Skip to main content
NCBI home page
Human Brain Mapping logoLink to Human Brain Mapping
. 2025 Nov 14;46(16):e70403. doi: 10.1002/hbm.70403

Diffusion MRI Microstructure Markers of Changes in the Human Brain Across the Lifespan Using Constrained Spherical Deconvolution and Fixel‐Based Techniques

Benjamin T Newman 1,2,✉, John D Van Horn 2,3, T Jason Druzgal 1
PMCID: PMC12616505  PMID: 41235744

ABSTRACT

Understanding how the brain develops, matures, ages, and declines is one of the fundamental questions facing neuroscience. This study uses 3‐tissue constrained spherical deconvolution (3T‐CSD) and more recently developed fixel‐based derivatives to examine the relationship between brain diffusion microstructure and chronological age. These metrics are able to quantify signal fraction measurements at the voxel‐wise level from six different tissue microenvironments found in the brain: extracellular free water, intracellular isotropic, intracellular anisotropic, fiber density and cross‐section, the g‐ratio, and aggregate conduction velocity. This study analyzes the Nathanial Kline Institute's Rockland cohort, a large‐scale community sample of brain MRI data across the lifespan (409 subjects, ages 5–85 years). Microstructural measurements were taken in a number of structures throughout the white matter, subcortical gray matter, and lobar cortical regions while additionally evaluating lateral differences in microstructural measurements. The general trajectory of signal fraction measurements was a positive relationship with age and extracellular signal fraction and g‐ratio, a negative relationship between age and intracellular isotropic signal fraction, fiber density and cross‐section, and conduction velocity, and an inverted U‐shaped trajectory for the intracellular anisotropic signal fraction. In individual sub‐areas, these trends tended to still be present, with some notable exceptions. However, there were large differences in microstructure measurements between individual structures, including significant lateral differences between hemispheres for each of the subcortical gray matter structures and for each of the cortical regions. These results demonstrate that 3T‐CSD and fixel‐based metrics are able to describe age‐related change across the brain and lifespan. By using a healthy population cohort this study can be used as a point of comparison for analysis of microstructure changes in the presence of pathology or with behavior. Finally, the detailed analysis of lateralized ROI results can inform diffusion microstructure studies examining cortical and subcortical regions.

Diffusion microstructure compartment and subject age across the lifespan. Whole brain tissue has the highest proportion of ICI signal fraction early in the lifespan; as development occurs the ICA signal fraction reaches its highest peak, and in later life the ECI signal fraction increases as the brain tissue declines.

graphic file with name HBM-46-e70403-g001.jpg

1. Introduction

Throughout the human lifespan, the structure of the brain changes dynamically in response to internal and external factors. A wealth of volumetric MRI studies have shown that early in life the brain dramatically increases in size while in older individuals the brain shows a remarkable decrease in volume (Driscoll et al. 2009; Fotenos et al. 2005; Good et al. 2001; Group 2012; Pomponio et al. 2020; Raz et al. 2005). Two recent studies using extremely large numbers of subjects have firmly established a general trajectory for changes in brain volume, with peaks in volume occurring around 12 years of age (for subcortical GM, surface area, and cerebrum volume) (Bethlehem et al. 2022). However these trajectories show comparatively little movement following development (especially after puberty, but with the greatest rates of change occurring before 2 years of age) (Dima et al. 2022). This leads to a long and slightly downward plateau before decline begins in earnest in advanced age, with especially noticeable increases in ventricular volume and declines in cortical GM volume (Bethlehem et al. 2022; Dima et al. 2022). This lifespan trajectory is interwoven with complex processes during development and decline, such as hormone levels following the onset of puberty in early adolescence (Juraska and Willing 2017) and changes in vascular health during aging (Jefferson et al. 2010). Concurrent with changes in gross anatomy are changes occurring at the cellular level, such as extensive synaptic pruning in development (Feinberg 1982; Selemon 2013) and excessive atrophy and cellular death in mild cognitive impairment and progressive decline (Stephan et al. 2012). Understanding the trajectory of these changes in the brain, as well as understanding what processes drive these changes, are important for research into neurological disease.

Recently, measures of cellular brain microstructure derived from diffusion MRI (dMRI) have proliferated in number and become more commonly applied in the analysis of development and pathology (Pasternak et al. 2009; Zhang et al. 2012). Each model has particularly defined functions applied to the raw diffusion signal, such as the separation of freely diffusing water from tissue in the bi‐tensor free water elimination model (Pasternak et al. 2009) the ‘ball and stick’ model of isotropic and anisotropic diffusion (Assaf and Basser 2005), or the detailed delineation of cellular somas from neurites (dendrites, axons, etc.) and extracellular water available with NODDI (Zhang et al. 2012). While these models have seen widespread application in recent years, most are still reliant on tensor‐based diffusion models or require sufficiently detailed acquisitions to accurately quantify all of their output metrics. Our lab has recently developed a model of diffusion microstructure known as 3‐tissue constrained spherical deconvolution (3T‐CSD) (Newman, Dhollander, et al. 2020). 3T‐CSD relies on more complex models of diffusion signal calculated using spherical harmonics (Tournier et al. 2007) and is able to measure both isotropic and anisotropic intracellular microstructure and extracellular water compartments derived from single‐shell clinical quality acquisitions with a high level of reliability and stability (Dhollander and Connelly 2016; Newman, Untaroiu, and Druzgal 2020). The ability of 3T‐CSD and other quantitative diffusion models to measure multiple microstructure compartments provides a far greater level of detail at the voxel‐wise level compared to well‐established volumetric MRI and voxel‐based morphometry measurements. These advantages mean that 3T‐CSD is well‐positioned to investigate changes in brain cellular microstructure across the lifespan. Within a specific age cohort, deviations from established age‐related trajectories might indicate accelerated or slowed aging or development, depending on the context being studied. Microstructural measurements may have increased sensitivity by detecting changes in cellular architecture before they become apparent as changes in volume or density, particularly the signal from extracellular freely diffusing cerebrospinal fluid (CSF) infiltrating into brain tissue (Blair et al. 2019).

3T‐CSD has several other advantages in examining changes in the brain across the lifespan. Importantly the intracellular isotropic component is flexible and nonspecific, and is intentionally not restricted to any specific cellular type or location within the brain. Changes in this signal fraction compartment are thought to arise from either neuronal soma or glial cell population changes. Voxels with the highest intracellular isotropic signal fraction occur in the typically defined gray matter areas (GM) in the cortex and subcortical structures. However nearly all voxels have a non‐trivial contribution from each signal fraction compartment, including in axonal areas (so called ‘deep’ white matter (WM)) where neuronal somas are unlikely to be found (Khan et al. 2021; Mito et al. 2020). This makes the intra‐ and extracellular isotropic signal fraction compartments excellent markers for changes in axonal white matter areas. 3T‐CSD can detect meaningful shifts from very high (~90% or more) intracellular anisotropic signal fraction in healthy axonal areas toward increasing intra‐ and extracellular isotropic signal fraction measurements in early development (Pietsch et al. 2019) or in the presence of pathology (Khan et al. 2021). But this specific pathological context is not always available and as 3T‐CSD begins to be applied to the study of broader conditions and developmental situations additional context is necessary to formulate hypotheses regarding the physiological change to the cellular microstructure that is observed in human studies. This context is especially necessary to interpret results from studies observing the relationship between cellular microstructure and variables with undetermined effects on neuronal physiology, such as pubertal hormones.

With the trajectory of volumetric MRI well established it is necessary to explore the cellular microstructure that underlies volumetric change with much greater detail, both with advanced dMRI models such as 3T‐CSD and with detailed analysis of specific regions of the brain at multiple scales and between hemispheres. The goal of this study is to analyze a large number of subjects from a publicly available population cohort in order to provide a lifespan trajectory for 3T‐CSD measurements of cellular microstructure from multiple areas of the brain. Establishing the relationship between chronological age and microstructural metrics will establish normal reference ranges and trajectories that are essential before investigating abnormal populations. This study will additionally investigate differences in cortical and subcortical areas and lateralized effects to inform future studies of significant regional and laterality differences that must be accounted for.

2. Methods

2.1. NKI/Rockland Cohort Data Acquisition

Data for this study was obtained from the Nathaniel Kline Institute's Rockland Study (NKI/Rockland) cohort (Nooner et al. 2012). NKI/Rockland is a large‐scale community sample of participants with ages across the lifespan gathered in Rockland County, New York. Rockland County was selected in part because its diverse ethnic and economic demographics resemble those of the United States as a whole, which aids generalizability to the broader population. A wide array of physiological, cognitive, genetic and neuroimaging assessments are collected and publicly released (available at: http://fcon_1000.projects.nitrc.org/indi/enhanced/index.html) (Nooner et al. 2012).

In this study, the 409 subjects from the NKI/Rockland study that passed quality control and were publicly available at the time proceeded to analysis. These subjects ranged in age from 6 to 85 years old (mean 42.67 ± 20.79 S.D.). There were 144 male and 265 female participants with average ages of 36.09 ± 21.22 S.D. and 46.25 ± 19.68 S.D., respectively. These ages were not equivalently distributed between sexes (Kolmogorov–Smirnov; D = 0.276, p < 0.001) which precluded performing analysis to compare trajectories between sexes.

2.2. Code Availability

All analysis performed in this study was done using freely available, open‐source software and resources. Code for generating the 3T‐CSD microstructural maps analyzed in this study are available at https://3tissue.github.io/ and the code for generating conduction velocity and g‐ratio estimates is available at https://github.com/btn6sb/Conduction_Velocity.

2.3. Anatomical Image Analysis

T1‐weighted MPRAGE sequences were collected on a Siemens MAGNETOM TrioTim with a field strength of 3T. Voxel size was 1.0 × 1.0 × 1.0 mm3 with an FOV of 256 × 256 × 200 mm3, a TR of 2500 ms and a TE of 3.5 ms. T1w images were analyzed using the recon‐all function in Freesurfer and the total brain volume calculated from surface fitting without ventricle volume (BrainSegNotVentSurf, mm3) was used as each subject's total brain volume. The average b = 0 from the preprocessed diffusion image was used as a T2‐weighted image and the T1‐weighted and T2‐weighted images were processed as described in the MICA‐MNI pipeline (Cruces et al. 2022) involving N4 bias correction from ANTs (Avants et al. 2009) and rescaling intensity between 0 and 100. The resulting T1w/T2w ratio image was used as the myelin volume fraction for the calculation of g‐ratio and aggregate conduction velocity as described previously (Newman, Jacokes, et al. 2024).

2.4. Diffusion Image Processing and Analysis

dMRI data from 521 subjects was acquired using a Siemens MAGNETOM TrioTim 3T scanner with an isotropic voxel size of 2.0 × 2.0 × 2.0 mm3, TE = 85 ms and TR = 2400 ms; 9 b = 0 s/mm2 images and 127 gradient directions at b = 1500 s/mm2. These images were processed through an automated pipeline with several manual quality control steps as follows. Each diffusion image set was analyzed using SS3T‐CSD (Dhollander and Connelly 2016; Jeurissen et al. 2014) implemented in the open‐source software MRtrix and Mrtrix3Tissue (Dhollander and Connelly 2016; Tournier et al. 2019). Several preprocessing steps utilized FSL (Jenkinson et al. 2012; Smith et al. 2004). Diffusion images were denoised (Veraart et al. 2016), corrected for Gibbs ringing (Kellner et al. 2016), susceptibility distortions (Smith et al. 2004), subject motion (Andersson et al. 2016), and eddy currents (Andersson and Sotiropoulos 2016). All images were upsampled to 1.3 × 1.3 × 1.3 mm3, and skull‐stripping was performed using the Brain Extraction Tool (Jenkinson et al. 2012). Response functions were generated using an unsupervised method (Dhollander et al. 2016) from each tissue type: white matter‐like, gray matter‐like, and cerebrospinal fluid‐like, using the mean spherical harmonic which was fit to the diffusion signal in voxels that were in the top 0.5% highest FA when fit using a typical diffusion tensor model for the white matter‐like response function, For the isotropic gray matter‐like and cerebrospinal fluid‐like response functions a signal decay metric was used to identify voxels that represent optimum thresholds when examining voxel intensity change between the b = 0 and the outer b‐value shell (here b = 1500 s/mm2). These response functions were then fit to the diffusion signal, represented by a spherical harmonic function, using an iterative linear least squares function termed single‐shell constrained spherical deconvolution (SS3T‐CSD) to generate fiber orientation distributions (FODs) (Dhollander and Connelly 2016; Jeurissen et al. 2014). Signal fractions were calculated directly from the FODs using 3‐tissue constrained spherical deconvolution (3T‐CSD), a method that uses the spherical harmonic coefficient of the FODs (a single value for gray matter‐like and cerebrospinal fluid‐like tissues, as these are isotropic, and the maximum harmonic coefficient from each WM fiber in the FOD) to measure cellular microstructure within each voxel fitting into intracellular anisotropic (ICA, white matter‐like), intracellular isotropic (ICI, gray matter‐like), and extracellular isotropic (ECI, cerebrospinal fluid‐like/Free Water) compartments (Newman, Dhollander, et al. 2020). This is performed by summing each of these spherical harmonic coefficients to 1 to calculate signal fractions, that is, the total signal that arises from each of the separate tissue compartments per voxel. This method has several distinct advantages; notably it is derived in an unsupervised manner from the diffusion image itself, which expands its potential applicability, and it requires only a single shell to resolve highly accurate multiple fiber populations within a single voxel. 3T‐CSD has further been demonstrated to be highly stable and reliable in measurement testing across several months (Newman, Untaroiu, and Druzgal 2020). The resulting ICA, ICI, and ECI signal fractions are flexible in application and context‐dependent in their interpretation, though they predominate in the white matter, gray matter, and ventricles/outer brain parenchyma, respectively. While each represents a specific type of tissue, the clinical context of these types can be highly varied, especially in regard to the ICI and ECI signal compartments. As a representative of freely diffusing water and extracellular fluid, the ECI signal fraction has been noted in atrophy due to aging, but also as a hallmark of tissue damage, such as in white matter hyperintensities (Khan et al. 2021; Mito et al. 2018; Newman et al. 2022) or edema (Dhollander et al. 2017; Khan et al. 2020). The ICI signal fraction compartment is more varied and less readily confirmable via other MRI imaging modalities compared to extracellular water. However evidence has suggested that the ICI signal fraction is primarily intracellular space that predominates in gray matter neuronal cell somas, but can also indicate microglial activation in response to injury (Newman, Untaroiu, and Druzgal 2020), neuroinflammation (Newman, Forman, et al. 2024), or glial cell conformational change, such as during myelin development (Newman et al. 2023). The ICA signal fraction is largely confined to highly anisotropic WM fibers. For additional discussion on interpreting the 3T‐CSD microstructural compartments see Dhollander et al. (2017). Whole brain measurements were calculated from native space signal fraction compartments. These metrics used a smaller whole brain mask that excluded the ventricles and subarachnoid space via an ECI threshold that restricted the voxels to only those containing a majority signal fraction from brain tissue compartments and not extracellular fluid.

White matter FODs were used to generate a cohort‐specific FOD template image from 50 individuals between ages 32 and 38 years old, and each subject's individual WM FOD image was registered to this template. Manual quality control of registration to the cohort‐specific FOD template was performed by visual inspection, excluding any subjects with obvious distortions or extreme shears. Following this procedure 409 subjects remained in the study for analysis. An FOD template in stereotaxic space was created using the NTU‐DSI‐122 template (https://www.nitrc.org/projects/ntu‐dsi‐122/), extracting the equivalent single b‐value shell and it was registered to the cohort‐specific template generated from this study (Hsu et al. 2015). This allowed for two separate atlases to be warped into the cohort template space to measure signal fraction averages within specific regions of interest.

2.5. Fixel‐Based Analysis

Fixels were derived from the WM FODs and used to generate three voxel‐wise metrics: aggregate g‐ratio, aggregate conduction velocity (ACV), and fiber density and cross‐section (FDC). FDC is a frequently used fixel‐based metric that provides fiber bundle specific density of intra‐axonal space and, using the Jacobian determinant perpendicular to the fiber direction of the registration to the cohort‐specific FOD template, is modified by the cross‐sectional area (Raffelt et al. 2015, 2017). The FDC was recentered around 1 to enforce positivity and was transformed from fiber‐wise to voxel‐wise via intra‐voxel summation. In this study we have summed the fiber density and cross‐section on a voxel‐wise basis prior to application as the axonal volume fraction for aggregate g‐ratio and aggregate conduction velocity analysis. The T1w/T2w ratio was utilized as the myelin volume fraction while the weighted sum FDC was used as the axonal volume fraction. Obtaining the AVF and MVF allows for the calculation of g‐ratio and conduction velocity, which retain their original definitions as formulated by Rushton (Berman et al. 2019; Drakesmith et al. 2019; Hodgkin and Huxley 1952; Mohammadi and Callaghan 2021; Rushton 1951) and were calculated on a voxel‐wise basis according to previously described methods (Berman et al. 2019; Mohammadi and Callaghan 2021; Newman, Jacokes, et al. 2024; Rushton 1951; Stikov et al. 2015). The g‐ratio is the ratio of the outer axon diameter (axon plus myelin sheath) to the inner axon diameter (intra‐axonal space alone), while conduction velocity represents the speed that action potentials are estimated to traverse the axonal fiber bundles in a particular voxel, and has been validated in humans using EEG (Coleman et al. 2025; Newman, Jacokes, et al. 2024). It is typically thought that lower g‐ratio values and higher ACV values are considered more desirable as they represent improved functioning and faster axonal action potential transmission. It should further be noted that for these fixel‐based measures of axonal quality, the metrics will appear lower than Rushton's ideal g‐ratio value of 0.6 in cortical ROIs due to the contribution of voxels with little to no axonal signal the further away from the deep WM. Values should thus be interpreted in relation to anatomically similar ROIs and not necessarily between ROIs as the quality of innervation and quantity of innervation are intertwined when taking the mean voxel‐wise values. An illustrated example of each of the microstructural metrics analyzed in this study is available in Figure 1.

FIGURE 1.

FIGURE 1

Open in a new tab

Illustration of the 6 different primary neuroimaging metrics analyzed in this study. The 3T‐CSD metrics (left; ICA, ICI, and ECI signal fractions) are derived directly from the FODs (only WM FODs are displayed above, but each compartment is derived from the coefficient of the isotropic GM FOD and CSF FOD for ICI and ECI signal fractions, respectively). The WM FODs are used to derive fixels, typically for fixel‐based analysis and the fixel‐based metrics (right). Here we use the FDC as the axonal volume fraction, and combined with the myelin volume fraction from T1w/T2w ratio images, we are able to calculate the voxel‐wise aggregate g‐ratio and then aggregate axonal conduction velocity.

2.6. Quality Control (QC)

QC was visually performed on all subjects and subjects with poor registration to the cohort‐specific template or low‐quality masks were excluded from the study cohort. Initial masking was performed on the diffusion images themselves which, alongside the wide range of ages involved in this study and the diffusion FOD‐based registration method employed, may have contributed to the high rate of rejected subjects in QC.

2.7. Regions of Interest

In order to cover a wide range of brain regions including cortex and deep white matter both the JHU‐DTI based ICBM‐DTI‐81 white matter atlas (available as part of FSL, hereafter referred to as the JHU WM atlas) including 48 ROIs (Hua et al. 2008; Jenkinson et al. 2012; Mori et al. 2005; Wakana et al. 2007), and the Destrieux atlas including 164 ROIs (Destrieux et al. 2010) were warped into the study space. The cortical ROIs from the Destrieux atlas were summarized in a whole cortical ribbon ROI, as well as a cerebellum GM, and a total subcortical GM ROI. The 48 ROIs from the JHU WM atlas were also combined to provide a summary of the WM skeleton.

Subcortical GM ROIs were utilized from the Destrieux atlas; the nucleus accumbens, amygdala, caudate, hippocampus, putamen, and thalamus were included without alteration. The Destrieux atlas cortical ROIs were also separately combined into one of 8 lobar cortical regions per hemisphere as specified by the atlas and packaged in Freesurfer (Destrieux et al. 2010; Fischl 2012). The cortical ROIs separately encompassed frontal, insular, limbic, motor, occipital, parietal, sensory, and temporal cortices. Fixel‐based metrics were not assessed within the cortical ROIs as the FDC and ACV displayed very low variability and values across the lifespan in the cortex.

2.8. Statistical Approach

Signal fraction values from each of the three tissue compartments (ICA, ICI, and ECI) were averaged within each of the ROIs for use in analysis. The average values for the three tissue compartments were plotted against age for each of the 409 subjects. Total age‐related change was examined both across these age groups and within each phase by the plotting of locally weighted scatterplot smoothing (loess) lines of best fit. This was performed in R using the default settings of a second order polynomial and span of 0.75 (meaning the localized slope reflects the closest 75% of data points). As a model‐free way of displaying general trends loess lines were selected for descriptive purposes to show overall change in each signal fraction and facilitate between‐ROI comparisons. LOESS was selected for the visualization in order to be more attentive toward local swings in the age trajectory. The linear models below the LOESS curves are intended to provide information on the typical age‐related slope in age groups that are commonly used for analysis so that other studies may be better able to interpret the directionality of effects and interactions with these microstructural metrics. To provide more precise information on rates of change simple linear models were calculated predicting signal fraction from subject age and the slope of that relationship was displayed for each ROI. For ease of comprehension as well as to delineate periods of particular interest across the lifespan (i.e., development, late age‐related decline) we have divided the subject cohort into 4 different life phases: (1) developmental phase before 20 years of age, (2) early adulthood phase between 20 and 40 years of age, (3) late adulthood phase between 40 and 60 years of age, and (4) senescence phase greater than 60 years of age. This division allows for an overview of the entire lifespan as well as comparisons to studies that only feature cohorts within a particular age range (e.g., development or advanced age).

The lateralized approach taken in this study for each of the subcortical and cortical ROIs allows for the comparison for each signal fraction between the left and right hemispheres. Measurements were taken from the left and right ROIs and compared across the lifespan using pairwise t‐tests to compare each lateralized region within subjects. p values were subsequently adjusted using a Benjamini and Hochberg correction for multiple comparisons.

3. Results

3.1. Whole Brain Shows Developmental Shift From ICI to ICA, Followed by Increasing ECI in Aging

As each whole brain signal fraction was individually collected in an equivalent native space they are displayed, colored by age, as a ternary plot to illustrate change across the brain, across the lifespan (Figure 2). It was observed that whole brain metrics begin with a higher percentage of signal from the ICI compartment early in development, then progress toward a greater signal fraction contribution from ICA in adulthood. Signal fraction contribution from ECI does not appear to change greatly during this time. During aging however, ECI increases at the expense of both ICA and ICI as the brain declines and atrophy occurs.

FIGURE 2.

FIGURE 2

Open in a new tab

Ternary plot showing whole brain relationship between each of the 3T‐CSD tissue compartments and subject age. Blue lines divide subjects into five 20%‐tile groups based on age to illustrate change in signal fraction values across the lifespan. Whole brain tissue has the highest proportion of ICI signal fraction early in the lifespan, as development occurs the ICA signal fraction reaches its highest peak, and in later life the ECI signal fraction increases as the brain tissue declines. As the 3T‐CSD signal fractions sum to 1 on a voxel‐wise basis changes in tissue cellular architecture induce relative changes in the metrics rather than absolute measurements of tissue compartments.

3.2. Large Anatomical Subregions

Loess lines of best fit (A) and the linear slope of change per year within each age group (B) are presented for the initial four generalized regions in Figure 3 for 3T‐CSD metrics and in Figure 4 for fixel‐based metrics. While the specific proportion of ICA, ICI, and ECI is different in each region, which is to be expected given that each region is largely defined by the predominance of a specific cell type, there is largely a consistent pattern of change in each across the lifespan. Matching the whole brain results, ECI showed low levels in development before 20 years of age; ECI even decreased in white matter areas covered by the JHU WM atlas and in subcortical GM structures. ECI remained largely flat in the early adulthood phase, between 20 and 40 years of age, but began to rise sharply during late adulthood between 40 and 60 years of age, with the sharpest increase occurring in the subcortical structures and cortical ribbon. In senescence greater than 60 years of age all major regions showed increasing levels of ECI reflecting widespread brain atrophy, loss of cellular integrity, and decline. The ICI signal fraction largely declined in all regions of the brain across the lifespan, though with important exceptions. The WM skeleton in the senescence phase showed a surprising increase in ICI signal fraction, especially considering that this is an area with very little ICI signal fraction. The cerebellum also remained relatively static in ICI signal fraction measurements until the senescence phase. The ICA signal fraction however, displays an inverted U shape across the lifespan, first rising greatly during development, slowing during the early adulthood phase, then increasing in decline during the early and senescence phases. The cerebellum is relatively consistent in ICA signal fraction in the development and late adulthood phases but otherwise each region is consistent in the trajectory of ICA measurements.

FIGURE 3.

FIGURE 3

Open in a new tab

Charts displaying the lifespan trajectories of each 3T‐CSD metric in 4 large anatomical brain subareas. The relationship between age and signal fraction is displayed either across the whole lifespan (A, C, and E for ECI, ICI, and ICA, respectively) or as the slope of the linear relationship during a limited age range (B, D, and F for ECI, ICI, and ICA, respectively). Overall, signal fraction compartment trajectories were relatively consistent between brain subareas, with a positive relationship between ECI and age, a negative relationship between ICI and age, and a positive relationship between ICA and age in the first two life phases, followed by a negative relationship between ICA and age in the later two life phases. There were however several interesting exceptions to this general trend. Within the WM skeleton there was a negative relationship between age and ECI signal fraction in the developmental phase, likely due to increased myelination reducing the extracellular space available for freely diffusing water. On the opposite end of the lifespan, in the senescence phase the WM skeleton displays an increase in ICI signal fraction.

FIGURE 4.

FIGURE 4

Open in a new tab

Charts displaying the lifespan trajectories of each fixel‐derived metric in 4 large anatomical brain subareas. The relationship between age and each metric is displayed either across the whole lifespan (A, C, and E for FDC, g‐ratio, and ACV, respectively) or as the slope of the linear relationship during a limited age range (B, D, and F for FDC, g‐ratio, and ACV, respectively). Overall, signal fraction compartment trajectories were relatively consistent between brain subareas, but with a greater drop in cerebellar cortex g‐ratio in the middle plateau phase and a greater drop in ACV in the WM skeleton in the senescence phases.

The fixel‐derived metrics of g‐ratio and ACV showed considerably more variability in the 20–40 age range than the relatively placid FDC. This is likely due to the influence of the T1w/T2w ratio as a myelin measurement, which can be noisy especially when a dedicated T2‐weighted image was not acquired. Interestingly, unlike the ICA signal fraction the FDC did not show a gradual U‐shaped pattern, declining immediately after spiking in the adolescent phase for the remainder of the lifespan. ACV and g‐ratio both improved during the middle plateau phase but had trends against speed in other age ranges.

3.3. Subcortical GM

Proceeding to a greater level of detail, different components of the subcortical GM were independently measured in each hemisphere and plotted similarly to the larger anatomical areas (Figure 5 for 3T‐CSD and Figure 6 for fixel‐based metrics). A majority of the ROIs in this sample followed the general pattern observed for the area as a whole, but several important deviations can be observed, both in absolute differences in tissue composition as well as different trajectories across the lifespan. For example, the caudate nucleus early in life has a relatively low ECI signal fraction closer to the more internally located nucleus accumbens, but by middle age this structure contains a higher ECI signal fraction than the hippocampus, and the rate of increase then continues to accelerate until the structure has the highest average signal fraction of any subcortical structure late in life. Other structures demonstrate substantial deviations from the trajectory of the subcortex as a whole, most notably the putamen, which appears largely resistant to large age‐related increases in ECI signal fraction, but in early development also the amygdala and hippocampus decrease in ECI signal fraction more akin to the WM skeletal trajectory than the cortical trajectory. The basal ganglia appear to have the least ECI signal fraction and have matching trajectories, but the ventral striatum nucleus accumbens and the dorsal striatum putamen show different trajectories in the other signal fraction compartments indicating different responses to age‐related change.

FIGURE 5.

FIGURE 5

Open in a new tab

Charts displaying the lifespan trajectories of each 3T‐CSD metric in 6 subcortical gray matter structures, including both left and right structures. The relationship between age and signal fraction is displayed either across the whole lifespan (A, C, and E for ECI, ICI, and ICA, respectively) or as the slope of the linear relationship during a limited age range (B, D, and F for ECI, ICI, and ICA, respectively). Trajectories largely follow the pattern established for the whole subcortical GM structure in Figure 2; however, some notable deviations: For the ECI signal fraction both amygdala, hippocampal, and amygdala drive the slight decrease in ECI signal fraction before 20 years of age, while caudate and thalamus surprisingly show steep increases during this time period and the hippocampus continues decreasing until reaching the late adulthood phase. In the ICI tissue compartment the thalamus, caudate, and putamen follow the whole subcortical trajectory, but the amygdala, nucleus accumbens, and to a near degree the hippocampus, are more resistant to age‐related decline. In the ICA tissue compartment the putamen is likewise resilient to age‐related decline.

FIGURE 6.

FIGURE 6

Open in a new tab

Charts displaying the lifespan trajectories of each fixel‐based metric in 6 subcortical gray matter structures, including both left and right structures. The relationship between age and signal fraction is displayed either across the whole lifespan (A, C, and E for ECI, ICI, and ICA, respectively) or as the slope of the linear relationship during a limited age range (B, D, and F for ECI, ICI, and ICA, respectively). The g‐ratio and ACV trends in the thalamus had large swings in value, likely due to noisy T2 estimates during early and late senescence. After an initial positive increase in early life, FDC tended to decrease throughout the lifespan in all subcortical areas examined. G‐ratio and ACV displayed characteristically opposite directional findings throughout the lifespan.

In the fixel‐based metrics, there was typically a flatter trajectory than the 3T‐CSD metrics. The nucleus accumbens in particular had very low or almost no axonal innervation, leading to very low values across all metrics. Interestingly, there was a large lateral difference between the right and left caudate, with the right having elevated FDC, g‐ratio, and ACV. Given the way in which these metrics are derived from the FDC it appears likely that this was the driving factor.

While the trajectory of each subarea largely appears more or less extreme in the ECI signal fraction, in the ICI and ICA signal fraction compartments some of the subareas deviate entirely from the established trajectory. The amygdala, nucleus accumbens, and to a near degree the hippocampus, are more resistant to age‐related decline and do not display substantial ICI signal fraction decrease in the early adulthood and late adulthood phases. These three regions approach the highest age included in this study with approximately 15 percentage points higher (a greater than 50% increase over) ICI signal fraction than the putamen and caudate which begin development at a reasonably similar ICI signal fraction. The hippocampus follows a trajectory more akin to the cerebral cortex, but this is not relevant for the deep internal structure of the amygdala and nucleus accumbens. While each of the subcortical GM structures shows a wide range of ICA signal fraction compositions, the largest outlier is the putamen, which shows an increase in ICA signal fraction throughout the lifespan even as every other region begins to decline in the early and senescence phases. Toward the end of the lifespan the putamen even approaches within 10 percentage points of the ICA signal fraction within the WM skeleton.

When these results were corrected for each subject's total brain volume (Figure S1) the ECI signal fraction increase was accelerated with an even steeper trajectory late in life, showing that the ECI signal fraction was increasing beyond total brain volume loss in late life. The ICI and ICA signal fractions, and fixel‐based metrics were largely flattened by total brain volume correction, with smaller changes observed throughout the lifespan. The change was sufficient to cause the ICI signal fraction to actually relatively increase in late senescence in the amygdala, nucleus accumbens, and hippocampus, suggesting that these areas become predominantly more ICI during this life phase. When corrected for total brain volume the fixel‐based metrics were largely flattened across the lifespan.

3.4. Subcortical GM Show Significant and Consistent Lateralized Differences

Results from lateralized comparisons are presented in Table 1. There was a high degree of lateralization in nearly all signal fractions with 15 out of 18 possible signal fraction/ROI combinations showing a significant difference between right and left hemispheres after correction for multiple comparisons. The ECI signal fraction was significantly higher in the right hemisphere in 4 of the 6 ROIs, as was the ICA signal fraction. The ICI signal fraction was higher in each left hemisphere ROI and was significantly so in all but the amygdala. The putamen, thalamus, and amygdala were observed to have the greatest laterality (at least one signal fraction compartment with a mean difference greater than 2%). The hippocampus displayed the smallest average difference between left and right hemispheres, but was still significantly different for 2 of the 3 signal fractions, displaying how prevalent lateralized differences in structure are in the brain.

TABLE 1.

Lateral differences between subcortical GM structures across the lifespan.

ROI Tissue type Higher signal fraction Mean difference t Adjusted p
N. Accumbens ECI Left 0.0005 0.359 0.720 n.s.
ICI Left 0.0106 2.634 0.012 *
ICA Right 0.0100 2.139 0.040 *
Amygdala ECI Right 0.0232 19.596 < 0.001 ***
ICI Left 0.0010 0.554 0.614 n.s.
ICA Left 0.0226 11.264 < 0.001 ***
Caudate ECI Right 0.0049 3.073 0.003 **
ICI Left 0.0156 10.608 < 0.001 ***
ICA Right 0.0106 7.158 < 0.001 ***
Hippocampus ECI Right 0.0029 2.517 0.016 *
ICI Left 0.0066 3.995 < 0.001 ***
ICA Right 0.0031 1.768 0.088 n.s.
Thalamus ECI Right 0.0169 21.483 < 0.001 ***
ICI Left 0.0324 32.87 < 0.001 ***
ICA Right 0.0155 12.574 < 0.001 ***
Putamen ECI Left 0.0021 5.632 < 0.001 ***
ICI Left 0.0578 22.783 < 0.001 ***
ICA Right 0.0598 22.542 < 0.001 ***
Open in a new tab

Note: Pairwise t‐tests were used to compare each lateralized region within subjects and p values were adjusted using a Benjamini and Hochberg correction for multiple comparisons. The ICI signal fraction was higher in each left hemisphere ROI and was significantly so in all but the amygdala. The putamen, thalamus, and amygdala were observed to have the greatest laterality (at least one signal fraction compartment with a mean difference greater than 2%). * p < 0.05, ** p < 0.01, *** p < 0.001.

3.5. Lobar Cortical Regions Show Similar Trajectories Across the Lifespan but Different Absolute Measurements and Inconsistent but Significant Lateral Differences

The cerebral cortex was then subdivided into eight regions of cortex separated by left and right hemisphere as defined by the Destrieux atlas and measurements were made as before (Figure 7). The signal fraction measurements in these regions generally followed the overall region trajectory displayed in Figure 3 but as with the subcortical GM, there were some exceptions. The ECI signal fraction increased in every ROI in both the early and senescence phases. In the ICI and ICA signal fractions the insular and limbic cortices displayed markedly higher solid tissue values than other cortical regions while still following relatively similar trajectories. There was a significant degree of hemispheric laterality between the left and right sides of the brain (Table 2), with little consistent differences between the signal fraction compartments. For the fixel‐based metrics, the different regions of the JHU WM atlas were instead used as they are more relevant to the intra‐axonal calculations used to derive the metrics (Figure 8).

FIGURE 7.

FIGURE 7

Open in a new tab

Charts displaying the lifespan trajectories of each 3T‐CSD metric in 8 cortical ROIs. The relationship between age and signal fraction is displayed either across the whole lifespan (A, C, and E for ECI, ICI, and ICA, respectively) or as the slope of the linear relationship during a limited age range (B, D, and F for ECI, ICI, and ICA, respectively). Several deviations from the average trajectory occur in the middle age phases in the cortex in the insular and limbic cortices in the ICI signal fraction but otherwise declines across the lifespan. The frontal cortex in the developmental age phase shows little to slightly positive change in the ICA signal fraction but otherwise follows a consistent upward trajectory during the initial two life phases followed by decreases in the next two life phases.

TABLE 2.

Lateral differences between cortical GM regions across the lifespan.

ROI Tissue type Higher signal fraction Mean difference t Adjusted p
Frontal Cortex ECI Right 0.0052 6.861 < 0.001 ***
ICI Right 0.0088 9.783 < 0.001 ***
ICA Left 0.0111 12.376 < 0.001 ***
Insular Cortex ECI Left 0.0121 16.632 < 0.001 ***
ICI Left 0.0207 15.656 < 0.001 ***
ICA Right 0.0339 30.966 < 0.001 ***
Limbic Cortex ECI Left 0.0021 4.365 < 0.001 ***
ICI Right 0.0012 1.896 0.067 n.s.
ICA Right 0.0036 4.665 < 0.001 ***
Motor Cortex ECI Left 0.0088 5.579 < 0.001 ***
ICI Left 0.0009 0.742 0.4583 n.s.
ICA Right 0.0021 1.425 0.1692 n.s.
Occipital Cortex ECI Left 0.0124 14.573 < 0.001 ***
ICI Left 0.0196 7.591 < 0.001 ***
ICA Right 0.0078 4.227 < 0.001 ***
Parietal Cortex ECI Right 0.0116 10.953 < 0.001 ***
ICI Right 0.0137 15.809 < 0.001 ***
ICA Right 0.0089 12.574 < 0.001 ***
Sensory Cortex ECI Right 0.0312 16.461 < 0.001 ***
ICI Left 0.0186 13.752 < 0.001 ***
ICA Left 0.0124 8.283 < 0.001 ***
Temporal Cortex ECI Right 0.0008 1.232 0.2282 n.s.
ICI Right 0.0209 14.775 < 0.001 ***
ICA Right 0.0308 19.764 < 0.001 ***
Open in a new tab

Note: Pairwise t‐tests were used to compare each lateralized region within subjects and p values were adjusted using a Benjamini and Hochberg correction for multiple comparisons. All cortical regions showed some significant degree of lateralization with the insular and temporal cortices each having a mean difference of greater than 2% ICA and ICI signal fraction measurements between left and right hemispheres. Conversely the limbic and motor cortices each had a mean difference of less than 0.5% excepting the ECI signal fraction in the motor cortex. Interestingly, as opposed to the subcortical areas which demonstrated relatively consistent right/left biases for the signal fraction compartments, there was a more balanced distribution of right/left biases for each of the signal fractions. * p < 0.05, ** p < 0.01, *** p < 0.001.

FIGURE 8.

FIGURE 8

Open in a new tab

Charts displaying the lifespan trajectories of each fixel‐based metric in all 48 JHU WM ROIs. The relationship between age and signal fraction is displayed either across the whole lifespan (A, C, and E for ECI, ICI, and ICA, respectively) or as the slope of the linear relationship during a limited age range (B, D, and F for ECI, ICI, and ICA, respectively). Trajectories largely were consistent across all WM structures and consistent with the overall WM ROI with some notable exceptions: The anterior corona radiata had the highest overall FDC and also the biggest variation in ACV, whereas the fornix had the lowest g‐ratio. G‐ratio was much more variable than FDC in middle age, which appeared to carry over to ACV. This suggests that much of the variability in ACV and WM development across age is due to changes in myelination and not in axonal diameter.

4. Discussion

This study examined a large population cohort using an advanced diffusion microstructure analysis technique to describe the relationship between chronological age and changes in brain cellular structure. The slope of these changes was steeper during the developmental phase before 20 years of age and in the senescence phase after 60 years of age with more gradual changes occurring between 20 and 60 years of age in the middle of the adult phases of the lifespan. The pattern established at the whole brain level, a shift toward increasing ICA signal fraction in the first half of the lifespan largely coupled with decreasing ICI signal fraction (as they sum to 1, any increase/decrease in one signal fraction compartment must be coupled with equivalent increases/decreases in the others), subsequently gives way to a decline pattern later in life characterized by increasing ECI signal fraction coupled with decreasing ICA signal fraction. In the fixel‐based analysis, FDC initially increased in adolescence, but there was a lagging decrease in g‐ratio and increase in ACV in the 20–40 age range. In early and late senescence both FDC and ACV dropped while g‐ratio increased, indicating lower axonal diameters and slower conduction velocity are a component of the aging process.

When the WM skeleton, cerebral cortex, cerebellular cortex, and subcortical GM were parcellated from the cohort it was observed that the ECI signal fraction increased precipitously with increasing age after 40 in all 4 ROIs but during development, it was stable in the subcortical GM and cerebellum, declined in the WM skeleton, and increased in the cortex. Despite three of the ROIs being thought of as predominantly GM areas there was observed a widespread negative correlation with age throughout the lifespan. During the developmental phase the ICA signal fraction increased dramatically in each of the 4 ROIs, matching the well‐established trajectories observed in volumetric data from this time period (Lebel and Beaulieu 2011; Lebel and Deoni 2018). Interestingly, the increased axonal innervation of the cortex during development can be observed in this trajectory as well as in the subcortical GM structures. This axonal innervation has been suggested to shift the WM/GM boundary line in volumetric studies (Natu et al. 2019) and this study certainly supports the cortex and subcortical GM structures becoming more similar to the WM skeleton in microstructural profile as development progresses. An interesting exception to this trend is in the frontal cortex, which has a flat slope for the relationship between ICA signal fraction and age during the developmental phase. This study also showed far more dramatic changes throughout the lifespan compared to large‐scale volumetric MRI studies (Bethlehem et al. 2022; Dima et al. 2022). Change in brain metrics in this study was not solely confined to early development and late senescence, and there were very large and distinct trajectories for each tissue compartment across different ROIs, suggesting that age‐related change is a highly complex process with different, distributed effects in different brain structures.

Amongst the fixel‐based metrics after correcting for total brain volume, the thalamus showed the largest midlife increases in g‐ratio of all the subcortical structures, with a corresponding decrease in conduction velocity. However, this essentially reverted to middle age values by the most extreme end of the age range included in this study. It is interesting to note that the thalamus also showed essentially no change across the lifespan in FDC, suggesting that myelin or very small diameter fiber changes drive the observed differences in g‐ratio and ACV. An additional interesting observation is that the right putamen had much higher ICA, FDC, and ACV across the lifespan with only a modest increase in g‐ratio, suggesting the diameter of axons innervating the right putamen may be much larger than the diameter of axons innervating the left putamen.

4.1. Addressing Directionality and Variation in G‐Ratio

The aggregate g‐ratio results described in this study are directionally inconsistent with previously published lifespan studies of neuroimaging data that described aggregate g‐ratio having a U‐shaped trajectory across the lifespan (higher values in adolescence and old age) (Bouhrara et al. 2021; Cortina et al. 2022). In this paper the necessity of using the T1‐weighted MPRAGE from an already collected dataset enforces constraints on the distribution of MVF values that are not easily resolved by post‐collection calculations or processing. Attempts by other groups have been made to derive a constant, k, that alters the range of myelin quantification into theoretically acceptable values, but the lack of a ground truth measure in any particular dataset, age range, or pathology, or even across the voxels of the brain containing different cellular composition, forces the introduction of further assumptions that may not be accurate (e.g., not all under 30 year‐olds have ideally healthy WM with no variations) (Cercignani et al. 2017). Calibrating the g‐ratio to the theoretical optimum of 0.6 was also demonstrated to lower metric reliability (Ellerbrock and Mohammadi 2018).

Importantly, the histological account of g‐ratio being lowest and closest to the optimal value within deep WM is frequently inverted in many MRI‐derived aggregate g‐ratio datasets, with the highest g‐ratio values occurring in deep WM and the lowest g‐ratio values spatially located at the WM/GM boundary and in the periphery (Campbell et al. 2018; Ellerbrock and Mohammadi 2018). This is suggestive of the AVF signal more heavily weighting the g‐ratio metric, and for example using some acquisition schemes has found that within multiple sclerosis lesions the g‐ratio is lower than in normal appearing healthy WM, which is opposite to the intuitive understanding of g‐ratio (Campbell et al. 2018). In this study, we have displayed results without calibration or specialized myelin imaging acquisitions, which appear to have led to a more AVF‐dependent g‐ratio and conduction velocity metrics, though with greater variation across the lifespan than the AVF metric by itself (FDC). We include these metrics here because they may prove valuable additional formulations (i.e., the near universally available T1w/T2w ratio for MVF and FDC for AVF) for interrogating regional differences or relative changes in the interplay between axonal fiber density and myelin. It is also essential for the correct interpretation of these metrics as they continue to proliferate in usage that precisely these deviations from an unknown ground truth are understood and appreciated by the neuroimaging community.

It was further observed that accounting for total brain volume largely flattened the inverted‐U shaped trajectory and reduced variability across the lifespan (Figure S1) suggesting that partial volume effects contributed. This would suggest that studies using aggregate g‐ratio and ACV should erode ROIs to reduce partial voluming, include total brain volume corrected values, or control for total brain volume in statistical models evaluating these metrics.

4.2. Laterality

Laterality was examined and was found to have a highly significant effect across multiple regions but this was typically without a consistent direction. The putamen, thalamus, amygdala, insular cortex, and temporal cortex were observed to have the greatest laterality (at least one signal fraction compartment with a mean difference greater than 2%). The hippocampus displayed the smallest average difference between left and right hemispheres, but the ECI and ICI signal fraction compartments were still significantly different, recapitulating previous work that found a significantly increased ECI signal fraction in the right hippocampus (Newman, Dhollander, et al. 2020). In the fixel‐based analysis, the caudate showed the largest lateral difference but the majority of subcortical regions did not show large deviations in lateral values.

4.3. Comparison to Other Microstructural Methodologies and Histology

Other advanced dMRI models have examined the relationship between their quantitative measurements and age across the lifespan. In a recent study NODDI outputs of isotropic volume fraction, roughly comparable to 3T‐CSD's ECI signal fraction, increased throughout the lifespan across the whole brain. Meanwhile intracellular volume fraction displayed an inverted U‐shaped pattern similar to 3T‐CSD's companion ICA signal fraction (Beck et al. 2021). Another NODDI study however found conflicting results using the orientation dispersion index, which does not have a direct 3T‐CSD parallel, with results from Nazeri et al., finding that orientation dispersion decreased across the lifespan in 4 cortical lobe ROIs, and Beck et al., finding that orientation dispersion increased across the whole brain (Beck et al. 2021; Nazeri et al. 2015). The current study found much more complicated relationships between many of the cortical lobes and age across the lifespan, and while this study did not directly compare results from different quantitative dMRI models, it is notable that NODDI and other advanced models did not perform better than traditional DTI metrics at predicting brain age (Beck et al. 2021).

There may be an expressed relationship in this study between signal fraction (particularly the relative proportion of ICA and ICI signal fractions) and neuronal cellular density as determined by histology. In particular the limbic cortex is known as having a particularly low neuronal density in humans (Semendeferi et al. 1998) and in this study was surprisingly found to have the highest level of ICI signal fraction (along with the insular cortex) throughout the lifespan. This suggests that a relatively low density of neurons may not cause a high ECI signal fraction but will instead be represented as an overly high ICI signal fraction, perhaps due to increased glial cell population or larger levels of intracellular space in comparison to areas of high neuronal and dendritic density, such as the cerebellum. This ROI did subsequently follow similar trajectories to other ROIs with differing levels of neuronal density and arrangement, but it may nonetheless hold promise as an effect replicable in pathological conditions affecting neuronal density. Decreased neuronal density in the prefrontal cortex has been implicated in major depressive and bipolar disorders, suggesting 3T‐CSD may be a potential mechanism for tracking alterations in these patients (Cotter et al. 2001; Rajkowska et al. 2001).

Other disorders may result from increased neuronal density including schizophrenia which may present a different signal fraction profile that is able to be detected using 3T‐CSD (Benes et al. 1986; Selemon et al. 1995). The developmental component of schizophrenia may allow for early detection if adolescents begin deviating from age‐related trajectories similar to those established in this study. Another dMRI microstructural analysis technique, termed free water elimination, has suggested that increased extracellular water may also contribute to the development of schizophrenia (Pasternak et al. 2012), however this method does not measure both extracellular and isotropic intracellular tissue compartments and thus makes an imperfect point of comparison to the 3T‐CSD technique used here.

4.4. Limitations

A known feature of the aging brain that was not accounted for in this study is the development of white matter hyperintensities (WMH). Previous work has demonstrated that developed and developing WMH are detectable using 3T‐CSD and have increased ECI and ICI signal fraction with correspondingly reduced ICA signal fraction compared to normal appearing areas of the WM skeleton (Mito et al. 2020). The presence of WMH not accounted for in this study is due to several reasons, including that the location of the WMH in the brain was largely determinative of its microstructural properties, with WMH located proximal to the ventricles having a greater amount of ECI signal fraction than WMH located within the deep WM (Mito et al. 2020). Additionally, there is evidence that 3T‐CSD is sensitive to developing WMH that subsequently appear after cerebral infarction (Khan et al. 2021). This created a situation where removing tissue identified via independent means as a WMH might contribute noise to estimates of subjects in the senescence phase and would be difficult to account for in the whole lifespan trajectories. Another feature not accounted for was subject sex due to the imbalance of women to men especially in the late adulthood and senescence phases, but this is undoubtedly a subject worth investigating for differences between sexes in future lifespan work.

5. Conclusion

In conclusion, these results describe the relationship between brain microstructure and age and illustrate how dMRI correlates of cellular microstructure changes throughout the lifespan. These results provide a powerful illustration of the dramatic shifts in microstructure metrics during development and aging and highlight the potential of 3T‐CSD and fixel‐based metrics as a tool to study development and aging. Establishing patterns of microstructure change across the lifespan will aid in studying and interpreting microstructure change in the presence of pathology. Studies analyzing dMRI microstructure in developing or aging cohorts should address shifts in anisotropic and isotropic signal in development and dramatically increased levels of extracellular water in aging.

Supporting information

Figure S1: Plots detailing each of the mean subcortical gray matter 3T‐CSD and fixel‐based metrics evaluated in this study corrected for each subject's total brain volume. In order, these are (A) ECI, (B) FDC, (C) ICI, (D) g‐ratio, (E) ICA, and (F) ACV. Loess lines are calculated identically to other charts presented in this paper.

HBM-46-e70403-s001.tiff (7.1MB, tiff)

Newman, B. T. , Van Horn J. D., and Druzgal T. J.. 2025. “Diffusion MRI Microstructure Markers of Changes in the Human Brain Across the Lifespan Using Constrained Spherical Deconvolution and Fixel‐Based Techniques.” Human Brain Mapping 46, no. 16: e70403. 10.1002/hbm.70403.

Data Availability Statement

The data that support the findings of this study are available in NKI/Rockland Cohort at https://rocklandsample.org/. These data were derived from the following resources available in the public domain: NKI/Rockland I, https://rocklandsample.org/.

References

  1. Andersson, J. L. , Graham M. S., Zsoldos E., and Sotiropoulos S. N.. 2016. “Incorporating Outlier Detection and Replacement Into a Non‐Parametric Framework for Movement and Distortion Correction of Diffusion MR Images.” NeuroImage 141: 556–572. [DOI] [PubMed] [Google Scholar]
  2. Andersson, J. L. , and Sotiropoulos S. N.. 2016. “An Integrated Approach to Correction for Off‐Resonance Effects and Subject Movement in Diffusion MR Imaging.” NeuroImage 125: 1063–1078. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Assaf, Y. , and Basser P. J.. 2005. “Composite Hindered and Restricted Model of Diffusion (CHARMED) MR Imaging of the Human Brain.” NeuroImage 27, no. 1: 48–58. [DOI] [PubMed] [Google Scholar]
  4. Avants, B. B. , Tustison N., and Song G.. 2009. “Advanced Normalization Tools (ANTS).” Insight Journal 2, no. 365: 1–35. [Google Scholar]
  5. Beck, D. , de Lange A.‐M. G., Maximov I. I., et al. 2021. “White Matter Microstructure Across the Adult Lifespan: A Mixed Longitudinal and Cross‐Sectional Study Using Advanced Diffusion Models and Brain‐Age Prediction.” NeuroImage 224: 117441. [DOI] [PubMed] [Google Scholar]
  6. Benes, F. M. , Davidson J., and Bird E. D.. 1986. “Quantitative Cytoarchitectural Studies of the Cerebral Cortex of Schizophrenics.” Archives of General Psychiatry 43, no. 1: 31–35. [DOI] [PubMed] [Google Scholar]
  7. Berman, S. , Filo S., and Mezer A. A.. 2019. “Modeling Conduction Delays in the Corpus Callosum Using MRI‐Measured g‐Ratio.” NeuroImage 195: 128–139. [DOI] [PubMed] [Google Scholar]
  8. Bethlehem, R. A. , Seidlitz J., White S. R., et al. 2022. “Brain Charts for the Human Lifespan.” Nature 604, no. 7906: 525–533. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Blair, J. C. , Newman B. T., and Druzgal T. J.. 2019. “Evaluating Lifespan Tissue Structure: Comparing CSD Signal Fraction and VBM Grey Matter Density.” In Proceedings of the ISMRM 27th General Meeting .
  10. Bouhrara, M. , Kim R. W., Khattar N., et al. 2021. “Age‐Related Estimates of Aggregate g‐Ratio of White Matter Structures Assessed Using Quantitative Magnetic Resonance Neuroimaging.” Human Brain Mapping 42, no. 8: 2362–2373. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Campbell, J. S. , Leppert I. R., Narayanan S., et al. 2018. “Promise and Pitfalls of g‐Ratio Estimation With MRI.” NeuroImage 182: 80–96. [DOI] [PubMed] [Google Scholar]
  12. Cercignani, M. , Giulietti G., Dowell N. G., et al. 2017. “Characterizing Axonal Myelination Within the Healthy Population: A Tract‐By‐Tract Mapping of Effects of Age and Gender on the Fiber g‐Ratio.” Neurobiology of Aging 49: 109–118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Coleman, C. R. , Nance M. G., Jacokes Z., et al. 2025. “Structural Determinants of Signal Speed: Estimated Axonal Latency and Its Multimodal Validation During Face Processing in Autism.” bioRxiv (2025): 3.
  14. Cortina, L. E. , Kim R. W., Kiely M., et al. 2022. “Cerebral Aggregate g‐Ratio Mapping Using Magnetic Resonance Relaxometry and Diffusion Tensor Imaging to Investigate Sex and Age‐Related Differences in White Matter Microstructure.” Magnetic Resonance Imaging 85: 87–92. 10.1016/j.mri.2021.10.019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Cotter, D. , Mackay D., Landau S., Kerwin R., and Everall I.. 2001. “Reduced Glial Cell Density and Neuronal Size in the Anterior Cingulate Cortex in Major Depressive Disorder.” Archives of General Psychiatry 58, no. 6: 545–553. [DOI] [PubMed] [Google Scholar]
  16. Cruces, R. R. , Royer J., Herholz P., et al. 2022. “Micapipe: A Pipeline for Multimodal Neuroimaging and Connectome Analysis.” NeuroImage 263: 119612. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Destrieux, C. , Fischl B., Dale A., and Halgren E.. 2010. “Automatic Parcellation of Human Cortical Gyri and Sulci Using Standard Anatomical Nomenclature.” NeuroImage 53, no. 1: 1–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Dhollander, T. , and Connelly A.. 2016. “A Novel Iterative Approach to Reap the Benefits of Multi‐Tissue CSD From Just Single‐Shell (+ b= 0) Diffusion MRI Data.” In Proc ISMRM, 24, 3010 .
  19. Dhollander, T. , Raffelt D., and Connelly A.. 2016. “Unsupervised 3‐Tissue Response Function Estimation From Single‐Shell or Multi‐Shell Diffusion MR Data Without a Co‐Registered T1 Image.” ISMRM Workshop on Breaking the Barriers of Diffusion MRI 5, no. 5. [Google Scholar]
  20. Dhollander, T. , Raffelt D., and Connelly A.. 2017. “Towards Interpretation of 3‐Tissue Constrained Spherical Deconvolution Results in Pathology.” In Proceedings of the International Society for Magnetic Resonance in Medicine 25: 1815.
  21. Dima, D. , Modabbernia A., Papachristou E., et al. 2022. “Subcortical Volumes Across the Lifespan: Data From 18,605 Healthy Individuals Aged 3–90 Years.” Human Brain Mapping 43, no. 1: 452–469. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Drakesmith, M. , Harms R., Rudrapatna S. U., Parker G. D., Evans C. J., and Jones D. K.. 2019. “Estimating Axon Conduction Velocity In Vivo From Microstructural MRI.” NeuroImage 203: 116186. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Driscoll, I. , Davatzikos C., An Y., et al. 2009. “Longitudinal Pattern of Regional Brain Volume Change Differentiates Normal Aging From MCI.” Neurology 72, no. 22: 1906–1913. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Ellerbrock, I. , and Mohammadi S.. 2018. “Four In Vivo g‐Ratio‐Weighted Imaging Methods: Comparability and Repeatability at the Group Level.” Human Brain Mapping 39, no. 1: 24–41. 10.1002/hbm.23858. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Feinberg, I. 1982. “Schizophrenia: Caused by a Fault in Programmed Synaptic Elimination During Adolescence?” Journal of Psychiatric Research 17, no. 4: 319–334. [DOI] [PubMed] [Google Scholar]
  26. Fischl, B. 2012. “FreeSurfer.” NeuroImage 62, no. 2: 774–781. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Fotenos, A. F. , Snyder A. Z., Girton L. E., Morris J. C., and Buckner R. L.. 2005. “Normative Estimates of Cross‐Sectional and Longitudinal Brain Volume Decline in Aging and AD.” Neurology 64, no. 6: 1032–1039. [DOI] [PubMed] [Google Scholar]
  28. Good, C. D. , Johnsrude I. S., Ashburner J., Henson R. N., Friston K. J., and Frackowiak R. S.. 2001. “A Voxel‐Based Morphometric Study of Ageing in 465 Normal Adult Human Brains.” NeuroImage 14, no. 1: 21–36. [DOI] [PubMed] [Google Scholar]
  29. Group, B. D. C . 2012. “Total and Regional Brain Volumes in a Population‐Based Normative Sample From 4 to 18 Years: The NIH MRI Study of Normal Brain Development.” Cerebral Cortex 22, no. 1: 1–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Hodgkin, A. L. , and Huxley A. F.. 1952. “A Quantitative Description of Membrane Current and Its Application to Conduction and Excitation in Nerve.” Journal of Physiology 117, no. 4: 500–544. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Hsu, Y.‐C. , Lo Y.‐C., Chen Y.‐J., Wedeen V. J., and Isaac Tseng W.‐Y.. 2015. “NTU‐DSI‐122: A Diffusion Spectrum Imaging Template With High Anatomical Matching to the ICBM‐152 Space.” Human Brain Mapping 36, no. 9: 3528–3541. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Hua, K. , Zhang J., Wakana S., et al. 2008. “Tract Probability Maps in Stereotaxic Spaces: Analyses of White Matter Anatomy and Tract‐Specific Quantification.” NeuroImage 39, no. 1: 336–347. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Jefferson, A. L. , Himali J. J., Beiser A. S., et al. 2010. “Cardiac Index Is Associated With Brain Aging: The Framingham Heart Study.” Circulation 122, no. 7: 690–697. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Jenkinson, M. , Beckmann C. F., Behrens T. E., Woolrich M. W., and Smith S. M.. 2012. “Fsl.” NeuroImage 62, no. 2: 782–790. [DOI] [PubMed] [Google Scholar]
  35. Jeurissen, B. , Tournier J.‐D., Dhollander T., Connelly A., and Sijbers J.. 2014. “Multi‐Tissue Constrained Spherical Deconvolution for Improved Analysis of Multi‐Shell Diffusion MRI Data.” NeuroImage 103: 411–426. 10.1016/j.neuroimage.2014.07.061. [DOI] [PubMed] [Google Scholar]
  36. Juraska, J. M. , and Willing J.. 2017. “Pubertal Onset as a Critical Transition for Neural Development and Cognition.” Brain Research 1654: 87–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Kellner, E. , Dhital B., Kiselev V. G., and Reisert M.. 2016. “Gibbs‐Ringing Artifact Removal Based on Local Subvoxel‐Shifts.” Magnetic Resonance in Medicine 76, no. 5: 1574–1581. [DOI] [PubMed] [Google Scholar]
  38. Khan, W. , Egorova N., Khlif M. S., Mito R., Dhollander T., and Brodtmann A.. 2020. “Three‐Tissue Compositional Analysis Reveals In‐Vivo Microstructural Heterogeneity of White Matter Hyperintensities Following Stroke.” NeuroImage 218: 116869. [DOI] [PubMed] [Google Scholar]
  39. Khan, W. , Khlif M. S., Mito R., Dhollander T., and Brodtmann A.. 2021. “Investigating the Microstructural Properties of Normal‐Appearing White Matter (NAWM) Preceding Conversion to White Matter Hyperintensities (WMHs) in Stroke Survivors.” NeuroImage 232: 117839. [DOI] [PubMed] [Google Scholar]
  40. Lebel, C. , and Beaulieu C.. 2011. “Longitudinal Development of Human Brain Wiring Continues From Childhood Into Adulthood.” Journal of Neuroscience 31, no. 30: 10937–10947. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Lebel, C. , and Deoni S.. 2018. “The Development of Brain White Matter Microstructure.” NeuroImage 182: 207–218. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Mito, R. , Dhollander T., Raffelt D., et al. 2018. “Investigating Microstructural Heterogeneity of White Matter Hyperintensities in Alzheimer's Disease Using Single‐Shell 3‐Tissue Constrained Spherical Deconvolution.” In Proceedings of the International Society for Magnetic Resonance in Medicine 135.
  43. Mito, R. , Dhollander T., Xia Y., et al. 2020. “In Vivo Microstructural Heterogeneity of White Matter Lesions in Healthy Elderly and Alzheimer's Disease Participants Using Tissue Compositional Analysis of Diffusion MRI Data.” NeuroImage: Clinical 28: 102479. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Mohammadi, S. , and Callaghan M. F.. 2021. “Towards In Vivo g‐Ratio Mapping Using MRI: Unifying Myelin and Diffusion Imaging.” Journal of Neuroscience Methods 348: 108990. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Mori, S. , Wakana S., Van Zijl P. C., and Nagae‐Poetscher L. M.. 2005. MRI Atlas of Human White Matter. Elsevier. [DOI] [PubMed] [Google Scholar]
  46. Natu, V. S. , Gomez J., Barnett M., et al. 2019. “Apparent Thinning of Human Visual Cortex During Childhood Is Associated With Myelination.” Proceedings of the National Academy of Sciences of the United States of America 116, no. 41: 20750–20759. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Nazeri, A. , Chakravarty M. M., Rotenberg D. J., et al. 2015. “Functional Consequences of Neurite Orientation Dispersion and Density in Humans Across the Adult Lifespan.” Journal of Neuroscience 35, no. 4: 1753–1762. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Newman, B. T. , Danoff J. S., Lynch M. E., et al. 2022. “Epigenetic Age Acceleration Predicts Subject‐Specific White Matter Degeneration in the Human Brain.” bioRxiv: 2022.11.14.516491. [DOI] [PMC free article] [PubMed]
  49. Newman, B. T. , Dhollander T., Reynier K. A., Panzer M. B., and Druzgal T. J.. 2020. “Test–Retest Reliability and Long‐Term Stability of Three‐Tissue Constrained Spherical Deconvolution Methods for Analyzing Diffusion MRI Data.” Magnetic Resonance in Medicine 84, no. 4: 2161–2173. 10.1002/mrm.28242. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Newman, B. T. , Forman R., Donahue E., et al. 2024. “Diffusion Signals of Glial Activation Correlate With Fitness Scores in Parkinson's Subcortical Gray Matter.” medRxiv, 2024–11.
  51. Newman, B. T. , Jacokes Z., Venkadesh S., et al. 2024. “Conduction Velocity, G‐Ratio, and Extracellular Water as Microstructural Characteristics of Autism Spectrum Disorder.” PLoS One 19, no. 4: e0301964. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Newman, B. T. , Patrie J. T., and Druzgal T. J.. 2023. “An Intracellular Isotropic Diffusion Signal Is Positively Associated With Pubertal Development in White Matter.” Developmental Cognitive Neuroscience 63: 101301. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Newman, B. T. , Untaroiu A., and Druzgal T. J.. 2020. “A Method for Post‐Surgical Evaluation of Targeting Accuracy in Transcranial Focused Ultrasound Thalamic Ablation.” In 28th International Society of Magnetic Resonance in Medicine 28: 4111.
  54. Nooner, K. B. , Colcombe S. J., Tobe R. H., et al. 2012. “The NKI‐Rockland Sample: A Model for Accelerating the Pace of Discovery Science in Psychiatry.” Frontiers in Neuroscience 6: 152. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Pasternak, O. , Sochen N., Gur Y., Intrator N., and Assaf Y.. 2009. “Free Water Elimination and Mapping From Diffusion MRI.” Magnetic Resonance in Medicine 62, no. 3: 717–730. [DOI] [PubMed] [Google Scholar]
  56. Pasternak, O. , Westin C.‐F., Bouix S., et al. 2012. “Excessive Extracellular Volume Reveals a Neurodegenerative Pattern in Schizophrenia Onset.” Journal of Neuroscience 32, no. 48: 17365–17372. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Pietsch, M. , Christiaens D., Hutter J., et al. 2019. “A Framework for Multi‐Component Analysis of Diffusion MRI Data Over the Neonatal Period.” NeuroImage 186: 321–337. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Pomponio, R. , Erus G., Habes M., et al. 2020. “Harmonization of Large MRI Datasets for the Analysis of Brain Imaging Patterns Throughout the Lifespan.” NeuroImage 208: 116450. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Raffelt, D. A. , Smith R. E., Ridgway G. R., et al. 2015. “Connectivity‐Based Fixel Enhancement: Whole‐Brain Statistical Analysis of Diffusion MRI Measures in the Presence of Crossing Fibres.” NeuroImage 117: 40–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Raffelt, D. A. , Tournier J.‐D., Smith R. E., et al. 2017. “Investigating White Matter Fibre Density and Morphology Using Fixel‐Based Analysis.” NeuroImage 144: 58–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Rajkowska, G. , Halaris A., and Selemon L. D.. 2001. “Reductions in Neuronal and Glial Density Characterize the Dorsolateral Prefrontal Cortex in Bipolar Disorder.” Biological Psychiatry 49, no. 9: 741–752. [DOI] [PubMed] [Google Scholar]
  62. Raz, N. , Lindenberger U., Rodrigue K. M., et al. 2005. “Regional Brain Changes in Aging Healthy Adults: General Trends, Individual Differences and Modifiers.” Cerebral Cortex 15, no. 11: 1676–1689. [DOI] [PubMed] [Google Scholar]
  63. Rushton, W. A. H. 1951. “A Theory of the Effects of Fibre Size in Medullated Nerve.” Journal of Physiology 115, no. 1: 101–122. [DOI] [PMC free article] [PubMed] [Google Scholar]
  64. Selemon, L. D. 2013. “A Role for Synaptic Plasticity in the Adolescent Development of Executive Function.” Translational Psychiatry 3, no. 3: e238. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Selemon, L. D. , Rajkowska G., and Goldman‐Rakic P. S.. 1995. “Abnormally High Neuronal Density in the Schizophrenic Cortex: A Morphometric Analysis of Prefrontal Area 9 and Occipital Area 17.” Archives of General Psychiatry 52, no. 10: 805–818. [DOI] [PubMed] [Google Scholar]
  66. Semendeferi, K. , Armstrong E., Schleicher A., Zilles K., and Van Hoesen G. W.. 1998. “Limbic Frontal Cortex in Hominoids: A Comparative Study of Area 13.” American Journal of Physical Anthropology 106, no. 2: 129–155. [DOI] [PubMed] [Google Scholar]
  67. Smith, S. M. , Jenkinson M., Woolrich M. W., et al. 2004. “Advances in Functional and Structural MR Image Analysis and Implementation as FSL.” NeuroImage 23: S208–S219. [DOI] [PubMed] [Google Scholar]
  68. Stephan, B. C. M. , Hunter S., Harris D., et al. 2012. “The Neuropathological Profile of Mild Cognitive Impairment (MCI): A Systematic Review.” Molecular Psychiatry 17, no. 11: 1056–1076. [DOI] [PubMed] [Google Scholar]
  69. Stikov, N. , Campbell J. S., Stroh T., et al. 2015. “In Vivo Histology of the Myelin g‐Ratio With Magnetic Resonance Imaging.” NeuroImage 118: 397–405. [DOI] [PubMed] [Google Scholar]
  70. Tournier, J.‐D. , Calamante F., and Connelly A.. 2007. “Robust Determination of the Fibre Orientation Distribution in Diffusion MRI: Non‐Negativity Constrained Super‐Resolved Spherical Deconvolution.” NeuroImage 35, no. 4: 1459–1472. [DOI] [PubMed] [Google Scholar]
  71. Tournier, J.‐D. , Smith R., Raffelt D., et al. 2019. “MRtrix3: A Fast, Flexible and Open Software Framework for Medical Image Processing and Visualisation.” NeuroImage 202: 116137. [DOI] [PubMed] [Google Scholar]
  72. Veraart, J. , Fieremans E., and Novikov D. S.. 2016. “Diffusion MRI Noise Mapping Using Random Matrix Theory.” Magnetic Resonance in Medicine 76, no. 5: 1582–1593. [DOI] [PMC free article] [PubMed] [Google Scholar]
  73. Wakana, S. , Caprihan A., Panzenboeck M. M., et al. 2007. “Reproducibility of Quantitative Tractography Methods Applied to Cerebral White Matter.” NeuroImage 36, no. 3: 630–644. [DOI] [PMC free article] [PubMed] [Google Scholar]
  74. Zhang, H. , Schneider T., Wheeler‐Kingshott C. A., and Alexander D. C.. 2012. “NODDI: Practical In Vivo Neurite Orientation Dispersion and Density Imaging of the Human Brain.” NeuroImage 61, no. 4: 1000–1016. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Figure S1: Plots detailing each of the mean subcortical gray matter 3T‐CSD and fixel‐based metrics evaluated in this study corrected for each subject's total brain volume. In order, these are (A) ECI, (B) FDC, (C) ICI, (D) g‐ratio, (E) ICA, and (F) ACV. Loess lines are calculated identically to other charts presented in this paper.

HBM-46-e70403-s001.tiff (7.1MB, tiff)

Data Availability Statement

The data that support the findings of this study are available in NKI/Rockland Cohort at https://rocklandsample.org/. These data were derived from the following resources available in the public domain: NKI/Rockland I, https://rocklandsample.org/.

Articles from Human Brain Mapping are provided here courtesy of Wiley

RESOURCES