Get Started
Home
Topics
Search
Library
Research questionHow can semi-supervised medical image segmentation prevent appearance variation from corrupting structural cues and pseudo-labels?Unlabeled medical images can contain appearance changes that obscure target structure. When these changes contaminate learned representations, unreliable pseudo-labels can reinforce segmentation errors during training.
AI
Computer Vision
Health
Image & Video Processing
Machine Learning
Research Paper
Statistical Machine Learning
Latest papersRecent research connected to this question, newest first.SAUF-Net: Structure--Appearance Representation Learning with Uncertainty Feedback for Semi-Supervised Medical Image SegmentationThe source concerns SAUF-Net for semi-supervised medical image segmentation, using structure–appearance decomposition, appearance-swapped consistency, and feature-level uncertainty feedback. Evidence comes from experiments on ISIC-2016 and Kvasir-SEG, including low-label settings.research paper · Sep 2, 2026
Related questions
How can semi-supervised hyperspectral classification use scarce labels while preserving boundaries and stabilizing pseudo-labels?How can medical image models segment lesions despite complex backgrounds and varied morphologies?How can reward-based fine-tuning preserve fine-grained fidelity and semantic consistency in conditional medical images?How should self-supervised visual learning combine objectives to prevent collapse while preserving semantic and spatial information?