Differential aging of cerebral white matter in middle-aged and older adults: A seven-year follow-up

Abstract

The few extant reports of longitudinal white matter (WM) changes in healthy aging, using diffusion tensor imaging (DTI), reveal substantial differences in change across brain regions and DTI indices. According to the "last-in-first-out" hypothesis of brain aging late-developing WM tracts may be particularly vulnerable to advanced age. To test this hypothesis we compared age-related changes in association, commissural and projection WM fiber regions using a skeletonized, region of interest DTI approach. Using linear mixed effect models, we evaluated the influences of age and vascular risk at baseline on seven-year changes in three indices of WM integrity and organization (axial diffusivity, AD, radial diffusivity, RD, and fractional anisotropy, FA) in healthy middle-aged and older adults (mean age=65.4, SD=9.0years). Association fibers showed the most pronounced declines over time. Advanced age was associated with greater longitudinal changes in RD and FA, independent of fiber type. Furthermore, older age was associated with longitudinal RD increases in late-developing, but not early-developing projection fibers. These findings demonstrate the increased vulnerability of later developing WM regions and support the "last-in-first-out" hypothesis of brain aging.

Keywords: Age; DTI; Diffusivity; Hypertension; Longitudinal; White matter.

Figure 1

Distribution of ages-at-measurement and intervals between measurement occasions for the 38 participants. Each row indicates a single participant, with lines denoting time between measurement occasions, and symbols representing each separate measurement occasion as follows: triangle = 1^st occasion; diamond = 2^nd occasion; circle = 3^rd occasion; square = 4^th occasion. See the Method section 2.1 for details on the differences in assessment intervals.

Figure 2

DTI processing pipeline. A. Common space determination across all measurement occasions is performed using the non-diffusion weighted b0 images which are hierarchically registered using a 6-degree of freedom, rigid registration, first between waves 1–2 and waves 3–4 to determine halfway, common space between these images; the resulting halfway 1–2 and 3–4 image spaces are subsequently registered and the halfway space from these provides a final, common space between all images. Native space diffusion tensors are fit to the data using FSL’s dtifit procedure, with the – save_tensor flag to retain the tensor components. Next, using the vecreg command in FSL, the saved tensor components are subsequently rotated and refit to the common space. B. Masking of white matter hyperintensities (WMH) and cerebrospinal fluid (CSF) is performed using FAST segmentation of the native b0 image into 6 classes based on voxel intensity. The two segmented masks that correspond primarily to WM are summed and binarized into a WM mask that omits areas of apparent WMH, CSF, and image noise. C. The FA images that were refit into common space in A., above are used as the basis for TBSS processing which nonlinearly transforms the data into standard space, and registers them to the 1mm³ FMRIB58 FA image before calculating mean FA across the sample and determining a mean WM skeleton. Using the tbss_deproject routine and inverse transformation matrices from the initial registration steps the JHU WM labels and tractography atlases (Mori et al., 2005; Wakana et al., 2007) are subsequently transformed back to individual native space, but are confined to the mean WM skeleton. Individual ROIs from the atlas are segmented and the DTI maps (FA, AD, RD), masked for WMH/CSF are sampled from those regions across participants.

Figure 3

Longitudinal changes over 1 to 4 occasions of measurement. The plots are organized by three fiber types (columns, left to right): Association, commissural, and projection fiber regions, and by three indices of WM micro-organization (rows, top to bottom): Axial diffusivity, radial diffusivity, and fractional anisotropy.