupervised Domain Adaptation for Medical Image Segmentation via Self-Training of Early Features

Rasha Sheikh, Thomas Schultz
Proceedings of The 5th International Conference on Medical Imaging with Deep Learning, PMLR 172:1096-1107, 2022.

Abstract

U-Net models provide a state-of-the-art approach for medical image segmentation, but their accuracy is often reduced when training and test images come from different domains, such as different scanners. Recent work suggests that, when limited supervision is available for domain adaptation, early U-Net layers benefit the most from a refinement. This motivates our proposed approach for self-supervised refinement, which does not require any manual annotations, but instead refines early layers based on the richer, higher-level information that is derived in later layers of the U-Net. This is achieved by adding a segmentation head for early features, and using the final predictions of the network as pseudo-labels for refinement. This strategy reduces detrimental effects of imperfection in the pseudo-labels, which are unavoidable given the domain shift, by retaining their probabilistic nature and restricting the refinement to early layers. Experiments on two medical image segmentation tasks confirm the effectiveness of this approach, even in a one-shot setting, and compare favorably to a baseline method for unsupervised domain adaptation.

Cite this Paper

BibTeX
@InProceedings{pmlr-v172-sheikh22a, title = {Unsupervised Domain Adaptation for Medical Image Segmentation via Self-Training of Early Features}, author = {Sheikh, Rasha and Schultz, Thomas}, booktitle = {Proceedings of The 5th International Conference on Medical Imaging with Deep Learning}, pages = {1096--1107}, year = {2022}, editor = {Konukoglu, Ender and Menze, Bjoern and Venkataraman, Archana and Baumgartner, Christian and Dou, Qi and Albarqouni, Shadi}, volume = {172}, series = {Proceedings of Machine Learning Research}, month = {06--08 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v172/sheikh22a/sheikh22a.pdf}, url = {https://proceedings.mlr.press/v172/sheikh22a.html}, abstract = {U-Net models provide a state-of-the-art approach for medical image segmentation, but their accuracy is often reduced when training and test images come from different domains, such as different scanners. Recent work suggests that, when limited supervision is available for domain adaptation, early U-Net layers benefit the most from a refinement. This motivates our proposed approach for self-supervised refinement, which does not require any manual annotations, but instead refines early layers based on the richer, higher-level information that is derived in later layers of the U-Net. This is achieved by adding a segmentation head for early features, and using the final predictions of the network as pseudo-labels for refinement. This strategy reduces detrimental effects of imperfection in the pseudo-labels, which are unavoidable given the domain shift, by retaining their probabilistic nature and restricting the refinement to early layers. Experiments on two medical image segmentation tasks confirm the effectiveness of this approach, even in a one-shot setting, and compare favorably to a baseline method for unsupervised domain adaptation.} }
Endnote
%0 Conference Paper %T Unsupervised Domain Adaptation for Medical Image Segmentation via Self-Training of Early Features %A Rasha Sheikh %A Thomas Schultz %B Proceedings of The 5th International Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2022 %E Ender Konukoglu %E Bjoern Menze %E Archana Venkataraman %E Christian Baumgartner %E Qi Dou %E Shadi Albarqouni %F pmlr-v172-sheikh22a %I PMLR %P 1096--1107 %U https://proceedings.mlr.press/v172/sheikh22a.html %V 172 %X U-Net models provide a state-of-the-art approach for medical image segmentation, but their accuracy is often reduced when training and test images come from different domains, such as different scanners. Recent work suggests that, when limited supervision is available for domain adaptation, early U-Net layers benefit the most from a refinement. This motivates our proposed approach for self-supervised refinement, which does not require any manual annotations, but instead refines early layers based on the richer, higher-level information that is derived in later layers of the U-Net. This is achieved by adding a segmentation head for early features, and using the final predictions of the network as pseudo-labels for refinement. This strategy reduces detrimental effects of imperfection in the pseudo-labels, which are unavoidable given the domain shift, by retaining their probabilistic nature and restricting the refinement to early layers. Experiments on two medical image segmentation tasks confirm the effectiveness of this approach, even in a one-shot setting, and compare favorably to a baseline method for unsupervised domain adaptation.
APA
Sheikh, R. & Schultz, T.. (2022). Unsupervised Domain Adaptation for Medical Image Segmentation via Self-Training of Early Features. Proceedings of The 5th International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 172:1096-1107 Available from https://proceedings.mlr.press/v172/sheikh22a.html.

Related Material