Spherical Deconvolution of High-Resolution 7T Whole-Head Diffusion Magnetic Resonance Images shows reduced radial anisotropic diffusion in human primary somatosensory cortex
Ralf L├╝tzkendorf1, Robin M. Heidemann2, Thorsten Feiweier2, Michael Luchtmann3, Sebastian Baecke1, Joern Kaufmann4, Joerg Stadler5, Eike Budinger5, and Johannes Bernarding1

1Biometry and Medical Informatics, University of Magdeburg, Magdeburg, Germany, 2Siemens Healthcare GmbH, Erlangen, Germany, 3Department of Neurosurgery, University of Magdeburg, Magdeburg, Germany, 4Department of Neurology, University of Magdeburg, Magdeburg, Germany, 5Leibniz Institute for Neurobiology, Magdeburg, Germany


Diffusion anisotropy in cortical gray matter (GM) and adjacent white matter (WM) provides microanatomic information about the course of the neuronal structures within GM and when connecting to other brain regions. However, interwoven neuronal fiber orientations and complex folded structures render the analysis difficult. Ultra-high-field diffusion MR imaging (dMRI) overcomes these limitations as the improved SNR allowed acquiring 1.4 mm isotropic voxel with increased diffusion-weighting. Applying constrained spherical deconvolution (1) enabled resolving radial and tangential anisotropic diffusion in cortical gray matter confirming recent reports of reduced radial anisotropy in primary somatosensory cortex as compared to other cortical areas (2).


We analyzed high-resolved whole-brain diffusion MRI maps acquired at 7T using constrained spherical deconvolution (CSD) for resolving radial diffusion anisotropy in cortical gray matter and adjacent white matter.


Data were measured on a research 7 Tesla whole-body MR scanner (Siemens Healthcare GmbH, Germany), equipped with a 70 mT/m gradient coil (slew rate of 200 T/m/s). A 32-channel phased-array head coil (Nova Medical, USA) was used for head imaging. The protocol consisted of a high-resolution anatomic scan (MPRAGE, 0.8 mm isotropic resolution, covering the whole head including the cerebellum), diffusion-weighted MR images (dMRI) using a prototype single-shot-EPI sequence employing a modified Stejskal Tanner diffusion encoding gradient scheme (3,4). Additionally, a Gradient Echo sequence was acquired serving for B0 field mapping (5). Wwe optimized the diffusion gradients by employing a web application for multiple-shell protocol design provided by Caruyer (http://www.emmanuelcaruyer.com/q-space-sampling.php), consisting of 128 diffusion gradients per shell and different gradients in each shell. The dMRI protocol compromised 137 volumes with 1.4 mm isotropic resolution. We acquired 128 diffusion-weighted data sets (b=3000 s/mm2) with different combinations of gradient directions (6), and nine non-diffusion-weighted data sets (b=0 s/mm2, b0 images) interspersed with every 17th diffusion-weighted data set. EPI acquisition was accelerated using GRAPPA factor 3, 36 reference lines, 6/8 partial Fourier mode . Other imaging parameters were;bandwidth 1526 Hz/Pixel, echo spacing of 0.76 ms, TE = 73 ms, base resolution 156*156, 98 slices, field of view 220 mm), dMRI measurements coverd the whole brain including the cerebellum. Duration of the measurement was 50 minutes.

Results and discussion:

Comparing anatomic maps and fODF maps confirmed that diffusion in GM could be clearly differentiated from diffusion in adjacent WM due to the spatially different diffusion characteristics (fig. 1): in GM fiber orientations vary strongly according to the complex three-dimensional folded structure while in WM the fibers appear more homogeneously directed along larger distances. Compared to standard direction-encoded color maps (DEC) where only the direction of the largest Eigenvector is depicted the fODF maps show more clearly crossing and bending fibers which allows following better complex fiber courses. In all volunteers, primary somatosensory cortex (fig.1; S1, fig. 2) exhibited strongly reduced radial anisotropy compared to opposite parts of primary motor cortex (fig. 1; M1, fig. 2) as well as to other cortical parts. Inspecting the data in all orthogonal views (fig. 2 a,b,c) revealed that this tissue characteristics is not an artifact due to digitizing the complex three-dimensional cortical surface but is an inherent feature of the central part of the primary somatosensory cortex probably reflecting the well-known different micro-anatomic tissue architecture of M1 as compared to S1. We found that the resolution of 1.4 mm isotropic was the optimum when acquiring whole head dMRI data with high diffusion-weighting of b=3000 s/mm2 in a single measurement (i.e., without averaging). Few studies were published with higher spatial resolution (1,7,8,9,10) but our study adds new evidence by analyzing isotropic whole head data of a larger cohort of 12 volunteers. The results suggest strongly that 1.4 mm isotropic resolution is sufficient to analyze GM diffusion characteristics with a sufficient degree of detail. The acquisition time of about 50 min renders the protocol acceptable for patients thus opening the application of the technique for clinical examinations. (Part of the results presented in the abstract was recently submitted for publication.)


No acknowledgement found.


[1] Tournier et al., Robust determination of the fibre orientation distribution in diffusion MRI: non-negativity constrained super-resolved spherical deconvolution.NeuroImage 35, 2007, 1459-72.

[2] McNab et al., Surface based analysis of diffusion orientation for identifying architectonic domains in the in vivo human cortex. NeuroImage 69, 2013, 87-100.

[3] Morelli JN et al., Evaluation of a modified Stejskal-Tanner diffusion encoding scheme, permitting a marked reduction in TE, in diffusion-weighted imaging of stroke patients at 3 T. Invest Rad 45, 29-35,

[4]Stejskal, E. O. & Tanner, J. E., Spin Diffusion Measurements: Spin Echoes in the Presence of a Time Dependent Field Gradient. J. Chem. Phys. 42, 288 (1965)

[5] Jones, D. K. & Cercignani, M.Twenty-five pitfalls in the analysis of diffusion MRI data. NMR in biomedicine 23, 803–820 (2010)

[6] Jones, D. K. et al. Isotropic resolution diffusion tensor imaging with whole brain acquisition in a clinically acceptable time. Human brain mapping 15, 216–230 (2002)

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

[8]Heidemann, Robin, M.; Anwander A.; Eichner, C.; Luetzkendorf, R.; Feiweier, T.; Knösche, T.R.; Bernarding, J.; Turner, R.; Isotropic Sub-Millimeter Diffusion MRI in Humans at 7T, Proceeding of the Organisation of Human Brain Mapping, June 26-30, Québec City (2011)

[9]Luetzkendorf, R.; Hertel, F.; Heidemann, RM.; Thiel, A.; Luchtmann, M.; Plaumann, M.; Stadler, J.; Baecke, S.; Bernarding, J.; Non-invasive high-resolution tracking of human neuronal pathways: Diffusion Tensor Imaging at 7T with 1.2 mm isotropic voxel size. Medical Imaging 2013: Physics of Medical Imaging, edited by Robert M. Nishikawa, Bruce R. Whiting, Christoph Hoeschen, Proc. of SPIE Vol. 8668, 866846 ·, 7 pages (2013).

[10]Anwander, A. Pampel, A. & Knösche, T. R. In vivo measurement of cortical anisotropy by diffusion-weighted imaging correlates with cortex type. ISMRM Joint Annual Meeting. Stockholm, Sweden. 2010-05-01 - 2010-05-07. 109 (2010).


Axial view of cortical GM and WM (a: MPRAGE, 0.8 isotropic; b: fODF, 1.4 mm isotropic, map overlaid onto b0 map with identical resolution). GM can be differentiated on the fODF maps according to strongly varying diffusion characteristics. Primary somatosensory cortex exhibits strongly reduced radial diffusion anisotropy compared to opposite parts of primary motor cortex or other cortical gray matter.

Orthogonal views of the fODF maps (overlaid onto b0 map). left: axial orientation; middle: coronal orientation; right: sagittal orientation. The differing radial diffusion anisotropy can be detected in all three orthogonal views.

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