Improved Constrained Spherical Deconvolution for Microstructural Imaging of Brain Gray Matter

Abstract

In diffusion magnetic resonance imaging (dMRI), conventional multi-shell constrained spherical deconvolution (CSD) methods usually model white matter microstructure as anisotropic and gray matter microstructure as isotropic. Nonetheless, investigations using high-resolution dMRI and histology have demonstrated that the diffusion process in gray matter exhibits clear anisotropic features. To better disclose the microstructural aspects of gray matter, this research suggests a gray matter multi-shell restricted spherical deconvolution method based on the SANDI (Soma And Neurite Density Imaging) model. We first simulate the signal of grey matter using the SpinDoctor with real neuronal data and then reconstruct the fibers in the grey matter portion of real brain data. Experimental results demonstrate that our algorithm effectively captures the anisotropic features of grey matter regions and estimates grey matter fiber orientations with greater accuracy, closely reflecting real-world scenarios.

Keywords

• diffusion magnetic resonance imaging (dmri)
• gray matter microstructural imaging
• constrained spherical deconvolution (csd)
• fiber orientation estimation