Representation learning from OCT images

2026-05-04

AI Summary

This study addresses the challenges of high reliance on expert annotations and poor diagnostic consistency across devices and populations in retinal OCT image analysis. It presents a systematic review and structured integration of representation learning approaches, spanning supervised, self-supervised, semi-supervised, generative, 3D modeling, and multimodal paradigms—from early deep learning to vision-language foundation models. The work innovatively introduces a unified mathematical framework and a comprehensive taxonomy, incorporating cutting-edge directions such as foundation models, uncertainty-aware learning, federated training, and fairness into a cohesive perspective for the first time. By clarifying the core contributions and limitations of existing methods, cataloging publicly available datasets and evaluation protocols, and identifying critical future avenues—including volumetric pretraining, privacy preservation, and interpretability—the paper provides a foundational roadmap for advancing robust and equitable OCT image analysis.

Abstract

Optical Coherence Tomography (OCT) has become one of the most used imaging modality in ophthalmology. It provides high-resolution, non-invasive visualization of retinal microarchitecture. The automated analysis of OCT images through representation learning has emerged as a central research frontier. This has mainly been driven by the clinical need to process large acquisition volumes. The objective is to reduce the reliance on expert annotation, and improve diagnostic consistency across devices and populations. This survey provides a comprehensive and structured review of representation learning methods for retinal OCT image analysis. It covers the period from early deep learning approaches to the most recent developments in foundation models and vision-language systems. We organize the literature along a principled taxonomy of learning paradigms, encompassing supervised learning with CNN-based and transformer-based architectures, self-supervised and semi-supervised methods, generative approaches, as well as 3D volumetric modeling, multimodal representation learning, and large-scale pretrained foundation models. For each paradigm, we analyze the core methodological contributions, identify persistent limitations, and trace the connections between successive approaches. We further provide a structured overview of publicly available OCT datasets, discuss evaluation protocol considerations, and present a unified problem formulation that situates each learning paradigm within a common mathematical framework. Building on this analysis, we identify and discuss the most pressing open research directions emerging in the literature. This includes volumetric foundation model pretraining, uncertainty-aware representation learning, federated and privacy-preserving training, fairness and bias mitigation, concept-based interpretability,...

Problem

• Representation learning
• Optical Coherence Tomography
• Automated analysis
• Expert annotation
• Diagnostic consistency

Innovation

• representation learning
• foundation models
• OCT image analysis
• self-supervised learning
• multimodal learning