UEPS: Robust and Efficient MRI Reconstruction

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Zhou, Xiang, Shang, Hong, Zhan, Zijian, He, Tianyu, Meng, Jintao, Liang, Dong
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908900727455744
author Zhou, Xiang
Shang, Hong
Zhan, Zijian
He, Tianyu
Meng, Jintao
Liang, Dong
author_facet Zhou, Xiang
Shang, Hong
Zhan, Zijian
He, Tianyu
Meng, Jintao
Liang, Dong
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2603_18572
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle UEPS: Robust and Efficient MRI Reconstruction
Zhou, Xiang
Shang, Hong
Zhan, Zijian
He, Tianyu
Meng, Jintao
Liang, Dong
Image and Video Processing
Computer Vision and Pattern Recognition
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.
title UEPS: Robust and Efficient MRI Reconstruction
topic Image and Video Processing
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.18572