UCAD: Uncertainty-guided Contour-aware Displacement for semi-supervised medical image segmentation

TL;DR

UCAD introduces an uncertainty-guided, contour-aware displacement framework that improves semi-supervised medical image segmentation by preserving anatomical boundaries and enhancing consistency learning, outperforming existing methods.

Contribution

The paper presents a novel UCAD framework that uses superpixels and uncertainty-guided displacement to better preserve anatomy and improve semi-supervised segmentation accuracy.

Findings

UCAD outperforms state-of-the-art methods on multiple datasets.

The dynamic uncertainty-weighted loss stabilizes training.

Superpixel-based regions enhance boundary preservation.

Abstract

Existing displacement strategies in semi-supervised segmentation only operate on rectangular regions, ignoring anatomical structures and resulting in boundary distortions and semantic inconsistency. To address these issues, we propose UCAD, an Uncertainty-Guided Contour-Aware Displacement framework for semi-supervised medical image segmentation that preserves contour-aware semantics while enhancing consistency learning. Our UCAD leverages superpixels to generate anatomically coherent regions aligned with anatomy boundaries, and an uncertainty-guided selection mechanism to selectively displace challenging regions for better consistency learning. We further propose a dynamic uncertainty-weighted consistency loss, which adaptively stabilizes training and effectively regularizes the model on unlabeled regions. Extensive experiments demonstrate that UCAD consistently outperforms state-of-the-art semi-supervised segmentation methods, achieving superior segmentation accuracy under limited annotation. The code is available at:https://github.com/dcb937/UCAD.