White matter (WM) evolves continually throughout human lifespan and therefore experiences changes in its biological state. The concept of biological state is very general, complex, and therefore challenging to define. MRI offers a wide array of measurable parameters that can be used as biological state descriptors. Some of these parameters are macroscopic geometrical measures –e.g. volume, spatial distribution, shape—, and other measures depict physico-physiological properties of tissue at the voxel level. Most of these WM measures have been shown to be age dependent.

Geometrical measures are commonly referred to as structural measures and these can be generated with any medical imaging modality, for example with high resolution x-ray CT, with the caveat that the exquisite contrast between white matter and gray matter (GM) as well as the high soft tissue sensitivity afforded by MRI makes it the preferred modality for structural WM characterization.

Physico-physiological tissue properties that are unique to MRI include: a) measures of intravoxel tissue hydration –e.g. 1H-proton density--, b) measures of microscopic and macroscopic motion –e.g. diffusion and perfusion parameters--, and c) measures of proton interactions with the microscopic magnetic environment as manifested in the relaxation times of the longitudinal magnetization (T1) and the transverse magnetization (T2 and T2*).

A large body of literature with a focus on WM evolution as a function of age is mounting at the intersections of several scientific disciplines including neuroscience, neurobiology, neuroimaging, and MRI science --image acquisition and image processing--. Early qualitative observations of age-dependent WM-to-GM contrast in T1- and T2-weighted images were followed by quantitative studies of age-dependent morphology and qMRI studies stablishing preliminary baseline age dependencies of (some) qMRI parameters throughout the human lifespan.

The purpose of this paper is to review publications that demonstrate aging patterns of healthy WM, from young adulthood to end of life, with structural-MRI at the macroscopic level and with qMRI at the spatial scale of the imaging voxel. In addition, simple model functions are used to graph idealized structural and qMRI aging patterns.

This section succinctly describes key imaging tools and methodologies used for generating structural measures and separately for generating qMRI measures of WM as functions of age.

Structural MRI

Subject-by-Subject Sequential Methods

Useful for structural MRI of WM are high spatial resolution T1-weighted pulse sequences such as three-dimensional magnetization prepared rapid acquisition with gradient echoes (3D MP-RAGE) (1) and dual-echo fast (turbo) spin echo (DE-FSE) (2) pulse sequences. Both of these can generate very high spatial resolution and high image quality datasets with marked WM-to-GM contrast. In addition, an increasing number of multispectral (PD, T1, T2 or T2*) qMRI pulse sequences that have been described in the literature (3-11) that can be used to generate T1 and PD maps. These maps offer high WM-to-GM contrast and are therefore useful for automated WM segmentation, which is the first step for studying WM in isolation from GM and CSF. The whole brain WM segment can be further divided into subsegments. Segmental volumes can be generated with pixel counting algorithms. In summary, sequential patient-by-patient WM, structural MRI is a viable yet laborious technique for studying the volumetric age dependencies of WM and its primary partitions as functions of age.

Statistical Methods

Alternatively, voxel-based morphometry (VBM) is a fundamentally different neuroimaging analysis technique for detecting focal differences in brain anatomy; it uses the approach of statistical parametric mapping (12,13). The VBM image processing protocol begins with registering every brain to a common template thus reducing most of the large differences in brain anatomy among subjects. Secondly, the brain images are smoothed so that each voxel represents the average of itself and its neighbors. Finally, the image volume is compared across brains at every voxel. VBM can be sensitive to various artifacts, which include misalignment of brain structures, misclassification of tissue types, differences in folding patterns and in cortical thickness.

Quantitative MRI

A qMRI technique consists of a pulse sequence and a matching mapping algorithm; these two components are designed to complement each other. The pulse sequence is used to generate the directly acquired (DA) images that have various weightings to the qMRI parameter of interest. The qMRI algorithm(s) are used to generate the quantitative MR images, which are known as qMRI maps.

Operational Principles and qMRI Algorithms

The principles of qMRI stem from the so-called pixel value equation, which is derived by solving the Bloch-Torrey-Stejskal equations (14-16) applied to the specific pulse sequence used to acquire the DA images. In a second step, a matching qMRI-mapping algorithm is constructed, by combining and/or transforming the pixel value equation. Three qMRI mapping principles can be readily formulated:

1) The principle of weighting reversal with pixel value calibration for quantifying the proton density (PD) of the liquid pool.

2) The principle of differential weighting for quantifying the liquid pool qMRI parameters. This is very useful because it leads to qMRI maps that are devoid of imperfections and artifacts in the directly acquired images; for example, spatial variations of the receiver sensitivity profile can be factored out by performing pixel value ratio. Hence, maps of T1, T2, T2*, and ADC are in general less prone to errors potentially caused from spatially dependent coil imperfections.

3) The principle of differential weighting for the semisolid pool parameters.

Hierarchical Organization of qMRI parameters

The qMRI parameters of the liquid pool can be grouped into three categories, specifically: 1) The proton density (PD), which quantifies the density of mobile 1H-protons in tissue and is independent of the strength B0. 2) The kinetic parameters of tissue diffusion and perfusion: a) apparent diffusion coefficient (ADC), b) the diffusion tensor (D), which leads to FA and MD by solving the eigenvalue problem for every voxel. 3) The relaxation times are indicative of the magnetic interactions between 1H-protons with each other and with the random magnetic fields present in the microenvironment.

The age dependencies of WM volume and qMRI parameters are illustrated in the following sections using approximate phenomenological mathematical expressions gained or developed from the literature. These functions of age are based on different cohorts of “normal” subjects studied with various MRI technologies that have not yet been fully standardized; presented results must therefore be regarded as approximate age tendencies.

Volumetry

As shown in the idealized graphs in Fig. 1 (17-21), the total intracranial volume (ICV) increases rapidly from birth to approximately eight years of age and remains constant thereafter. The total volume of brain matter (BM=WM+GM) of the combined cerebrum and cerebellum, also increases rapidly and peaks in early adolescence; it decreases gradually and approximately linearly during adulthood and quadratically during senescence. The volumetric age dependencies of males and females are very similar with smaller adult ICV by about 150cm³.

Proton density (mobile water)

PD is difficult to map accurately because a qMRI technique based on the principle of differential weighting has not yet been devised. Therefore, PD maps are modulated by coil sensitivity profile and in addition require the use of a calibration tissue/substance, which can be internal –intraventricular CSF is commonly used as PD reference-- or external –typically a vial with water attached to the subject’s forehead--. Nevertheless for brain tissue in general and WM in particular, an alternative and indirect approach based on the empirical relationship proposed by Fatouros et al. (22) is available (see equation insert in Fig. 2). This PD-T1 relationship has been assumed here as valid for all ages. Hence, graphs of Fig. 2 must be interpreted as an approximation that needs confirmation with a direct PD mapping technique. To the best of our knowledge, such measurements have not been published.

Diffusion tensor (DTI)

A recent single-institution cross-sectional study demonstrated four distinct periods of life based on the age dependence of the peak ADC values of brain tissue (WM+GM), as derived from sequential patient-by-patient histogram analysis (21). It was shown that the peak ADC could be accurately modeled as the sum of two decaying exponential functions plus a constant term and a linearly increasing term. The two decaying exponential terms model the very fast ADC drop from birth to two years of age (maturation period) and the slower ADC decrease from two-to-twenty years (developmental period). The constant term is dominant during adulthood from twenty-to-sixty at which age, the positive linear term models the very gradual ADC increase observed during senescence.

The literature reporting WM age dependencies of DTI derived parameters is without doubt the most abundant in the context of qMRI research, a small sample includes the following references (23-34) and an exhaustive listing is beyond the scope of this short syllabus/paper. A recent DTI study showed (see Fig. 3) that twelve white matter tracts followed similar general fractional anisotropy (FA) age dependent trajectories (35). FA increased from childhood to adulthood, reached a peak in early to mid-adulthood and then decreased during later adulthood at a rate slower than the initial increases. Such peak FA value ages are fiber specific (see Fig. 3: peak age for FA of the corpus callosum genu is approximately ten years younger than that of the anterior limb). In contradistinction, the mean diffusivity MD age trends (not shown) were opposite, decreasing initially, reaching a minimum, and then increasing at a slower rate.

Perfusion

Reported age-dependent qMRI measures of the normal brain perfusion are relatively scarce mainly because the current clinical MRI standard involves an injection of a gadolinium based contrast agent, which have some associated risks. Hence, the perfusion technique of choice for research with healthy volunteers as well as with vulnerable populations has been arterial spin labeling (ASL), which provides noninvasively measures of cerebral blood flow (CBF) (36-43). Global CBF measures as a function of age reveal (37) a pronounced decrease in CBF for GM and WM at the onset of adulthood (e.g. twenty years of age) as well as linear decline in CBF (with a slope of−0.38% per year) for cortical GM and relative CBF stability for subcortical GM.

The longitudinal magnetization relaxation time (T1)

Of all known qMRI parameters, T1 is the clearest WM-to-GM discriminator because of the marked T1 shortening effects of myelin (6,20,44-47). As shown in Fig. 4, at birth the T1s of WM and GM are very high and approximately equal. The T1 values of both tissues decrease and begin to differentiate during maturation. WM-to-GM differentiation continues during the developmental period peaking at about twenty years of age. During adulthood the WM-to-GM T1 difference is comparatively stable but decreases slowly. During adulthood and senescence, the T1s of WM and GM increase and decrease respectively. Interestingly, by extrapolating these T1 age curves, the GM curve crosses the WM curve at about 120 years of age (Fig. 4). This coincides approximately with the reported ages of the oldest human beings; the longest unambiguously documented human lifespan is 122 years, and 164 days (https://en.wikipedia.org/wiki/Oldest_people).

The transverse magnetization relaxation time (T2)

As shown in Fig. 5, the T2 vs. age curves of healthy GM and WM show four periods of life (20,48): the maturation and developmental periods, during which T2 decreases exponentially and reaches minimum GM and WM T2 values at the onset of adulthood (20 years of age). T2 values remain comparatively stable during adulthood and slowly increase during senescence. The T2 differences between GM and WM are much smaller than for T1 because the T2 relaxation effects are dominated by the pure spin-spin dephasing interactions, the strength of which is much less myelin specific: see insert formula in Fig. 5.

White matter is a highly specialized and morphologically complex tissue of the central nervous system with distinct MR properties. WM may be thought of as a grid of electrical conductors for action potentials, and as a network of roads for the bi-directional transport of biological molecules and organelles between the neuron cell bodies and synapses (49). It primarily consists of axons and glial cells. Axons, which can be myelinated or unmyelinated, transmit action potentials and therefore information from one region of the brain to another, and between the brain and lower brain centers. Axons are long and thin semi-cylindrical projections of neurons that conduct electrical impulses away from the neuron's cell body. Glial cells are non-neuronal cells that maintain homeostasis, produce myelin, and provide support and protection to neurons in the central and peripheral nervous systems.

Structural-MRI and qMRI can provide uniquely informative insights into the macrostructure and microstructure of white matter. In the present review of healthy WM, the age dependencies of global volumetric measures and of several key qMRI parameters –specifically, PD, D, CBF, T1, and T2-- have been compiled. Additionally, these results were exemplified using idealized mathematical models gained from the scientific literature.

The body of experimental and theoretical MR information accumulated over the past three decades on this multifaceted topic is already very large and informative; it is also rapidly growing, still incomplete, and highly compartmentalized. The development of new image acquisition technologies, including powerful MRI hardware and highly specialized pulse sequences for the Human Connectome project (50,51), as well as the emergence of axon diameter mapping techniques (e.g. CHARMED, AxCaliber, ActiveAx (52-54)) at the forefront of neuroimaging, will be instrumental for establishing more complete white matter aging norms (e.g. (55)). Availability of such norms is indispensable for characterizing WM biological states in health and most importantly, in disease.

Herein, published MRI aging patterns of healthy white matter from young adulthood to end of life have been reviewed. At the confluence of these WM studies, four distinct periods of life become clearly demarcated: 1) maturation from birth to two years, 2) development from two to twenty years of age, 3) adulthood from twenty to fifty-five years, and 4) a very long senescence period from fifty-five to end of life, which as documented, can extend up to 122 years.