Dataset: In-vivo characterization of magnetic inclusions in the subcortex from non-exponential transverse relaxation decay

This repository includes the data used to compile the results presented in the scientific publication: "In-vivo characterization of magnetic inclusions in the subcortex from non-exponential transverse relaxation decay". Rita Oliveira, Antoine LuttiLaboratory for Research in Neuroimaging (LREN)Department of Clinical Neuroscience, Lausanne University Hospital and University of LausanneMont-Paisible 16, CH-1011 Lausanne, SwitzerlandClassically, the MRI transverse relaxation decay is analyzed by fitting the signal decay over echo time voxel-wise with a monoexponential function (Exp), for which a decay rate R2∗ is estimated. However, the presence of magnetic material within the tissue, such as iron-loaded cells, myelin, or blood vessels, introduces variations in the magnetic field, which can modify the exponential behaviour of the decay (1,2). In such inhomogeneous magnetic fields, the theory predicts a transient regime starting with a Gaussian behaviour at short echo times and approaching a monoexponential relaxation at long echo times (1,3–6).We highlight three different analytical descriptions of the signal decay that account for the transient regime of the transverse relaxation decay: i) the Anderson and Weiss, 1953 model (AW); ii) the Jensen and Chandra, 2000 model/Sukstanskii and Yablonskiy, 2003 model (SY; in the article is called JC); iii) and following a Padé approximation (Padé) of the transition from Gaussian to exponential decay.This repository includes transverse relaxation decay data that enables the observation of the non-exponential MRI transverse relaxation. The data was acquired from 5 healthy volunteers at 3T. AW, SY, Padé, and Exp are the different methods that we used to fit the data with. Here we focus on the analysis of subcortical brain regions: Substantia Nigra, Pallidum, Putamen, Caudate, and Thalamus. Data DescriptionThe necessary files to compile the results presented in the scientific publication can be found in the ‘multiecho’ folder. There are three different folders corresponding to three repetitions of the acquisition (‘rep1’ to ‘rep3’). The data consists of:• resc_den_ subject_name_N.nii: magnitude image file corresponding to echo N. These files were previously denoised and rescaled (resc_den). The description field of the header of the images contains the corresponding TE at which the image was acquired, which will be needed in the fitting routine. Since we focus on the analysis of subcortical brain regions (Substantia Nigra, Pallidum, Putamen, Caudate, and Thalamus), the multi-echo data is masked within this region.• nf: value of the noise floor level. Corresponds to the noncentrality parameter of a Rician distribution fitted to the background signal. In the ‘anat’ folder the user has access to:• MT: Magnetization Transfer map (MTsat) that serves as a reference anatomical image.• ROI folder: contains masks of each of the 5 regions of interest analyzed in the scientific paper: Substantia Nigra, Pallidum, Putamen, Caudate, and Thalamus. The ‘modelfits’ folder contains pre-computed results for each subject analyzed. If the user uses the analysis code that comes along with this dataset (https://github.com/LREN-physics/TransverseRelaxation), this folder will be overwritten with the new results. For each method (‘AW’, ‘SY’, ‘Pade’, ‘Exp’) there is a folder containing the corresponding resulting maps. These maps are:• R2s.nii: map of R2,micro∗ [ms-1] for ‘AW’, ‘SY’, and ‘Pade’ options. Map of R2∗ [ms-1] for ‘Exp’ fit.• OmegaSq.nii: map of 〈\(\Omega^2\) [rad2 ms-2]. Not available for ‘Exp’ fit.• TE0signal.nii: map of the initial signal amplitude S0.• T2mol.nii: map of the inverse of effective transverse relaxation rate resulting from processes on the nanoscale [ms]• AIC.nii: map of Akaike information criterion regarding the fitting procedure.• MSE.nii: maps of the mean square error of the fitting procedure.• DataMatrix.mat: matrix containing the data used for the fitting procedure.• VoxelIndices.mat: vector containing the indices of the voxels corresponding to the analyzed data, which is restricted to the subcortical regions.• Params.mat: structure containing the parameters used for the analysis.Inside ‘modelfits’ there are also two folders corresponding to two different regimes that can describe the transverse relaxation decay: static dephasing regime (‘SDR’) or diffusion narrowing regime (‘DNR’). Under the assumption of SDR, we computed:• ki_ppm.nii: maps of 𝛥𝜒, which is the difference in susceptibility of the magnetic inclusions to the surrounding tissue [addimentional, in ppm and in SI units]• zeta.nii: maps of 𝜁, which is the volume fraction of the magnetic inclusions [addimentional]Under the assumption of DNR, we computed:• alpha.nii: 𝛼=𝜏〈\(\sqrt{\Omega^2}\)〉 [addimentional]• tau_ms.nii: maps of 𝜏, which is the time scale for water molecules to diffuse away from magnetic inclusions [ms]Please refer to the corresponding article for a complete description of the methods and corresponding estimated parameters.本仓库包含用于撰写下述科学论文的实验数据：《基于非指数横向弛豫衰减的皮层下脑区磁性内含物体内表征》。 丽塔·奥利维拉（Rita Oliveira）、安托万·吕蒂（Antoine Lutti） 神经成像研究实验室（Laboratory for Research in Neuroimaging, LREN） 临床神经科学系，洛桑大学医院与洛桑大学 瑞士洛桑市Mont-Paisible 16号，邮编CH-1011 传统上，磁共振成像（MRI）横向弛豫衰减的分析通常采用单指数函数（monoexponential function，简称Exp）对每个体素的回波时间信号衰减进行拟合，由此估算弛豫速率R2∗。然而，组织内存在的磁性物质（如载铁细胞、髓鞘或血管）会引发磁场不均匀性，进而改变弛豫衰减的指数特性（1,2）。在这类不均匀磁场中，理论预测会存在一段过渡态：在短回波时间下表现为高斯型衰减，而在长回波时间下逐渐趋近于单指数弛豫（1,3–6）。 本研究提出三种可描述横向弛豫衰减过渡态的信号解析模型：① 1953年Anderson与Weiss提出的模型（Anderson and Weiss, 1953，简称AW）；② 2000年Jensen和Chandra/2003年Sukstanskii和Yablonskiy提出的模型（Jensen and Chandra, 2000/Sukstanskii and Yablonskiy, 2003，简称SY，本文中记作JC）；③ 基于高斯型到指数型衰减过渡的帕德近似（Padé approximation，简称Padé）模型。 本仓库包含可用于观测非指数型磁共振成像横向弛豫衰减的实验数据，该数据采集自5名健康志愿者的3特斯拉（3T）磁共振扫描结果。本研究采用AW、SY、Padé与Exp四种方法对数据进行拟合，分析重点聚焦于皮层下脑区：黑质（Substantia Nigra）、苍白球（Pallidum）、壳核（Putamen）、尾状核（Caudate）以及丘脑（Thalamus）。 ### 数据说明 用于复现本文研究结果的必要文件均存放于`multiecho`文件夹中，该文件夹下包含三次采集重复的对应子文件夹（`rep1`至`rep3`）。本数据集包含以下内容： • `resc_den_<subject_name>_<N>.nii`：对应第N个回波的幅值图像文件。此类文件已完成降噪与重缩放处理（resc_den），图像头文件的描述字段中存储了该图像的采集回波时间（TE），该信息将用于后续拟合流程。由于本研究聚焦皮层下脑区分析，多回波数据已完成该区域的掩膜处理。 • `nf`：噪声基底水平值，对应拟合背景信号的莱斯分布（Rician distribution）非中心参数。 在`anat`文件夹中，用户可获取以下文件： • `MT`：磁化传递图（Magnetization Transfer map，MTsat），用作解剖参考图像。 • `ROI`文件夹：包含本文分析的5个目标脑区的掩膜文件，分别为黑质（Substantia Nigra）、苍白球（Pallidum）、壳核（Putamen）、尾状核（Caudate）以及丘脑（Thalamus）。 `modelfits`文件夹包含所有分析受试者的预计算拟合结果。若用户使用本数据集附带的分析代码（https://github.com/LREN-physics/TransverseRelaxation），该文件夹内的内容将被新的计算结果覆盖。针对每种拟合方法（`AW`、`SY`、`Pade`、`Exp`），均设有独立子文件夹存储对应结果图，具体包括： • `R2s.nii`：针对`AW`、`SY`与`Pade`方法，该文件存储微观横向弛豫速率R2,micro∗ [ms⁻¹]分布图；针对`Exp`方法，则存储R2∗ [ms⁻¹]分布图。 • `OmegaSq.nii`：⟨Ω²⟩ [rad² ms⁻²]分布图，`Exp`方法无对应结果。 • `TE0signal.nii`：初始信号振幅S0分布图。 • `T2mol.nii`：纳米尺度过程对应的有效横向弛豫速率倒数分布图，单位为ms。 • `AIC.nii`：拟合流程的赤池信息准则（Akaike information criterion）分布图。 • `MSE.nii`：拟合流程的均方误差（mean square error）分布图。 • `DataMatrix.mat`：存储拟合流程所用数据的矩阵文件。 • `VoxelIndices.mat`：存储分析数据对应体素索引的向量文件，仅包含皮层下脑区的体素。 • `Params.mat`：存储本次分析所用参数的结构体文件。 `modelfits`文件夹内还包含两个子文件夹，分别对应描述横向弛豫衰减的两种不同机制：静态失相机制（static dephasing regime，简称SDR）与扩散窄化机制（diffusion narrowing regime，简称DNR）。 基于静态失相机制假设，本研究计算得到以下结果： • `ki_ppm.nii`：磁化率差Δ𝜒分布图，即磁性内含物与周围组织的磁化率差值，单位为ppm与国际单位制（SI），无量纲。 • `zeta.nii`：磁性内含物体积分数𝜁分布图，无量纲。 基于扩散窄化机制假设，本研究计算得到以下结果： • `alpha.nii`：参数𝛼=𝜏⟨√Ω²⟩分布图，无量纲。 • `tau_ms.nii`：水分子从磁性内含物扩散离开的特征时间𝜏分布图，单位为ms。 有关方法与估算参数的完整说明，请参阅本文对应的学术论文。