English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Applying pre-trained medical segmentation models on out-of-domain imagesoften yields predictions of insufficient quality. Several strategies have beenproposed to maintain model performance, such as finetuning or unsupervised- andsource-free domain adaptation. These strategies set restrictive requirementsfor data availability. In this study, we propose to combine domaingeneralization and test-time adaptation to create a highly effective approachfor reusing pre-trained models in unseen target domains. Domain-generalizedpre-training on source data is used to obtain the best initial performance inthe target domain. We introduce the MIND descriptor previously used in imageregistration tasks as a further technique to achieve generalization and presentsuperior performance for small-scale datasets compared to existing approaches.At test-time, high-quality segmentation for every single unseen scan is ensuredby optimizing the model weights for consistency given different imageaugmentations. That way, our method enables separate use of source and targetdata and thus removes current data availability barriers. Moreover, thepresented method is highly modular as it does not require specific modelarchitectures or prior knowledge of involved domains and labels. We demonstratethis by integrating it into the nnUNet, which is currently the most popular andaccurate framework for medical image segmentation. We employ multiple datasetscovering abdominal, cardiac, and lumbar spine scans and compose severalout-of-domain scenarios in this study. We demonstrate that our method, combinedwith pre-trained whole-body CT models, can effectively segment MR images withhigh accuracy in all of the aforementioned scenarios. Open-source code can befound here: https://github.com/multimodallearning/DG-TTA

DG-TTA: Out-of-domain medical image segmentation through Domain Generalization and Test-Time Adaptation