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

arXiv PyPI GitHub Paper MIT License
Toufiq Musah · Salvatore Calcagno · Federica Proietto Salanitri · Xiaomeng Li · Maruf Adewole · Marawan Elbatel

Abstract

Ultra-low-field (ULF) MRI enables portable and accessible neuroimaging, but suffers from low signal-to-noise ratio and limited spatial resolution relative to high-field (HF) systems. Acquiring paired ULF–HF data for supervised enhancement is often infeasible, particularly in resource-limited settings. We introduce ULF-Synth, a framework combining: (i) acquisition-based synthesis of realistic ULF images from HF volumes for large-scale paired training, and (ii) a spatial-frequency domain objective that prioritizes recovery of high-frequency anatomical detail. The formulation is architecture-agnostic, consistently improving structural similarity and perceptual fidelity across encoder-decoder, adversarial, and diffusion-based translation models. Trained exclusively on synthetic data, our models generalize to real 64 mT ULF acquisitions, improving multiclass brain segmentation and achieving higher radiologist preference and diagnostic acceptability in a blinded reader study.

Method

ULF-Synth method overview
Figure 1: ULF-Synth framework overview. (a) HF volumes are degraded via sequential image- and k-space operations to generate synthetic ULF pairs. (b) Real ULF scans are enhanced and passed to a segmentation model for downstream evaluation. (c) Framework components.

The ULF synthesis module models the key physical phenomena distinguishing ULF from HF acquisitions:

01

Signal scaling

(BULF/BHF)2 polarization ratio

02

T2* decay & B0 inhomogeneity

Spatially-varying exponential decay from random B0 field maps

03

Thermal noise

Gaussian noise scaled to SNR 15–50

04

k-space cropping

Reduced resolution (45–55%)

05

k-space undersampling

Accelerated acquisition (2×–3×) with center-out sampling

06

B0 off-resonance distortion

Phase distortion from random B0 field maps

Results

Qualitative results
Figure 2: Qualitative comparison of ULF image enhancement on synthetic and real clinical ULF acquisitions.

Usage

Installation

Terminal 
# Install ULF-Synth
pip install ulfsynth

Enhancement

Python 
from ulfsynth.enhance import enhance_file, enhance_folder

# Single file (CPU)
enhance_file("ulf_scan.nii.gz", "enhanced.nii.gz", device="cpu")

# Batch folder (GPU)
enhance_folder("ulf_scans/", "enhanced_scans/", device="cuda")

Simulation

Python 
from ulfsynth.simulate import simulate_ulf, simulate_file, simulate_folder

# Get the ULF volume + metadata
ulf_vol, affine, header, params = simulate_ulf("hf_scan.nii.gz", seed=42)

# Save directly
params = simulate_file("hf_scan.nii.gz", "ulf_scan.nii.gz", seed=42)

# Process a folder
results = simulate_folder("hf_scans/", "ulf_scans/", seed=42)

Citation

@article{musah2026ulfsynth, title = {ULF-Synth: Physics-Guided Ultra-Low-Field MRI Enhancement for Pediatric Neuroimaging}, author = {Toufiq Musah and Salvatore Calcagno and Federica Proietto Salanitri and Xiaomeng Li and Maruf Adewole and Marawan Elbatel}, year = {2026}, journal = {arXiv preprint arXiv:2605.24625}, eprint = {2605.24625}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2605.24625} }