The Myelin Water Fraction Serves as a Marker for Age-Related Myelin Alterations in the Cerebral White Matter - A Multiparametric MRI Aging Study

Abstract

Quantitative MRI modalities, such as diffusion tensor imaging (DTI) or magnetization transfer imaging (MTI) are sensitive to the neuronal effects of aging of the cerebral white matter (WM), but lack the specificity for myelin content. Myelin water imaging (MWI) is highly specific for myelin and may be more sensitive for the detection of changes in myelin content inside the cerebral WM microstructure. In this multiparametric imaging study, we evaluated the performance of myelin water fraction (MWF) estimates as a marker for myelin alterations during normal-aging. Multiparametric MRI data derived from DTI, MTI and a novel, recently-proposed MWF-map processing and reconstruction algorithm were acquired from 54 healthy subjects (aged 18-79 years) and region-based multivariate regression analysis was performed. MWFs significantly decreased with age in most WM regions (except corticospinal tract) and changes of MWFs were associated with changes of radial diffusivity, indicating either substantial alterations or preservation of myelin content in these regions. Decreases of fractional anisotropy and magnetization transfer ratio were associated with lower MWFs in commissural fiber tracts only. Mean diffusivity had no regional effects on MWF. We conclude that MWF estimates are sensitive for the assessment of age-related myelin alterations in the cerebral WM of normal-aging brains.

Keywords: T2-relaxometry; diffusion imaging; magnetization transfer imaging; myelin water imaging; normal aging.

FIGURE 1

Work-flow of post-acquisition processing steps. Figure presents subjects’ characteristics and visualizes distinct steps in the work-flow of image processing.

FIGURE 2

Progression of MRI parameters with age in different brain regions. Figure displays age-related regional progressions of Diffusion Tensor scalars [fractional anisotropy (FA); mean diffusivity (MD); radial diffusivity (RD); magnetization transfer ratio (MTR); myelin water fraction (MWF)]. CST, corticospinal tract; GCC, genu of corpus callosum; SCC, splenium of corpus callosum.

FIGURE 3

Region-based relationships of MWF metrics with diffusion-based scalars, MTR and age in the frontal WM. The Effect plots in figure show the marginal effects of the multivariate regression model. For each of the five independent variables [age: gray plot; Radial Diffusivity: blue plot: Fractional Anisotropy: orange plot: Mean Diffusivity: green plot: Magnetization Transfer Ratio: purple plot] we show the model-based effect on the MWF measured in the frontal white matter. The regression coefficient (β), confidence interval (CI, 2.5–97.5%) and p-value for the multivariate model are shown above the respective plot region. For the variable “age,” the model’s regression coefficient is estimated using age in decades. Additionally, we show the bivariate unadjusted correlation (points, regression line, and CI in light gray in the background). The respective regression coefficient for the bivariate model, the related CI and p-value are shown below the respective plot region in light gray.

FIGURE 4

Region-based relationships of MWF metrics with diffusion-based scalars, MTR and age in the parietal WM. The Effect plots in figure show the marginal effects of the multivariate regression model. For each of the five independent variables [age: gray plot; Radial Diffusivity: blue plot: Fractional Anisotropy: orange plot: Mean Diffusivity: green plot: Magnetization Transfer Ratio: purple plot] we show the model-based effect on the MWF measured in the parietal white matter. The regression coefficient (β), confidence interval (CI, 2.5–97.5%) and p-value for the multivariate model are shown above the respective plot region. For the variable “age,” the model’s regression coefficient is estimated using age in decades. Additionally, we show the bivariate unadjusted correlation (points, regression line, and CI in light gray in the background). The respective regression coefficient for the bivariate model, the related CI and p-value are shown below the respective plot region in light gray.

FIGURE 5

Region-based relationships of MWF metrics with diffusion-based scalars, MTR and age in the occipital WM. The Effect plots in figure show the marginal effects of the multivariate regression model. For each of the five independent variables [age: gray plot; Radial Diffusivity: blue plot: Fractional Anisotropy: orange plot: Mean Diffusivity: green plot: Magnetization Transfer Ratio: purple plot] we show the model-based effect on the MWF measured in the occipital white matter. The regression coefficient (β), confidence interval (CI, 2.5–97.5%) and p-value for the multivariate model are shown above the respective plot region. For the variable “age,” the model’s regression coefficient is estimated using age in decades. Additionally, we show the bivariate unadjusted correlation (points, regression line, and CI in light gray in the background). The respective regression coefficient for the bivariate model, the related CI and p-value are shown below the respective plot region in light gray.

FIGURE 6

Region-based relationships of MWF metrics with diffusion-based scalars, MTR and age in the CST. The Effect plots in figure show the marginal effects of the multivariate regression model. For each of the five independent variables [age: gray plot; Radial Diffusivity: blue plot: Fractional Anisotropy: orange plot: Mean Diffusivity: green plot: Magnetization Transfer Ratio: purple plot] we show the model-based effect on the MWF measured in the corticospinal tract (CST). The regression coefficient (β), confidence interval (CI, 2.5–97.5%) and p-value for the multivariate model are shown above the respective plot region. For the variable “age,” the model’s regression coefficient is estimated using age in decades. Additionally, we show the bivariate unadjusted correlation (points, regression line, and CI in light gray in the background). The respective regression coefficient for the bivariate model, the related CI and p-value are shown below the respective plot region in light gray.

FIGURE 7

Region-based relationships of MWF metrics with diffusion-based scalars, MTR and age in the SCC. The Effect plots in figure show the marginal effects of the multivariate regression model. For each of the five independent variables [age: gray plot; Radial Diffusivity: blue plot: Fractional Anisotropy: orange plot: Mean Diffusivity: green plot: Magnetization Transfer Ratio: purple plot] we show the model-based effect on the MWF measured in the splenium of the corpus callosum (SCC). The regression coefficient (β), confidence interval (CI, 2.5–97.5%) and p-value for the multivariate model are shown above the respective plot region. For the variable “age,” the model’s regression coefficient is estimated using age in decades. Additionally, we show the bivariate unadjusted correlation (points, regression line, and CI in light gray in the background). The respective regression coefficient for the bivariate model, the related CI and p-value are shown below the respective plot region in light gray.

FIGURE 8

Region-based relationships of MWF metrics with diffusion-based scalars, MTR and age in the GCC. The Effect plots in figure show the marginal effects of the multivariate regression model. For each of the five independent variables [age: gray plot; Radial Diffusivity: blue plot: Fractional Anisotropy: orange plot: Mean Diffusivity: green plot: Magnetization Transfer Ratio: purple plot] we show the model-based effect on the MWF measured in the genu of the corpus callosum (GCC). The regression coefficient (β), confidence interval (CI, 2.5–97.5%) and p-value for the multivariate model are shown above the respective plot region. For the variable “age,” the model’s regression coefficient is estimated using age in decades. Additionally, we show the bivariate unadjusted correlation (points, regression line and CI in light gray in the background). The respective regression coefficient for the bivariate model, the related CI and p-value are shown below the respective plot region in light gray.