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 Video

Updated: Jun 13, 2026

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
04:48

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

Optic Disc Fundus Images Retain Biometric Identity Signals Under Deep Learning.

Ali Azizi, Rafael Scherer, Aaron S Rabinowitz

    Research Square
    |June 12, 2026
    PubMed
    Summary
    This summary is machine-generated.

    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

    Extraction of Glaucoma Diagnosis, Type, and Severity from Clinical Notes using Secure Cloud-based Large Language Models.

    medRxiv : the preprint server for health sciences·2026
    Same author

    GLLaucoMed: A Secure LLM-Powered Agentic Workflow for Automated Medication Extraction from Free-Text Glaucoma Clinical Notes.

    medRxiv : the preprint server for health sciences·2026
    Same author

    Development and Qualification of a Nipah Virus Glycoprotein-Specific IgG ELISA for the Assessment of Human Antibody Responses.

    Vaccines·2026
    Same author

    Development and Pilot Testing of a Mobile App Psychosocial Intervention for Psychological Distress in Individuals with Glaucoma.

    medRxiv : the preprint server for health sciences·2026
    Same author

    Reply to Comment on "RNFL Thickness in a Population-Based Cohort: The Canadian Longitudinal Study on Aging M2M (Machine-to-Machine) Study".

    American journal of ophthalmology·2026
    Same author

    Identification of Risk Factors for Glaucoma Progression in Free-Text Clinical Notes Using a Local Large Language Model.

    Translational vision science & technology·2026
    Same journal

    HIV Transmission Dynamics in Greater Mexico City are Shaped by Dense Spatial Mixing.

    Research square·2026
    Same journal

    A UCP1-IRES-Cre Knock-In Mouse Enables Specific Brown Adipocyte Targeting Without CNS Off-Target Expression.

    Research square·2026
    Same journal

    Precision RNAi for Fibrodysplasia Ossificans Progressiva: a combinatorial, unimolecular, allele selective approach.

    Research square·2026
    Same journal

    Perceptions of end-of-life care quality among bereaved closest contacts of community-dwelling older Australians: a cross-sectional survey of the ASPREE cohort.

    Research square·2026
    Same journal

    Heavy-chain immune repertoire sequencing enables language-model prediction of antigen-specific antibodies.

    Research square·2026
    Same journal

    25+ Years of TRPV4: From Discovery to Translational Horizons.

    Research square·2026
    See all related articles

    Deep learning models using optic disc-centered fundus images retain subject-specific information for biometric verification. Models trained on images including the peripapillary rim performed comparably to full-field images.

    Area of Science:

    • Ophthalmology
    • Computer Science
    • Biometrics

    Background:

    • Deep learning models are increasingly used for analyzing medical images.
    • The potential for using cropped fundus images for biometric verification is underexplored.
    • Subject-specific information in fundus images is crucial for identification and privacy.

    Purpose of the Study:

    • To evaluate the efficacy of deep learning models trained on optic disc-centered fundus images for biometric verification.
    • To compare the performance of models trained on full-field, disc-region, and disc-only fundus images.
    • To assess the retention of subject-specific information in reduced retinal image regions.

    Main Methods:

    • Trained Siamese convolutional neural network models using triplet loss on three image types: full-field fundus, optic disc region, and disc-only.

    More Related Videos

    Using Retinal Imaging to Study Dementia
    09:17

    Using Retinal Imaging to Study Dementia

    Published on: November 6, 2017

    Related Experiment Videos

    Last Updated: Jun 13, 2026

    Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
    04:48

    Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

    Published on: November 30, 2022

    Using Retinal Imaging to Study Dementia
    09:17

    Using Retinal Imaging to Study Dementia

    Published on: November 6, 2017

  • Utilized 30,836 fundus photographs from 4,500 subjects, partitioned into training, validation, and test sets.
  • Evaluated biometric verification performance using cosine similarity on an independent test set.

    • Main Results:

    • All image representations retained measurable subject-specific biometric signals.
    • The full-fundus model achieved the highest performance (AUC, 0.992; EER, 4.4%), followed by the disc-region (AUC, 0.989; EER, 5.5%) and disc-only models (AUC, 0.969; EER, 10.5%).
    • Disc-region models showed performance comparable to full-fundus models, while disc-only models performed significantly worse.

    Conclusions:

    • Deep learning models trained on optic disc-centered fundus images retain meaningful subject-specific information for biometric verification.
    • Including a narrow peripapillary rim in training data yields performance comparable to full-field images.
    • Reduced retinal image regions retain subject-specific features, emphasizing cautious data-sharing practices.