SpinDoctor-IVIM：一种用于体素内无序运动MRI的虚拟成像框架|文献速递-生成式模型与transformer在医学影像中的应用

Title

题目

SpinDoctor-IVIM: A virtual imaging framework for intravoxel incoherent motion MRI

SpinDoctor-IVIM：一种用于体素内无序运动MRI的虚拟成像框架

01

文献速递介绍

体素内无序运动（IVIM）成像是一种扩散磁共振成像（dMRI）技术，能够评估组织中的血液灌注。IVIM指的是水分子在MRI体素内的旋转位移过程，这一过程发生在测量时间内。水分子的旋转速度会呈现出一个分布，影响旋转速度的方向和/或幅度（Le Bihan et al., 1986）。这一分布源自血液在微血管网络中的流动，如图1.a所示，模拟了一个伪扩散过程。在扩散编码梯度存在的情况下，伪扩散过程会导致dMRI信号更快衰减，如图1.b所示（Jerome et al., 2021）。作为一种内源性对比度技术，IVIM成像提供了关于血流灌注（激发和读出在同一平面上进行）和组织微观结构（通过水分子在血管外空间中的扩散）的局部信息，这些优点使其适用于临床应用（Federau, 2017）。

Abatract

摘要

Intravoxel incoherent motion (IVIM) imaging is increasingly recognised as an important tool in clinical MRI,where tissue perfusion and diffusion information can aid disease diagnosis, monitoring of patient recovery,and treatment outcome assessment. Currently, the discovery of biomarkers based on IVIM imaging, similar toother medical imaging modalities, is dependent on long preclinical and clinical validation pathways to linkobservable markers derived from images with the underlying pathophysiological mechanisms. To speed up thisprocess, virtual IVIM imaging is proposed. This approach provides an efficient virtual imaging tool to design,evaluate, and optimise novel approaches for IVIM imaging. In this work, virtual IVIM imaging is developedthrough a new finite element solver, SpinDoctor-IVIM, which extends SpinDoctor, a diffusion MRI simulationtoolbox. SpinDoctor-IVIM simulates IVIM imaging signals by solving the generalised Bloch–Torrey partialdifferential equation. The input velocity to SpinDoctor-IVIM is computed using HemeLB, an established LatticeBoltzmann blood flow simulator. Contrary to previous approaches, SpinDoctor-IVIM accounts for volumetricmicrovasculature during blood flow simulations, incorporates diffusion phenomena in the intravascular space,and accounts for the permeability between the intravascular and extravascular spaces. The above-mentionedfeatures of the proposed framework are illustrated with simulations on a realistic microvasculature model.

体素内无序运动（IVIM）成像越来越被认为是临床MRI中的一个重要工具，它可以通过评估组织灌注和扩散信息来辅助疾病诊断、患者恢复监测以及治疗效果评估。目前，基于IVIM成像发现生物标志物的过程与其他医学影像技术类似，需要通过长期的临床前和临床验证路径，将从影像中得到的可观察标志物与潜在的病理生理机制联系起来。为了加速这一过程，提出了虚拟IVIM成像。这种方法提供了一种高效的虚拟成像工具，用于设计、评估和优化IVIM成像的新方法。

在本研究中，虚拟IVIM成像通过一个新的有限元求解器——SpinDoctor-IVIM得以实现，该求解器是扩散MRI模拟工具箱SpinDoctor的扩展。SpinDoctor-IVIM通过求解广义Bloch-Torrey偏微分方程来模拟IVIM成像信号。SpinDoctor-IVIM的输入速度是通过HemeLB计算的，HemeLB是一个成熟的格子玻尔兹曼血流模拟器。与以往的方法不同，SpinDoctor-IVIM在血流模拟中考虑了体积微血管，结合了血管内空间中的扩散现象，并考虑了血管内外空间之间的通透性。上述框架的特性通过对一个真实微血管模型的仿真进行了展示。

Method

方法

4.1. Finite element solution of generalised bloch–torrey partial differentialequationThis section presents a mathematical contribution to extend theoriginal version of SpinDoctor to the SpinDoctor-IVIM version. ThegBT equation, Eq. (15), is spatially discretised using the finite elementmethod (FEM), similar to the BT in Li et al. (2019),

4.1. 广义Bloch–Torrey偏微分方程的有限元解法 本节介绍了一个数学方法，用于将原始版本的SpinDoctor扩展到SpinDoctor-IVIM版本。广义Bloch-Torrey方程（gBT方程），公式(15)，采用有限元法（FEM）进行空间离散化，类似于Li等人（2019）中的BT方程，

Conclusion

结论

In conclusion, this paper introduces a significant advancementin the field of IVIM imaging through the development of a virtualimaging tool, SpinDoctor-IVIM. By leveraging an innovative finiteelement solver and integrating with HemeLB for blood flow simulations, this tool allows for the efficient design, evaluation, andoptimisation of novel approaches in IVIM imaging. Unlike previousmethods, SpinDoctor-IVIM uniquely considers the volumetric microvasculature during blood flow simulations, incorporates diffusion phenomena in the intravascular space, and represents the permeabilitybetween intravascular and extravascular spaces. The proposed virtualIVIM imaging approach offers a promising avenue to expedite the validation process of IVIM-related quantitative models by bridging the gapbetween observable markers in imaging and underlying pathophysiological mechanisms. Additionally, it can provide recommendations forIVIM acquisition parameters to the IVIM imaging community. Therefore, SpinDoctor-IVIM presents a valuable tool to propel advancementsin clinical MRI and enhance our understanding of tissue perfusion anddiffusion dynamics.

总之，本文通过开发虚拟成像工具 SpinDoctor-IVIM，介绍了 IVIM 成像领域的重要进展。该工具通过采用创新的有限元求解器，并与 HemeLB 血流模拟结合，能够高效地设计、评估和优化 IVIM 成像中的新方法。与以前的方法不同，SpinDoctor-IVIM 独特地考虑了血流模拟中的体积微血管，结合了血管内空间的扩散现象，并表示了血管内外空间之间的渗透性。所提出的虚拟 IVIM 成像方法为加速 IVIM 相关定量模型的验证过程提供了一个有前景的途径，弥合了成像中可观察标记与潜在病理生理机制之间的差距。此外，它还可以为 IVIM 成像社区提供有关 IVIM 获取参数的建议。因此，SpinDoctor-IVIM 是推动临床 MRI 进展并增强我们对组织灌注和扩散动力学理解的宝贵工具。

Figure

图

Fig. 1. IVIM model. (a) A schematic illustrating water molecules’ transitional movement in the intravascular space (depicted in red) and Brownian motion in theextravascular space (depicted in blue) within a tissue voxel. (b) A comparison of theobserved dMRI signal decays based on the Brownian movement of water molecules(diffusion) and transitional movement of water molecules (pseudo diffusion). Pseudodiffusion contributes significantly faster signal decays than diffusion and is thusobserved only at low b-values. Image reproduced with permission (Jerome et al., 2021)

Fig. 1. IVIM模型。 (a) 示意图展示了水分子在组织体素内的过渡运动（红色表示血管内空间）和布朗运动（蓝色表示血管外空间）。 (b) 比较了基于水分子的布朗运动（扩散）和水分子的过渡运动（伪扩散）所观测到的dMRI信号衰减。伪扩散导致比扩散更快速的信号衰减，因此仅在低b值下观察到。图像已获Jerome等（2021）许可转载。

Fig. 2. Schematic diagram of the proposed integrated tool for virtual IVIM imaging

图2. 提议的虚拟IVIM成像集成工具的示意图

Fig. 3. A schematic of the relative position of a lattice versus tetrahedron.

图.3.六方晶格与四面体相对位置的示意图。

Fig. 4. Mesh generation and registration. (a) Generating the tetrahedron mesh, to be used as the SpinDoctor IVIM input, from the input surface mesh of HemeLB; (b) Registrationbetween the tetrahedron mesh and the output lattice of HemeLB.

Fig. 4. 网格生成与配准。 (a) 从 HemeLB 输入的表面网格生成四面体网格，作为 SpinDoctor IVIM 的输入； (b) 四面体网格与 HemeLB 输出格点的配准。

Fig. 5. Sub-set of murine retinal vascular plexus (red) surrounded by extra-vascularspace (blue)

Fig. 5. 小鼠视网膜血管丛的子集（红色），被血管外空间（蓝色）包围。

Fig. 6. Excluding the effect of the outlet boundaries on the simulated IVIM imagingsignal by setting zero initial magnetisation at max{𝒗(𝒓, 𝑡)} × 𝑇𝑒 vicinity of outletboundaries. At the end of the simulation, the magnetisation in this region is dismissedin the computation of the final signal.

Fig. 6. 通过在出口边界的最大速度 v(r,t)\mathbf{v}(\mathbf{r}, t) × TeT_e 附近设置初始磁化强度为零，排除出口边界对模拟 IVIM 成像信号的影响。在模拟结束时，该区域的磁化强度在最终信号的计算中被排除。

Fig. 7. Effect of different ROIs, i.e., max{𝒗(𝒓, 𝑡)} × 𝑇𝑒 , and maximum velocitiesmax{𝒗(𝒓, 𝑡)} for each ROI on normalised IVIM imaging signal.

Fig. 7. 不同感兴趣区域（ROI）的影响，即 max{v(r,t)}×Te{\mathbf{v}(\mathbf{r}, t)} \times T_e，以及每个 ROI 的最大速度 max{v(r,t)}{\mathbf{v}(\mathbf{r}, t)} 对归一化 IVIM 成像信号的影响。

Fig. 8. Comparison between magnetisation resulting from SpinDoctor-IVIM and COMSOL for a cylinder with the radius of 5 μm and length of 100 μm: (a) Averaged magnitude ofmagnetisation, (b) Difference in the normalised magnitude of magnetisation.

Fig. 8. 通过 SpinDoctor-IVIM 和 COMSOL 对比磁化结果，对于半径为 5 μm、长度为 100 μm 的圆柱体：(a) 磁化的平均幅度，(b) 正常化磁化幅度的差异。

Fig. 9. Comparison between signal intensity resulting from SpinDoctor-IVIM and COMSOL for a cylinder with the radius of 5 μm and length of 100 μm: (a) Normalised signal, (b)Difference in the normalised signal. The subscripts ’s’ and ’c’ indicate the signals of SpinDoctor-IVIM and COMSOL, respectively

Fig. 9. 通过 SpinDoctor-IVIM 和 COMSOL 对半径为 5 μm、长度为 100 μm 的圆柱体得到的信号强度比较： (a) 归一化信号， (b) 归一化信号的差异。下标 's' 和 'c' 分别表示 SpinDoctor-IVIM 和 COMSOL 的信号。

Fig. 10. (a) The distributions of vessel segments in the ROI with an average of 60.28 μm and median of 43 μm; (b) The distribution of magnitudes of velocities at 𝑣𝑚𝑎𝑥 = 0.01 m s −1with an average of 450 μm s −1 and median of 96.7 μs−1; (c) The distribution of magnitudes of velocities at 𝑣𝑚𝑎𝑥 = 0.02 m s −1 with an average of 900 μm s −1 and median of 193.45 μm s −1 .

Fig. 10. (a) ROI 中血管段的分布，平均值为 60.28 μm，中位数为 43 μm； (b) 在最大速度 vmax=0.01v{\text{max}} = 0.01 m/s 下，速度幅度的分布，平均值为 450 μm/s，中位数为 96.7 μm/s； (c) 在最大速度 vmax=0.02v{\text{max}} = 0.02 m/s 下，速度幅度的分布，平均值为 900 μm/s，中位数为 193.45 μm/s。

Fig. 11. Effect of changing 𝐷𝑏 and 𝑃 on (a) IVIM imaging signal originated from both intravascular and extravascular spaces at the velocity of 0.01 m s −1 , (b) IVIM imaging signaloriginated from intravascular spaces at the velocity of 0.01 m s −1 , (c) IVIM imaging signal originated from both intravascular and extravascular spaces at the velocity of 0.02 m s −1 ,and (d) IVIM imaging signal originated from intravascular spaces at the velocity of 0.02 m s −1 .

Fig. 11. 改变 DbD_b 和 PP 对以下内容的影响： (a) 速度为 0.01 m/s 时，来自血管内外空间的 IVIM 成像信号； (b) 速度为 0.01 m/s 时，来自血管内空间的 IVIM 成像信号； (c) 速度为 0.02 m/s 时，来自血管内外空间的 IVIM 成像信号； (d) 速度为 0.02 m/s 时，来自血管内空间的 IVIM 成像信号。

Fig. 12. The estimated value of 𝐷∗ at the different 𝑣, 𝐷𝑏 and 𝑃 using the different models.

Fig. 12. 使用不同模型估算的 D∗D^* 值，在不同的 vv、DbD_b 和 PP 下。

Fig. 13. The estimated value of 𝐷 at the different 𝑣, 𝐷𝑏 and 𝑃 using the different models.

Fig. 13. 使用不同模型估算的 DD 值，在不同的 vv、DbD_b 和 PP 下。

Fig. 14. The estimated value of 𝑓 at the different 𝑣, 𝐷𝑏 and 𝑃 using the different models.

Fig. 14. 使用不同模型估算的 ff 值，在不同的 vv、DbD_b 和 PP 下。

Fig. 15. The estimated value of 𝑣𝑒 at the different 𝑣, 𝐷𝑏 and 𝑃 using the different models.

Fig. 15. 使用不同模型估算的 vev_e 值，在不同的 vv、DbD_b 和 PP 下。

Table

表

Table 1Computational cost of the simulation.

Table 1 计算模拟的计算成本.