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

Segmentation tasks in medical imaging are inherently ambiguous: the boundaryof a target structure is oftentimes unclear due to image quality and biologicalfactors. As such, predicted segmentations from deep learning algorithms areinherently ambiguous. Additionally, "ground truth" segmentations performed byhuman annotators are in fact weak labels that further increase the uncertaintyof outputs of supervised models developed on these manual labels. To date, mostdeep learning segmentation studies utilize predicted segmentations withoutuncertainty quantification. In contrast, we explore the use of Monte Carlodropout U-Nets for the segmentation with additional quantification ofsegmentation uncertainty. We assess the utility of three measures ofuncertainty (Coefficient of Variation, Mean Pairwise Dice, and Mean VoxelwiseUncertainty) for the segmentation of a less ambiguous target structure (liver)and a more ambiguous one (liver tumors). Furthermore, we assess how the utilityof these measures changes with different patch sizes and cost functions. Ourresults suggest that models trained using larger patches and the weightedcategorical cross-entropy as cost function allow the extraction of moremeaningful uncertainty measures compared to smaller patches and soft dice loss.Among the three uncertainty measures Mean Pairwise Dice shows the strongestcorrelation with segmentation quality. Our study serves as a proof-of-conceptof how uncertainty measures can be used to assess the quality of a predictedsegmentation, potentially serving to flag low quality segmentations from agiven model for further human review.

Give me (un)certainty -- An exploration of parameters that affect segmentation uncertainty