Atlas is Your Perfect Context: One-Shot Customization for Generalizable Foundational Medical Image Segmentation

TL;DR

AtlasSegFM is a novel framework that enables one-shot customization of foundation models for medical image segmentation by leveraging atlas-guided prompts and test-time adaptation, improving accuracy especially for small structures.

Contribution

The paper introduces AtlasSegFM, a new atlas-guided approach that adapts foundation models to specific clinical contexts with a single annotated example, enhancing segmentation performance.

Findings

Consistently improves segmentation accuracy across multiple datasets and modalities.

Enhances performance on small and delicate structures in medical images.

Provides a lightweight, deployable solution for clinical workflows.

Abstract

Accurate medical image segmentation is essential for clinical diagnosis and treatment planning. While recent interactive foundation models (e.g., nnInteractive) enhance generalization through large-scale multimodal pretraining, they still depend on precise prompts and often perform below expectations in contexts that are underrepresented in their training data. We present AtlasSegFM, an atlas-guided framework that customizes available foundation models to clinical contexts with a single annotated example. The core innovations are: 1) a pipeline that provides context-aware prompts for foundation models via registration between a context atlas and query images, and 2) a test-time adapter to fuse predictions from both atlas registration and the foundation model. Extensive experiments across public and in-house datasets spanning multiple modalities and organs demonstrate that AtlasSegFM consistently improves segmentation, particularly for small, delicate structures. AtlasSegFM provides a lightweight, deployable solution one-shot customization of foundation models in real-world clinical workflows. The code will be made publicly available.