MIRAGE: Multi-scale Lesion-Informed Representation with Auxiliary Guidance for MRI Contrast Enhancement

2026-07-21

AI Summary

This work addresses the challenge of synthesizing contrast-enhanced breast MRI from a single non-contrast slice, where uncertainty in lesion enhancement complicates the simultaneous preservation of global realism and lesion fidelity. To this end, the authors propose MIRAGE, a method built upon a residual 2D U-Net architecture that integrates global reconstruction and perceptual losses, and introduces three lesion-aware supervision mechanisms tailored to distinct training phases: asymmetric tumor enhancement deficiency penalty, multi-scale auxiliary tumor segmentation, and frozen nnU-Net guidance. Evaluated on a multi-center MAMA-SYNTH dataset of 301 cases, MIRAGE outperforms baseline approaches—including pix2pix and conditional diffusion models—on six out of eight quantitative metrics and demonstrates superior performance in downstream lesion localization tasks.

Abstract

Inferring contrast enhancement from one pre-contrast breast MRI slice is underdetermined: post-contrast appearance contains physiological information that is not uniquely encoded in baseline anatomy. Optimizing only paired pixel fidelity can suppress uncertain lesion enhancement, whereas adversarial or stochastic generative objectives can favor realistic post-contrast appearance without guaranteeing patient-specific lesion fidelity. We introduce MIRAGE, a residual 2D U-Net that combines global reconstruction and perceptual losses with three forms of lesion-aware supervision available only during training: an asymmetric penalty for missed tumor enhancement, multi-scale auxiliary tumor segmentation, and guidance through a frozen post-contrast tumor segmentation nnU-Net. We evaluate the method on 301 cases from the multi-centre MAMA-SYNTH data using eight complementary image-, region-, radiomics-, and segmentation-based metrics. MIRAGE ranks first on six metrics and markedly improves downstream lesion localization over tuned pix2pix, conditional diffusion, and latent bridge-matching baselines. The generative alternatives retain advantages in LPIPS or contrast classification, revealing a clear fidelity-utility trade-off. Leave-one-in and leave-one-out ablations show that the losses are partly redundant for lesion localization but exert distinct effects on appearance, radiomics, and boundary accuracy. These results support task-aware synthesis while also showing that its apparent optimality is conditional on the downstream models and metrics used to define utility.

Problem

• MRI contrast enhancement
• lesion fidelity
• underdetermined inference
• generative synthesis
• breast MRI

Innovation

• lesion-informed representation
• multi-scale supervision
• auxiliary guidance
• contrast enhancement synthesis
• task-aware MRI generation