H2AL: Hyperbolic Hierarchy-aware Aggregative Learning for Registration-based Few-shot Medical Image Segmentation

2026-08-07

AI Summary

This work addresses the limitations of existing registration-based few-shot medical image segmentation methods, which operate in Euclidean space and neglect the hierarchical relationships inherent in anatomical structures, often yielding low-quality pseudo-labels and blurred boundaries. To overcome this, the authors propose H2AL, a novel framework that introduces hyperbolic space to this task for the first time. H2AL models anatomical hierarchies through hierarchy-aware contrastive learning and a gated injection mechanism, while employing a dual-task collaborative learning strategy that jointly optimizes registration and segmentation via gradient aggregation in an end-to-end manner. Extensive experiments across two anatomical regions and five evaluation settings demonstrate that H2AL significantly improves both registration accuracy and segmentation performance, confirming its effectiveness and efficiency.

Abstract

Registration-based Few-shot medical image segmentation (RFMIS) aims to generate pseudo-labels for unlabeled images by warping a labeled image through registration. However, existing methods primarily perform pixel-level optimization and inference in Euclidean space, treating anatomical structures as flat and disjoint. This neglect of inherent hierarchies degrades pseudo-label quality and weakens the discrimination of ambiguous regions, limiting the segmentation performance. To overcome this challenge, we propose a Hyperbolic Hierarchy-aware Aggregative Learning framework for RFMIS, termed H2AL, that enhances both deformation plausibility and anatomical discrimination for dual-task learning. Specifically, we introduce a Hyperbolic Hierarchy-aware Infusion (H2I) module, which leverages the hierarchical modeling capability of hyperbolic space to learn precise hierarchy-aware representations via transformation-guided supervised hyperbolic contrastive learning, and injects such hierarchical priors into Euclidean space through a gated infusion block while preserving semantic richness. Furthermore, we propose an end-to-end joint optimization algorithm by gradient aggregation, where the gradients from the registration and segmentation decoders, embedding semantic and hierarchical cues, are aggregated to update the shared encoder to promote collaborative learning across tasks. Extensive experiments on two anatomical regions, with five experimental settings, demonstrate the effectiveness and efficiency of our method in both registration and segmentation. The code is publicly available at https://github.com/JiamingCai469/H2AL.

Problem

• Few-shot medical image segmentation
• Registration-based segmentation
• Hierarchical structure
• Pseudo-label quality
• Anatomical discrimination

Innovation

• Hyperbolic space
• Hierarchy-aware learning
• Few-shot segmentation
• Registration-based segmentation
• Gradient aggregation