SHTA: Semantic Hard Token Correction and Center Alignment for Semi-Supervised Medical Image Segmentation

2026-07-08

AI Summary

Existing semi-supervised medical image segmentation methods struggle to learn semantically consistent intermediate representations in challenging regions, limiting performance gains. This work proposes SHTA—a lightweight, training-time semantic enhancement branch that explicitly improves semantic consistency in difficult areas through semantic assignment, hard-label refinement, and semantic centroid alignment, all without modifying the original prediction pathway. As the first approach to directly optimize intermediate semantic consistency during training, SHTA incurs no additional inference overhead and integrates seamlessly into mainstream frameworks such as GA-CPS, CPS, URPC, and MagicNet. Experiments demonstrate that SHTA significantly boosts segmentation accuracy and weak-organ recovery on the Synapse and AMOS datasets, effectively mitigates semantic ambiguity, and introduces only minimal computational cost during training.

Abstract

Recent advances in semi-supervised medical image segmentation have achieved remarkable performance through prediction consistency, pseudo-label supervision, and hard-region supervision. However, these methods primarily improve supervision quality rather than explicitly enforcing semantic consistency in the learned representations of hard regions. Consequently, even under increasingly stronger prediction-level supervision, difficult regions exhibiting unstable semantic assignment often fail to establish semantically consistent representations during training, thereby limiting further segmentation improvement. To address this issue, we propose SHTA (Semantic Hard Token Correction and Center Alignment), a lightweight training-time semantic representation branch. Instead of introducing additional prediction supervision, SHTA refines intermediate semantic representations through Semantic Assignment, Hard Token Refinement, and Semantic Center Alignment, thereby improving semantic consistency in hard regions while preserving the original prediction pathway and introducing no additional inference cost. We integrate SHTA into representative semi-supervised segmentation frameworks, including GA-CPS, CPS, URPC, and MagicNet, and conduct evaluations on the Synapse and AMOS datasets. Experimental results demonstrate that SHTA delivers consistent paired improvements across frameworks, with especially clear gains in segmentation accuracy, weak-organ recovery, and semantic ambiguity reduction, while incurring only training-time overhead. The code is available at https://anonymous.4open.science/r/release_SHTA-42D5/.

Problem

• semi-supervised medical image segmentation
• semantic consistency
• hard regions
• semantic representation
• prediction supervision

Innovation

• Semantic Consistency
• Hard Region Refinement
• Token Correction
• Center Alignment
• Semi-Supervised Segmentation