Bézier Meets Diffusion: Robust Generation Across Domains for Medical Image Segmentation

2025-09-26

AI Summary

Cross-modal medical image segmentation suffers from substantial domain shifts and the inability of conventional GAN-based style transfer methods to accurately model mappings in highly variable anatomical regions. Method: This paper proposes a unified cross-domain image generation framework: (1) Bézier curve modeling for nonlinear, fine-grained cross-modal style transfer to improve mapping fidelity in high-variability regions; and (2) an uncertainty-guided score-matching strategy to enhance the robustness of conditional diffusion models (CDMs) against noisy pseudo-labels, enabling high-fidelity synthesis of labeled target-domain images. Contribution/Results: Extensive experiments on multiple public benchmarks demonstrate that the generated images achieve both realistic appearance and accurate annotations, effectively augmenting target-domain supervision. The synthesized data significantly boosts cross-domain segmentation performance, achieving state-of-the-art results.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Natural Language Processing: Generation

Application Category

• 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
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Training robust learning algorithms across different medical imaging modalities is challenging due to the large domain gap. Unsupervised domain adaptation (UDA) mitigates this problem by using annotated images from the source domain and unlabeled images from the target domain to train the deep models. Existing approaches often rely on GAN-based style transfer, but these methods struggle to capture cross-domain mappings in regions with high variability. In this paper, we propose a unified framework, Bézier Meets Diffusion, for cross-domain image generation. First, we introduce a Bézier-curve-based style transfer strategy that effectively reduces the domain gap between source and target domains. The transferred source images enable the training of a more robust segmentation model across domains. Thereafter, using pseudo-labels generated by this segmentation model on the target domain, we train a conditional diffusion model (CDM) to synthesize high-quality, labeled target-domain images. To mitigate the impact of noisy pseudo-labels, we further develop an uncertainty-guided score matching method that improves the robustness of CDM training. Extensive experiments on public datasets demonstrate that our approach generates realistic labeled images, significantly augmenting the target domain and improving segmentation performance.

Problem

• Reducing domain gaps across medical imaging modalities for segmentation
• Generating realistic labeled images using diffusion models
• Improving segmentation robustness with uncertainty-guided training

Innovation

• Bezier-curve style transfer reduces domain gap
• Conditional diffusion model synthesizes labeled target images
• Uncertainty-guided score matching improves training robustness