AGGR-Mamba: Mamba-based Network with Attention-Guided Gated Residual Refinement for Undersampled MRI Reconstruction

Yue Zhuo; Wenbin Deng; Lingjun Liu; Xuhang Chen; Zhonghua Xie

cs.CVOct 20, 2026

AGGR-Mamba: Mamba-based Network with Attention-Guided Gated Residual Refinement for Undersampled MRI Reconstruction

Authors: Yue Zhuo, Wenbin Deng, Lingjun Liu, Xuhang Chen, Zhonghua Xie

Organizations: School of Mathematics and Statistics, Huizhou University, Huizhou, China · School of Computer Science and Engineering, Huizhou University, Huizhou, China

Abstract

Undersampled magnetic resonance imaging (MRI) reconstruction aims to recover high-quality images from partially acquired k-space data, thereby reducing scan time. However, missing frequency-domain measurements often lead to residual aliasing artifacts, blurred edges, and loss of local structural details. In this paper, we propose AGGR-Mamba, a lightweight output-stage residual refinement method built upon a Mamba backbone for undersampled brain MRI reconstruction. The Mamba backbone first produces an initial reconstruction, and the proposed refinement modules then correct the remaining local errors and high-frequency artifacts in a controlled residual manner. Specifically, an Attention Refine Head is introduced to recalibrate local anatomical structures, edge textures, and residual-error regions through channel and spatial attention. In addition, a GDFN-style gated residual refinement branch is designed to selectively preserve informative residual features while suppressing redundant or unstable responses. By combining attention-based recalibration with gated residual selection, AGGR-Mamba improves local structure recovery without substantially disturbing the backbone reconstruction. Experiments are conducted on the IXI brain MRI dataset with T1- and T2-weighted images under acceleration factors of R=4 and R=8. The results show that AGGR-Mamba achieves competitive PSNR and SSIM performance across different reconstruction settings, demonstrating the effectiveness of the proposed output-stage refinement strategy.

Figures & tables

Fig. 1: Illustration of the undersampled MRI reconstruction task: undersampled input (left) and AGGR-Mamba reconstruction (right) on IXI-T2 with $R=4$ . The two rows highlight different regions of the same slice; blue boxes and lines indicate the corresponding enlarged regions.

Fig. 2: Schematic illustration of the proposed AGGR-Mamba method.

Fig. 3: Reconstruction results and error maps of different methods on IXI-T2 $R=4$ . The zoomed-in regions and error maps are placed below each reconstructed image for visual comparison.

Method	IXI-T1 $R=4$		IXI-T1 $R=8$		IXI-T2 $R=4$		IXI-T2 $R=8$
Method	PSNR $\uparrow$	SSIM(%) $\uparrow$	PSNR $\uparrow$	SSIM(%) $\uparrow$	PSNR $\uparrow$	SSIM(%) $\uparrow$	PSNR $\uparrow$	SSIM(%) $\uparrow$
UNet	37.01	94.07	31.82	90.57	38.06	94.85	33.86	91.98
E2E-VarNet	41.94	98.56	35.72	97.21	42.96	98.69	37.23	97.34
SwinUNET	34.77	94.73	30.59	91.75	37.54	95.11	33.92	92.44
SwinMR	38.30	96.05	33.76	93.65	40.03	96.17	35.75	93.89
ReconFormer	42.36	99.24	33.87	97.51	46.71	99.53	38.34	98.45

TABLE I: Quantitative results of different methods on the IXI dataset. The best performance is marked in bold, while the second-best performance is underlined.

Variant	Attention Refine	GDFN	Joint Fine-tuning	PSNR(dB) $\uparrow$	SSIM(%) $\uparrow$	$\Delta$ PSNR
Mamba Backbone	$\times$	$\times$	$\times$	46.4907	99.38	–
Attention Refine	$\checkmark$	$\times$	$\times$	46.9393	99.44	+0.4486
Attention Refine + GDFN	$\checkmark$	$\checkmark$	$\times$	46.9930	99.44	+0.5023
AGGR-Mamba	$\checkmark$	$\checkmark$	$\checkmark$	46.9945	99.45	+0.5038

TABLE II: Module-wise ablation study on IXI-T2

R=4

\Delta

indicates the PSNR gain relative to the Mamba backbone.

Fig. 4: PSNR comparison of different module combinations on IXI-T2

R=4

Setting	Late-stage GDFN Fine-tuning	Joint Fine-tuning	$\Delta$ PSNR
IXI-T1 $R=4$	41.7266 / 99.11	41.9360 / 99.11	+0.2094
IXI-T1 $R=8$	35.3777 / 97.04	35.3879 / 97.04	+0.0102
IXI-T2 $R=4$	46.9930 / 99.44	46.9945 / 99.45	+0.0015
IXI-T2 $R=8$	38.4408 / 97.87	38.4789 / 97.88	+0.0381

TABLE III: Comparison between late-stage GDFN fine-tuning and joint fine-tuning on the four IXI test settings. Each cell reports PSNR(dB) / SSIM(%).

Accelerated MRI Reconstruction

Accelerated MRI reconstruction recovers diagnostic images from undersampled k-space data to reduce scan time while preserving anatomical and quantitative information.

See all 85 papers

Variational Network with Wavelet-based UNET in Accelerated MRI Reconstruction from Under Sampled K-space Data
Jun 13, 2026Yasir Arafat Prodhan, Shaikh Anowarul FattahU-Net Wavelet Methods
Department of Electrical and Electronic Engineering, Bangladesh University of Engineering and Technology, Dhaka, Bangladesh · This paper was produced by the IEEE Publication Technology Group. They are in Piscataway, NJ.
Next-Acceleration-Scale Prediction for Autoregressive MRI Reconstruction
May 19, 2026Yilmaz Korkmaz, Vishal M. PatelVisual Autoregressive Models Computational Imaging
Johns Hopkins University, Baltimore MD 21218, USA
Direct low-field MRI super-resolution using undersampled k-space
Feb 28, 2026Daniel Tweneboah Anyimadu, Mohammed M. Abdelsamea, Ahmed Karam EldalyMagnetic Resonance Imaging Medical Image Super-Resolution
Department of Computer Science, University of Exeter, Exeter, United Kingdom · UCL Hawkes Institute, University College London, London, United Kingdom

Jun 13, 2026J/K move · Enter open · S save

Variational Network with Wavelet-based UNET in Accelerated MRI Reconstruction from Under Sampled K-space Data

Yasir Arafat Prodhan, Shaikh Anowarul Fattah

Department of Electrical and Electronic Engineering, Bangladesh University of Engineering and Technology, Dhaka, Bangladesh · This paper was produced by the IEEE Publication Technology Group. They are in Piscataway, NJ.

U-Net Wavelet Methods Accelerated MRI Reconstruction Magnetic Resonance Imaging

Open paper AskPDF arXiv

Fully sampled MRI requires dense k-space acquisition, leading to long scan times, reduced clinical throughput, and increased sensitivity to patient motion. Accelerated MRI addresses this by acquiring undersampled k-space data and reconstructing the missing information computationally. However, reconstruction from undersampled measurements is highly ill-posed and can introduce aliasing artifacts, noise amplification, and loss of anatomical detail. Although conventional parallel imaging and compressed sensing methods mitigate these issues, and deep learning methods have further improved reconstruction quality, preserving high-frequency structures under aggressive undersampling remains challenging. In this work, we propose a Variational Network with a Wavelet-based U-Net (W-UNet) for accelerated MRI reconstruction. The framework combines physics-guided iterative reconstruction with learnable multi-scale frequency representations. Standard pooling operations are replaced with Discrete Wavelet Transform and Inverse Wavelet Transform modules, enabling lossless downsampling while preserving low-frequency structure and high-frequency edge details. Integrated into the refinement and sensitivity map estimation stages, the proposed design improves artifact suppression, feature preservation, and reconstruction fidelity in both single-coil and multi-coil settings. Experiments on fastMRI knee and M4Raw brain datasets show state-of-the-art performance. Ablation studies further confirm the effectiveness of wavelet-based feature decomposition for accelerated MRI reconstruction.

Abstract

Accelerated MRI Reconstruction

Explore similar work

Variational Network with Wavelet-based UNET in Accelerated MRI Reconstruction from Under Sampled K-space Data

Next-Acceleration-Scale Prediction for Autoregressive MRI Reconstruction

Direct low-field MRI super-resolution using undersampled k-space