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

Organ at risk (OAR) segmentation in computed tomography (CT) imagery is adifficult task for automated segmentation methods and can be crucial fordownstream radiation treatment planning. U-net has become a de-facto standardfor medical image segmentation and is frequently used as a common baseline inmedical image segmentation tasks. In this paper, we develop a multiple decoderU-net architecture where a noisy auxiliary decoder is used to generate noisysegmentation. The segmentation from the main branch and the noisy segmentationfrom the auxiliary branch are used together to estimate the attention. Ourcontribution is the development of a new attention module which derives theattention from the softmax probabilities of two decoder branches. The union andintersection of two segmentation masks from two branches carry the informationwhere both decoders agree and disagree. The softmax probabilities from regionsof agreement and disagreement are the indicators of low and high uncertainty.Thus, the probabilities of those selected regions are used as attention in thebottleneck layer of the encoder and passes only through the main decoder forsegmentation. For accurate contour segmentation, we also developed a CTintensity integrated regularization loss. We tested our model on two publiclyavailable OAR challenge datasets, Segthor and LCTSC respectively. We trained 12models on each dataset with and without the proposed attention model andregularization loss to check the effectiveness of the attention module and theregularization loss. The experiments demonstrate a clear accuracy improvement(2\% to 5\% Dice) on both datasets. Code for the experiments will be madeavailable upon the acceptance for publication.

Uncertainty Driven Bottleneck Attention U-net for OAR Segmentation