Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Light-induced retinal degeneration is prevented by zinc, a component in the age-related eye disease study formulation.

Photochemistry and photobiology·2012
Same author

Light induced and circadian effects on retinal photoreceptor cell crystallins.

Photochemistry and photobiology·2010
Same author

Direct observation of a widely tunable bandgap in bilayer graphene.

Nature·2009
Same author

Investigation of second-order nonlinearity in poled-polymer during photobleaching.

Optics express·2009
Same author

A novel dual PI3Kalpha/mTOR inhibitor PI-103 with high antitumor activity in non-small cell lung cancer cells.

International journal of molecular medicine·2009
Same author

Axonal regeneration of optic nerve after crush in Nogo66 receptor knockout mice.

Neuroscience letters·2009

Related Experiment Video

Updated: Mar 24, 2026

Quantification of Vascular Parameters in Whole Mount Retinas of Mice with Non-Proliferative and Proliferative Retinopathies
12:28

Quantification of Vascular Parameters in Whole Mount Retinas of Mice with Non-Proliferative and Proliferative Retinopathies

Published on: March 12, 2022

4.4K

Few-shot annotation correction for lightweight retinal vessel image segmentation.

Huazhang Li1,2, Yueming Sun2, Daniel Organisciak2

  • 1Department of Ophthalmology, Harbin Medical University, Harbin, China.

Frontiers in Medicine
|March 23, 2026
PubMed
Summary

Manual annotations for retinal vessel segmentation can be inaccurate. This study introduces a robust framework that corrects positional label noise, improving segmentation accuracy and demonstrating practical resilience in ophthalmology diagnostics.

Keywords:
U-netdeep learningmedical image analysisnoisy annotationsretinal vessel segmentation

More Related Videos

Using Retinal Imaging to Study Dementia
09:17

Using Retinal Imaging to Study Dementia

Published on: November 6, 2017

22.4K
In Vivo Vascular Injury Readouts in Mouse Retina to Promote Reproducibility
07:35

In Vivo Vascular Injury Readouts in Mouse Retina to Promote Reproducibility

Published on: April 21, 2022

2.7K

Related Experiment Videos

Last Updated: Mar 24, 2026

Quantification of Vascular Parameters in Whole Mount Retinas of Mice with Non-Proliferative and Proliferative Retinopathies
12:28

Quantification of Vascular Parameters in Whole Mount Retinas of Mice with Non-Proliferative and Proliferative Retinopathies

Published on: March 12, 2022

4.4K
Using Retinal Imaging to Study Dementia
09:17

Using Retinal Imaging to Study Dementia

Published on: November 6, 2017

22.4K
In Vivo Vascular Injury Readouts in Mouse Retina to Promote Reproducibility
07:35

In Vivo Vascular Injury Readouts in Mouse Retina to Promote Reproducibility

Published on: April 21, 2022

2.7K

Area of Science:

  • Ophthalmology
  • Medical Image Analysis
  • Computer Vision

Background:

  • Retinal vessel segmentation is crucial for diagnosing eye diseases.
  • Manual annotations are prone to errors, especially for thin vessels.
  • The impact of positional label noise on segmentation accuracy is not well understood.

Purpose of the Study:

  • To develop a robust framework for retinal vessel segmentation that addresses positional label noise.
  • To quantify the effect of annotation errors on segmentation performance.
  • To evaluate the proposed method's effectiveness across different datasets.

Main Methods:

  • A lightweight, few-shot U-Net-based framework was developed for annotation correction and noise-robust learning.
  • Performance was analyzed on the DRIVE dataset with varying degrees of label displacement.
  • Cross-dataset validation was performed on CHASE_DB1 and STARE datasets.

Main Results:

  • Performance significantly degraded with increasing label displacement.
  • The proposed method achieved high Accuracy (96.51% on CHASE_DB1, 97.54% on STARE), AUC (98.01% on CHASE_DB1, 98.45% on STARE), and F1 scores (83.55% on CHASE_DB1, 83.11% on STARE).
  • The framework demonstrated competitive performance compared to state-of-the-art methods.

Conclusions:

  • Retinal vessel segmentation is sensitive to positional annotation errors.
  • The developed framework offers practical robustness against noisy labels in ophthalmology.
  • This work highlights the importance of addressing annotation quality for reliable diagnostic tools.