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

Owing to the success of transformer models, recent works study theirapplicability in 3D medical segmentation tasks. Within the transformer models,the self-attention mechanism is one of the main building blocks that strives tocapture long-range dependencies. However, the self-attention operation hasquadratic complexity which proves to be a computational bottleneck, especiallyin volumetric medical imaging, where the inputs are 3D with numerous slices. Inthis paper, we propose a 3D medical image segmentation approach, named UNETR++,that offers both high-quality segmentation masks as well as efficiency in termsof parameters, compute cost, and inference speed. The core of our design is theintroduction of a novel efficient paired attention (EPA) block that efficientlylearns spatial and channel-wise discriminative features using a pair ofinter-dependent branches based on spatial and channel attention. Our spatialattention formulation is efficient having linear complexity with respect to theinput sequence length. To enable communication between spatial andchannel-focused branches, we share the weights of query and key mappingfunctions that provide a complimentary benefit (paired attention), while alsoreducing the overall network parameters. Our extensive evaluations on fivebenchmarks, Synapse, BTCV, ACDC, BRaTs, and Decathlon-Lung, reveal theeffectiveness of our contributions in terms of both efficiency and accuracy. OnSynapse, our UNETR++ sets a new state-of-the-art with a Dice Score of 87.2%,while being significantly efficient with a reduction of over 71% in terms ofboth parameters and FLOPs, compared to the best method in the literature. Code:https://github.com/Amshaker/unetr_plus_plus.

UNETR++: Delving into Efficient and Accurate 3D Medical Image Segmentation