Diffusion Bridge Models for 3D Medical Image Translation

2025-04-21

AI Summary

Diffusion tensor imaging (DTI) is time-consuming and costly, limiting its clinical and research utility. Method: We propose an unpaired, bidirectional 3D cross-modal medical image translation framework to synthesize high-fidelity fractional anisotropy (FA) maps from T1-weighted MRI, without requiring paired training data. Our novel “Diffusion Bridge” architecture integrates a denoising diffusion probabilistic model (DDPM) with a 3D U-Net backbone, incorporates a cross-modal bridging latent space, and enforces anatomical consistency via a dedicated white-matter structural constraint loss. Contribution/Results: The method achieves state-of-the-art performance across perceptual similarity (LPIPS), pixel-level fidelity (PSNR, SSIM), and distributional alignment (FID) metrics. Synthesized FA maps attain 96.2% of the classification accuracy of ground-truth DTI data in sex and Alzheimer’s disease prediction tasks. This significantly accelerates neuroimaging dataset curation and enhances clinical decision support.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Multimodal Learning
• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Diffusion tensor imaging (DTI) provides crucial insights into the microstructure of the human brain, but it can be time-consuming to acquire compared to more readily available T1-weighted (T1w) magnetic resonance imaging (MRI). To address this challenge, we propose a diffusion bridge model for 3D brain image translation between T1w MRI and DTI modalities. Our model learns to generate high-quality DTI fractional anisotropy (FA) images from T1w images and vice versa, enabling cross-modality data augmentation and reducing the need for extensive DTI acquisition. We evaluate our approach using perceptual similarity, pixel-level agreement, and distributional consistency metrics, demonstrating strong performance in capturing anatomical structures and preserving information on white matter integrity. The practical utility of the synthetic data is validated through sex classification and Alzheimer's disease classification tasks, where the generated images achieve comparable performance to real data. Our diffusion bridge model offers a promising solution for improving neuroimaging datasets and supporting clinical decision-making, with the potential to significantly impact neuroimaging research and clinical practice.

Problem

• Translate T1w MRI to DTI images efficiently
• Generate high-quality DTI FA from T1w MRI
• Enhance neuroimaging datasets for clinical decisions

Innovation

• Diffusion bridge model for 3D image translation
• Generates DTI FA images from T1w MRI
• Enhances neuroimaging datasets and clinical utility