Construction of a diffusion MRI brain template using Human Connectome Project database
Yung-Chin Hsu1 and Wen-Yih Isaac Tseng1

1Institute of Medical Device and Imaging, National Taiwan University College of Medicine, Taipei, Taiwan

Synopsis

The present study uses the diffusion data of the human connectome project (HCP) as well as a state-of-the-art registration strategy to develop a diffusion template in the ICBM-152 space. A total of 489 diffusion datasets were included in the template construction. The HCP diffusion template matches to the ICBM-152 space well, in both the deep and superficial white matter regions. In addition, this template is capable of revealing detailed diffusion-dependent structures, such as the cortical-depth dependence of general fractional anisotropy (GFA). The HCP diffusion template is potentially useful in segmenting fine fiber tracts and parceling thalamic nuclei or hippocampal subfields.

Introduction

The ICBM-152 space is one of the most popular coordinate systems in the neuroimaging community. Several diffusion templates have been developed in the ICBM-152 space, including diffusion tensor imaging (DTI) templates1-3, high angular resolution diffusion imaging (HARDI) templates4,5, and a diffusion spectrum imaging (DSI) template6. For each of these templates, diffusion data of a large amount of healthy subjects were registered altogether and averaged to produce the final template. With the advance in imaging technology and registration algorithm, new diffusion templates will emerge to reveal more detailed white matter (WM) structures than the conventional clinically-feasible diffusion data. In the present study, therefore, we used a set of high quality diffusion data from the human connectome project (HCP)7 to build a diffusion template in the ICBM-152 space. Specifically, we used the strategy suggested by Hsu et al.6 to build the HCP diffusion template, as this registration strategy has been demonstrated to achieve better alignment in anatomical features. In addition, the HCP diffusion template was saved in the raw diffusion-weighted image (DWI) format, allowing various possible post-processing approaches.

Materials and methods

The HCP database comprised more than 500 subjects, wherein the diffusion data of $$$N=489$$$ subjects were available. Each of the HCP diffusion data was acquired using a 3-shell q-space sampling scheme, 90 points per shell, where the b-values of the shells were 1000, 2000, and 3000 (s/mm2). The spatial resolution of the DWI was 1.25 mm3. The HCP diffusion data were preprocessed8 to correct for the eddy current-induced distortions, susceptibility-induced distortions, and motion-related artifacts, and then rigidly registered to the T1w images. With the preprocessed data, the procedures of building the HCP diffusion template were as follow.

(1) The individual T1w and T2w images were segmented9 through SPM12, resulting in the individual tissue probability maps (TPMs).

(2) The Shoot10 toolbox of SPM12 was employed to estimate the Karcher mean of the TPMs and the associated deformation maps, $$$Φ^j$$$, $$$j=1...N$$$.

(3) The DARTEL11 toolbox of SPM12 was used to normalize the mean TPM to the ICBM-152 template, resulting in the deformation map, $$$Φ^0$$$.

(4) Combining the results of steps (2) and (3), produced the deformation maps $$$Θ^j=Φ^j\circΦ^0$$$, $$$j=1...N$$$, which mapped the individual spaces the ICBM-152 space.

(5) The individual diffusion data were transformed to the ICBM-152 space according to $$$Θ^j$$$, $$$j=1...N$$$, where the q-space signals of each voxel were reoriented through the local rotation matrixes. For the image space signals, the 2nd order b-spline interpolation was used; for the q-space signals, the 4th order spherical harmonic interpolation was performed for each q-shell.

(6) The transformed diffusion data were then averaged to produce the final HCP diffusion template.

Results

The final HCP diffusion template consisted of 271 DWI at spatial resolution of 0.7 mm3, corresponding to the 3 q-shells in the q-space. Figure 1 displays the color-coded general fractional anisotropy (GFA) map on top of the ICBM-152 T1w image, where regions with GFA>0.04 were shown. In the deep WM regions, the orientations coincide with the known anatomy, and in the superficial WM regions, the WM contours spatially conform to the gyrus-sulcus patterns of the ICBM-152 template, suggesting that the HCP diffusion template matches well to the ICBM-152 space. Figure 2 shows the GFA map at the boundaries between the gray matter (GM) and WM. It reveals the cortical-depth dependence of the GFA values12. More specifically, the cerebral cortex can be divided into two layers, the one close to the GW-WM boundary presents lower GFA values, leading to dark bands in the cortical regions, such as those indicated by the red arrowheads. Figure 3 illustrates the U-shape tracts which connected the right precentral and caudal-middle-frontal gyri, demonstrating the capability of the HCP diffusion template to estimate the small fibers.

Discussion

The present study uses the HCP diffusion data to develop a diffusion template in the ICBM-152 space. Thanks to the high quality HCP diffusion data and the state-of-the-art registration strategy, the HCP diffusion template is capable of revealing detailed diffusion-dependent structures. This template can be used to build a more comprehensive tract atlas, particularly the tracts that are very difficult to be reconstructed, such as the U-shape tracts, the tracts passing through the brain stem, and the cerebellar tracts. Another potential application of the template is tract-based parcellations for, say, thalamic nuclei and hippocampi subfields.

Acknowledgements

No acknowledgement found.

References

1. Mori, S. et al. Stereotaxic white matter atlas based on diffusion tensor imaging in an ICBM template. Neuroimage 40, 570-582 (2008).

2. Peng, H. et al. Development of a human brain diffusion tensor template. Neuroimage 46, 967-980 (2009).

3. Zhang, S., Peng, H., Dawe, R. J. & Arfanakis, K. Enhanced ICBM diffusion tensor template of the human brain. Neuroimage 54, 974-984 (2011).

4. Yeh, F. C. & Tseng, W. Y. NTU-90: a high angular resolution brain atlas constructed by q-space diffeomorphic reconstruction. Neuroimage 58, 91-99 (2011).

5. Varentsova, A., Zhang, S. & Arfanakis, K. Development of a high angular resolution diffusion imaging human brain template. Neuroimage 91, 177-186 (2014).

6. Hsu, Y. C., Lo, Y. C., Chen, Y. J., Wedeen, V. J. & Isaac Tseng, W. Y. NTU-DSI-122: A diffusion spectrum imaging template with high anatomical matching to the ICBM-152 space. Hum Brain Mapp 36, 3528-3541 (2015).

7. Van Essen, D. C. et al. The WU-Minn Human Connectome Project: an overview. Neuroimage 80, 62-79 (2013).

8. Sotiropoulos, S. N. et al. Advances in diffusion MRI acquisition and processing in the Human Connectome Project. Neuroimage 80, 125-143 (2013).

9. Ashburner, J. & Friston, K. J. Unified segmentation. Neuroimage 26, 839-851 (2005).

10. Ashburner, J. & Friston, K. J. Diffeomorphic registration using geodesic shooting and Gauss-Newton optimisation. Neuroimage 55, 954-967 (2011).

11. Ashburner, J. A fast diffeomorphic image registration algorithm. Neuroimage 38, 95-113 (2007).

12. Truong, T. K., Guidon, A. & Song, A. W. Cortical depth dependence of the diffusion anisotropy in the human cortical gray matter in vivo. PloS one 9, e91424 (2014).

Figures

Figure 1. The GFA map of the HCP diffusion template overlaid with the ICBM-152 T1w image. Regions with GFA>0.04 are shown and color-coded according to the principal vectors.

Figure 2. Cortical-depth dependence of GFA values. This map shows that the cerebral cortex can be divided into two layers, the one close to the GW-WM boundary presents lower GFA values, leading to dark bands in the cortical regions, as indicated by the red arrowheads.

Figure 3. The U-shape tracts connecting the right precentral and caudal-middle-frontal gyri.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3063