[论文解读] Unrolled Reconstruction with Integrated Super-Resolution for Accelerated 3D LGE MRI

Md Hasibul Husain Hisham, Shireen Elhabian+6 more|arXiv (Cornell University)|Mar 18, 2026

Advanced Image Processing Techniques被引用 0

一句话总结

本工作将增强型深度超分辨率（EDSR）网络嵌入为模型驱动展开重建的近端算子，以实现加速3D LGE MRI 的联合超分辨与数据一致性约束，从而提升重建质量与左心房分割效果。

ABSTRACT

Accelerated 3D late gadolinium enhancement (LGE) MRI requires robust reconstruction methods to recover thin atrial structures from undersampled k-space data. While unrolled model-based networks effectively integrate physics-driven data consistency with learned priors, they operate at the acquired resolution and may fail to fully recover high-frequency detail. We propose a hybrid unrolled reconstruction framework in which an Enhanced Deep Super-Resolution (EDSR) network replaces the proximal operator within each iteration of the optimization loop, enabling joint super-resolution enhancement and data consistency enforcement. The model is trained end-to-end on retrospectively undersampled preclinical 3D LGE datasets and compared against compressed sensing, Model-Based Deep Learning (MoDL), and self-guided Deep Image Prior (DIP) baselines. Across acceleration factors, the proposed method consistently improves PSNR and SSIM over standard unrolled reconstruction and better preserves fine cardiac structures, leading to improved LA (left atrium) segmentation performance. These results demonstrate that integrating super-resolution priors directly within model-based reconstruction provides measurable gains in accelerated 3D LGE MRI.

研究动机与目标

• 通过从欠采样数据中恢复薄壁心房结构来推动加速的3D LGE-MRI。
• 开发一种混合的展开重建，其中 EDSR 取代传统的近端算子。
• 演示端到端训练以同时提升像素保真度和下游分割性能。
• 与 CS、MoDL 及自引导型 DIP 基线进行比较，并通过 LA 分割评估临床相关性。

提出的方法

• 在每次展开迭代中用一个在各迭代之间共享的 EDSR 网络替换近端算子。
• 以零填充的伴随重建初始化，通过共轭梯度法实现数据一致性。
• 以完整采样的真值作为 L2 损失进行端到端训练。
• 对三维多线圈 k 空间数据进行切片级二维处理，采用 SENSE 编码。
• 包括一个仅仅作为消融的普通展开变体，使用 U-Net 去噪。

Figure 1: Proposed Unrolled EDSR framework. The EDSR Net serves as the learned proximal operator at each of the $N$ unrolled iterations, with weights $\phi$ shared across iterations. The initial reconstruction $\mathbf{x}^{(0)}=\mathbf{A}^{H}\mathbf{y}$ is the zero-filled adjoint, and data consisten

实验结果

研究问题

• RQ1 将每次展开迭代中嵌入的超分先验是否能在保持数据保真度的同时提升高频细节？
• RQ2 Unrolled+EDSR 是否在 PSNR/SSIM 方面优于 CS、MoDL 与自引导 DIP，在加速的3D LGE MRI 中？
• RQ3 与基线相比，该方法是否提升下游 LA 分割的性能？

主要发现

Method	PSNR (R=4)	SSIM (R=4)	PSNR (R=6)	SSIM (R=6)
Compressed Sensing	31.6±1.0	0.865±0.019	28.6±1.4	0.824±0.020
Self-Guided DIP	32.8±1.2	0.894±0.018	29.3±1.1	0.855±0.015
Unrolled	35.1±1.2	0.906±0.015	32.4±1.5	0.878±0.012
EDSR-Unrolled	35.6±1.1	0.915±0.014	32.9±1.6	0.889±0.015

• Unrolled+EDSR 在加速因子 R=4 和 R=6 时，PSNR/SSIM 高于基线。
• 在 R=4 时，Unrolled+EDSR 获得 PSNR 35.6 dB、SSIM 0.915，优于 Unrolled（35.1 dB、0.906）和 CS（31.6 dB、0.865）。
• 在 R=6 时，Unrolled+EDSR 获得 PSNR 32.9 dB、SSIM 0.889，优于 Unrolled（32.4 dB、0.878）。
• 对重建切片的 LA 分割显示 DSC：Unrolled+EDSR 为 0.893，Unrolled 为 0.884。
• 消融分析表明 9 次展开迭代在质量与内存之间取得平衡；超过该值收益递减。

Figure 2: Qualitative comparison of reconstruction methods at acceleration factors $R=4$ and $R=6$ . Each column shows a representative axial slice from a preclinical subject. At $R=4$ , Unrolled+EDSR and Unrolled both recover fine cardiac structures faithfully, while Compressed Sensing introduces v