UHR-Net: An Uncertainty-Aware Hypergraph Refinement Network for Medical Image Segmentation

2026-04-30

AI Summary

This work addresses the challenges in medical image segmentation caused by ambiguous lesion boundaries, low contrast with surrounding tissues, and the dilution of small lesions during multi-scale feature extraction. To mitigate these issues, the authors propose a novel network architecture that integrates uncertainty-aware learning with hypergraph refinement. Specifically, they introduce an uncertainty-guided instance contrastive pretraining strategy to enhance feature discriminability and design a hypergraph refinement module that decouples and models high-order relationships between foreground and background regions. Additional components include geometry-aware copy-paste augmentation, hard negative mining, and entropy-based uncertainty map generation. Extensive experiments demonstrate that the proposed method significantly outperforms strong baselines across five public medical image segmentation benchmarks, achieving notable improvements in boundary delineation and small lesion segmentation accuracy.

Abstract

Accurate lesion segmentation is crucial for clinical diagnosis and treatment planning. However, lesions often resemble surrounding tissues and exhibit ill-defined boundaries, leading to unstable predictions in boundary/transition regions. Moreover, small-lesion cues can be diluted by multi-scale feature extraction, causing under- or over-segmentation. To address these challenges, we propose an Uncertainty-Aware Hypergraph Refinement Network (UHR-Net). First, we introduce an Uncertainty-Oriented Instance Contrastive (UO-IC) pretraining strategy that couples geometry-aware copy-paste augmentation with hard-negative mining of lesion-like background regions to improve instance-level discrimination for small and visually ambiguous lesions. Second, we design an Uncertainty-Guided Hypergraph Refinement (UGHR) block, which derives an entropy-based uncertainty map from a coarse probability map to guide hypergraph refinement. By splitting hyperedge prototypes into foreground and background groups, UGHR decouples higher-order interactions and improves refinement in ambiguous regions. Experiments on five public benchmarks demonstrate consistent gains over strong baselines. Code is available at: https://github.com/CUGfreshman/UHR-Net.

Problem

• medical image segmentation
• lesion boundary ambiguity
• small lesion segmentation
• prediction uncertainty
• multi-scale feature dilution

Innovation

• Uncertainty-Aware
• Hypergraph Refinement
• Instance Contrastive Learning
• Medical Image Segmentation
• Entropy-Based Uncertainty