White matter multi-scale dataset: Diffusion weighted MRI and synchrotron x-ray scans of vervet monkey-, healthy mouse-, and cuprizone mouse brains

Sample details:

Monkey brain: All tissue samples originate from a 32-month-old female perfusion-fixated vervet (Chlorocebus aethiops) monkey brain, obtained from the Montreal Monkey Brain Bank. In 2017, biopsies were extracted, stained, epon-embedded, and scanned at both DESY and ESRF synchrotron facilities (denoted XNH in the file names – abbrev. of X-ray Nano Holotomography). Included in this repository: One sample from the Corpus Callosum (CC), and one from Centrum Semiovale (CS). The XNH scans are reoriented to appropriate anatomical orientations and stored as unsigned 8-bit volumes. The ESRF tomograms are “stitched” (by header information), but no further co-registration has been applied.

Mice brains: The tissue samples originate from C57BL/6 mice. Included in repository: One healthy brain used for MRI. One healthy control (HC) brain and one cuprizone-fed (CPZ) brain from which corpus callosum (CC) regions were extracted, stained, epon-embedded, and scanned at both DESY and ESRF synchrotron facilities in 2018. The XNH scans are co-registered (either through a provided parameter file or by header information) and stored as unsigned 8-bit volumes.

Data files details:

MRI results: The diffusion weighted MRI data was processed using Matlab. The following types of files are shared.

• Diffusion tensor FA volumes: The fractional anisotropy (FA) relationship between the eigenvalues of the diffusion tensor in each voxel. Stored as 16-bit floating-point volumes in range [0, 1] (filename: *DT_FA*)
• Diffusion tensor eigenvector volumes: A 4-D volume – having 3 channels representing the x,y,z component of the eigenvector in a floating point format for each voxel (filename: *DT_eigVectors*)
• MicroFA volumes: The fractional anisotropy (FA) relationship between the eigenvalues of the microFA model in each voxel. Stored as 16-bit floating-point volumes (filename: *microFA_FA*)
• Segmentation masks: Primarily a mask of the brain region (filename: *mask_brain*), but also segmentation masks that show which voxel corresponds to the XHN sample locations (filename: *mask_XNHloc*)

XNH structure tensor results: For the publication most of the files were processed using MatLab. For re-calculation of the structure tensors (ST), we would now recommend using this python implementation. The structure tensors were always converted to a diffusion like tensor, such that the largest eigenvector corresponds to the direction of fiber / anisotropic structures (see the main paper). Related to the structure tensor analysis, the following types of files are shared:

• Primary eigenvector volumes: A 4-D volume – having 3 channels representing the x,y,z component of the main eigenvector in a floating point format for each voxel (filename: *ST_dirXYZ*). 
• Primary direction colormap volumes: For easy visualization the dominant eigenvector direction is mapped onto a spherical colormap and stored as an unsigned 8-bit RGB volume (filename: *ST_dirRGB*). NB: Different spherical colormaps have been used for different samples (look-up the figures in the main paper).
• Voxel-wise FA volumes: The fractional anisotropy (FA) relationship between the eigenvalues in each voxel. Stored as 16-bit floating-point volumes in range [0, 1] (filename *ST_FA*).

XNH tractography results: Tractography was done with MRTrix, using the FACT algorithm. The output .tck files are not shared, as they are not easy to visualize in standard viewers. The streamlines were clustered (QuickBundles method) and stored as labelled integer masks (filename: *QBcentroid*) for easy visualization akin to a segmentation mask.