Do We Really Need Diffusion? A Fast U-Net for Paired Medical Image Translation

2026-06-16

AI Summary

This study addresses the challenge of accurately estimating magnetic resonance imaging–derived fat fraction (MRI-SFF) from conventional T2-weighted MRI without relying on specialized imaging sequences. To this end, the authors propose a lightweight four-layer U-Net architecture trained on over 230,000 paired 2D medical images for image-to-image translation, benchmarked against state-of-the-art diffusion models such as DDPM and DDIM. Experimental results demonstrate that the proposed U-Net achieves superior performance in both accuracy—exhibiting a Pearson correlation coefficient of 0.975 and a mean absolute error of 0.014—and inference speed, processing each image in 25.2 milliseconds, which is 208 times faster than DDPM. These findings not only meet clinical requirements for real-time deployment but also challenge the prevailing assumption that complex generative models are indispensable for medical image translation tasks.

Abstract

Magnetic resonance imaging-signal fat fraction (MRI-SFF) quantifies tissue fat and serves as an established biomarker for metabolic and musculoskeletal disorders. The acquisition requires, however, specialized MRI sequences, which are not available routinely. We investigate whether SFF can be estimated from widely available T2-weighted (T2w) MRI via image-to-image translation (I2I). We further compare a lightweight 4-level U-Net to a state-of-the-art Denoising Diffusion Probabilistic Model (DDPM) using a dataset of 230 048 paired 2D images (183 517 train, 23 621 val, 22 910 test) from the German National Cohort (NAKO). Both models clearly outperform the identity baseline (Pearson correlation r = 0.769, mean absolute error MAE = 0.070 +/- 0.054), which confirms that the models learn a non-trivial cross-modal mapping. Interestingly, the lightweight U-Net outperforms the DDPM in both correlation (r = 0.975 vs. 0.962) and error (MAE = 0.014 +/- 0.015 vs. 0.019 +/- 0.019), while reducing inference time by a factor of 208 (25.2 ms vs. 5 227.2 ms per image using 50 Denoising Diffusion Implicit Model (DDIM) steps). The strong clinical performance at substantially reduced computational cost enables real-time clinical use.

Problem

• MRI-SFF
• T2-weighted MRI
• image-to-image translation
• medical image translation
• fat fraction quantification

Innovation

• U-Net
• image-to-image translation
• diffusion models
• medical image analysis
• real-time inference