ULF-Synth: Physics-Guided Ultra-Low-Field MRI Enhancement for Pediatric Neuroimaging

2026-05-23

AI Summary

This work addresses the challenges of ultra-low-field (ULF) MRI in pediatric neuroimaging, where inherently low signal-to-noise ratio and limited resolution hinder clinical utility, compounded by the absence of real paired high- and low-field data for supervised learning. To overcome this, the authors propose a synthetic supervision paradigm that leverages a physics-based model to generate realistic ULF images from high-field MRI scans, thereby constructing large-scale paired training datasets without requiring actual ULF–high-field image pairs. They further introduce a joint spatial–frequency domain optimization objective to effectively recover high-frequency anatomical details. The framework is compatible with encoder–decoder, adversarial, and diffusion architectures, demonstrates strong generalization on real 64 mT ULF data, significantly improves brain segmentation accuracy, and achieves higher diagnostic acceptability and radiologist preference in blinded evaluations.

Abstract

Ultra-low-field (ULF) MRI offers portable and accessible neuroimaging but suffers from reduced signal-to-noise ratio and limited spatial resolution compared to high-field (HF) systems. Acquiring paired ULF-HF data for supervised enhancement is often difficult, particularly in resource-limited settings. We introduce ULF-Synth, a framework that combines: (i) acquisition-based synthesis of realistic ULF images from HF volumes to create large-scale paired training data, (ii) a spatial-frequency domain objective that prioritizes recovery of high-frequency anatomical detail. This formulation is architecture-agnostic, consistently improving structural similarity and perceptual fidelity across encoder-decoder, adversarial, and diffusion-based translation models. When trained exclusively on synthetic data, the resulting models generalize effectively to real 64mT ULF acquisitions, improving downstream multiclass brain segmentation and achieving higher radiologist preference and diagnostic acceptability in a blinded reader study. These findings demonstrate that synthetic paired supervision provides a practical and scalable pathway for enhancing ULF MRI without requiring real paired acquisitions. Code, Models and Dataset: https://github.com/toufiqmusah/ULF-Synth

Problem

• Ultra-low-field MRI
• Pediatric neuroimaging
• Paired data acquisition
• Image enhancement
• Signal-to-noise ratio

Innovation

• ULF-MRI
• physics-guided synthesis
• spatial-frequency domain loss
• paired data synthesis
• architecture-agnostic enhancement