Anatomy-Aware Unsupervised Detection and Localization of Retinal Abnormalities in Optical Coherence Tomography

2026-04-23

AI Summary

This work addresses the limited generalizability of optical coherence tomography (OCT) anomaly detection methods stemming from their reliance on costly expert annotations by proposing an unsupervised framework. It is the first to incorporate retinal anatomical priors into unsupervised learning, training a discrete latent model exclusively on normal B-scans and integrating hierarchical perceptual supervision with structured triplet learning to enable both image-level and pixel-level anomaly detection without lesion annotations. Experimental results demonstrate that the model achieves AUROC scores of 0.799 and 0.884 on the Kermany and cross-dataset Srinivasan benchmarks, respectively, and obtains a Dice score of 0.200 and mIoU of 0.117 on the RETOUCH benchmark, significantly outperforming existing baselines and confirming its strong generalization across devices and patient populations.

Technology Category

• Computer Vision: Other Foundations of Computer Vision
• Machine Learning: Unsupervised & Self-Supervised Learning
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs

Abstract

Reliable automated analysis of Optical Coherence Tomography (OCT) imaging is crucial for diagnosing retinal disorders but faces a critical barrier: the need for expensive, labor-intensive expert annotations. Supervised deep learning models struggle to generalize across diverse pathologies, imaging devices, and patient populations due to their restricted vocabulary of annotated abnormalities. We propose an unsupervised anomaly detection framework that learns the normative distribution of healthy retinal anatomy without lesion annotations, directly addressing annotation efficiency challenges in clinical deployment. Our approach leverages a discrete latent model trained on normal B-scans to capture OCT-specific structural patterns. To enhance clinical robustness, we incorporate retinal layer-aware supervision and structured triplet learning to separate healthy from pathological representations, improving model reliability across varied imaging conditions. During inference, anomalies are detected and localized via reconstruction discrepancies, enabling both image and pixel-level identification without requiring disease-specific labels. On the Kermany dataset (AUROC: 0.799), our method substantially outperforms VAE, VQVAE, VQGAN, and f-AnoGAN baselines. Critically, cross-dataset evaluation on Srinivasan achieves AUROC 0.884 with superior generalization, demonstrating robust domain adaptation. On the external RETOUCH benchmark, unsupervised anomaly segmentation achieves competitive Dice (0.200) and mIoU (0.117) scores, validating reproducibility across institutions.

Problem

• unsupervised anomaly detection
• retinal abnormalities
• Optical Coherence Tomography
• annotation efficiency
• domain generalization

Innovation

• unsupervised anomaly detection
• anatomy-aware learning
• retinal layer-aware supervision
• structured triplet learning
• cross-domain generalization