Whole-brain white matter volume shrinkage is small, estimated at no more than 2% per decade²⁷ (but see ^31,44). Yet postmortem investigations, in contrast to macrostructural approaches, have provided fruitful leads to age-related changes in white matter by revealing microstructural white matter compromise⁴⁵ including degradation of microtubules, myelin and axons^46-48. In vivo, white matter microstructure is detectable with DTI, which has successfully revealed evidence of regional white matter disruption in normal ageing even in regions appearing normal on bulk volume imaging.

Quantification of DTI Data

Water molecules in the brain are in constant Brownian motion. Although movement of water protons affects conventional structural imaging to some degree, DTI allows quantification of this microscopic movement within each voxel, providing a depiction of local water molecular movement in terms of the preferred orientation of the motion. Along a white matter fibre, the path of a water molecule is constrained and this movement is called ‘anisotropic’, because diffusion along the long axis of a fibre (axial diffusion) is greater than diffusion across the fibre (radial diffusion⁴⁹). Conversely, regions with few or no constraints imposed by physical boundaries such as CSF in the ventricles allow water movement to be random in every direction; this is therefore ‘isotropic’. Disruption of white matter microstructure detectable with DTI can reflect compromise of cytoskeletal structure, myelin or axon density^50,51.

With advancing age and senescence, a consensus maintains a decline in fractional anisotropy in brain white matter^56,60-65 (but see ^66,67). Furthermore, consistent evidence supports an anterior-posterior gradient of fractional anisotropy decline with age^{57,58,60,61,65,68-74} that has been confirmed in a monkey model of ageing⁷⁵. TBSS likewise substantiates an anterior-posterior gradient, namely by identifying lower fractional anisotropy values in the frontal, parietal and temporal lobes, corpus callosum (particularly the genu and body) and the internal capsule⁷⁶.

In complement to white matter fractional anisotropy decline in ageing, diffusivity in white matter increases^{58,65,67,77-80}. As with fractional anisotropy, profile analysis reveals that the primary locus of the age effect is in anterior white matter⁸⁰. This pattern of age-related decline of fractional anisotropy and increase in diffusivity mirrors postmortem observations with predominant age-related findings in anterior white matter systems^45,81.

The degree to which the diffusion orientation of a voxel is similar to its neighbours, or in other words, coherence of diffusivity measures between voxels^82,83, serves the conceptual basis for quantitative fibre tracking^55,84-87. Connectivity and coherence between different brain regions on vector and fibre tracking maps is readily apparent on visual inspection⁸⁸. Methods for quantitative analysis of structural connectivity of white matter include fibre-tract trajectories^89-91 and maps of the degree of ‘alignment’ among neighbouring vectors, on a voxel-to-voxel basis, resulting in a measure of intervoxel coherence⁸². Advantages of quantitative fibre tracking over focal ROI analysis include the ability to measure the entire extent of a fibre tract and to characterize its integrity along its full extent.

Longitudinal studies of normal, healthy men and women across the adult age range document normative age-related, DTI-detectable changes in brain white matter that are requisite for interpreting DTI data from individuals with neuropsychiatric disorders and other conditions affecting the brain. However, longitudinal studies have their own problems with respect to measurement reliability within and across research centres, selective attrition, and hardware and software system modifications and drift¹⁰⁰. For example, a study of five normal adults imaged five times each resulted in 5% across-subject variation and a mean variation of 2.3 ± 1.2 SD% in diffusivity⁷⁸. Another study measured the reliability of global and regional fractional anisotropy and diffusivity measurements in ten healthy young adults, imaged three times on two different 1.5-T scanners made by the same manufacturer¹⁰¹. When reliability was measured on a voxel-by-voxel basis or on a slice-by-slice basis, fractional anisotropy and diffusivity values were equivalently and significantly higher within than across scanners. In addition to reliability of a common measurement approach, substantial differences can arise when different analysis schemes are applied to the same data set¹⁰².

The functional ramifications of the DTI metrics have been regularly verified with observations of correlations between regionally specific low fractional anisotropy or high diffusivity and poor cognitive^{61,62,71-73,106,107} or motor⁶⁸ test performance in humans and also a monkey model of ageing⁷⁵. Lower whole-brain fractional anisotropy and higher mean diffusivity correlated with working memory performance^106,108. The alternating finger-tapping task, a test of interhemispheric information transfer, revealed a relationship between splenium and parietal pericallosal white matter fractional anisotropy and finger-tapping output that was selective to the alternating condition, requiring interhemispheric motor coordination, and not unimanual conditions, free of transcallosal information exchange⁶⁸. Task switching^61,109 and lower scores on tests of executive functions^63,109 correlated with low fractional anisotropy and greater diffusivity in anterior brain regions; lower verbal fluency scores correlated with lower fractional anisotropy in central samples of white matter⁶¹.

DTI studies indicate widely varying fractional anisotropy across brain regions, with lower anisotropy and higher in diffusivity in older than younger healthy adults, that is prominent in the anterior supratentorial white matter but minimal to none in posterior cortical and corticospinal tracts. A caveat to this generalization is that selective deletion of uniformly orientated white matter fibres from a tissue sample of crossing fibres, as can occur with Wallerian degeneration, can cause abnormally high anisotropy¹¹⁰. Therefore, interpretation of fractional anisotropy results requires guidance by knowledge of the underlying regional architecture, especially white matter¹¹¹ and supplemented with non-human primate studies, affording the opportunity for in vivo examination and postmortem verification of imaging results (cf.⁸⁸).

Furthermore, the diameters of axons, for example in the corpus callosum ranging from 0.4 to 5μm, with most less than 1μm, means that a cross-section of the genu of the human corpus callosum contains approximately 400 000 fibres per mm² and a single DTI voxel using relatively high resolution (2 × 2 × 2 mm³) for current standards could contain 1.6 million fibres in a cross-section¹¹². Consequently, caution must be used when interpreting the anatomical meaning of observations made with radiological measures, which are coarse on the size scale of the underlying fibres.

Finally, the diffusion-weighted signal can also be influenced by iron accumulation with age, and the resulting effect is different among deep grey matter structures and white matter and needs to be considered, especially when using DTI to characterize structures known to accumulate iron (in vivo¹¹³, postmortem¹¹⁴) or in conditions, such as stroke, that produce high ferritin deposition (cf.¹¹⁵).