Scribble-Supervised Medical Image Segmentation with Dynamic Teacher Switching and Hierarchical Consistency

2026-01-21

AI Summary

This work addresses the challenges of pseudo-label noise and difficulty in learning anatomical boundaries caused by sparse scribble annotations in medical image segmentation. To this end, the authors propose SDT-Net, a semi-supervised framework featuring a novel Dynamic Teacher Switching (DTS) mechanism within a dual-teacher single-student architecture. The approach integrates high-confidence pixel selection (PRP) and multi-level feature consistency constraints (HiCo) to effectively fuse reliable supervisory signals. By dynamically leveraging complementary teacher predictions and enforcing consistency across hierarchical features, SDT-Net significantly improves pseudo-label quality and anatomical plausibility. Extensive experiments on the ACDC and MSCMRseg datasets demonstrate state-of-the-art performance, yielding more accurate and anatomically consistent segmentation results.

Technology Category

• Machine Learning: Semi-Supervised Learning
• Computer Vision: Segmentation
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

Application Category

• Semantics and Knowledge: Representation, semantic annotation, enhancement, enrichments, access and/or integration of a variety of data on the Web
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Scribble-supervised methods have emerged to mitigate the prohibitive annotation burden in medical image segmentation. However, the inherent sparsity of these annotations introduces significant ambiguity, which results in noisy pseudo-label propagation and hinders the learning of robust anatomical boundaries. To address this challenge, we propose SDT-Net, a novel dual-teacher, single-student framework designed to maximize supervision quality from these weak signals. Our method features a Dynamic Teacher Switching (DTS) module to adaptively select the most reliable teacher. This selected teacher then guides the student via two synergistic mechanisms: high-confidence pseudo-labels, refined by a Pick Reliable Pixels (PRP) mechanism, and multi-level feature alignment, enforced by a Hierarchical Consistency (HiCo) module. Extensive experiments on the ACDC and MSCMRseg datasets demonstrate that SDT-Net achieves state-of-the-art performance, producing more accurate and anatomically plausible segmentation.

Problem

• scribble-supervised
• medical image segmentation
• annotation ambiguity
• pseudo-label noise
• anatomical boundaries

Innovation

• Scribble-supervised segmentation
• Dynamic Teacher Switching
• Hierarchical Consistency
• Pseudo-label refinement
• Medical image segmentation