ESRVS: Extreme Semi-Supervised Retinal Vessel Segmentation with a Single Annotated Image
2026-07-27 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionArtificial IntelligenceMachine Learning
AI summaryⓘ
The authors address the challenge of segmenting blood vessels in retinal images using very few labeled examples—just one annotated image and many unlabeled ones. They created ESRVS, which picks a key reference image to label and then uses smart features called DINOv3 to transfer vessel information to other images. By combining prototype maps and a physics-inspired model, they generate initial guesses for vessel locations and improve these guesses with special training methods. Tested on eight datasets, ESRVS performed better or close to methods that use much more labeled data, showing promise for efficient medical image analysis.
retinal vessel segmentationsemi-supervised learningDINOv3 featurespseudo-labelingprototype learningadversarial refinementDice scoreclDiceMask2Formerlabel propagation
Authors
Mingzhi Xu, Yizhe Zhang
Abstract
Learning from minimal human supervision is a long-standing goal in medical image analysis, where dense expert annotations are costly. We study retinal vessel segmentation in an extreme semi-supervised setting with one annotated image and a pool of unlabeled images. We propose ESRVS, which selects a representative reference image for manual annotation and transfers vessel cues using target-domain-adapted DINOv3 features. ESRVS constructs a multi granular vessel prototype, combines prototype-similarity maps with a physics-inspired prior to generate initial pseudo-labels, and refines the transferred supervision through weighted pseudo-label training and adversarial refinement. Across eight public datasets, ESRVS achieves the best Dice and clDice on six datasets, and the best HD95 on all eight datasets among the compared semi-supervised methods, although those methods use 10 to 20% labeled data. With Mask2Former, ESRVS retains on average 93.7% of fully supervised Dice and 95.1% of fully supervised clDice. These results demonstrate the potential of foundation-model label propagation for highly label-efficient retinal vessel segmentation. Code is available at https://github.com/IAANNH/ESRVS.