A paired dataset of multi-modal MRI at 3 Tesla and 7 Tesla with manual hippocampal subfield segmentations on 7T T2-weighted images

We disseminate a dataset comprising paired 3T and 7T MRI scans from 20 healthy volunteers, with manual hippocampal subfield annotations on 7T T2-weighted images. This dataset is designed to support the development and evaluation of both 3T-to-7T MR image synthesis models and automated hippocampal segmentation algorithms on 3T images. We assessed the image quality using MRIQC. The dataset is freely accessible on IEEE DataPort, a data repository created by IEEE and can be found at the following URL: https://ieeexplore.ieee.org/document/10218394/algorithms?tabFilter=dataset. The shared dataset comprises four principal directories.The first directory contains raw MRI data in <i>.ima</i> format within <i>rawdata_DICOM</i>. Additionally, the acquired MRI scans were converted from DICOM to the Neuroimaging Informatics Technology Initiative (NIfTI) format and organized in accordance with the Brain Imaging Data Structure (BIDS) format by employing the BIDScoin Python application (version 4.3.0) and stored in <i>rawdata_BIDS</i> directory.The third directory pertains to hippocampal subfield segmentation and includes two subdirectories: '<i>hippo_subfield\7T_T2w_0.7_for_subfield_delineation</i>', featuring 7T T2w MRI data downsampled to a 0.7 mm slice thickness through B-spline interpolation, post Gaussian smoothing denoising and N4 bias field correction using Advanced Normalization Tools (ANTs); and <i>'hippo_subfield\hippo_label</i>', which contains the manual segmentation labels for hippocampal subregions for each subject. The fourth directory,<i> \MRIQC</i>, designated for the results of quality control assessments. For each participant, the<i> \MRIQC </i>directories contain <i>\anat</i> and <i>\func</i> subdirectories, which hold image quality metric reports for T1w, T2w, and resting-state functional scans. These quality metrics, available in both <i>.html</i> and <i>.json</i> formats, aid in evaluating data quality and provide estimates of motion, signal-to-noise ratios, and intensity non-uniformities, supplemented with visual reports.It is noteworthy that, due to detectable head motion during the original scans, the 3T T2w images for two participants were subject to rescanning. Subsequently, only the datasets from these supplementary sessions have been preserved within the <i>rawdata\BIDS</i> directory for further quality evaluation. Additionally, Diffusion Weighted Imaging (DWI) sequences included in the <i>rawdata\DICOM</i> directory for 3T MRI were not matched with 7T MRI sequences and, thus, are excluded from the BIDS-formatted shared dataset.<br>

本数据集公开了20名健康志愿者的配对3T与7T磁共振成像（MRI）扫描数据，并对7T T2加权图像完成了海马亚区的人工标注。本数据集旨在支撑3T到7T磁共振图像合成模型，以及3T图像上的海马亚区自动分割算法的研发与性能评估。本研究采用MRIQC工具对图像质量进行了评估。本数据集可通过IEEE创建的数据存储库IEEE DataPort免费获取，访问地址为：https://ieeexplore.ieee.org/document/10218394/algorithms?tabFilter=dataset。本次共享数据集包含四个主要目录：第一个目录`rawdata_DICOM`内存储了格式为.ima的原始MRI数据。此外，本研究使用Python工具BIDScoin（版本4.3.0）将采集的MRI扫描数据从DICOM格式转换为神经影像信息技术倡议（Neuroimaging Informatics Technology Initiative, NIfTI）格式，并按照脑成像数据结构（Brain Imaging Data Structure, BIDS）规范进行整理，最终存储于`rawdata_BIDS`目录中。第三个目录与海马亚区分割相关，包含两个子目录：`hippo_subfield7T_T2w_0.7_for_subfield_delineation`与`hippo_subfieldhippo_label`。前者存储了经B样条插值下采样至0.7mm层厚，并通过高级归一化工具（Advanced Normalization Tools, ANTs）完成高斯平滑去噪与N4偏置场校正的7T T2加权MRI数据；后者则存储了所有受试者的海马亚区人工分割标注结果。第四个目录为`MRIQC`，用于存储质量控制评估结果。针对每位受试者，`MRIQC`目录下包含`anat`与`func`两个子目录，分别存储T1加权、T2加权以及静息态功能扫描的图像质量指标报告。这些质量指标以.html和.json两种格式提供，可辅助数据质量评估，涵盖了头动、信噪比以及强度非均匀性的估算结果，并附带可视化报告。值得注意的是，由于原始扫描过程中存在可检测到的头动，两名受试者的3T T2加权图像被重新扫描。因此，仅将这两次补充扫描得到的数据保留在`rawdataBIDS`目录中，用于后续质量评估。此外，`rawdataDICOM`目录中存储的3T MRI弥散加权成像（Diffusion Weighted Imaging, DWI）序列未匹配对应的7T MRI序列，因此未被纳入BIDS格式的共享数据集中。