Comparing Diffusion MRI with the Fiber Architecture and Tract Density of Gyral Blades
Kurt Schilling1, Vaibhav Janve1, Yurui Gao1, Iwona Stepniewska1, Bennett Landman1, and Adam Anderson1

1Vanderbilt University, Nashville, TN, United States


It has been reported that diffusion tractography has a tendency for streamlines to terminate preferentially on gyral crowns rather than on sulcal walls or fundi. Rather than anatomical reality, it has been suggested that this is a bias associated with tractography. To better understand this issue, we compare histology to diffusion MRI of the same specimen. We measure the trajectories and density of axons crossing the gray matter/white matter boundary and compare to diffusion tensor measures and deterministic tractography. The results of this study lead to a better understanding of gyral anatomy and potential limitations of fiber tractography.


The ability of diffusion MRI (dMRI) fiber tractography to non-invasively map the structural connectivity of the brain has proven a valuable neuroimaging tool. However, diffusion tractography has several potential limitations, which may prevent it from faithfully representing true axonal connections of the brain. Specifically, it has been shown that streamlines terminate preferentially on gyral crowns rather than on sulcal walls (1,2). These results, however, are believed not to be a true difference in anatomical connectivity, but rather, a confound of tractography (3,4). Thus, there is a need to better understand the true trajectories of axons in the gyral blades. Here, we use light microscopy to determine the ground truth fiber orientation distribution near the cortex and make quantitative comparisons to the distributions obtained with deterministic tractography of the same brain. This allows us to study: (A) the trajectories of axons near the white matter/gray matter (WM/GM) border; (B) how this varies with position along the gyrus; (C) how the density of axons entering the GM varies with position; and (D) whether axons terminate primarily on the crowns of gyri, or is this a bias of tractography.


dMRI: dMRI of an ex vivo squirrel monkey brain was acquired on a 9.4T Agilent scanner using a PGSE multi-shot spin-echo sequence (gradient directions=32, b≈1000s/mm2, voxel size=300μm isotropic). Tractography: MRTrix (5) and Diffusion Toolkit (6) were utilized for tractography, using default settings. MRTrix used the diffusion tensor to perform deterministic tracking (7) (step-size: 0.1*voxel size, minimum length:5*voxel size, FA-cutoff: .1, seed: whole brain). Diffusion Toolkit used the diffusion tensor and FACT for fiber propagation (same parameters). Histology: The brain was sectioned coronally (50um) and every 6th slice stained using Gallyas-silver stain. Photomicrographs of 27 slides were obtained on a Leica Brightfield microscope at a resolution of 1um/pixel. Image Registration/Processing: Individual slices were registered to dMRI data using the methods described in (8). 32 Gyral blades were selected for further processing, which included determination of WM/GM border and structure tensor (ST) analysis (9), resulting in an orientation estimate for every pixel in the image. For each gyrus, the crown was defined as the voxels on the WM/GM interface with greatest curvature, while two walls were determined where curvature was minimum. Fiber densities: The number of fibers entering/leaving the WM (at a distance of 300um into the cortex, equivalent to an MRI voxel) was automatically detected and counted to determine the histological fiber density. Corresponding tractography densities were determined at the same locations using the tract-density images generated by MRTrix and Trackvis. We then calculate the ratio of axons (or streamlines) leaving the gyral crown to those leaving at the sulcal walls.

Results and Discussion

Figure 1 shows the results of ST analysis with fibers color-coded based on orientation. Magnified views show fibers leaving the crown with little to no change in orientation from WM to GM (blue), as well as the typical sharp bend in fibers near the walls upon entering the cortex (red). Figure 2 displays the same slice, where the crown and walls are shown extending from 400um in WM, to 1000um into the cortex. ST orientations are shown in magenta (left), and registered diffusion tensor primary eigenvectors in green (middle). Also shown is the Diffusion Toolkit tract-density map. For this gyrus, there is a clear penchant of fibers to terminate on the crown. Figure 3 summarizes the results of all 32 gyral blades. The plot (left) shows the average fiber orientation relative to the normal to the WM/GM boundary. At the walls, fibers go from nearly orthogonal to the surface normal in WM (77°±11°), to within 28° from the normal only 600um into the cortex, while those of the crown stay nearly parallel (<20°) throughout. Finally, the ratio of fiber density at the crown to those at the walls is shown for histology compared to Diffusion Toolkit (middle) and MRTrix (right). The average histological ratio of crown:walls was 1.52±0.69, while those of tractography were higher, 2.51±1.9 and 2.1±1.8, respectively. A paired t-test resulted in p<0.05 for Diffusion toolkit, indicating a significant bias towards gyral crowns.


We find a preference for dMRI streamlines to terminate on gyral crowns, although this bias is algorithm-dependent and lower than in previously reported studies (1). Comparisons with histology are necessary to better understand axonal trajectories near the cortex and to improve the anatomical accuracy of tractography. Future studies will analyze how streamline orientations compare to histology at these locations, how specific anatomical features may account for this tractography bias, and the effects of other diffusion models or tracking parameters.


NIH 2R01NS058639-05/08


1. Nie et al. Cereb Cortex 2012;22(12):2831-9. 2. Chen et al. Cereb Cortex 2013;23(5):1208-17. 3. Van Essen et al. Ch 16. Diffusion MRI. 2014. p 337-358. 4. Kleinnijenhuis et al. Neuroimage 2015;109:378-87. 5. Tournier et al. International Journal of Imaging Systems and Technology 2012;22(1):53-66. 6. Wang et al. Proc. Intl. Soc. Mag. Reson. Med. (15) 2007. p 3720. 7. Basser et al. Magn Reson Med 2000;44(4):625-32. 8. Choe et al. Magn Reson Imaging 2011;29(5):683-92. 9. Budde et al. Neuroimage 2012;63(1):1-10.


Figure 1. Structure tensor analysis of representative myelin stained gyrus where orientation is displayed as directionally encoded color image. Magnified views show fibers of the gyral crown (blue) and sulcal wall (red). The ability to distinguish individual fibers is readily apparent.

Figure 2. Primary axon orientations derived from ST analysis (left, magenta) and the registered DTI data (middle, green) are displayed as lines. Gyral crown and two walls are also indicated, along with WM/GM boundary and the boundary normal (yellow sticks). A registered tract density image is shown at right.

Figure 3. Summary statistics of 32 gyral blades. The average angle from WM/GM boundary normal is shown for all sulcal walls and gyral crowns (left). The histological fiber density is compared to tract density images and results displayed as paired box plots for Diffusion Toolkit (middle) and MRTrix (right) tractography.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)