标题： 可重复可见的双域自监督深度展开网络用于MRI重建 显示英文标题

标题： Re-Visible Dual-Domain Self-Supervised Deep Unfolding Network for MRI Reconstruction

摘要： 磁共振成像（MRI）在临床实践中被广泛应用，但其采集时间较长。尽管已经提出了深度学习方法来加速采集并表现出有前景的性能，但它们依赖于高质量的完全采样数据集进行监督训练。然而，这样的数据集收集耗时且成本高昂，限制了它们的更广泛应用。另一方面，自监督方法通过仅从欠采样数据中学习提供了一种替代方案，但大多数现有方法依赖于进一步分割的欠采样k空间数据作为模型的输入进行训练，导致有价值的信息丢失。此外，它们的模型并未充分结合图像先验，导致重建性能下降。在本文中，我们提出了一种新颖的可重新可见的双域自监督深度展开网络，以解决仅在可用欠采样数据集时的这些问题。具体而言，通过结合可重新可见的双域损失，在训练过程中利用所有欠采样k空间数据以减轻进一步分割引起的信息丢失。这种设计使模型能够隐式地适应所有欠采样k空间数据作为输入。此外，我们设计了一个基于Chambolle和Pock近似点算法的深度展开网络（DUN-CP-PPA），以实现端到端重建，结合成像物理和图像先验来指导重建过程。通过采用空间-频率特征提取（SFFE）模块来捕捉全局和局部特征表示，我们提高了模型学习全面图像先验的效率。在fastMRI和IXI数据集上进行的实验表明，我们的方法在重建性能方面显著优于最先进的方法。 显示英文摘要

摘要： Magnetic Resonance Imaging (MRI) is widely used in clinical practice, but suffered from prolonged acquisition time. Although deep learning methods have been proposed to accelerate acquisition and demonstrate promising performance, they rely on high-quality fully-sampled datasets for training in a supervised manner. However, such datasets are time-consuming and expensive-to-collect, which constrains their broader applications. On the other hand, self-supervised methods offer an alternative by enabling learning from under-sampled data alone, but most existing methods rely on further partitioned under-sampled k-space data as model's input for training, resulting in a loss of valuable information. Additionally, their models have not fully incorporated image priors, leading to degraded reconstruction performance. In this paper, we propose a novel re-visible dual-domain self-supervised deep unfolding network to address these issues when only under-sampled datasets are available. Specifically, by incorporating re-visible dual-domain loss, all under-sampled k-space data are utilized during training to mitigate information loss caused by further partitioning. This design enables the model to implicitly adapt to all under-sampled k-space data as input. Additionally, we design a deep unfolding network based on Chambolle and Pock Proximal Point Algorithm (DUN-CP-PPA) to achieve end-to-end reconstruction, incorporating imaging physics and image priors to guide the reconstruction process. By employing a Spatial-Frequency Feature Extraction (SFFE) block to capture global and local feature representation, we enhance the model's efficiency to learn comprehensive image priors. Experiments conducted on the fastMRI and IXI datasets demonstrate that our method significantly outperforms state-of-the-art approaches in terms of reconstruction performance.