Understanding the correlation between brain functions and tissues microstructure¹ is crucial to establish novel atlases of the cortex in vivo. Diffusion MR microscopy^2,3 has proven to be a useful tool to investigate the cytoarchitecture of brain tissues, and recent models like NODDI⁴ give the opportunity to probe the dendrite density within the cortex in vivo. Moreover, few studies have been published to demonstrate its potential to probe the modifications occurring during normal development^5,6 or progression of diseases^7,8. In this work, we establish a novel atlas of the dendrite density in the healthy volunteer and we investigate their asymmetries in cortical areas between left and right hemispheres.

Acquisition - 71 right-handed healthy volunteers were scanned on a 3T MRI system using a dedicated protocol⁹ including: a) a MPRAGE sequence to obtain a T1-weighted anatomy (1mm isotropic; TE=2.98ms; TR=2.3s); b) a multiple-shell diffusion-weighted SE-EPI sequence (1.7mm isotropic; TE=117ms; TR=14s; 10 b-values (300-3000s/mm2) along 20 directions + 10 scans at b=0s/mm2); c) a fieldmap calibration to remove distorsions.

Individual NODDI maps – Individual DW dataset were preprocessed using Connectomist⁹ to remove imaging artifacts, thus enabling accurate matching of DW and T1-weighted data using rigid registration. Individual NODDI maps were computed using Connectomist (fig1): the intracellular fraction (f_intra) referring to the space bounded by the membranes of neurites, their orientation dispersion (OD), the Bingham concentration parameter (K), and the CSF fraction (f_iso).

Individual and group pial and white surfaces – T1-weighted MR scans were processed using FreeSurfer et aparc.a2009s¹⁰ to extract individual pial and white matter surfaces. Homologue vertices between pial/white surfaces and between individuals allow to easily navigate between surfaces. Cortex parcelation was done using the aparc.a2009s Destrieux atlas¹¹ defining 152 regions.

Dendrite density maps – The correspondence of vertices allowed to extract a distribution of f_intra for each vertex by sampling the matched f_intra map along the segment defined by each pair of ( pial, white ) vertices (fig2). Maps of local distributions of f_intra were computed and statistics were inferred to build maps of mean, standard deviation (stddev) and median of f_intra at the vertex level (fig2). To obtain distributions of f_intra at the scale of cortical areas, the distributions obtained for vertices included in the area were merged together to form a global distribution from which statistics were established, providing a microstructural signature of this area (fig3).

Evaluation of left/right asymmetries - the signed differences of mean f_intra between left and right cortical areas were computed for all the areas and subjects, and a t-test was performed to evaluate the L-R asymmetry of f_intra for each region.

Fig.1 depicts T1-weighted and NODDI individual maps, clearly showing a lower f_intra and a higher OD in the cortex. Fig.2 illustrates the pipeline used to extract surface maps of f_intra at the vertex scale at individual and group levels, showing that f_intra patterns remain similar on both hemispheres, but present differences within the same hemisphere. Fig.3 shows the mean/median/stddev maps of f_intra obtained for cortical areas at individual and group levels. Fig.4 provides a table of t-tests performed on L-R mean differences for all the regions. False Discovery Rate correction was applied to correct for multiple comparisons, yielding 44 significant areas: 26 corresponding to a higher dendrite density on the left hemisphere and 18 on the right hemisphere. Fig.5 provides renderings of the significant regions: 1) the primary sensorimotor cortex depicts a higher dendrite density in the left hemisphere, which is in perfect adequation with the right handedness of the group; 2) the superolateral part of the temporal lobe supporting the auditory and langage networks depicts a higher dendrite density in the left hemisphere, which fits with the left localization of this function; 3) on the contrary, the occipital lobe supporting the visual cortex depicts a higher dendrite density in the right hemisphere to be confronted to the comparisons done according to the gender¹², as well as the fusiform area processing color/face/body/word recognition; 4) the superior frontal gyrus, involved in self-awareness also depicts a higher dendrite density in the right hemisphere. In this work, we demonstrated that diffusion MR microscopy, using the NODDI model, opens promising perspectives for the investigation of the cortical areas dendritic composition in vivo. For a population of right-handed subjects, we showed that the established asymmetries between L-R hemispheres are in good agreement with the lateralization of the underlying brain functions.