Graph-Theoretic Consistency for Robust and Topology-Aware Semi-Supervised Histopathology Segmentation
This addresses the challenge of costly and limited dense annotations in computational pathology, offering a robust and topology-aware solution for semi-supervised segmentation.
The paper tackled the problem of noisy pseudo-labels and fragmented masks in semi-supervised semantic segmentation for histopathology by proposing Topology Graph Consistency (TGC), which integrates graph-theoretic constraints to enforce global topology, resulting in state-of-the-art performance under 5-10% supervision and significantly narrowing the gap to full supervision on GlaS and CRAG datasets.
Semi-supervised semantic segmentation (SSSS) is vital in computational pathology, where dense annotations are costly and limited. Existing methods often rely on pixel-level consistency, which propagates noisy pseudo-labels and produces fragmented or topologically invalid masks. We propose Topology Graph Consistency (TGC), a framework that integrates graph-theoretic constraints by aligning Laplacian spectra, component counts, and adjacency statistics between prediction graphs and references. This enforces global topology and improves segmentation accuracy. Experiments on GlaS and CRAG demonstrate that TGC achieves state-of-the-art performance under 5-10% supervision and significantly narrows the gap to full supervision.