Principal Investigator National Institutes of Health
Abstract Text: Introduction
We can comprehensively describe subvoxel microstructural heterogeneity using an ensemble of microscopic diffusion tensors, or a diffusion tensor distribution (DTD), corresponding to distinct compartments, with different sizes, shapes, and orientations. DTD-derived parameters provide improved sensitivity to microstructural alterations in brain tumors, and potentially in traumatic brain injury (TBI). Nevertheless, measuring arbitrary DTDs requires long scans with multiple diffusion-encoded (MDE) measurements that are infeasible for clinical applications.
We propose a practical new framework, called constrained reference frame diffusion tensor spectroscopic (CORTECS) MRI that drastically simplifies the measurement of arbitrary DTDs using much smaller single diffusion encoded (SDE) data sets which can be obtained more efficiently than MDE measurements. We demonstrate its application in a fixed macaque monkey brain and compare the results with histology. CORTECS MRI advances the clinical translation of DTD-MRI enabling applications in TBI, cancer, and neurodegeneration.
Methods
In tissues with consistent, well-defined architecture, such as the cortex, if the voxel size is significantly smaller than the cortical folding radius of curvature, the intravoxel orientational dispersion becomes negligible. Under these circumstances we can constrain the microscopic diffusion tensors to be diagonalized by the same orthogonal references of the voxel, defined by the principal axes of diffusion measured from the signal in the entire voxel. In each voxel, the constrained DTD (cDTD) is completely determined by the 3D correlation spectrum of the microscopic principal diffusivities along with these orientations and can be estimated using only SDE measurements. Moreover, if the tissue architecture varies along a single dominant orientation, we can describe the cDTDs more efficiently as the 2D correlation spectrum of radial and tangential diffusivities.
We performed Monte Carlo (MC) simulations to assess the cDTD reconstruction accuracy at different signal-to-noise ratio (SNR) levels. We acquired high-resolution dMRI data from a perfusion-fixed macaque monkey brain using 200µm cubic voxels, TE/TR=50/650ms, 112 SDE DWIs with multiple b-values and orientations. After post-processing, we estimated non-parametric 2D cDTDs in each voxel and derived the marginal distributions of radial and tangential diffusivities. To characterize the size-shape distributions of the cDTDs we derived intravoxel 2D correlation spectra of microscopic FA and MD values, along with the corresponding marginal distributions of these important microstructural parameters. Finally, we quantified the concentration of water pools with different diffusion properties by integrating the cDTD across different spectral domains.
Results
MC simulations demonstrate that, from datasets containing a clinically feasible number of SDE acquisitions, we can disentangle multiple microscopic diffusion tensors aligned along the same voxel reference frame based on differences in their principal diffusivities. The mean normalized spectra reconstructed at various SNR levels one quantifies the concentrations and diffusion tensor properties of all microscopic tissue compartments.
Maps of 2D cDTDs in cortical GM reveal diffusion processes with distinct joint radial and tangential diffusivities and different specificities across cortical domains and layers. cDTDs with different mixtures of isotropic and anisotropic microscopic diffusion components show high specificity to cortical layers and are in good agreement with fiber orientation distributions (FOD) maps and corresponding histology.
The 2D µFA-MD spectra provide a tally of the shape-size diffusion tensor characteristics across all subvoxel tissue microenvironments as a new means to characterize microstructure. The largest concentrations of isotropic microscopic diffusion processes (µFA < 0.18) were observed in the upper cortical layers. The most anisotropic diffusion processes (µFA>0.35) were localized in the mid-cortical layers and in the subcortical white matter. The spatial distribution and concentrations of various tissue compartments obtained by integrating the 2D µFA-MD spectra over empirically defined spectral domains show prominent layer-specific motifs.
Discussion
In tissues with consistent, well-defined architecture, CORTECS MRI greatly simplifies the data acquisition and spectral reconstruction requirements for high-resolution DTD MRI. It significantly reduces the dimensionality of non-parametric DTD MRI, allowing the robust estimation of non-parametric cDTDs without the need for statistical reconstruction methods or MDE measurements which are difficult in clinical practice. Moreover, the fixed voxel reference frame allows the consistent assessment of the distribution of DTD eigenvalues and parameters (µFA, MD) without confounds from sorting bias. Taken together, CORTECS MRI provides a practical and feasible approach to non-parametric quantitation of microstructural tissue heterogeneity.