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 Experiment Videos

Advancing AMD Detection: Dataset Design and Deep Learning Optimization for Unconstrained Retinal Images.

Hala Nafie Fathee1, Reyhan Babayev2, Shaaban Sahmoud3

  • 1College of Physical Education and Sport Sciences, University of Mosul, Mosul 41002, Iraq.

Vision (Basel, Switzerland)
|May 27, 2026
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

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

Sort by
Same journal

Subtitle Engagement Varies with Audio-Subtitle Language-Script Pairing: Evidence from Hindi-English Bilinguals with an English-Medium Instruction Background.

Vision (Basel, Switzerland)·2026
Same journal

Ultra-Early OCT Changes After Intravitreal Injection: Evidence Consistent with Transient Mechanical Compression.

Vision (Basel, Switzerland)·2026
Same journal

A Cerebral Basis for Visual Discomfort and Visual Stress.

Vision (Basel, Switzerland)·2026
Same journal

Practice Effects and the Lanthony D15.

Vision (Basel, Switzerland)·2026
Same journal

Wouldn't It Be Nice to Not Fall for It Twice? Prior Experience Does Not Abolish the Impact of Expectancy Violations on Attention Capture.

Vision (Basel, Switzerland)·2026
Same journal

Evaluation of the Efficacy of Treatment for Convergence Insufficiency with a New Digital Mobile Platform: A Comparative Preliminary Study.

Vision (Basel, Switzerland)·2026

A new dataset for age-related macular degeneration (AMD) detection under real-world conditions improves deep learning models. Optimized VGG16 achieved 88% accuracy, showing the value of realistic data for clinical AI.

Area of Science:

  • Ophthalmology
  • Computer Science
  • Artificial Intelligence

Background:

  • Age-related macular degeneration (AMD) is a primary cause of vision loss globally.
  • Early detection of AMD is crucial for effective treatment and management.
  • Current deep learning models for AMD detection often lack generalizability due to reliance on controlled datasets.

Purpose of the Study:

  • To develop a novel dataset simulating unconstrained retinal imaging conditions for AMD detection.
  • To evaluate the performance of various deep learning architectures on this realistic dataset.
  • To identify the most robust deep learning model for AMD detection in clinical settings.

Main Methods:

  • A new dataset was created incorporating noise, luminance variations, and artifacts typical of real-world retinal scans.
Keywords:
AMD datasetAMD detectionage-related macular degenerationdeep learningunconstrained retinal images

Related Experiment Videos

  • Six deep learning architectures (VGG16, VGG19, InceptionV3, MobileNetV2, ResNet50, DenseNet) were comparatively evaluated.
  • The best-performing model (VGG16) was further optimized using targeted training and fine-tuning.

    • Main Results:

    • Significant performance variations were observed among the evaluated deep learning models, indicating differences in robustness to image degradation.
    • VGG16 demonstrated superior overall performance compared to other architectures on the unconstrained dataset.
    • The optimized VGG16 model achieved an accuracy of 88% for AMD detection.

    Conclusions:

    • Realistic, unconstrained datasets are essential for developing reliable deep learning tools for clinical AMD diagnosis.
    • The optimized VGG16 architecture shows significant promise for practical, real-world AMD detection.
    • This work highlights the importance of dataset diversity in advancing AI for medical imaging.