3777
Ex vivo mapping of the cyto- and the myeloarchitecture of the human cerebral cortex using ultra-high field MRI (7T and 11.7T)1CEA - NeuroSpin, Gif-sur-Yvette, France, 2Université Paris-Saclay, Orsay, France, 3CentraleSupélec, Gif-sur-Yvette, France, 4Imaging and Brain laboratory (iBrain), Université de Tours - INSERM, Tours, France
Synopsis
The investigation of the human cerebral cortex at the mesoscopic scale remains challenging but promising to better understand brain pathologies associated with cortex damage. In this ex vivo study, samples of occipital cortex from both hemispheres of an unique subject have been delineated using their microstructural and myeloarchitectural information inferred from ultra-high field anatomical, quantitative and diffusion MRI. The high-resolved UHF-MRI dataset enabled to perform an automatic segmentation of the cortical layers within the primary and secondary visual cortices. The segmentations highlighted their commonalities and differences between the two hemispheres.
Introduction
Imaging the cortical thickness has become a well-established approach to investigate neurodegenerative, neurodevelopmental or psychiatric diseases such as Alzheimer’s, cortical malformations or schizophrenia1. Segmentation of its laminar structure at a mesoscopic scale would allow to develop novel atlases of the cortex anatomy and thus help to better identify slight cortical damage in pathological cases. Since it offers improved spatial resolution and new contrast sources such as susceptibility-weighted imaging, ultra-high field (UHF) MRI (>=7T) is a suitable candidate to probe the cortex laminar structure2. Quantitative T1-weighted MRI related to myelination3 and diffusion MRI related to cytoarchitecture4 have also proven to provide complementary insights about cortical lamination. In this work, we propose to investigate various quantitative (T1, T2, T2*, ρ and myelin water fraction (MWF)) and diffusion UHF MRI contrasts to characterize the laminar structure of the occipital lobe at the mesoscopic scale.Materials and methods
Human samples – Two formalin-fixed postmortem human brain samples containing the primary and secondary visual cortices (V1 and V2) of both hemispheres from the same donor were rehydrated in a 0.1M phosphate-buffered saline solution before being scanned using two preclinical Bruker 11.7T/7T MRI systems with strong gradients (respectively 780mT/m and 750mT/m), as well as 60mm proton volume coils.MRI protocols – A first 11.7T MRI protocol was designed to provide anatomical (aMRI) and diffusion (dMRI) MRI datasets including: a 3D T2-weighted MSME sequence (isotropic resolution 100μm; TE/TR=20/500ms), a 2D T2-weighted SE sequence (isotropic resolution 150μm; TE/TR=16/6647ms) and a series of 3D segmented EPI PGSE sequences (isotropic resolution 200μm ; TE/TR=24.3/250ms; 30 segments; δ/∆=5/12.3ms; b=1500/4500/8000s/mm² along 25/60/90 directions).
A second 7T MRI protocol was designed to provide a quantitative (qMRI) MRI dataset including: a series of 3D variable flip angle (VFA) FLASH sequences5 (65 angles; flip angles=3-45°; TE/TR=4.99/15ms), a 3D T2-weighted MSME sequence (isotropic resolution 200μm; TE/TR=5.56-166.8/1000ms; 30 echoes), a 3D T2*-weighted GRE EPI sequence (TE/TR=3-69.04/9000ms; 8 echoes) and a dual flip-angle EPI B1 mapping sequence6,7 (TE/TR=13.41/1500ms; flip angles=30/60°) to compute actual flip angle and correct B1 inhomogeneity.
Pre-processing – All qMRI and dMRI data were denoised using a non-local means filter, considering the Rician nature of the noise8.
Post-processing – Quantitative ρ (qPD), T1 (qT1), T2 (qT2) and T2* (qT2*) maps were computed using a robust fit of the VFA FLASH, T2-weighted and T2*-weighted signal equations using a non-linear programming Nelder-Mead-based algorithm. The MWF map was obtained using a model fitting procedure9 based on the VFA FLASH and the T2-weighted MSME sequences. The dMRI dataset enabled to infer the NODDI model10 providing proxies of the neurite density (fintra) and their orientation dispersion (OD). The local principal direction was initialized from the jointly estimated DTI model and the parallel diffusivity of neurites set to d// =4.8.10-10m²/s. 7T and 11.7T datasets were registered using 3D affine transformation. Accurate cortices’ masks were obtained from qT1 and qPD maps that exhibit a clear contrast between PBS residuals, gray and white matters.
Clustering – By defining at each voxel a N-dimensional feature combining the pre-computed qT1, qT2, MWF, fintra and kappa maps, N-dimension cluster maps and their posterior probability maps were established using a Gaussian mixture model for a number of components corresponding to the minimum of the Bayesian Information Criterion (BIC), thus defining a coherent spatial parcellation of the cortical mantel. The resulting cluster maps were enhanced using the Potts regularization model to correct outliers.
Results and discussion
A high SNR was obtained for the quantitative (>20) and the diffusion (>10) datasets presented in Figure 1. The Gennari line (layer 4 of V1 in myeloarchitectonics) is clearly visible on anatomical scans and the qT2 map in Figure 2 (green arrow). Subcortical U-shaped white matter bundles (blue arrows) and cortical lamination are also noticeable on this same map. Figure 3 displays the cortices’ masks of both hemispheres defining the computing domain for the cyto- and myelo- Gaussian mixture models. Figure 4 depicts the laminar segmentation of the cortex and the Gaussian distributions' parameters.The left and right hemisphere samples present similar contrasts and quantitative values, both being coherent with literature11,12. Cortical myelination differs within the calcarine sulcus where a subcluster (the green zone) is found for the left hemisphere. These discrepancies may implicate a hemispherical specificity of the myelinated components and even ocular dominance13. The cyto-clustering maps reveal that the left hemisphere displays one more cellular layer than the right hemisphere. The latter also presents an inhomogeneous spatial distribution of its cyto-clusters. Further works will consist in investigating the observed cellular disparities. As for the subcortical white matter, it exhibits two distinct contrasts on the qT2 map as presented in Figure 2, inferring a varying spatial organization for these U-shaped fibers.
Conclusion
Through this study, we showed that UHF quantitative and diffusion MRI enabled the mapping of the cytoarchitecture and the myeloarchitecture of the primary and secondary visual cortices of both hemispheres from an unique human sample. The clusterings pointed out the cellular and myelinated environments of the occipital cortex that are usually observed at the microscopic scale.This work paves the way to a complete structural and quantitative mapping of the human cerebral cortex that would unveil laminar features at the mesoscopic scale.
Acknowledgements
This work received funding from the European Union’s Horizon 2020 Framework Program for Research and Innovation (Grant Agreement No 720270, Human brain Project SGA2). The authors acknowledge the donor who gave his body to science for this study.References
1. Thompson, P. M. & Toga, A. W. Cerebral Cortex Diseases and Cortical Localization. in eLS (American Cancer Society, 2003). doi:10.1038/npg.els.0002195.
2. Roebroeck, A., Miller, K. L. & Aggarwal, M. Ex vivo diffusion MRI of the human brain: Technical challenges and recent advances. NMR Biomed. 0, e3941 (2018).
3. Lifshits, S. et al. Resolution considerations in imaging of the cortical layers. NeuroImage 164, 112–120 (2018).
4. Beaujoin, J., Popov, A., Yebga Hot, R., Poupon, F., Mangin, J.-F., Destrieux, C., Poupon, C. CHENONCEAU : towards a novel mesoscopic (100/200µm) postmortem humain brain MRI atlas at 11.7T. ISMRM 2019.
5. Trzasko, J. D., Mostardi, P. M., Riederer, S. J. & Manduca, A. Estimating T1 from Multichannel Variable Flip Angle SPGR Sequences. Magn. Reson. Med. Off. J. Soc. Magn. Reson. Med. Soc. Magn. Reson. Med. 69, (2013).
6. Cunningham, C. H., Pauly, J. M. & Nayak, K. S. Saturated double-angle method for rapid B1+ mapping. Magn. Reson. Med. 55, 1326–1333 (2006).
7. Boudreau, M. et al. B1 mapping for bias-correction in quantitative T1 imaging of the brain at 3T using standard pulse sequences. J. Magn. Reson. Imaging 46, 1673–1682 (2017).
8. Buades, A., Coll, B. & Morel, J.-M. Non-Local Means Denoising. Image Process. Line 1, 208–212 (2011).
9. Kulikova, S., Hertz-Pannier, L., Dehaene-Lambertz, G., Poupon, C. & Dubois, J. A New Strategy for Fast MRI-Based Quantification of the Myelin Water Fraction: Application to Brain Imaging in Infants. PLOS ONE 11, e0163143 (2016).
10. Zhang, H., Schneider, T., Wheeler-Kingshott, C. A. & Alexander, D. C. NODDI: Practical in vivo neurite orientation dispersion and density imaging of the human brain. NeuroImage 61, 1000–1016 (2012).
11. Birkl, C. et al. Effects of formalin fixation and temperature on MR relaxation times in the human brain. NMR Biomed. 29, 458–465 (2016).
12. Sengupta, S. et al. High resolution anatomical and quantitative MRI of the entire human occipital lobe ex vivo at 9.4T. NeuroImage 168, 162–171 (2018).
13. Menon, RS. et al. Ocular Dominance in Human V1 Demonstrated by Functional Magnetic Resonance Imaging. Journal of Neurophysiology 77, 2780-2787 (1997).
Figures
A=anterior; P=posterior; S=superior; I=inferior.
Quantitative relaxometric and diffusion maps at 200μm isotropic resolution in sagittal view of right-hemisphere C1: (a) proton density; (b) T1 relaxation time; (c) T2* relaxation time; (d) T2 relaxation time; (e) myelin water fraction; (f) intracellular volume fraction; (g) kappa; (h) orientation dispersion; (i) color-encoded diffusion directions map from the DTI model. T2 map close-up (red square) highlighting cortical lamination, the Gennari line (green arrow) and subcortical U-shaped white matter bundles (blue arrows).
A=anterior; P=posterior; S=superior; I=inferior.
Gaussian distributions' parameters (mean values of the clusters and BIC scores) related to the clustering maps of both hemispheres issued from a Gaussian mixture model: the cyto-clustering (combination of kappa, intracellular volume fraction and quantitative T2) and myelo-clustering (combination of myelin water fraction, quantitative T1 and T2) maps.
The clusterings revealed 8 cyto-cluster classes and 6 myelo-cluster classes. The numbers of Gaussian distributions were set to be a trade-off between a minimum BIC score and an anatomically plausible clustering of the cortex.