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

Medical image segmentation plays a crucial role in advancing healthcaresystems for disease diagnosis and treatment planning. The u-shapedarchitecture, popularly known as U-Net, has proven highly successful forvarious medical image segmentation tasks. However, U-Net's convolution-basedoperations inherently limit its ability to model long-range dependencieseffectively. To address these limitations, researchers have turned toTransformers, renowned for their global self-attention mechanisms, asalternative architectures. One popular network is our previous TransUNet, whichleverages Transformers' self-attention to complement U-Net's localizedinformation with the global context. In this paper, we extend the 2D TransUNetarchitecture to a 3D network by building upon the state-of-the-art nnU-Netarchitecture, and fully exploring Transformers' potential in both the encoderand decoder design. We introduce two key components: 1) A Transformer encoderthat tokenizes image patches from a convolution neural network (CNN) featuremap, enabling the extraction of global contexts, and 2) A Transformer decoderthat adaptively refines candidate regions by utilizing cross-attention betweencandidate proposals and U-Net features. Our investigations reveal thatdifferent medical tasks benefit from distinct architectural designs. TheTransformer encoder excels in multi-organ segmentation, where the relationshipamong organs is crucial. On the other hand, the Transformer decoder proves morebeneficial for dealing with small and challenging segmented targets such astumor segmentation. Extensive experiments showcase the significant potential ofintegrating a Transformer-based encoder and decoder into the u-shaped medicalimage segmentation architecture. TransUNet outperforms competitors in variousmedical applications.

3D TransUNet: Advancing Medical Image Segmentation through Vision Transformers