A diffusion tensor brain template for rhesus macaques

Abstract

Diffusion tensor imaging (DTI) is a powerful and noninvasive imaging method for characterizing tissue microstructure and white matter organization in the brain. While it has been applied extensively in research studies of the human brain, DTI studies of non-human primates have been performed only recently. The growing application of DTI in rhesus monkey studies would significantly benefit from a standardized framework to compare findings across different studies. A very common strategy for image analysis is to spatially normalize (co-register) the individual scans to a representative template space. This paper presents the development of a DTI brain template, UWRMAC-DTI271, for adolescent Rhesus Macaque (Macaca mulatta) monkeys. The template was generated from 271 rhesus monkeys, collected as part of a unique brain imaging genetics study. It is the largest number of animals ever used to generate a computational brain template, which enables the generation of a template that has high image quality and accounts for variability in the species. The quality of the template is further ensured with the use of DTI-TK, a well-tested and high-performance DTI spatial normalization method in human studies. We demonstrated its efficacy in monkey studies for the first time by comparing it to other commonly used scalar-methods for DTI normalization. It is anticipated that this template will play an important role in facilitating cross-site voxelwise DTI analyses in Rhesus Macaques. Such analyses are crucial in investigating the role of white matter structure in brain function, development, and other psychopathological disorders for which there are well-validated non-human primate models.

Figure 1

The pipeline for generating the template: After the data are acquired, the DWI images are corrected for eddy current distortions and field in-homogeneities. Then, brain tissue is extracted from the images so further processing is done only on the relevant regions of the images. Tensors are estimated by non-linear optimization. An initial bootstrap template is then computed using the Log-Euclidean mean approach. Finally, the bootstrap template is iteratively improved using three layers: rigid registration, then affine registration and lastly by diffeomorphic registration.

Figure 2

Qualitative comparisons of the mean FA templates for each of the four spatial normalization strategies applied to DTI from 30 monkeys (DTI30). Spatial normalization using T1-weighted images co-registered to FA and B0 (FA-T1 and BO-T1, respectively) generated considerably more blurry FA templates. The full-tensor (DTI-TK) and FA-ANTS yielded similar and sharper FA templates; however, some of the WM structures are better delineated on the DTI-TK map, such as the separation between the internal and external capsule as indicated by the red arrow. The color bar indicates the FA intensity scale (unitless).

Figure 3

Comparisons of the overall performance of the four spatial normalization methods in terms of the empirical CDF of both normalized standard deviation of FA, σ̄_FA (left), and the dyadic coherence, κ (right), computed for the voxels within the WM. The corresponding histograms are shown as insets in each plot. In both cases, the CDFs and histograms for DTI-TK demonstrated greatest intersubject consistency. The σ̄_FA shows similarity of the performance (left shift) of FA-ANTS and DTI-TK while dyadic coherence shows the better performance (right shift) of DTI-TK in preserving white matter orientations. The improvement in performance is statistically significant (p < 1e-10) as per the two sample KS tests.

Figure 4

Qualitative comparisons of the normalized standard deviation of FA (left) and TR (right) maps in the white matter for each of the normalization methods. Both lower σ̄_FA and lower σ̄_TR are desirable and clearly better for both DTI-TK and FA-ANTS relative to the other approaches.

Figure 5

Qualitative comparisons of the dyadic coherence, κ, maps for the whole brain for each of the normalization methods. Highest κ was observed for DTI-TK, followed by FA-ANTS, which suggests that the tensor-based normalization best preserves the orientation information.

Figure 6

From left to right, empirical CDFs and histograms (insets) of cross correlations with respect to number of subjects: cross correlation of WM FA, cross correlation of WM TR and eigenvalue-eigenvector pair overlap with the template. DTI-TK performs better (indicated by right shift in the plots) with statistical significance of p < 1e-09 according to KS tests.

Figure 7

Empirical CDFs of tensor distances with respect to WM voxels. Left: Euclidean distance (ED) of the tensors to the template. Right: Deviatoric distance (DD) of the tensors to the template. The corresponding histograms are shown in inset plots. DTI-TK has lowest distances indicated by the left-shift of the curves. Euclidean distances (ED) are shown on the left and deviatoric distances (DD) are shown on the right for all the four registration methods. DTI-TK shows better performance as can be seen from the left shift. The two sample KS tests reveal that the shifts are statistically significant with p < 1e-10.

Figure 8

Population averaged, 3D fractional anisotropy (FA), trace, axial diffusivity and radial diffusivity maps of corresponding slices in sagittal (slice 125: position −1.0 mm), coronal (slice 106: position 11.5 mm) and axial (slice 156: position 20.0 mm) planes of the UWRMAC-DTI271 in Saleem and Logothetis (McLaren et al., 2009) atlas space. Each of these represents different average properties of the diffusion tensor and are useful in voxelwise analyses to localize individual differences. The units for trace, axial diffusivity and radial diffusivity are mm²/sec and indicate the rate of water diffusion, while the FA ranges between 0 and 1 and is unitless.

Figure 9

Axial slices (every 5^th slice from slice 123 through 183) of eigenvector color maps from UWRMAC-DTI271 in Saleem and Logothetis atlas space (McLaren et al., 2009). The last row shows corresponding (to the second row: slices from 143 through 163) axial slices of a T1-weighted template (McLaren et al., 2009). The top left frame shows the color mapping of the WM orientation: medial/lateral (right/left) is mapped to red, inferior/superior to blue and anterior/posterior (right/left) to green. The scale of the image is shown on bottom left. The positions of the slices relative to the origin in mm are shown on the bottom right of each slice.

Figure 10

White matter tracts reconstructed on the UWRMAC-DTI271 template by adapting strategies described in Catani et al. (2008) and Mori et al. (2002) for human white matter tracts. The tracts were obtained using the tensor deflection (TEND) tractography algorithm (Lazar et al., 2003) with a step size of 0.025 mm, stopping criteria of FA < 0.1–0.15 and a curvature threshold > 45°–60°. The tracts are overlaid on the slices of the Paxinos T1 atlas (Paxinos et al., 2009). Several additional 3D renderings of the tracts are available online as supplementary material.

Figure 11

Five corresponding slices (every second one from slice 28 through 36) for six different subjects (one subject per row) and the corresponding slices in the template (bottom most row). The normalized images show high anatomical consistency across subjects.