FALCON: Few-Shot Adversarial Learning for Cross-Domain Medical Image Segmentation

2026-01-04

arXiv.org

AI Summary

This work addresses the challenges of scarce annotations, high inter-subject variability, privacy constraints, and substantial computational costs in 3D medical image segmentation by proposing the FALCON framework. FALCON first acquires a generalizable segmentation prior through meta-learning on natural images, then leverages adversarial fine-tuning and boundary-aware learning to enable effective cross-domain transfer. It further employs support-sample-guided 2D slice inference to efficiently produce high-quality 3D segmentations. Notably, this is the first approach to integrate meta-learning with adversarial fine-tuning for few-shot medical image segmentation, achieving significant improvements in boundary accuracy and model adaptability without relying on data augmentation or extensive annotations. Evaluated on four benchmark datasets, FALCON attains the lowest Hausdorff distance and Dice scores comparable to state-of-the-art methods, substantially reducing both annotation requirements and computational overhead.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Data transparency and provenance
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Precise delineation of anatomical and pathological structures within 3D medical volumes is crucial for accurate diagnosis, effective surgical planning, and longitudinal disease monitoring. Despite advancements in AI, clinically viable segmentation is often hindered by the scarcity of 3D annotations, patient-specific variability, data privacy concerns, and substantial computational overhead. In this work, we propose FALCON, a cross-domain few-shot segmentation framework that achieves high-precision 3D volume segmentation by processing data as 2D slices. The framework is first meta-trained on natural images to learn-to-learn generalizable segmentation priors, then transferred to the medical domain via adversarial fine-tuning and boundary-aware learning. Task-aware inference, conditioned on support cues, allows FALCON to adapt dynamically to patient-specific anatomical variations across slices. Experiments on four benchmarks demonstrate that FALCON consistently achieves the lowest Hausdorff Distance scores, indicating superior boundary accuracy while maintaining a Dice Similarity Coefficient comparable to the state-of-the-art models. Notably, these results are achieved with significantly less labeled data, no data augmentation, and substantially lower computational overhead.

Problem

• few-shot learning
• cross-domain segmentation
• medical image segmentation
• 3D annotation scarcity
• computational overhead

Innovation

• few-shot learning
• cross-domain segmentation
• adversarial fine-tuning
• boundary-aware learning
• meta-learning