School of Artificial Intelligence, Beijing University of Post and Telecommunication, Beijing, 100876, China.
Institute of Biophysics, Chinese Academy of Sciences, Beijing, 100101, China.
Comput Biol Med. 2024 Jun;175:108508. doi: 10.1016/j.compbiomed.2024.108508. Epub 2024 Apr 23.
Multi-dimensional diffusion-relaxation correlation (DRC) magnetic resonance imaging (MRI) techniques have recently been developed to investigate tissue microstructures. Sub-voxel tissue heterogeneity is resolved from the local correlation distributions of relaxation time and molecular diffusivity. However, the implementation of these techniques considerably increases the total acquisition time, and simply reducing the scan time may be at the expense of detailed structural resolution. To overcome these limitations, an optimised framework was proposed for acquiring microstructural maps of the human brain on a clinically feasible timescale. First, the acquisition parameters of the multi-dimensional DRC MRI method were sparsely optimised using a genetic algorithm with a fitness function according to the spectral resolution of the correlation map, hardware requirements, and total scan time. Next, the acquired DRC MRI data were processed using a proposed numerical algorithm based on the dynamic inverse Laplace transform (ILT). Prior knowledge from one-dimensional data was then utilised in the iterative procedure to improve the spectral resolution. Finally, the proposed framework was validated using Monte Carlo simulations and experimental data acquired from healthy participants on an MRI scanner. The results demonstrated that the suggested approach is feasible for offering high-resolution DRC maps that correspond to distinct microstructures with a limited amount of optimised acquisition data from two-dimensional DRC sampling space. By significantly reducing scan time while retaining structural resolution, this approach may enable multi-dimensional DRC MRI to be more widely used for quantitative evaluation in biological and medical settings.
多维扩散-弛豫相关（DRC）磁共振成像（MRI）技术最近已被开发出来用于研究组织微结构。亚体素组织异质性是从弛豫时间和分子扩散率的局部相关分布中解析出来的。然而，这些技术的实现大大增加了总采集时间，简单地减少扫描时间可能会以牺牲详细的结构分辨率为代价。为了克服这些限制，提出了一种优化框架，以便在临床可行的时间范围内获取人脑的微结构图谱。首先，使用遗传算法根据相关图的谱分辨率、硬件要求和总扫描时间对多维 DRC MRI 方法的采集参数进行稀疏优化。接下来，使用基于动态逆拉普拉斯变换（ILT）的提出的数值算法处理采集的 DRC MRI 数据。然后，在迭代过程中利用一维数据的先验知识来提高谱分辨率。最后，使用蒙特卡罗模拟和从 MRI 扫描仪上的健康参与者获得的实验数据验证了所提出的框架。结果表明，所提出的方法对于提供高分辨率 DRC 图谱是可行的，这些图谱对应于不同的微结构，只需从二维 DRC 采样空间中采集优化的有限数量的数据。通过显著减少扫描时间而保持结构分辨率，这种方法可能使多维 DRC MRI 能够更广泛地用于生物和医学环境中的定量评估。
