In-vivo parcellation of human subcortex by multi-modal MRI

Introduction: Human subcortex comprises multiple subcortical grey matter (SGM) structures, many with several nuclei and specialised sub-regions dedicated to highly specific functions. Histology-driven brain atlases provide detailed delineation of these sub-regions; however, those cannot be directly applied to in-vivo MRI studies. Here, we integrate the information from anatomy, diffusion micro-environment, and directions of white matter (WM) fibres within SGM, from multi-contrast MRI, to segregate the nuclei and specialised sub-regions of human subcortex.

Methods: Minimally pre-processed T1w, T2w, and Diffusion MRI (dMRI) data obtained at 3T were downloaded from Human Connectome Project [1-3] for 50 healthy subjects. For each subject, fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), radial diffusivity (RD), and myelin index (T1w/T2w) were computed following the processing pipeline in [4], orientation of the WM fibre tracts were computed within each voxel [5]. The parametric maps were combined by computing track-weighted imaging maps [6,7] with 0.7mm isotropic super-resolution. SGM structures were masked using FSL [8], data corresponding to SGM structures were first analysed by principal component analysis and then used by k-means clustering to parcellate SGM.

Results and Conclusions: We have identified and mapped 56 newly delineated nuclei, sub-nuclei, and sub-regions within human SGM. Our SydSGM parcellation demonstrated remarkable resemblance to the gold-standard histology-based atlas by Mai et al. [9]. SydSGM parcellation can be directly applied to subject-specific and/or group average MRI dataset as a standalone atlas or by incorporating it into existing brain atlases. We demonstrate SydSGM parcellation’s advantage by revealing declined structural connectivity of highly resolved SGM subnuclei in patients with early Parkinson’s disease.

References: [1] Van Essen et al. 2013. NeuroImage; [2] Glasser et al. 2013. NeuroImage; [3] Sotiropoulos et al. 2013. NeuroImage; [4] Ali et al. 2022. Magnetic Resonance in Medicine; [5] Dhollander et al. 2015. Proc Intl Soc Mag Reson Med; [6] Calamante et al. 2012. NeuroImage; [7] Calamante et al. 2017. MAGMA; [8] Patenaude et al. 2011. NeuroImage; [9] Mai et al., 2015.