Unifying Active Learning and Semi-Supervised Learning for Medical Image Segmentation

2026-07-27

AI Summary

This work addresses the conflicting objectives and training instability between active learning and semi-supervised learning in medical image segmentation under extremely limited annotation budgets. To resolve this, the authors propose RegAL, a framework that unifies both objectives through topology-aware Pareto optimization. RegAL selects informative samples by jointly considering voxel-wise uncertainty and feature diversity, and leverages the same criteria to guide diffeomorphic registration-based data augmentation, thereby enabling stable training from cold start. Evaluated on BraTS 2021, dHCP, and ProstateX datasets, RegAL consistently outperforms existing methods, demonstrating particularly robust performance at very low annotation rates across key metrics including Dice score, Average Surface Distance (ASD), and Hausdorff Distance at 95% (HD95).

Abstract

In practical settings, medical image segmentation models are often developed with limited annotated data rather than fully labeled datasets. Training frequently begins in ultra-low labeled regimes where only a small number of volumes are annotated. In such scenarios, practitioners must simultaneously decide which cases to annotate and how to best use the remaining unlabeled data. Although active learning (AL) and semi-supervised learning (SSL) both target annotation scarcity, they are typically designed and optimized independently, resulting in objective mismatch and unstable training during early-stage "cold start" conditions. We propose RegAL, a unified active semi-supervised framework governed by a shared topology-aware Pareto optimization that couples sample acquisition with unlabeled data utilization. RegAL evaluates images along three complementary axes, voxel-wise uncertainty, feature diversity, and a novel topological consistency metric, to select anatomically informative edge cases for annotation. On the other hand, the same criteria are used to identify geometrically stable atlas candidates for diffeomorphic registration-guided augmentation to train a self-supervised Mean Teacher segmentation network. Across BraTS 2021, dHCP, and ProstateX, RegAL remains stable with few labeled volumes and consistently outperforms state-of-the-art AL, SSL, and active semi-supervised baselines across Dice and boundary-distance (ASD, HD95) metrics under extreme annotation scarcity.

Problem

• medical image segmentation
• active learning
• semi-supervised learning
• annotation scarcity
• cold start

Innovation

• active learning
• semi-supervised learning
• topological consistency
• diffeomorphic registration
• medical image segmentation