Mask-HybridGNet: Graph-based segmentation with emergent anatomical correspondence from pixel-level supervision

2026-02-24

AI Summary

This work addresses the challenge of graph-based medical image segmentation, which typically relies on scarce point-to-point anatomical landmark annotations. To overcome this limitation, the authors propose Mask-HybridGNet, a framework that trains a fixed-topology graph model using only standard pixel-level segmentation masks. By aligning deformable object boundaries with fixed landmark nodes via Chamfer distance and incorporating edge regularization with differentiable rasterization, the method implicitly learns anatomical correspondences without requiring explicit landmark supervision. The approach enables temporal tracking, cross-sectional reconstruction, and population-level morphological analysis, while extracting stable anatomical graphs from any high-quality segmentation. Evaluated on chest X-ray, cardiac ultrasound, cardiac MRI, and fetal imaging tasks, the method achieves segmentation accuracy comparable to state-of-the-art pixel-level approaches while preserving boundary connectivity and anatomical plausibility.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Graph-based Machine Learning
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Graph-based medical image segmentation represents anatomical structures using boundary graphs, providing fixed-topology landmarks and inherent population-level correspondences. However, their clinical adoption has been hindered by a major requirement: training datasets with manually annotated landmarks that maintain point-to-point correspondences across patients rarely exist in practice. We introduce Mask-HybridGNet, a framework that trains graph-based models directly using standard pixel-wise masks, eliminating the need for manual landmark annotations. Our approach aligns variable-length ground truth boundaries with fixed-length landmark predictions by combining Chamfer distance supervision and edge-based regularization to ensure local smoothness and regular landmark distribution, further refined via differentiable rasterization. A significant emergent property of this framework is that predicted landmark positions become consistently associated with specific anatomical locations across patients without explicit correspondence supervision. This implicit atlas learning enables temporal tracking, cross-slice reconstruction, and morphological population analyses. Beyond direct segmentation, Mask-HybridGNet can extract correspondences from existing segmentation masks, allowing it to generate stable anatomical atlases from any high-quality pixel-based model. Experiments across chest radiography, cardiac ultrasound, cardiac MRI, and fetal imaging demonstrate that our model achieves competitive results against state-of-the-art pixel-based methods, while ensuring anatomical plausibility by enforcing boundary connectivity through a fixed graph adjacency matrix. This framework leverages the vast availability of standard segmentation masks to build structured models that maintain topological integrity and provide implicit correspondences.

Problem

• graph-based segmentation
• anatomical correspondence
• landmark annotation
• medical image segmentation
• pixel-level supervision

Innovation

• graph-based segmentation
• anatomical correspondence
• pixel-level supervision
• implicit atlas learning
• differentiable rasterization