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

Fully convolutional networks (FCNs), including UNet and VNet, are widely-usednetwork architectures for semantic segmentation in recent studies. However,conventional FCN is typically trained by the cross-entropy or Dice loss, whichonly calculates the error between predictions and ground-truth labels forpixels individually. This often results in non-smooth neighborhoods in thepredicted segmentation. To address this problem, we propose a two-stageframework, with the first stage to quickly localize the prostate region and thesecond stage to precisely segment the prostate by a multi-task UNetarchitecture. We introduce a novel online metric learning module throughvoxel-wise sampling in the multi-task network. Therefore, the proposed networkhas a dual-branch architecture that tackles two tasks: 1) a segmentationsub-network aiming to generate the prostate segmentation, and 2) a voxel-metriclearning sub-network aiming to improve the quality of the learned feature spacesupervised by a metric loss. Specifically, the voxel-metric learningsub-network samples tuples (including triplets and pairs) in voxel-levelthrough the intermediate feature maps. Unlike conventional deep metric learningmethods that generate triplets or pairs in image-level before the trainingphase, our proposed voxel-wise tuples are sampled in an online manner andoperated in an end-to-end fashion via multi-task learning. To evaluate theproposed method, we implement extensive experiments on a real CT image datasetconsisting of 339 patients. The ablation studies show that our method caneffectively learn more representative voxel-level features compared with theconventional learning methods with cross-entropy or Dice loss. And thecomparisons show that the proposed method outperforms the state-of-the-artmethods by a reasonable margin.

MetricUNet: Synergistic Image- and Voxel-Level Learning for Precise CT Prostate Segmentation via Online Sampling