IntroductionDMD-augmented Unpaired Neural Schrödinger Bridge for Ultra-Low Field MRI Enhancement

2026-03-04

AI Summary

This work addresses the challenge of poor image quality and anatomical distortion in ultra-low-field (64 mT) brain MRI, which stems from the absence of paired data with high-field (3T) images. To overcome this, the authors propose an unpaired 64 mT-to-3T MRI enhancement framework that synergistically combines diffusion-guided distribution matching (DMD²) with an anatomy-preserving regularizer (ASP). The approach leverages a frozen 3T diffusion teacher model, PatchNCE loss, a boundary-aware foreground–background consistency constraint, and multi-step unpaired Neural Schrödinger Bridge (UNSB) optimization to jointly enforce global structural fidelity and distribution alignment. Experiments on two independent cohorts demonstrate that the method significantly improves photorealism under unpaired settings and achieves superior structural preservation compared to existing unpaired techniques.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Multimodal Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Bridging structured and unstructured data
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Ultra Low Field (64 mT) brain MRI improves accessibility but suffers from reduced image quality compared to 3 T. As paired 64 mT - 3 T scans are scarce, we propose an unpaired 64 mT $\rightarrow$ 3 T translation framework that enhances realism while preserving anatomy. Our method builds upon the Unpaired Neural Schrödinge Bridge (UNSB) with multi-step refinement. To strengthen target distribution alignment, we augment the adversarial objective with DMD2-style diffusion-guided distribution matching using a frozen 3T diffusion teacher. To explicitly constrain global structure beyond patch-level correspondence, we combine PatchNCE with an Anatomical Structure Preservation (ASP) regularizer that enforces soft foreground background consistency and boundary aware constraints. Evaluated on two disjoint cohorts, the proposed framework achieves an improved realism structure trade-off, enhancing distribution level realism on unpaired benchmarks while increasing structural fidelity on the paired cohort compared to unpaired baselines.

Problem

• Ultra-Low Field MRI
• Image Enhancement
• Unpaired Translation
• Anatomical Structure Preservation
• Distribution Alignment

Innovation

• Unpaired Neural Schrödinger Bridge
• Diffusion-guided Distribution Matching
• Anatomical Structure Preservation
• Ultra-Low Field MRI
• Multi-step Refinement