MPFlow: Multi-modal Posterior-Guided Flow Matching for Zero-Shot MRI Reconstruction

TL;DR

MPFlow is a novel zero-shot MRI reconstruction method that leverages auxiliary modalities through cross-modal guidance, significantly reducing hallucinations and improving efficiency without retraining the generative prior.

Contribution

Introduces MPFlow, a multi-modal reconstruction framework utilizing a self-supervised pretraining strategy for cross-modal guidance, enhancing zero-shot MRI reconstruction fidelity and efficiency.

Findings

Matches diffusion baselines with 80% fewer steps

Reduces tumor hallucinations by over 15%

Improves anatomical fidelity in reconstructed images

Abstract

Zero-shot MRI reconstruction relies on generative priors, but single-modality unconditional priors produce hallucinations under severe ill-posedness. In many clinical workflows, complementary MRI acquisitions (e.g. high-quality structural scans) are routinely available, yet existing reconstruction methods lack mechanisms to leverage this additional information. We propose MPFlow, a zero-shot multi-modal reconstruction framework built on rectified flow that incorporates auxiliary MRI modalities at inference time without retraining the generative prior to improve anatomical fidelity. Cross-modal guidance is enabled by our proposed self-supervised pretraining strategy, Patch-level Multi-modal MR Image Pretraining (PAMRI), which learns shared representations across modalities. Sampling is jointly guided by data consistency and cross-modal feature alignment using pre-trained PAMRI, systematically suppressing intrinsic and extrinsic hallucinations. Extensive experiments on HCP and BraTS show that MPFlow matches diffusion baselines on image quality using only 20% of sampling steps while reducing tumor hallucinations by more than 15% (segmentation dice score). This demonstrates that cross-modal guidance enables more reliable and efficient zero-shot MRI reconstruction.