Cobre for machine learning

COBRE dataset, preprocessed and functional connectivity features extracted at 7 resolutions (7,12,20,36,64,122,197,325,444). Pearson correlation was used to compute functional connectivity between time series. The resolution are based on a partition using Cambridge dataset availlable at http://dx.doi.org/10.6084/m9.figshare.1285615 ### Content This work is a derivative from the COBRE sample found in the [International Neuroimaging Data-sharing Initiative (INDI)](http://fcon_1000.projects.nitrc.org/indi/retro/cobre.html), originally released under Creative Commons -- Attribution Non-Commercial. It includes preprocessed resting-state functional magnetic resonance images for 72 patients diagnosed with schizophrenia (58 males, age range = 18-65 yrs) and 74 healthy controls (51 males, age range = 18-65 yrs). The fMRI dataset for each subject are single nifti files (.nii.gz), featuring 150 EPI blood-oxygenation level dependent (BOLD) volumes were obtained in 5 mns (TR = 2 s, TE = 29 ms, FA = 75°, 32 slices, voxel size = 3x3x4 mm3 , matrix size = 64x64, FOV = mm2 ). * cobre_model_group.csv A comma-separated value file, with the sz (1: patient with schizophrenia, 0: control), age, sex, and FD (frame displacement, as defined by Power et al. 2012) variables. Each column codes for one variable, starting with the label, and each line has the label of the corresponding subject. * cobre_resolution_xx.mat: a .mat (octave/matlab) structure with two variables: data a NxF (N subjects x F features) and a subj_idx the subject id of each row. The features are a vetororized for of the connectome. ### Preprocessing The datasets were analysed using the NeuroImaging Analysis Kit (NIAK https://github.com/SIMEXP/niak) version 0.12.14, under CentOS version 6.3 with Octave(http://gnu.octave.org) version 3.8.1 and the Minc toolkit (http://www.bic.mni.mcgill.ca/ServicesSoftware/ServicesSoftwareMincToolKit) version 0.3.18.<br>Each fMRI dataset was corrected for inter-slice difference in acquisition time and the parameters of a rigid-body motion were estimated for each time frame. Rigid-body motion was estimated within as well as between runs, using the median volume of the first run as a target. The median volume of one selected fMRI run for each subject was coregistered with a T1 individual scan using Minctracc (Collins and Evans, 1998), which was itself non-linearly transformed to the Montreal Neurological Institute (MNI) template (Fonov et al., 2011) using the CIVET pipeline (Ad-Dabbagh et al., 2006). The MNI symmetric template was generated from the ICBM152 sample of 152 young adults, after 40 iterations of non-linear coregistration. The rigid-body<br>transform, fMRI-to-T1 transform and T1-to-stereotaxic transform were all combined, and the functional volumes were resampled in the MNI space at a 3 mm isotropic resolution. The “scrubbing” method of (Power et al., 2012), was used to remove the volumes with excessive motion (frame displacement greater than 0.5 mm). A minimum number of 60 unscrubbed volumes per run, corresponding to ~180 s of acquisition, was then required for further analysis. For this reason, 16 controls and 29 schizophrenia patients were rejected from the subsequent analyses. The following nuisance parameters were regressed out from the time series at each voxel: slow time drifts (basis of discrete cosines with a 0.01 Hz high-pass cut-off), average signals in conservative masks of the white matter and the lateral ventricles as well as the first principal components (95% energy) of the six rigid-body motion parameters and their squares (Giove et al., 2009). The fMRI volumes were finally spatially smoothed with a 6 mm isotropic Gaussian blurring kernel. ### References Ad-Dab’bagh, Y., Einarson, D., Lyttelton, O., Muehlboeck, J. S., Mok, K., Ivanov, O., Vincent, R. D., Lepage, C., Lerch, J., Fombonne, E., Evans, A. C., 2006. The CIVET Image-Processing Environment: A Fully Automated Comprehensive Pipeline for Anatomical Neuroimaging Research. In: Corbetta, M. (Ed.), Proceedings of the 12th Annual Meeting of the Human Brain Mapping Organization. Neuroimage, Florence, Italy. Bellec, P., Rosa-Neto, P., Lyttelton, O. C., Benali, H., Evans, A. C., Jul. 2010. Multi-level bootstrap analysis of stable clusters in resting-state fMRI. Neu-<br>roImage 51 (3), 1126–1139. URL http://dx.doi.org/10.1016/j.neuroimage.2010.02.082 Collins, D. L., Evans, A. C., 1997. Animal: validation and applications of nonlinear registration-based segmentation. International Journal of Pattern Recognition and Artificial Intelligence 11, 1271–1294. Fonov, V., Evans, A. C., Botteron, K., Almli, C. R., McKinstry, R. C., Collins, D. L., Jan. 2011.Unbiased average age-appropriate atlases for pediatric studies. NeuroImage 54 (1), 313–327.<br>URL http://dx.doi.org/10.1016/j.neuroimage.2010.07.033 Giove, F., Gili, T., Iacovella, V., Macaluso, E., Maraviglia, B., Oct. 2009. Images-based suppression of unwanted global signals in resting-state functional connectivity studies. Magnetic resonance imaging 27 (8), 1058–1064. URL http://dx.doi.org/10.1016/j.mri.2009.06.004 Power, J. D., Barnes, K. A., Snyder, A. Z., Schlaggar, B. L., Petersen, S. E., Feb. 2012. Spurious but systematic correlations in functional connectivity MRI networks arise from subject motion. NeuroImage 59 (3), 2142–2154. URL http://dx.doi.org/10.1016/j.neuroimage.2011.10.018

### 数据集概述
本数据集为经过预处理的COBRE数据集，在9个分辨率（7、12、20、36、64、122、197、325、444）下提取了功能连接（functional connectivity）特征。采用皮尔逊相关系数（Pearson correlation）计算时间序列间的功能连接。该分辨率方案基于剑桥数据集的分区，相关数据集可从http://dx.doi.org/10.6084/m9.figshare.1285615获取。

### 内容
本工作衍生自国际神经影像数据共享倡议（International Neuroimaging Data-sharing Initiative, INDI）开源的COBRE样本，原始授权协议为知识共享署名非商业性使用许可（Creative Commons -- Attribution Non-Commercial）。数据集包含72例精神分裂症患者（58名男性，年龄范围18~65岁）与74例健康对照者（51名男性，年龄范围18~65岁）的预处理静息态功能磁共振成像（resting-state functional magnetic resonance images）数据。每名受试者的fMRI数据集为单份NIfTI（.nii.gz）文件，包含150个血氧水平依赖（blood-oxygenation level dependent, BOLD）回波平面成像（EPI） volumes，采集时长约5分钟（重复时间TR=2s，回波时间TE=29ms，翻转角FA=75°，共32层切片，体素尺寸3×3×4 mm³，矩阵尺寸64×64，视场FOV未明确标注）。

#### 数据文件说明
1. `cobre_model_group.csv`：逗号分隔值文件，包含精神分裂症患病状态（sz：1为患者，0为健康对照）、年龄、性别以及帧位移（frame displacement，定义参考Power等2012年的研究）变量。每一列对应一个变量，首行为变量标签，每一行对应一名受试者的标签信息。
2. `cobre_resolution_xx.mat`：Matlab/Octave格式的结构文件，包含两个变量：`data`为N×F矩阵（N为受试者数量，F为特征数量），`subj_idx`为每行对应受试者的ID。特征为连接组的向量化结果。

### 预处理流程
本数据集采用神经影像分析工具包（NeuroImaging Analysis Kit, NIAK）0.12.14版本进行分析，运行环境为CentOS 6.3，搭配Octave 3.8.1版本与Minc工具包0.3.18版本。
1. 对每份fMRI数据集进行切片采集时间差校正，并估计每一时间帧的刚体运动（rigid-body motion）参数。采用首次运行的中位数体积作为基准，分别估计运行内与运行间的刚体运动。
2. 将每名受试者选定的fMRI运行的中位数体积，通过Mincracc工具（Collins与Evans, 1998）与个体T1扫描图像配准；随后通过CIVET流水线（Ad-Dabbagh等, 2006）将T1图像非线性配准至蒙特利尔神经研究所（Montreal Neurological Institute, MNI）模板（Fonov等, 2011），该MNI对称模板由152名青年成年人的ICBM152样本经过40次非线性配准迭代生成。
3. 合并刚体变换、fMRI到T1的变换以及T1到立体定向空间的变换，将功能影像重采样至3mm各向同性分辨率的MNI空间。
4. 采用Power等（2012）提出的剔除（scrubbing）方法，移除帧位移超过0.5mm的影像 volumes。要求每份运行至少保留60个未被剔除的 volumes（对应约180秒采集时长），因此最终有16名健康对照与29名精神分裂症患者被排除在后续分析之外。
5. 从每个体素的时间序列中回归去除以下混杂参数：缓慢时间漂移（采用截止频率0.01Hz的离散余弦基）、白质（white matter）与侧脑室（lateral ventricles）保守掩膜内的平均信号，以及6个刚体运动参数及其平方项的前95%能量主成分（principal components, PC）（Giove等, 2009）。
6. 最后使用6mm各向同性高斯模糊核（Gaussian blurring kernel）对fMRI影像进行空间平滑。

### 参考文献
1. Ad-Dab’bagh Y, Einarson D, Lyttelton O, Muehlboeck JS, Mok K, Ivanov O, Vincent RD, Lepage C, Lerch J, Fombonne E, Evans AC. CIVET图像处理环境：面向解剖神经影像研究的全自动综合流水线[C]//Corbetta M. 第12届人类脑图谱组织年会论文集. 神经影像学, 意大利佛罗伦萨, 2006.
2. Bellec P, Rosa-Neto P, Lyttelton OC, Benali H, Evans AC. 静息态fMRI稳定簇的多水平bootstrap分析[J]. NeuroImage, 2010, 51(3): 1126-1139. URL: http://dx.doi.org/10.1016/j.neuroimage.2010.02.082
3. Collins DL, Evans AC. 基于非线性配准的分割方法：验证与应用[J]. 国际模式识别与人工智能期刊, 1997, 11: 1271-1294.
4. Fonov V, Evans AC, Botteron K, Almli CR, McKinstry RC, Collins DL. 适用于儿科研究的无偏平均年龄适配图谱[J]. NeuroImage, 2011, 54(1): 313-327. URL: http://dx.doi.org/10.1016/j.neuroimage.2010.07.033
5. Giove F, Gili T, Iacovella V, Macaluso E, Maraviglia B. 静息态功能连接研究中无关全局信号的图像抑制方法[J]. Magnetic Resonance Imaging, 2009, 27(8): 1058-1064. URL: http://dx.doi.org/10.1016/j.mri.2009.06.004
6. Power JD, Barnes KA, Snyder AZ, Schlaggar BL, Petersen SE. 受试者运动导致功能连接MRI网络中出现虚假但系统性的相关性[J]. NeuroImage, 2012, 59(3): 2142-2154. URL: http://dx.doi.org/10.1016/j.neuroimage.2011.10.018

搜集汇总

背景与挑战

背景概述

该数据集是基于COBRE样本的预处理功能连接特征集合，包含72名精神分裂症患者和74名健康对照者的静息态fMRI数据，并提取了7个不同分辨率的功能连接特征。数据集提供了用于机器学习的特征矩阵和受试者元数据，特别适用于精神分裂症相关神经影像学研究和生物标志物分析。