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/

Zendy Plus

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

Zendy Tools

Zendy Open

Automated optic disc (OD) and optic cup (OC) segmentation in fundus images isrelevant to efficiently measure the vertical cup-to-disc ratio (vCDR), abiomarker commonly used in ophthalmology to determine the degree ofglaucomatous optic neuropathy. In general this is solved using coarse-to-finedeep learning algorithms in which a first stage approximates the OD and asecond one uses a crop of this area to predict OD/OC masks. While this approachis widely applied in the literature, there are no studies analyzing its realcontribution to the results. In this paper we present a comprehensive analysisof different coarse-to-fine designs for OD/OC segmentation using 5 publicdatabases, both from a standard segmentation perspective and for estimating thevCDR for glaucoma assessment. Our analysis shows that these algorithms notnecessarily outperfom standard multi-class single-stage models, especially whenthese are learned from sufficiently large and diverse training sets.Furthermore, we noticed that the coarse stage achieves better OD segmentationresults than the fine one, and that providing OD supervision to the secondstage is essential to ensure accurate OC masks. Moreover, both the single-stageand two-stage models trained on a multi-dataset setting showed results in pairor even better than other state-of-the-art alternatives, while ranking first inREFUGE for OD/OC segmentation. Finally, we evaluated the models for vCDRprediction in comparison with six ophthalmologists on a subset of AIROGSimages, to understand them in the context of inter-observer variability. Wenoticed that vCDR estimates recovered both from single-stage and coarse-to-finemodels can obtain good glaucoma detection results even when they are not highlycorrelated with manual measurements from experts.

Assessing Coarse-to-Fine Deep Learning Models for Optic Disc and Cup Segmentation in Fundus Images