ESRVS method achieves high accuracy in retinal vessel segmentation with minimal supervision

Researchers have developed ESRVS, a novel method for retinal vessel segmentation that requires only a single annotated image and a collection of unlabeled images. This approach leverages foundation model label propagation, using DINOv3 features and a physics-inspired prior to generate initial pseudo-labels. ESRVS then refines this supervision through weighted pseudo-label training and adversarial refinement, achieving state-of-the-art results on multiple public datasets and retaining a high percentage of performance compared to fully supervised methods. AI

IMPACT Demonstrates significant potential for label-efficient medical image segmentation, reducing reliance on costly expert annotations.

RANK_REASON Academic paper detailing a new method for medical image analysis. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• DINOv3
• ESRVS
• Mask2Former

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →