Patch-Based Diffusion for Data-Efficient, Radiologist-Preferred MRI Reconstruction

2025-09-25

AI Summary

MRI reconstruction faces challenges including prolonged acquisition times, motion artifacts, and scarcity of high-quality training data. To address these, we propose PaDIS-MRI—the first patch-based diffusion model explicitly designed for complex-valued, multi-coil MRI reconstruction. It instantiates the Patch-based Diffusion Inverse Solver (PaDIS) framework for severely undersampled k-space recovery—a novel application of patch-based diffusion to inverse MRI problems. Remarkably, PaDIS-MRI achieves competitive performance with only 25 k-space samples for training, substantially outperforming the full-image diffusion baseline FastMRI-EDM: it attains superior PSNR, SSIM, and NRMSE on the FastMRI brain dataset. In blinded clinical evaluation, radiologists rated 91.7% of reconstructions as diagnostically superior. Moreover, PaDIS-MRI provides robust uncertainty quantification and demonstrates strong cross-contrast generalization. This work enables high-fidelity, clinically trustworthy MRI reconstruction at minimal data cost.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Multimodal Learning
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Magnetic resonance imaging (MRI) requires long acquisition times, raising costs, reducing accessibility, and making scans more susceptible to motion artifacts. Diffusion probabilistic models that learn data-driven priors can potentially assist in reducing acquisition time. However, they typically require large training datasets that can be prohibitively expensive to collect. Patch-based diffusion models have shown promise in learning effective data-driven priors over small real-valued datasets, but have not yet demonstrated clinical value in MRI. We extend the Patch-based Diffusion Inverse Solver (PaDIS) to complex-valued, multi-coil MRI reconstruction, and compare it against a state-of-the-art whole-image diffusion baseline (FastMRI-EDM) for 7x undersampled MRI reconstruction on the FastMRI brain dataset. We show that PaDIS-MRI models trained on small datasets of as few as 25 k-space images outperform FastMRI-EDM on image quality metrics (PSNR, SSIM, NRMSE), pixel-level uncertainty, cross-contrast generalization, and robustness to severe k-space undersampling. In a blinded study with three radiologists, PaDIS-MRI reconstructions were chosen as diagnostically superior in 91.7% of cases, compared to baselines (i) FastMRI-EDM and (ii) classical convex reconstruction with wavelet sparsity. These findings highlight the potential of patch-based diffusion priors for high-fidelity MRI reconstruction in data-scarce clinical settings where diagnostic confidence matters.

Problem

• Reducing MRI acquisition time through data-efficient reconstruction
• Overcoming small dataset limitations with patch-based diffusion models
• Achieving radiologist-preferred diagnostic quality in undersampled MRI

Innovation

• Patch-based diffusion models for MRI reconstruction
• Training on small datasets with 25 images
• Outperforming baselines in radiologist diagnostic preference