LUMOS: Universal Semi-Supervised OCT Retinal Layer Segmentation with Hierarchical Reliable Mutual Learning

TL;DR

LUMOS is a semi-supervised framework for OCT retinal layer segmentation that effectively leverages cross-granularity supervision and enhances generalization across datasets.

Contribution

It introduces a dual-decoder network with a hierarchical prompting strategy and a reliable progressive multi-granularity learning approach for improved segmentation.

Findings

LUMOS outperforms existing methods on six OCT datasets.

It demonstrates strong cross-domain and cross-granularity generalization.

The proposed approach effectively suppresses pseudo-label noise.

Abstract

Optical Coherence Tomography (OCT) layer segmentation faces challenges due to annotation scarcity and heterogeneous label granularities across datasets. While semi-supervised learning helps alleviate label scarcity, existing methods typically assume a fixed granularity, failing to fully exploit cross-granularity supervision. This paper presents LUMOS, a semi-supervised universal OCT retinal layer segmentation framework based on a Dual-Decoder Network with a Hierarchical Prompting Strategy (DDN-HPS) and Reliable Progressive Multi-granularity Learning (RPML). DDN-HPS combines a dual-branch architecture with a multi-granularity prompting strategy to effectively suppress pseudo-label noise propagation. Meanwhile, RPML introduces region-level reliability weighing and a progressive training approach that guides the model from easier to more difficult tasks, ensuring the reliable selection of cross-granularity consistency targets, thereby achieving stable cross-granularity alignment. Experiments on six OCT datasets demonstrate that LUMOS largely outperforms existing methods and exhibits exceptional cross-domain and cross-granularity generalization capability.