Sample comparison.

A) Rendering of the vervet monkey brain structural MRI close to the mid-sagittal plane. B) Large field-of-view (FOV) scanning of the biopsy, DESY. The small orange box near the red arrow in A) indicates the location and size of the FOV within the MRI data. C) The stack of 4 small FOVs scans obtained at ESRF. The cylinder in B) indicates the relative size of the FOV within the DESY scan. D) Relative size comparison of one MRI voxel, the biopsy sample prior to staining and fixation, and various synchrotron FOVs. For cylindrical FOVs, the first number indicates the diameter, and the second number is the height. Sample orientations are related to the whole brain in A): R: Right, L: Left, I: Inferior, S: superior, A: Anterior, P: posterior.

dMRI-based orientations and tensor shapes.

A) The diffusion tensor model showing principal directions in a sagittal slice of the monkey corpus callosum (CC). The voxel corresponding to the biopsy sampling location (the synchrotron FOVs) is outlined in a white box. B) The fibre orientation distribution (FOD) of the CSD model in the selected diffusion MRI voxel is represented as a spherical polar histogram. C) The corresponding glyph representations of the diffusion tensor and CSD.

Structure tensor shape for corpus callosum (CC) sample.

A-B) 3D renderings from respectively the DESY and ESRF CC biopsy sample with structure tensor main direction colour coding (in accordance with the colour sphere). The white markings illustrate bands where axons are oriented at an angle compared to their surroundings, indicating a laminar organisation. The red cylinder on (A) shows the scale difference between the DESY and ESRF data. The position of the sample within the CC is marked on the MR image in Figures 1A and 2A. C-D) Selected coronal slice overlaid with a regularly spaced subset of structure tensor glyphs, coloured according to their predominant direction. E-F) Spherical polar histograms of the DESY structure tensor main directions (FOD) and the corresponding glyph. G) Kernel density estimates of structure tensor FA values. H-I) Spherical histogram of the ESRF structure tensor main directions (FOD) and the corresponding glyph. NB: We exclude contributions from the voxel of blood vessels in FA histograms and FODs.

Results in complex monkey deep white matter (WM) region.

A) Anatomy and CSD glyphs seen in the volume surrounding the local neighbourhood of the sampled biopsy location (white circle). B and C) Results in the matching MRI voxel, showing FOD (CSD) and the corresponding glyph. D) The directional colour map used throughout the figure. E) Example of rendering from an ESRF volume with colouring corresponding to the structure tensor main direction. Marking with dashed lines indicates clear fasciculi. F and G) Structure tensor directional statistics from the stacked FOV, showing the structure tensor FOD and corresponding glyph. H) Structure tensor shape statistics from the stacked FOV, showing the kernel density estimate of FA values. The red curve is a copy of the FA distribution from the ESRF CC region (Fig. 3G).

Structure tensor-based tractography for the monkey samples.

A) Streamlines in the CC sample from DESY. Purple streamlines have a strong upwards directional component, unlike the cyan streamlines (assuming that streamlines travel from right to left). B) The statistics of streamlines quantification using tortuosity index and maximum deviation. C) A selection of streamlines in the CC within a single ESRF scan. Cyan streamlines have R-L as the strongest directional component, whereas red streamlines do not. The orange structure is a segmented blood vessel. D) Streamlines in the complex centrum semiovale region (CS) within a single ESRF scan. Streamlines are coloured according to their local main direction. Orange structures represent segmented blood vessels.

Overview of the mouse datasets.

A) Location of the sample within the mouse brain and a coronal slice from the DESY dataset, with indications of anatomic regions: cc = corpus callosum, cg = cingulum, ctx = cortex, and ccp = cortical projections. The blue box indicates the size of a DWI voxel. B) The co-registered slice from the ESRF volume (position indicated by the red frame). Labels indicating blood vessels (bv) and cells (c). C) Coronal slice from the DESY dataset of a CPZ-treated mouse, with indication of anatomic regions: cc = corpus callosum, cg = cingulum, alv = alveus, hps = hippocampal striatum, hpp = hippocampal pyramidal layer, ccp = cortical projections. The dashed purple line shows a demyelinated region of the CC. D) The co-registered slice from the ESRF volume (indicated by the red frame). Labels indicate blood vessels (bv), cells (c), and myelin ‘debris’/macrophages (m). E) 3D rendering of a local segmentation in the corresponding region in D). The blood vessel is coloured in red, cells in blue, and ‘myelin debris’ in green. This segmentation was done manually using ITK-SNAP (RRID: SCR_002010).

Data samples included in this study.

All voxel sizes are isotropic. The FOVs of ID-2 and ID-3 are given as four stitched scans (marked with *). Individual FOVs were 0.21 x 0.21 x 0.21 mm.

Mouse diffusion tensor and structure tensor results

A) and B): The directional components of the diffusion tensor and the structure tensor in a coronal slice from a healthy mouse brain. The white frame in (A) indicates the approximate size and location of the DESY FOV, i.e., the black frame in (B). C) and D) The main structure tensor direction from the DESY data overlaid on a slice from a healthy mouse brain (C) and a CPZ-treated mouse (D). E) Structure tensor FODs from a healthy mouse along with corresponding glyphs. The glyph colouring indicates whether the FOD contribution is from the CC (red/blue) or cingulum (yellow/cyan). F) structure tensor FODs from the CPZ-treated mouse CC (DESY), split into contributions from a normal appearing region and a demyelinated area. G) and H) structure tensor FA distributions from various regions of healthy and lesioned mouse samples.

Structure tensor parameter values used for the samples in this study (see Table 1 for sample ID). The kernel size represents the width of the ρ-kernel converted to physical distance in accordance with the voxel size of the specific dataset.

Mouse tractography results.

A) Tractography visualisation from a healthy mouse brain sample. Red and Blue: Streamlines in CC. Yellow and Cyan: Streamlines in cingulum. Pink streamlines are extrapolations of CC streamlines that represent plausible cortical projections through the cingulum. B) Tractography visualisation from DESY data of a CPZ-treated mouse brain. C) and E) Tractography streamline centroid statistics from the healthy mouse brain synchrotron volumes. The curves are kernel density estimates of the tortuosity and maximum deviation, respectively. D) and F) Statistics for the streamlines in the corpus callosum of the DESY healthy and CPZ mouse brain samples.

HCP dataset diffusion tensor modelling.

The RGB colour-coded first, second, and third eigenvectors (ordered by decreasing eigenvalues) for the four subjects in both sagittal and coronal crop-out views.

Example of applying the scale-space structure tensor. (Top row)

A region of a slice from the ESRF monkey CC sample, where the FA is estimated from 8 scales (see Table 2), here showing only the scales 2, 6, and 8. The red transparent overlay shows the thresholded FA response (FA > 0.7). Notice how large, medium, and small cross-sectional axons respond differently at the different scales (represented by the blue, light blue, and orange circle, respectively). (Bottom row, left): The final thresholded dominant FA response (dFA > 0.7), shows high FA values for all axons regardless of the diameter. (Bottom row, mid and right): The same subset of voxels coloured according to the dominant scale index for the same 2D slice region and in a 3D rendering respectively. Notice how the low scales (large kernels) are selected for large axons, and similarly high scales (small kernels) selected for the small axons.

ST-derived statistics from the individual stacked FOVs in the monkey ESRF CC and CS sample. (Rows 1-4)

Individual FODs showing minor variation as the FOV placement changes. Across the CC, the most notable change is the anisotropic shape of the peak. Across the CS, the prominence of the L-R directed peak (near the red points) represents the largest variation. (Row 5): FA distributions of the four individual FOVs plotted together, showing almost no difference in the statistics.