UEPS: Robust and Efficient MRI Reconstruction

2026-03-19

AI Summary

This work proposes UEPS, a novel architecture for accelerated MRI reconstruction that addresses the domain shift and degraded generalization performance commonly caused by inaccurate coil sensitivity map estimation in deep unrolling models. UEPS eliminates reliance on coil sensitivity maps by employing per-coil independent reconstruction, progressive resolution refinement from k-space to image space, and an MRI-specific sparse attention mechanism tailored for 1D undersampling patterns. The method demonstrates superior robustness across ten out-of-distribution test sets encompassing diverse anatomies, views, contrasts, scanner vendors, field strengths, and coil configurations. Furthermore, UEPS enables low-latency inference, offering both computational efficiency and strong potential for clinical deployment.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Deep unrolled models (DUMs) have become the state of the art for accelerated MRI reconstruction, yet their robustness under domain shift remains a critical barrier to clinical adoption. In this work, we identify coil sensitivity map (CSM) estimation as the primary bottleneck limiting generalization. To address this, we propose UEPS, a novel DUM architecture featuring three key innovations: (i) an Unrolled Expanded (UE) design that eliminates CSM dependency by reconstructing each coil independently; (ii) progressive resolution, which leverages k-space-to-image mapping for efficient coarse-to-fine refinement; and (iii) sparse attention tailored to MRI's 1D undersampling nature. These physics-grounded designs enable simultaneous gains in robustness and computational efficiency. We construct a large-scale zero-shot transfer benchmark comprising 10 out-of-distribution test sets spanning diverse clinical shifts -- anatomy, view, contrast, vendor, field strength, and coil configurations. Extensive experiments demonstrate that UEPS consistently and substantially outperforms existing DUM, end-to-end, diffusion, and untrained methods across all OOD tests, achieving state-of-the-art robustness with low-latency inference suitable for real-time deployment.

Problem

• MRI reconstruction
• domain shift
• robustness
• coil sensitivity map
• generalization

Innovation

• coil sensitivity map-free
• unrolled expanded architecture
• progressive resolution
• sparse attention
• zero-shot MRI reconstruction