Deep learning models have shown potential for reconstructing undersampled, multi-channel magnetic resonance (MR) image acquisitions. Recently proposed methods, however, have not leveraged information from prior subject-specific MR imaging sessions. Such data are often readily available through a picture archiving and communication system (PACS). We propose a flexible three-step method to incorporate this prior information into an enhanced deep-learning-based reconstruction process. The method consists of Step 1: an initial reconstruction; Step 2: registration of the previous scan to the initial reconstruction; and Step 3: an enhancement network. Training and testing used longitudinally acquired, three-dimensional, T1-weighted brain images acquired with different acquisition parameters. We tested our networks using data from $\mathbf {2808}$ images (obtained in 18 subjects) under four different acceleration factors ($\mathbf {R=\lbrace 5,10,15,20\rbrace }$). Our enhanced reconstruction (Steps 1-3) produced higher-quality images: structural similarity and peak signal-to-noise ratio increased, and normalized root mean squared error decreased on average by $\mathbf {16.5\%}$, $\mathbf {7.0\%}$ and $\mathbf {21.1\%}$, respectively, compared to the non-enhanced reconstruction (Step 1 only) under the same network capacity as the enhanced reconstruction model. These differences were statistically significant ($\boldsymbol{p< 0.001}$, Wilcoxon signed-rank test). Further volumetric analysis performed on key brain regions (brain, white matter, gray matter and cortex) indicated that our enhanced images had better volume agreement with the fully sampled reference images compared to the non-enhanced images. Our enhanced images for $\mathbf {R=20}$ were comparable to the non-enhanced images for $\mathbf {R=10}$ demonstrating that our proposed method can use prior scan information to further accelerate MR examinations.

中文翻译：

---

通过利用先前特定于受试者的脑成像增强基于深度学习的磁共振图像重建：使用一组假定正常受试者的概念验证

深度学习模型已显示出重建欠采样、多通道磁共振 (MR) 图像采集的潜力。然而，最近提出的方法没有利用来自先前特定于主题的 MR 成像会话的信息。此类数据通常可通过图片存档和通信系统 (PACS) 轻松获得。我们提出了一种灵活的三步法，将这些先验信息整合到增强的基于深度学习的重建过程中。该方法包括步骤 1：初始重建；步骤2：将先前的扫描配准到初始重建；第三步：增强网络。训练和测试使用纵向获取的、三维的、T1 加权的大脑图像，这些图像通过不同的采集参数获得。我们使用来自的数据测试了我们的网络$\mathbf {2808}$ 四种不同加速因子下的图像（在 18 位受试者中获得）（$\mathbf {R=\lbrace 5,10,15,20\rbrace }$）。我们的增强重建（步骤 1-3）产生了更高质量的图像：结构相似性和峰值信噪比增加，归一化均方根误差平均减少了$\mathbf {16.5\%}$, $\mathbf {7.0\%}$ 和 $\mathbf {21.1\%}$，分别与在与增强重建模型相同的网络容量下的非增强重建（仅步骤 1）进行比较。这些差异具有统计学意义（$\boldsymbol{p< 0.001}$, Wilcoxon 符号秩检验）。对关键大脑区域（大脑、白质、灰质和皮层）进行的进一步体积分析表明，与非增强图像相比，我们的增强图像与完全采样的参考图像具有更好的体积一致性。我们的增强图像$\mathbf {R=20}$ 与非增强图像相比 $\mathbf {R=10}$ 证明我们提出的方法可以使用先前的扫描信息来进一步加速 MR 检查。