Influence of Age on Global and Regional Brain Stiffness in Young and Middle-Aged Adults

Abstract

Background:

An understanding of potential age-related changes in brain stiffness and its regional variation is important for further clinical application of MR elastography.

Purpose:

To investigate the effect of age on global and regional brain stiffness in young and middle-aged adults.

Study Type:

Prospective.

Subjects:

Fifty subjects with normal brains and aged in their 20s, 30s, 40s, 50s, or 60s (five men, five women per decade).

Field Strength/Sequence:

3.0T MRI and elastography with a vibration frequency of 60 Hz.

Assessment:

Stiffness was measured in nine brain regions (cerebrum, temporal lobes, sensorimotor areas, frontotemporal composite region, deep gray matter and white matter (deep GM/WM), parietal lobes, occipital lobes, frontal lobes, and cerebellum) using an atlas-based region-of-interest approach. The influence of age on regional brain stiffness was evaluated.

Statistical Tests:

Multiple linear regression analysis, followed by Dunnett’s multiple comparisons test, using subjects in their 20s as controls.

Results:

Following adjustment for sex, multiple linear regression revealed a significant negative correlation between age and stiffness of the cerebrum (P < 0.0001), temporal lobes (P < 0.0001), sensorimotor areas (P < 0.0001), frontotemporal composite region (P < 0.0001), deep GM/WM (P = 0.0028), parietal lobes (P < 0.0001), occipital lobes (P = 0.0055), and frontal lobes (P < 0.0001). Dunnett’s multiple comparison test showed that the stiffness of the sensorimotor areas, frontotemporal composite region, and frontal lobes was significantly decreased in subjects in their 40s (P < 0.0367), 50s (P < 0.0001), and 60s (P < 0.0001), while that of the cerebrum, temporal lobes, and parietal lobes was significantly decreased only in subjects in their 50s (P < 0.0012) and 60s (P < 0.0031) when compared with the controls.

Data Conclusion:

There is an age-related decrease in brain stiffness that varies across the different regions.

AGING CAUSES HISTOPATHOLOGICAL and structural changes in the human brain. Numerous microstructural changes occur in response to normal aging, including shrinkage of neurons¹ and changes in the numbers of dendric spines and synapses.² Several neuroimaging studies have also reported that aging is associated with widespread changes in the structure and perfusion alterations of the brain, including loss of volume and/or thinning of the cortex, with some predominance in the prefrontal cortex, lateral parietal and lateral temporal association areas,^3,4 and somatosensory and motor regions,^4–6 including a decrease in cerebral blood flow (CBF) in the frontal-temporal region.⁷ It is reasonable to expect that these microscopic or macroscopic structural alterations are likely to result in changes in brain stiffness.

Early experiments in elastography started with ultrasonographic imaging.⁸ The brain is protected and isolated by the skull, so the obstacle of wave propagation in response to the lack of an appropriate acoustic window limits the accuracy of estimation of brain stiffness. Therefore, brain stiffness has only been estimated in vivo by intraoperative ultrasound elastography⁹; this has prevented widespread application of measurement of brain stiffness. Magnetic resonance elastography (MRE) is an evolving method for noninvasive measurement of the mechanical properties of soft tissue in vivo.^10–12 Since the advent of MRE, the mechanical properties of the brain have been well studied.¹³ More recently, several MRE studies have revealed reproducible changes in stiffness in patients with Alzheimer’s disease,^14–16 Parkinson’s disease,¹⁷ multiple sclerosis,¹⁸ frontotemporal dementia,¹⁶ and normal pressure hydrocephalus¹⁶ when compared with normal subjects.

It is necessary to understand age-related changes in brain stiffness as well as its regional patterns when considering the clinical application of MRE, given that the normal alterations occurring with aging and their geometric preferences could have confounding effects on alterations in brain stiffness that selectively target specific regions of the brain in patients with neurological disorders. This should be recognized particularly when assessing disease progression longitudinally or the response of a neurological disorder to treatment. Although some researchers have concluded that elasticity of the brain decreases with age,^16,19–22 the regional dependence of brain stiffness on age remains less well defined. Sack et al reported that the elasticity of the entire brain and some brain regions, including the white matter, gray matter, frontal, and posterior areas, show a linear decline with aging.^20,21 However, these measurements were not obtained from the lobes of the brain using an atlas-based method. More recently, Arani et al reported a significant linear decline in the stiffness of the cerebrum, frontal lobes, posterior lobes, parietal lobes, and temporal lobes in older adults (aged 56–89 years); however, the authors did not include young and middle-aged adults in their study.¹⁹

The effect of the sex of the subject on brain stiffness is an important concern when considering clinical application of brain stiffness measurements, given that numerous mental, psychiatric, and neurological conditions, including anxiety and depressive disorders, autism spectrum disorders, cognitive decline and Alzheimer’s disease, and multiple sclerosis, show sexual dimorphism.²³ However, this finding has been inconsistent across studies. Sack et al reported higher global brain stiffness in female subjects,²⁰while a later study from the same group did not find a sex difference.²¹ Arani et al reported that the effect of sex on brain stiffness was only significant in the occipital and temporal lobes in older adults.¹⁹

The purpose of this study was to investigate the effect of age and sex on global and regional brain stiffness in young and middle-aged adults.

Materials and Methods

This prospective study was approved by our Institutional Review Board. All subjects provided written informed consent before recruitment and examination by MRE.

Subjects

Fifty healthy subjects aged 20–69 years and without apparent neurological or psychiatric conditions were included in this study. We recruited 10 sex-matched subjects (five men, five women) from each of the five 10-year age groups between their 20s and 60s. The mean age was 42.5 ± 15.0 (range 20–69) years in the 25 men, 42.6 ± 15.4 (range 20–69) years in the 25 women, and 42.4 ± 14.9 years in all 50 subjects. The median age of the study population (n = 50) was 41 years.

Data Acquisition

MRE was performed using a superconducting magnet operating at 3.0T (Discovery 750; GE Medical Systems, Waukesha, WI) with a 32-channel phased-array coil. The images were obtained while the patients were in the supine position with a soft pillow-like passive driver placed under the subject’s head. A vibration source (Resoundant, Rochester, MN)²⁴ placed outside of the MR examination room produced pneumatic pressure waves that were delivered to the passive driver via a plastic tube. 3D wavefield imaging, with six repetitions for the motion encoding in the positive and negative x, y, and z directions was performed using a spin-echo echo-planar imaging (SE-EPI) MRE sequence,^{14–16,19,25,26} and a 3D inversion algorithm²⁴ was used to process the data to obtain elasticity images. In order to cover the entire brain, the number of slices was individually determined for each subject within the range of 48–50 slices. A section thickness of 3 mm with no gap was applied. The repetition time (TR) and echo time (TE) were 3600 msec and 62 msec, respectively (flip angle, 90;numberofexcita-tions, 1; field of view [FOV], 24 × 24 cm; matrix size, 128 × 128; ASSET acceleration, 3; acquisition time, ~5 min; and axes of motion-sensitizing gradient pulses, x, y, and z). The frequency of the driver was 60 Hz with eight phase offsets sampled over one period of 60 Hz motion.

Atlas regions for each subject were obtained using a separately acquired 3D IR-SPGR T₁-weighted image. In order to cover the entire brain, the number of slices was individually determined for each subject within the range of 128–186 slices. A section thickness of 1.2 mm with no gap was applied. The TR and TE were 5.0 msec and 1.5 msec, respectively (sagittal orientation; flip angle, 25°; inversion time [TI], 400 msec; FOV, 25 × 25 cm; matrix size, 256 × 256; ASSET acceleration, 1.75; acquisition time, ~3 min).

Image Processing

The MR scanner automatically generated stiffness maps by processing the acquired propagating shear wave images using an algorithm similar to that used by Murphy et al.²⁴ First, the original complex-valued images for each slice, phase offset, and motion-encoding direction were reconstructed to a 128 × 128 matrix. Second, the complex images for the positive motion-encoding directions were multiplied by the complex-conjugate of the negative motion-encoding directions to produce complex-valued images, the phase of which is the phase difference between the positive and negative motion-encoding directions. Third, these volumes were smoothed in the Z direction using a 1D, 4th-order, Butterworth, low-pass filter with a cutoff frequency of (Nz/4) cycles per FOVz (where Nz is the number of slices) to reduce any potential slice-to-slice phase jitter that may have been in the data. Fourth, the curl of the smooth, wrapped, phase-difference data was calculated to remove any longitudinal wave information from the data without unwrapping the phase data using the technique nique previously shown by Glaser et al²⁷ and central-difference derivative kernels. Fifth, the curl data were smoothed using quartic smoothing kernels of the form (1-x²)²*(1-y²)²*(1-z²)², where x, y, and z are linearly spaced from –1 to 1 over the chosen window size of 7 × 7 × 5 voxels. Sixth, the first temporal harmonic of the smoothed curl data was calculated using a temporal Fourier transform to isolate the motion information occurring at the primary frequency of vibration and to remove static and higher-frequency information from the data. Seventh, the first-harmonic curl wave information was inverted using a direct inversion of the Helmholtz wave equation to calculate the complex-valued shear modulus everywhere in the volume. Eighth, the complex shear modulus values were median-filtered using a 3 × 3 × 3 medianfilter to remove possible extreme outliers from the results. Finally, images (elastograms) of the shear stiffness (ie, the absolute value of the complex shear modulus) were created on the scanner from which regional stiffness information could be measured.

A T₁-weighted image was segmented to create a region of interest (ROI) in the gray matter (GM), white matter (WM), and cerebrospinal fluid (CBF) content for each voxel as previously described.^24,28 To create the global and regional ROI, a lobar atlas in a standard template space provided by WFU PickAtlas (Talairach Brain Atlas Theory)²⁹ was warped into the subject’s T₁-weighted image using a unified segmentation algorithm implemented in SPM12.³⁰ Nine ROIs were included for calculation of brain stiffness: the cerebrum (entire brain excluding the cerebellum), temporal lobes, sensorimotor areas (precentral gyrus and postcentral gyri), frontotemporal composite region (frontal and temporal lobes excluding the precentral and postcentral gyri), deep GM/WM (insula, deep gray nuclei, and white matter tracts), parietal lobes, occipital lobes, frontal lobes, and cerebellum. The T₁-weighted image was then registered into the magnitude image of the MRE examination along with the segmented ROIs and the warped atlas. The brain and regional masks were calculated by including any voxels where the GM content plus the WM content was greater than the CBF content. The ROI used for reporting brain stiffness was eroded by 3 voxels from every edge to remove edge artifacts, which has previously shown as a pipeline to minimize partial volume and edge-related bias.²⁴

Statistical Analysis

The results are expressed as the mean ± standard deviation. The brain volume computed in the subject’s magnitude image-space (total brain, total GM, total WM, and brain parenchymal fraction [BPF]) and brain stiffness in the cerebrum, temporal lobes, sensorimotor areas, frontotemporal composite region, deep GM/WM, parietal lobes, occipital lobes, frontal lobes, and cerebellum were calculated for each subject. The brain GM and WM volumes in each global and regional ROI in the subject’s magnitude image space were also calculated.

Initially, a multivariate linear regression analysis was performed to evaluate the influence of age and sex on brain volume and brain stiffness. The differences in brain stiffness were evaluated using Dunnett’s multiple comparisons test. Measurements from the subjects in their 20s were considered to be the reference for comparison with those from the subjects in their 30s, 40s, 50s, and 60s. P < 0.05 was considered statistically significant. All statistical analyses were performed using commercial software (JMP v. 13.0.0, SAS Institute, Cary, NC).

Results

The mean total brain volume and BPF values for all subjects were 1182.63 ± 116.2 cm³ and 0.78 ± 0.06, respectively (Table 1). The results of the multiple linear regression analysis are summarized in Table 2. A significant sex difference was observed in all brain volume parameters (total brain volume, GM volume, WM volume, and BPF), while a significant linear correlation was found between brain volume parameters and age only for total brain volume, GM volume, and BPF (Table 2). Following correction for sex, the stiffness of the cerebrum (P < 0.0001), temporal lobes (P < 0.0001), sensorimotor areas (P < 0.0001), frontotemporal composite region (P < 0.0001), deep GM/WM (P = 0.0028), parietal lobes (P < 0.0001), occipital lobes (P = 0.0055), and frontal lobes (P < 0.0001) decreased significantly with age (Table 2). The multiple linear regression model estimated the annual decline in brain stiffness in these significant regions to be – 0.0049 kPa to –0.0136 kPa per year (R² = 0.15–0.60). No significant correlation between brain stiffness and age was only found in the cerebellum (P = 0.0761). When corrected for age, the multiple linear regression model revealed no significant sex differences in stiffness for any of the brain regions.

TABLE 1.

Brain Volume and Brain Stiffness in the Entire Cohort

	All (n = 5)	Men (n = 25)	Women (n = 25)
Brain volume
Total brain volume, cm3	1182.63 ± 116.2	1232.48 ± 127.09	1132.77 ± 78.98
GM volume, cm3	678.07 ± 79.86	710.58 ± 93.11	645.55 ± 46.35
WM volume, cm3	504.56 ± 93.11	521.90 ± 48.79	487.22 ± 50.41
BPF	0.78 ± 0.06	0.75 ± 0.07	0.80 ± 0.05
Brain stiffness
Cerebrum, kPa	2.34 ± 0.19	2.35 ± 0.19	2.33 ± 0.19
Temporal lobes, kPa	2.61 ± 0.17	2.63 ± 0.17	2.59 ± 0.17
Sensorimotor areas, kPa	2.12 ± 0.26	2.13 ± 0.29	2.10 ± 0.24
Frontotemporal composite region, kPa	2.39 ± 0.17	2.42 ± 0.16	2.37 ± 0.18
Deep GM/WM, kPa	2.26 ± 0.29	2.27 ± 0.31	2.25 ± 0.26
Parietal lobes, kPa	2.13 ± 0.21	2.11 ± 0.22	2.15 ± 0.19
Occipital lobes, kPa	2.44 ± 0.19	2.45 ± 0.19	2.43 ± 0.20
Frontal lobes, kPa	2.22 ± 0.21	2.25 ± 0.20	2.19 ± 0.21
Cerebellum, kPa	1.79 ± 0.13	1.78 ± 0.12	1.80 ± 0.14

The data are shown as the mean ± standard deviation. BPF, brain parenchymal fraction; GM, gray matter; WM, white matter.

TABLE 2.

Age and Sex Dependency of Brain Volume and Stiffness Estimated by Using Multiple Linear Regression Models

Parameter	Annual change (P-value)	Sex bias (P-value)	Predicted value at age 41	Overall F-test P-value	R2
Brain volume	cm3/year	cm3	cm3
Total brain volume	−2.5584 ± 0.9472 (P = 0.0096*)	−50.1629 ± 14.0691 (P = 0.0008*)	1186.4139 ± 14.1383	0.0003*	0.2970
Total GM volume	−2.8628 ± 0.5714 (P < 0.0001*)	−32.8568 ± 8.4864 (P = 0.0003*)	682.3047 ± 8.5282	< 0.0001*	0.4584
WM volume	0.3044 ± 0.4753 (P = 0.5249)	−17.3061 ± 7.0591 (P = 0.0180*)	504.1092 ± 7.0938	0.0488*	0.1206
BPF	−0.0026 ± 0.0004 (P < 0.0001*)	0.0226 ± 0.0059 (P = 0.0004*)	0.7801 ± 0.0060	< 0.0001*	0.5540
Brain stiffness	kPa/year	kPa	kPa
Cerebrum	−0.0080 ± 0.0014 (P < 0.0001*)	−0.0104 ± 0.0209 (P = 0.6223)	2.3545 ± 0.021	< 0.0001*	0.4054
Temporal lobes	−0.0065 ± 0.0013 (P < 0.0001*)	−0.0213 ± 0.0196 (P = 0.2827)	2.6175 ± 0.0197	< 0.0001*	0.3528
Sensorimotor areas	−0.0136 ± 0.0016 (P < 0.0001*)	−0.0167 ± 0.0241 (P = 0.4908)	2.1353 ± 0.0242	< 0.0001*	0.6010
Frontotemporal composite region	−0.0072 ± 0.0013 (P < 0.0001*)	−0.0301 ± 0.0187 (P = 0.1824)	2.4049 ± 0.0188	< 0.0001*	0.4301
Deep GM/WM	−0.0080 ± 0.0025 (P = 0.0028*)	−0.0107 ± 0.0377 (P = 0.7768)	2.2694 ± 0.0379	0.0106*	0.1761
Parietal lobes	−0.0089 ± 0.0015 (P < 0.0001*)	0.0187 ± 0.0226 (P = 0.4115)	2.1414 ± 0.0227	< 0.0001*	0.4240
Occipital lobes	−0.0049 ± 0.0017 (P = 0.0055*)	−0.0082 ± 0.0252 (P = 0.7465)	2.4487 ± 0.0253	0.0196*	0.1541
Frontal lobes	−0.0093 ± 0.0015 (P < 0.0001*)	−0.0303 ± 0.0217 (P = 0.1690)	2.2326 ± 0.0218	< 0.0001*	0.4734
Cerebellum	−0.0022 ± 0.0012 (P = 0.0761)	0.0105 ± 0.0183 (P = 0.5688)	1.7972 ± 0.0184	0.1735	0.0718

Brain stiffness = (predicted value at age 41) + (annual change)* (age- 41) + (sex bias)*(sex), where +1 is for female sex and −1 is for male sex. BPF, brain parenchymal fraction; GM, gray matter; WM, white matter.

The results of the Dunnett’s multiple comparison test are shown in Fig. 1. The stiffness of the sensorimotor areas, frontotemporal composite region, and frontal lobes was significantly decreased in subjects in their 40s (P < 0.0367), 50s (P < 0.0001), and 60s (P < 0.0001), while that of the cerebrum, temporal lobes, and parietal lobes was significantly decreased only in subjects in their 50s (P < 0.0012) and 60s (P < 0.0031). No significant decrease was found in the deep GM/WM, occipital lobes, or cerebellum in subjects of any age group when compared with those in their 20s. The GM volume in the frontal lobes showed a significant decrease in subjects in their 40s (P = 0.0300), 50s (P = 0.0013), and 60s (P = 0.0007). No significant change in WM volume was observed in any of the brain regions.

FIGURE 1:

Bar charts showing measurements for brain stiffness, gray matter volume, and white matter volume. Dunnett’s multiple comparison tests were performed using the measurements for subjects in their 20s as the control. The data are shown as the mean. The error bars represent the 95% confidence intervals. ***P < 0.0001, **P < 0.001, *P < 0.05.

Color-coded stiffness maps at the level of the cerebrum are shown for two subjects in Fig. 2. These maps clearly demonstrate softening of stiffness throughout the entire brain of a 65-year-old man in comparison with that of a 23-year-old man (top row). A marked decrease in stiffness is evident on the stiffness maps of the cerebrum eroded by 3 voxels from every edge when overlaid on the corresponding T₁-weighted images (bottom row).

FIGURE 2:

Color-coded stiffness maps of the cerebrum for a 23-year-old man and a 65-year-old man showing stiffness of 2.57 kPa and 1.94 kPa, respectively.

Discussion

In this study, the stiffness of the cerebrum, temporal lobes, sensorimotor areas, frontotemporal composite region, deep GM/WM, parietal lobes, occipital lobes, and frontal lobes was found to decrease significantly in a linear manner with age, whereas that of the posterior lobes and cerebellum did not.

Previous studies have reported a significant correlation between age and stiffness in the entire brain and the temporal, parietal, occipital and frontal lobes but not in the cerebellum,^19,21 which is consistent with our results. The stiffness of the cerebrum decreased with age by –0.0080 kPa/year in our study. This rate is equivalent to the decline of –0.011 kPa per year reported by Arani et al¹⁹ and that of –0.0075kPa per year in a storage module with a drive frequency of 62.5 Hz reported by Sack et al.²⁰ However, Arani et al found no significant age-related change in the sensorimotor areas and deep GM/WM. Given that the subjects in the previous studies were older (56–89 years) than those in our study (20–69 years), it is possible that the change in stiffness in the sensorimotor areas and deep GM/WM is more prominent in the younger to middle-aged groups.

We found no significant sex-related difference in brain stiffness in any of the brain regions, which is in agreement with the findings of Sack et al,²¹ but not with those of Arani et al,¹⁹ who found a sex-related difference in stiffness of the temporal and occipital lobes. The effect of sex on brain stiffness is not fully understood and requires investigation in a larger study in the future.

The mechanism by which brain stiffness declines with age has not been fully elucidated. In our study, the entire GM volume decreased linearly with age at a rate of ~2.9 cm³/year, which is compatible with the finding in some other studies that GM volume decreases linearly with age (from – 2.5 cm³/year to –3.9 cm³/year).^5,6 However, we found no significant decline in WM volume with aging. Some studies have shown an age-associated decline in WM volume,^3,31 whereas others found that the WM volume remained stable with aging,32–34 which is in agreement with our results. Several publications have reported nonlinearly age-related changes in the rate of decline of WM volume, demonstrating an inverted U-shaped trajectory of change across the lifespan, with a volume increase in young adulthood, a plateau in middle age, and a precipitous decline in old age.^35,36 The main age range examined in this study (young to middle-aged adults) can be included in this volume increase/plateau phase, which might account for the discrepancy. Other possible reasons may include the difference in accuracy of the segmentation methods used, or methods for handling white matter signal elevation.

The age at which the GM volume decreased in comparison with that in the subjects in their 20s was similar to the age at which brain stiffness declined, with no significant alteration of WM volume. We hypothesized that these findings indicate the possibility of a contribution of the change in cortical composition related to loss of GM volume to the maintenance of brain stiffness. Klein et al found increased shear elasticity in mice as a result of reactively generated neurons induced by a dopamine defect.³⁷ Another group reported that brain stiffness was decreased because of loss of neurons in a mouse model of stroke.³⁸ Both these studies indicate that the neurons in the brain have a role as supporting structures with mechanical properties and complement our hypothesis. Another possible explanation is the potential contribution of changes in CBF. The decline in CBF with advancing age is well known.⁸ Chatelin and colleagues observed a parallel rapid decrease in CBF and brain elasticity in a short time in a mouse model of drug-induced hypotension, which supports this idea.⁴⁰ However, recent publications have also reported an association of glial cells³⁹ and functional connectivity¹⁵ to the viscoelastic properties of the brain. Further investigation of the relationships between tissue structure and the mechanical properties of the brain is required to elucidate the mechanism of the effect of age on brain stiffness.

Our study has several limitations. First, we did not apply voxel-based analysis (VBA) in this study. VBA is sensitive to registration accuracy and is unreliable when based on MRE, which does not contain as much structural information as a 3D T₁-weighted image. Furthermore, MRE is heavily processed and smoothed through the inversion algorithm and is limited because of edge-related artifact²⁴; therefore, whether or not the value of a particular voxel accurately represents brain stiffness in the normalized-space when VBA is applied would be uncertain. The lack of an established processing VBA pipeline in MRE to handle these problems adequately limited the exploratory analysis of our data. Second, our sample size was relatively small. A large cohort study is needed to determine the distribution of brain stiffness according to sex and race in the general population. Finally, although all the subjects were of the same race, we did not investigate or adjust for other possible confounders, such as socioeconomic status and education level.

In conclusion, brain stiffness decreases overall with aging. However, this age-related change is not uniform across the brain regions.