Evaluation of diffusion kurtosis imaging in ex vivo hypomyelinated mouse brains

Abstract

Diffusion tensor imaging (DTI), diffusion kurtosis imaging (DKI), and DKI-derived white matter tract integrity metrics (WMTI) were experimentally evaluated ex vivo through comparisons to histological measurements and established magnetic resonance imaging (MRI) measures of myelin in two knockout mouse models with varying degrees of hypomyelination. DKI metrics of mean and radial kurtosis were found to be better indicators of myelin content than conventional DTI metrics. The biophysical WMTI model based on the DKI framework reported on axon water fraction with good accuracy in cases with near normal axon density, but did not provide additional specificity to myelination. Overall, DKI provided additional information regarding white matter microstructure compared with DTI, making it an attractive method for future assessments of white matter development and pathology.

Keywords: Diffusion; Kurtosis; MRI; Myelin; Quantitative histology; White matter.

FIG. 1

Quantitative histology metrics derived from EM. (a) T₂-weighted sagittal cross-section showing locations of white matter histology sections. (b) Typical 15000x EM image of the splenium. (c) EM image after segmentation with myelin pixels shown in white. (d) EM image showing axons after region growing segmentation is performed to derive axon area and diameter. (e) Typical axon area and z_m measurements.

FIG. 2

Representative EM images of white matter. (Top row) EM images of the splenium from control, Rictor and TSC mouse brains. (Bottom row) Processed and segmented EM images corresponding to the images shown on the top row. Compared to the control brain, Rictor brain WM shows a slight decrease in myelin thickness and myelinated axon density. TSC brain WM shows a significant decrease in myelinated axon density.

FIG. 3

Representative coronal DTI parameter maps. The rows represent the three groups: control (top), Rictor (middle), and TSC (bottom). The columns represent the DTI parameters FA, MD, AD, and RD. FA is unitless and diffusivities are in units of μm²/ms.

FIG. 4

ROI group comparison of DTI parameters FA, MD, AD, and RD. Bar heights represent the ROI parameter means across each group, with error bars spanning the mean ± SEM. The three groups are represented with different colors: control = blue, Rictor = green, and TSC = red. 6 WM ROIs are indicated on the x-axis: MidCC = midbody of the corpus callosum, GCC = genu, SCC = splenium, AC = anterior commissure, EC = external capsule, and IC = internal capsule. FA is unitless and diffusivities are in units of μm²/ms. * denotes p < 0.05 for comparison between the control group and each CKO group.

FIG. 5

Representative coronal DKI parameter maps. The rows represent the three groups: control (top), Rictor (middle), and TSC (bottom). The columns represent the DKI parameters MK, AK, and RK. Kurtosis is unitless.

FIG. 6

ROI group comparison of DKI parameters MK, AK, and RK. Bar heights represent the ROI parameter means across each group, with error bars spanning the mean ± SEM. The three groups are represented with different colors: control = blue, Rictor = green, and TSC = red. 6 WM ROIs are indicated on the x-axis: MidCC = midbody of the corpus callosum, GCC = genu, SCC = splenium, AC = anterior commissure, EC = external capsule, and IC = internal capsule. Kurtosis is unitless. * denotes p < 0.05 for comparison between the control group and each CKO group.

FIG. 7

Representative coronal WMTI parameter maps. WMTI parameter values for white matter voxels are overlaid on the b=0 image. The rows represent the three groups: control (top), Rictor (middle), and TSC (bottom). The columns represent the WMTI parameters F_a, D_a, D_e,||, and D_e,⊥. F_a is unitless and diffusivities are in units of μm²/ms.

FIG. 8

ROI group comparison of WMTI parameters F_a, D_a, D_e,||, and D_e,⊥. Bar heights represent the ROI parameter means across each group, with error bars spanning the mean ± SEM. The three groups are represented with different colors: control = blue, Rictor = green, and TSC = red. 6 WM ROIs are indicated on the x-axis: MidCC = midbody of the corpus callosum, GCC = genu, SCC = splenium, AC = anterior commissure, EC = external capsule, and IC = internal capsule. F_a is unitless and diffusivities are in units of μm²/ms. * denotes p < 0.05 for comparison between the control group and each CKO group.

FIG. 9

Linear correlation plots between DTI metrics FA, MD, and RD (y-axis) and myelin measures MWF, PSR, and f_m (x-axis). Each point is an ROI mean from an individual mouse brain. The shape represents the group: control = circle, Rictor = square, and TSC = x and the color represents the WM region: MidCC = blue, GCC = green, SCC = red, AC = cyan, EC = magenta, IC = yellow. For significant correlations, r² is shown in the lower right-hand corner of each plot. FA, MWF, PSR, and f_m are unitless, MD and RD are in units of μm²/ms.

FIG. 10

Linear correlation plots between DTI metrics FA, MD, and RD (y-axis) and the histologic measures d_a, ρ_a, f_a, and g (x-axis). Each point is an ROI mean from an individual mouse brain. The shape represents the group: control = circle, Rictor = square, and TSC = x and the color represents the WM region: MidCC = blue, GCC = green, SCC = red, AC = cyan. For significant correlations, r² is shown in the lower right-hand corner of each plot. FA, f_a, and g are unitless, MD and RD are in units of μm²/ms, d_a is in units of μm, and ρ_a is in units of axons/μm².

FIG. 11

Linear correlation plots between DKI metrics MK and RK (y-axis) and myelin measures MWF, PSR, and f_m (x-axis). Each point is an ROI mean from an individual mouse brain. The shape represents the group: control = circle, Rictor = square, and TSC = x and the color represents the WM region: MidCC = blue, GCC = green, SCC = red, AC = cyan, EC = magenta, IC = yellow. For significant correlations, r² is shown in the lower right-hand corner of each plot. MK, RK, MWF, PSR, and f_m are unitless.

FIG. 12

Linear correlation plots between DKI metrics MK and RK (y-axis) and the histologic measures d_a, ρ_a, f_a, and g (x-axis). Each point is an ROI mean from an individual mouse brain. The shape represents the group: control = circle, Rictor = square, and TSC = x and the color represents the WM region: MidCC = blue, GCC = green, SCC = red, AC = cyan. For significant correlations, r² is shown in the lower right-hand corner of each plot. MK, RK, f_a, and g are unitless, d_a is in units of μm, and ρ_a is in units of axons/μm².

FIG. 13

Linear correlation plots between WMTI metrics F_a and D_e,⊥ (y-axis) and myelin measures MWF, PSR, and f_m (x-axis). Each point is an ROI mean from an individual mouse brain. The shape represents the group: control = circle, Rictor = square, and TSC = x and the color represents the WM region: MidCC = blue, GCC = green, SCC = red, AC = cyan, EC = magenta, IC = yellow. For significant correlations, r² is shown in the lower right-hand corner of each plot. F_a, MWF, PSR, and f_m are unitless, D_e,⊥ is in units of μm²/ms.

FIG. 14

Linear correlation plots between WMTI metrics F_a and D_e,⊥ (y-axis) and the histologic measures d_a, ρ_a, f_a, and g (x-axis). Each point is an ROI mean from an individual mouse brain. The shape represents the group: control = circle, Rictor = square, and TSC = x and the color represents the WM region: MidCC = blue, GCC = green, SCC = red, AC = cyan. For significant correlations, r² is shown in the lower right-hand corner of each plot. F_a, f_a, and g are unitless, D_e,⊥ is in units of μm²/ms, d_a is in units of μm, and ρ_a is in units of axons/μm².