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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Convolutional neural networks (CNNs) are a promising technique for automatedglaucoma diagnosis from images of the fundus, and these images are routinelyacquired as part of an ophthalmic exam. Nevertheless, CNNs typically require alarge amount of well-labeled data for training, which may not be available inmany biomedical image classification applications, especially when diseases arerare and where labeling by experts is costly. This article makes twocontributions to address this issue: (1) It extends the conventional Siamesenetwork and introduces a training method for low-shot learning when labeleddata are limited and imbalanced, and (2) it introduces a novel semi-supervisedlearning strategy that uses additional unlabeled training data to achievegreater accuracy. Our proposed multi-task Siamese network (MTSN) can employ anybackbone CNN, and we demonstrate with four backbone CNNs that its accuracy withlimited training data approaches the accuracy of backbone CNNs trained with adataset that is 50 times larger. We also introduce One-Vote Veto (OVV)self-training, a semi-supervised learning strategy that is designedspecifically for MTSNs. By taking both self-predictions and contrastivepredictions of the unlabeled training data into account, OVV self-trainingprovides additional pseudo labels for fine-tuning a pre-trained MTSN. Using alarge (imbalanced) dataset with 66,715 fundus photographs acquired over 15years, extensive experimental results demonstrate the effectiveness of low-shotlearning with MTSN and semi-supervised learning with OVV self-training. Threeadditional, smaller clinical datasets of fundus images acquired under differentconditions (cameras, instruments, locations, populations) are used todemonstrate the generalizability of the proposed methods.

One-Vote Veto: Semi-Supervised Learning for Low-Shot Glaucoma Diagnosis