MAGO-SP: Detection and Correction of Water-Fat Swaps in Magnitude-Only VIBE MRI

TL;DR

This work tackles water-fat swaps in magnitude-only VIBE Dixon MRI, which bias proton density fat fraction ($PDFF$) estimates in large population datasets. It introduces MAGO-SP, a three-part pipeline that (i) detects swaps via segmentation trained on synthetic Perlin-noise-swapped data, (ii) generates a water signal prior with aPalette Conditional Denoising Diffusion Network, and (iii) performs physics-constrained reconstruction to resolve water and fat signals, applicable to both 2-point and 6-point VIBE. The approach yields improved solution selection (e.g., $ ext{Fraction Correct} ightarrow 0.911 ext{ (±0.020)}$) and higher image fidelity (SSIM and PSNR) while revealing BMI-dependent swap biases that the correction mitigates. By enabling reliable PDFF estimation across large cohorts, MAGO-SP provides a scalable foundation for automated population imaging analyses and downstream metabolic-health research, with code and models publicly available as noted.

Abstract

Volume Interpolated Breath-Hold Examination (VIBE) MRI generates images suitable for water and fat signal composition estimation. While the two-point VIBE provides water-fat-separated images, the six-point VIBE allows estimation of the effective transversal relaxation rate R2* and the proton density fat fraction (PDFF), which are imaging markers for health and disease. Ambiguity during signal reconstruction can lead to water-fat swaps. This shortcoming challenges the application of VIBE-MRI for automated PDFF analyses of large-scale clinical data and of population studies. This study develops an automated pipeline to detect and correct water-fat swaps in non-contrast-enhanced VIBE images. Our three-step pipeline begins with training a segmentation network to classify volumes as "fat-like" or "water-like," using synthetic water-fat swaps generated by merging fat and water volumes with Perlin noise. Next, a denoising diffusion image-to-image network predicts water volumes as signal priors for correction. Finally, we integrate this prior into a physics-constrained model to recover accurate water and fat signals. Our approach achieves a < 1% error rate in water-fat swap detection for a 6-point VIBE. Notably, swaps disproportionately affect individuals in the Underweight and Class 3 Obesity BMI categories. Our correction algorithm ensures accurate solution selection in chemical phase MRIs, enabling reliable PDFF estimation. This forms a solid technical foundation for automated large-scale population imaging analysis.

Figures (4)

• Figure 1: Example of two and six-point Dixon data where the reconstruction from MRI-device vendor failed. The wrong result is selected during the solution selection due to signal ambiguity. We highlighted the liver in the 6-point Dixon in purple.
• Figure 2: Our proposed pipeline. We detect water-fat swaps by segmenting them into the water and fat volume. If a swap is detected, we generate a signal prior from the raw data and then use the physically constrained method to reconstruct the water and fat images. Our proposed steps are agnostic towards the number of gradient echos and the type of the applied physics model.
• Figure 3: In Vivo water reconstruction of 2-point Dixon VIBE. Our method retrospectively corrected the swaps. Swaps are marked with red arrows.
• Figure 4: Known good in vivo water reconstruction of 6-point Dixon VIBE with compared reconstruction methods. False solution selection leads to black spots in the baselines. Signal prior is generated from an image-to-image network, and MAGO variants add the physical constraints.