Higher spatial resolution at ultra-high field (≥7T) magnetic resonance imaging (UHF MRI), with consequent reduction of partial volume effects, enables diffusion tensor imaging (DTI) to visualize sharp turns of the white matter (WM) fibers entering into grey matter cortex (CTX), which are challenging to see in lower magnetic field data ^1-3. Moreover, UHF DTI data also uncover rather high fractional anisotropy (FA) in several CTX regions, which may help to minimize the strong gyral bias in tractography that was found even in the relatively high resolution 3T DTI ⁴. In this work we analyse statistical properties of the major UHF DTI eigenvector direction at the WM-CTX interface and inside the CTX to understand fiber orientation characteristics in the whole human brain in vivo and allow connectivity analysis from WM into the CTX.

Data acquisition. A male volunteer was scanned at 7T (Siemens Healthcare, Germany). Navigated point-spread function (PSF) scans ⁵ without (b-value = 0) and with 12 different diffusion-weighting (DW) gradients (b-value = 800 s/mm²) were acquired with Stejskal-Tanner diffusion encoding using a 32-channel head coil (Nova Medical, Wilmington, USA). Imaging protocol parameters were: TR/TE=3560/49 ms, 41 slices (covering 3.28cm), slice thickness=0.8 mm, field of view (FOV)=192² mm², matrix size=240² (=0.8² mm², isotropic resolution). Each PSF dataset was acquired with an acceleration factor of 3 in the PSF-PE dimension (corresponding to 80 repetitions or averages) and with a reduced resolution factor of 4 in the EPI-PE dimension (resulting in a matrix size of 60 in the EPI-PE coordinate). No cardiac gating was used and the total scan time for all PSF scans was about 55 minutes. This experiment was repeated four times to cover the entire brain volume. Anatomical imaging included MP-RAGE and 3D GE acquisitions with an isotropic resolution of 0.6 mm.

Image processing. Anatomical images were intensity corrected as described in ⁶. The brain was segmented (CTX, WM, and CSF ) and a series of lamellae was created using the surface expansion method ⁷. The normalized absolute scalar product (AbsScalarProd) of the major DTI eigenvectors and normal vectors perpendicular to the lamellae was calculated. AbsScalarProd was mapped onto the inflated CTX followed by statistical analysis separated in CTX banks and CTX gyri and sulci. CTX banks were identified by a WM-CTX surface curvature within the interval [-0.15, 0.15] 1/mm², and gyri and sulci were characterized by curvature values outside this interval.

Fig.1 demonstrates the anatomical image with overlaid lamella contours and color-coded major DTI eigenvectors. Lamella#1 and #3 were determined by WM-CTX and CTX-CSF interfaces, and lamella#2 was built in the medial CTX depth. Fig.2a and Fig.2b show the distribution of AbsScalarProd on banks, gyri and sulci over the whole human brain in vivo. Lamella#1 was used for creation of surface normal vectors in this case. We observe that the peak of the distribution is located near 0 (banks) or 1 (gyri and sulci). If lamella#2 is taken to form the surface normal vectors then peaks of both distributions of AbsScalarProd are shifted to 1 (Fig.3a,b). The spatial specificity of AbsScalarProd is given in Fig.4. As can be seen on WM-CTX interface, thresholded AbsScalarProd is located on borders (banks) of gyri and sulci (pointed by arrows). On the other hand if medial CTX lamella is used to calculate normal vectors then the marginal distribution of AbsScalarProd is shifted to the centers of gyri and sulci (pointed by arrows).We had shown that major DTI eigenvectors at UHF can exhibit patterns of sharp turning of fibers in the human CTX in vivo. The analysis of DTI is facilitated by the AbsScalarProd parameter which reflects the curvature of the CTX and building of lamellae in the CTX depth. Thus UHF DTI with resolution 0.8iso mm³ robustly probes structural differences along the CTX and in cortical depth. Potentially AbsScalarProd can be used as a metric to measure the fiber bending in different parts of the WM-CTX interface, which may become relevant for detecting diseases and brain malformation⁹. Further validation of the correspondence of AbsScalarProd and mikrostructure using ex vivo samples together with histology is required.