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

Glaucoma is one of the leading causes of irreversible blindness. Segmentationof optic disc (OD) and optic cup (OC) on fundus images is a crucial step inglaucoma screening. Although many deep learning models have been constructedfor this task, it remains challenging to train an OD/OC segmentation model thatcould be deployed successfully to different healthcare centers. Thedifficulties mainly comes from the domain shift issue, i.e., the fundus imagescollected at these centers usually vary greatly in the tone, contrast, andbrightness. To address this issue, in this paper, we propose a novelunsupervised domain adaptation (UDA) method called Reconstruction-drivenDynamic Refinement Network (RDR-Net), where we employ a due-path segmentationbackbone for simultaneous edge detection and region prediction and design threemodules to alleviate the domain gap. The reconstruction alignment (RA) moduleuses a variational auto-encoder (VAE) to reconstruct the input image and thusboosts the image representation ability of the network in a self-supervisedway. It also uses a style-consistency constraint to force the network to retainmore domain-invariant information. The low-level feature refinement (LFR)module employs input-specific dynamic convolutions to suppress thedomain-variant information in the obtained low-level features. Theprediction-map alignment (PMA) module elaborates the entropy-driven adversariallearning to encourage the network to generate source-like boundaries andregions. We evaluated our RDR-Net against state-of-the-art solutions on fourpublic fundus image datasets. Our results indicate that RDR-Net is superior tocompeting models in both segmentation performance and generalization ability

Reconstruction-driven Dynamic Refinement based Unsupervised Domain Adaptation for Joint Optic Disc and Cup Segmentation