Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Videos

Multi-OCT-SelfNet: integrating self-supervised learning with multi-source data fusion for enhanced multi-class

Fatema E Jannat¹, Sina Gholami¹, Jennifer I Lim²

¹Department of Electrical and Computer Engineering, University of North Carolina at Charlotte, Charlotte, NC, United States.

Frontiers in Systems Biology

|June 17, 2026

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 author

Sequential Deep Learning to Predict Non-Central to Central Geographic Atrophy Progression from OCT Imaging.

medRxiv : the preprint server for health sciences·2026

Same author

Explicit Inclusion of Diabetes Mellitus Without Retinopathy Within Diabetic Retinopathy Prediction.

Translational vision science & technology·2026

Same author

Deep transfer learning for detection of placenta-mediated diseases from ultrasound images.

Scientific reports·2026

Same author

Summary of Research: Aflibercept 8 mg versus Faricimab Treat‑and‑Extend for Diabetic Macular Edema or Neovascular Age‑Related Macular Degeneration: A Bayesian Fixed‑Effect Network Meta‑analysis of Clinical Trials.

Ophthalmology and therapy·2026

Same author

Semaglutide and Neovascular Age-Related Macular Degeneration Among Adults with Type 2 Diabetes: An OHDSI Network Study.

Ophthalmology·2026

Same author

Retinal imaging and artificial intelligence analysis.

Handbook of clinical neurology·2026

Same journal

Computational analysis and validation of UGT1A1/4 missense variants impacting tecovirimat metabolism in monkeypox patients.

Frontiers in systems biology·2026

Same journal

Glioma identification from microRNA biomarkers using machine learning.

Frontiers in systems biology·2026

Same journal

Prioritizing long COVID related single nucleotide polymorphisms by mining genome-wide association studies of COVID-19 susceptibility and hospitalization.

Frontiers in systems biology·2026

Same journal

Multilayer network approaches to omics data integration in digital twins for cancer research.

Frontiers in systems biology·2026

Same journal

Quorum-quenching for bacterial pathogen control and health management in aquaculture: mechanisms, applications, current status and future prospects.

Frontiers in systems biology·2026

Same journal

Duty of Care Governance: a conceptual framework for aligning health system scale with the clinician-patient relationship.

Frontiers in systems biology·2026

See all related articles

This summary is machine-generated.

Multi-OCT-SelfNet enhances deep learning for retinal disease classification by using self-supervised pre-training on diverse optical coherence tomography (OCT) datasets. This improves model generalization, especially in data-limited scenarios.

Area of Science:

Ophthalmology
Medical Imaging
Artificial Intelligence

Background:

Acquiring large, diverse medical imaging datasets for deep learning is challenging due to privacy, cost, and variability.
Limited datasets reduce the generalization ability of deep learning models in retinal disease classification.

Purpose of the Study:

To propose Multi-OCT-SelfNet, a self-supervised framework using a SwinV2 transformer for multi-class retinal disease classification from optical coherence tomography (OCT) images.
To improve the generalization and robustness of AI models in retinal disease classification, particularly under data-limited and domain-shifted conditions.

Main Methods:

Developed a self-supervised framework (Multi-OCT-SelfNet) with a SwinV2 transformer backbone.
Employed masked autoencoder-based self-supervised pre-training on multi-source OCT datasets.

Keywords:

AI OCT SwinV2 data fusion retinal disease classification self-supervised learning transfer learning transformer

Related Experiment Videos

Fine-tuned the model on individual downstream datasets for supervised classification.

Main Results:

Achieved competitive AUC-ROC scores (0.97 on DS1, 0.97 on DS2, 0.89 on DS3) in on-domain evaluations.
Significantly improved cross-dataset performance, e.g., AUC-ROC increased from 0.61 to 0.90 when training on DS2 and testing on DS3.
Demonstrated enhanced robustness in limited-data settings (50% training samples), maintaining higher AUC-ROC scores compared to baselines.

Conclusions:

Multi-OCT-SelfNet learns more transferable OCT representations than conventional supervised methods.
The proposed framework shows promise for robust AI-assisted retinal disease classification in clinical settings with limited or shifted data.
Multi-source data fusion and self-supervised pre-training are key to improving generalization for OCT image analysis.