Purpose

Multi-echo T₂ (ME-T₂) imaging enables gauging the contribution of water signal from different cellular and extracellular compartments in subcortical white matter (WM). A relatively short T₂ relaxation component is associated with the water trapped between myelin sheaths and its magnitude is commonly expressed as a myelin water fraction. A much larger water signal from the intracellular and extracellular (IE) compartment of WM with intermediate T₂ relaxation times, however, remains relatively less explored. The geometric mean T₂ of water from the IE space (geomT_2IEW) in WM is relatively longer¹ and axonal density (AxD) lower² in the splenium of the corpus callosum (CC) compared to the genu. Thus, geomT_2IEW may reflect the influence of AxD and axonal size³ (AxS). To elucidate the complex relationship between the geomT_2IEW and WM microstructure, we explored the effects of age on geomT_2IEW in subcortical WM tracts, in a sample spanning the developmental and adult age range.

Methods

The participants were paid volunteers recruited from a longitudinal study of healthy aging, and control subjects from a study of OCD (see Figure 1 for sample description). All images were collected on 3T Siemens Verio system using either a 12 or 32-channel receive-only head coil. The ME-T₂ was acquired using a 3D-GRASE sequence (32 echoes, 11 ms inter-echo spacing, TR = 1,110 ms, FOV = 220×165 mm2, matrix size= 192×144 in the axial plane and 24 5-mm thick slices). Six white matter tracts or regions of interest (ROI) were examined, which included the anterior limb of internal capsule (ALIC), posterior limb of internal capsule (PLIC), genu and splenium of the CC, superior longitudinal fasciculus (SLF), and the inferior fronto-occipital fasciculus (IFOF). The ROIs were identified using the JHU White Matter Atlas in standard space, projected into the subjects’ space, and used to extract ME-T₂ data (Figure 2). Regularized NNLS was used to fit the ME-T₂ data with 200 logarithmically spaced T₂ values from 10-2,000 ms. The IE T₂ component was defined as T₂ values between 40 and 200 ms. A repeated measures general linear model (RM-GLM) was fit to the data. In the GLM, geomT_2IEW was the dependent variable, ROI - a six-level within-subject factor, sex - a between-subject factor, and age, as well as age², centered at the sample mean - continuous independent variables. Bootstrapped 95% confidence intervals, using 5,000 samples, were generated for coefficients corresponding to age and age² terms.

Results

After discarding non-significant ROI × age² × sex and ROI × age × sex interactions (F <1; F =1.764, p =.127, respectively), we found significant interactions: ROI × age [F(5,405) = 14.559, p < .001] and ROI × age² [F(5,405) = 5.997, p < .001]. In the post-hoc analysis (see Figures 3 and 4) both age² and age were initially entered in the regression for each ROI, and dropped if non-significant.

Conclusions

We observed quadratic effects of age in the ALIC and PLIC and linear age differences in the SLF, IFOF and the Genu and Splenium of CC. Given the heterogeneous effect of age on geomT_2IEW, future studies should directly evaluate the relationships between geomT_2IEW of the IE compartment, AxD and AxS. Because this sample consisted of adolescents, young and older adults it is likely that age differences in geomT_2IEW reflect age differences in AxD and AxS. Nonetheless, we show that the geomT_2IEW provides characterization of subcortical WM microstructure that is complementary to MWF, and that the magnitude of age differences varies across WM tracts.

Acknowledgements

This work was supported by a grant from the National Institute on Aging [R37-AG011230] to Naftali Raz and the Lycaki-Young Funds from the State of Michigan.