Assessment of microstructural changes in white matter hyperintensities in aging and mild cognitive impairment revealed by advanced diffusion MRI

Abstract

Background

White matter hyperintensities (WMHs) are common in older adults and appear as abnormal signal on T2-weighted or fluid-attenuated inversion recovery (FLAIR) magnetic resonance imaging (MRI). They are often linked to demyelination, axonal damage, and gliosis, as well as vascular changes. WMHs occur in both normal aging and mild cognitive impairment (MCI), where they may contribute to cognitive decline.

Objective

The study objectives were to determine whether advanced diffusion MRI (dMRI) techniques can detect distinct microstructural changes within WMHs in individuals with MCI compared to healthy controls and to evaluate relationships between these measures and cognitive performance.

Methods

Advanced dMRI techniques were used to assess WMH microstructural changes in normal aging (n = 55) and MCI (n = 46) participants from the ADNI database. WMHs and their surrounding penumbra were identified using an automated approach. Microstructural characteristics, derived from free-water (FW) diffusion tensor imaging and diffusion kurtosis imaging, were evaluated between groups and white matter regions. Associations between these measures and cognitive performance (assessed by the Mini-Mental State Exam) were examined.

Results

Across both groups, WMHs showed higher FW and lower FW-fractional anisotropy and kurtosis metrics compared to normal-appearing white matter, indicating widespread microstructural alterations. No groupwise microstructural differences were observed within corresponding tissue types. In the MCI group, kurtosis metrics within WMHs correlated with cognitive performance.

Conclusions

These findings highlight the complexity of WMH-related microstructural changes and suggest that advanced dMRI biomarkers may offer valuable insights into the role of white matter changes in aging and cognitive decline.

Keywords

Alzheimer's disease magnetic resonance imaging mild cognitive impairment white matter hyperintensities

Introduction

In older individuals, white matter hyperintensities (WMHs) are commonly observed using magnetic resonance imaging (MRI).¹ These WMHs represent regions of abnormally elevated signal intensity within the white matter and are often visualized on T2-weighted or fluid-attenuated inversion recovery (FLAIR) MRI sequences. The underlying pathological causes of WMHs are complex and likely multifactorial, with histological studies demonstrating associations with demyelination, axonal damage, and gliosis.² Previous studies have further linked WMHs with both hemodynamic impairments^3–5 and tissue edema.^2,6,7 Regardless of the underlying pathology, the presence and extent of WMHs have been associated with an increased risk of stroke and dementia.⁸ Risk factors for developing WMHs include age, hypertension, diabetes, and cerebrovascular disease.^9,10

Previous studies have demonstrated increased prevalence of WMHs in neurodegenerative diseases, such as Alzheimer's disease (AD), vascular dementia, and cognitively impaired Parkinson's disease.¹¹ In mild cognitive impairment (MCI), WMHs may serve as biomarkers for underlying cerebrovascular pathology and small vessel disease.¹² WMHs in MCI may also contribute to cognitive decline and could increase the risk of progression to dementia. More specifically, WMHs are thought to disrupt neural connectivity and information processing within white matter tracts, thereby impairing cognitive function.¹³ Importantly, WMHs capture vascular contributions to cognitive impairment that may occur alongside or even independently of hallmark AD pathology. Beyond WMHs, disruptions in white matter microstructure may occur before the onset of cognitive changes, and as such, biomarkers of overall white matter integrity may be early indicators of both neurovascular and neurodegenerative processes.¹⁴

Diffusion MRI (dMRI) is widely used to study white matter microstructure, providing insights into white matter tract integrity and brain structural connectivity.^15,16 In healthy older adults, decreases in fractional anisotropy (FA) and increases in mean diffusivity (MD) have been observed in WMHs compared to normal appearing white matter (NAWM).¹⁷ These changes in dMRI biomarkers have been shown to extend progressively to regions adjacent to WMHs in both healthy aging^18,19 and stroke²⁰ cohorts, creating a perilesional penumbra that appears normal on conventional MRI. These microstructural alterations may reflect processes like demyelination, axonal loss, or chronic ischemia and have been linked to cognitive decline.²¹ Understanding the WMH penumbra provides insight into the broader impact of white matter injury and may offer a novel target for early intervention and treatment.

Both FA and MD are obtained from standard diffusion tensor imaging (DTI), which models the Gaussian diffusion of water molecules. However, the limitations of this model include an inability to account for both partial volume effects (PVEs) and non-Gaussian diffusion processes. Correcting for PVEs using a free-water DTI (FW-DTI) model²² has been shown to produce more accurate dMRI metrics, particularly in the context of aging and cognitive decline;²³ additionally, the resulting FW index has been shown to be a robust biomarker for neurodegenerative diseases^24–26 Non-Gaussian diffusion patterns exhibited by biological tissue can be adeptly modeled via diffusion kurtosis imaging (DKI), which incorporates higher-order diffusion moments to provide insights into tissue microstructural complexity.^27,28 Lower diffusion kurtosis values reflect more unrestricted water diffusion and may be indicative of reduced microstructural complexity, which may be associated with neuron cell loss, demyelination, or other changes in cellular and tissue architecture.²⁹ As for FW-DTI, DKI changes have been observed in MCI and dementia cohorts, thus contributing to our understanding of these conditions.^23,26,^30–32

The present study builds upon our prior work investigating whole-brain microstructural changes in MCI, which employed a voxel-based analysis approach.³³ While the prior work focused on identifying widespread, diffuse patterns across the brain, the current study adopts a region-of-interest (ROI) approach to specifically examine WMHs and their surrounding penumbra. We compared cognitively normal (CN) and MCI groups using advanced dMRI techniques, including FW-DTI and DKI metrics. The primary objective of this study was to characterize the microstructural properties of WMHs and their penumbra in CN and MCI participants. In addition to diffusion-based measurements, we also evaluated group differences in normalized WMH and penumbra volumes to provide a more comprehensive assessment of the structural burden associated with cognitive decline. We hypothesized that: (1) individuals with MCI would exhibit both volumetric and microstructural alterations within WMH and their penumbra compared to CN participants; and (2) these microstructural changes would be associated with cognitive performance, as assessed by standard neuropsychological tests.

Methods

Participants

All data were downloaded from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database³⁴ (https://adni.loni.usc.edu/). The ADNI was launched in 2003 as a public-private partnership, led by Principal Investigator Michael W. Weiner, MD. The original goal of ADNI was to test whether serial MRI, positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of MCI and early AD. The current goals include validating biomarkers for clinical trials, improving the generalizability of ADNI data by increasing diversity in the participant cohort, and to provide data concerning the diagnosis and progression of Alzheimer's disease to the scientific community. For up-to-date information, see adni.loni.usc.edu.

Subjects included participants in the ADNI-defined CN and MCI cohorts with multi-shell dMRI data collected in ADNI-3 between February 2017 and August 2020; additionally, magnetization-prepared rapid gradient-echo (MPRAGE) and T2-weighted-fluid-attenuated inversion recovery (T2-FLAIR) images collected on the same day as the multi-shell dMRI acquisition were required for inclusion in this study. No a priori power analysis was performed, as the study leveraged the complete available dataset to maximize statistical power. Additionally, the current study uses the same sample of participants described previously.³³ All imaging data were reanalyzed using the ROI-based methods focusing on WMH, as outlined below.

This study included data from CN (n = 55, 39 females, 76.1 ± 7.0 years old) and MCI (n = 46, 16 females, 74.2 ± 7.6 years old) cohorts (Table 1). Additionally, Mini-Mental State Examination (MMSE) scores are reported in Table 1, and a cardiovascular risk score (CRS, range 0–6) was calculated as the sum of six binary variables: obesity (body mass index ≥ 30), current smoking status, and history of hypertension, diabetes, heart disease, and stroke.³⁵

Table 1.

Complete subject characteristics.

Group	N (#F)	Age (SD) year	MMSE [available]	CRS (SD)	ABS motion (SD) (mm)	REL. motion (SD) (mm)	Outliers (SD) (%)
CN	55 (39)	76.1 (7.0)	29.04 (1.43) [55]	2.69 (1.40)	0.76 (0.32)	0.56 (0.18)	0.37 (0.18)
MCI	45* (15)	74.4 (7.6)	27.34 (4.00) [44]	2.36 (1.77)	0.86 (0.40)	0.66 (0.32)	0.42 (0.28)
Shapiro-Wilk		W = 0.986; p = 0.450	W = 0.528; p < 0.001	W = 0.909; p < 0.001	W = 0.881; p < 0.001	W = 0.800; p < 0.001	W = 0.844; p < 0.001
t-test		t = −1.162; p = 0.248	-	─
Mann–Whitney U test		-	W = 1621; p = 0.003	W = 1357.5; p = 0.396	W = 1028; p = 0.150	W = 1054; p = 0.205	W = 1145; p = 0.524

*one female removed due to motion.

SD: standard deviation; MMSE: Mini-Mental State Examination; CRS: cardiovascular risk score; #F = number of females.

MRI acquisition

The ADNI data were acquired at 3 Tesla (Siemens Prisma). Images were collected using a sagittal MPRAGE sequence with 1.0 mm isotropic resolution, TR: 2.3 s, TE/TI: 2.98/900 ms, and flip angle: 9°, while T2-FLAIR images were collected with 0.86 × 0.86 mm² in-plane resolution, 5 mm thick slices, TR: 9.0 s, and TE: 90.0 ms. Multi-shell dMRI data were collected using spin-echo diffusion-weighted EPI with 127 gradient directions, b-values of 500, 1000, and 2000s/mm², and 11 b = 0 (b0) images. Other image acquisition parameters include TR: 3.4 s, TE: 71 ms, multi-band = 3, and an isotropic voxel size of 2.0 mm. All DICOM data obtained from the ADNI database were converted into NIFTI format using dcm2niix (https://github.com/rordenlab/dcm2niix).

Data pre-processing: WMH and WMH penumbra

WMH volume (in milliliters, ml) was segmented using the Lesion Segmentation Tool (LST version 3.0.0 via SPM12 in MATLAB v. 2023b; https://www.applied-statistics.de/lst.html), which quantifies total WMH burden across the whole brain. LST inputs included T2-weighted FLAIR and MPRAGE images, with an initial threshold value of 0.3, the Markov Random Field (MRF) parameter set to 1, and a maximum of 50 iterations were used. Segmentation outputs were visually inspected for quality control to confirm the absence of gross errors; however, no manual adjudication or rater-dependent adjustments were applied. This approach preserved the objectivity and reproducibility of the automated method, while avoiding operator bias. Subregional classification into periventricular and deep WMHs was not performed, as the segmentation method did not reliably distinguish between these subtypes.

The WMH penumbra was defined as the area surrounding the WMH lesions, marked by subtle signal intensity changes that may reflect early or less severe tissue alterations. To delineate the penumbra, a dilation of 1 mm was applied to the segmented WMH masks using modal dilation (fslmaths). The penumbra region was subsequently refined by subtracting the WMH lesion, gray matter, and cerebrospinal fluid masks to prevent overlap between these regions. WMH and penumbra volumes were then normalized by intracranial volume.

Data pre-processing: dMRI

The NIFTI data were processed using different software tools: Mrtrix3 (https://www.mrtrix.org/),³⁶ FMRIB Software Library (FSL v.6.0),³⁷ and Advanced Normalization Tools (ANTs) (https://github.com/ANTsX/ANTs; v2.3.5). Preprocessing steps for dMRI data included noise reduction with dwidenoise³⁸ (MRtrix3), image alignment, and correction for eddy currents using the eddy tool³⁹ (FSL). For quality control, eddy QC tools were used. Slices impacted by signal loss due subject motion were substituted with predictions derived from a Gaussian process. Inclusion criteria for dMRI data in this study required less than 3 mm average absolute volume-to-volume head motion and less than 5% of total outliers. Brain extraction was performed on the b0 images, and all dMRI images were rescaled to 1.25 mm isotropic voxel sizes (dwi2mask and mrgrid, respectively, Mrtrix3). To analyze dMRI metrics within WMH and NAWM ROIs, the ANTs Symmetric Image Normalization (SyN) co-registration algorithm⁴⁰ registered MPRAGE to the dMRI native space. ROI analysis of dMRI-derived metrics inside WMH and NAWM was performed in individual dMRI native space.

FW-DTI and DKI

FW-DTI and DKI metrics were computed using DIPY (FW-DTI: fwdti.FreeWaterTensorModel class object and fwdtimodel.fit function; DKI: dki.DiffusionKurtosisModel class object and dkimodel.fit function, as well as MeanDiffusionKurtosisModel class and the msdki_model.fit function), along with a custom Python script (version 3.11.2).

The FW-DTI model implemented in this study aims to reduce the adverse effects of PVEs on diffusion measurements by employing a two-compartment diffusion model using multi-shell dMRI data.⁴¹ The derived FW-DTI metrics include FW-corrected fractional anisotropy (FW-FA) and the free-water index (f).

The DKI model quantifies deviation from a Gaussian diffusion distribution, represented through kurtosis metrics such as mean kurtosis tensor (MKT), kurtosis fractional anisotropy (KFA), mean kurtosis (MK), radial kurtosis (RK), and axial kurtosis (AK). To reduce the impact of noise and image artifacts, the signal intensities were also averaged over all diffusion directions for each b-value before calculating the kurtosis, via the mean signal DKI (MSDKI) model.⁴² The MSDKI model quantifies the scalar mean signal diffusion (MSD) and mean signal kurtosis (MSK) metrics. MSD represents the average diffusion coefficient occurring in all directions within a voxel, while MSK represents average diffusion kurtosis across all diffusion directions. Higher MSD values suggest more unrestricted water diffusion, while the MSK metric is indicative of the tissue microstructural complexity and heterogeneity.

Statistical analyses

For each group, demographic information, cognitive scores, and dMRI motion parameters are presented as mean and standard deviation. Following Shapiro-Wilk (SW) test for normality (W = 0.987, p_SW = 0.452), group differences in age were assessed using a Student's t-test. Due to non-normality (p_SW ≤ 0.001), group differences in MMSE, CRS, motion, and outliers were analyzed using the Mann-Whitney U test.

Differences in normalized WMH volume and penumbra volume between groups were assessed using a linear model in R (version 3.6.3) and RStudio (version 1.3.1093), with age and sex included as covariates. Effect sizes (Cohen's d) were also calculated between the CN and MCI groups for WMH, penumbra, and NAWM ROIs. For this dataset, a Cohen's d value of 0.65 represents a large effect size, corresponding to a statistical power of 0.90 at an alpha level of 0.05 for our sample size and study design.

To evaluate group differences between CN and MCI groups, linear models were used with age and sex as covariates to compare the mean values for NAWM, WMH, and penumbra ROIs. Statistical significance was set at p < 0.05, and corrections for multiple comparisons were applied using the False Discovery Rate (FDR) method. In addition, effect sizes were estimated using eta squared (η²), which represents the proportion of variance in the dependent variable explained by group membership. For this dataset, η² values greater than 0.33 were interpreted as large effects (statistical power of 0.90 and α = 0.05 for our sample size and study design).

To assess within-group differences in mean diffusion metrics across WMH, penumbra, and NAWM ROIs, linear mixed-effect models were employed to account for repeated measures within subjects. Pairwise comparisons were performed using Tukey's method, which adjusts for multiple comparisons. Effect sizes for fixed effects were estimated using marginal R² (R_m²), which quantifies the proportion of variance explained by the fixed effects alone, excluding variance attributed to random effects. In this study, R_m² greater than 0.33 was considered a large effect.

The relationship between MMSE scores and both normalized WMH volume and dMRI metrics within WMH was assessed using partial Spearman correlation coefficients (ρ), controlling for age and sex as covariates. Correlation analyses were performed using the ppcor package in R, both on the combined group of CN and MCI participants, as well as separately within the MCI group. Statistical significance was set at p < 0.05. Effect sizes were assessed via the absolute value of Spearman's rank correlation coefficient (ρ), with a large effect defined as ρ > 0.32 in the combined CN and MCI groups analyses and ρ > 0.46 in the only MCI analyses (corresponding to statistical power of 0.90 and α = 0.05).

Results

Due to excessive motion (absolute motion = 4.50 mm), we excluded one female MCI participant from the final analysis. No significant group differences were observed for age (t = −1.162; p = 0.248), CRS (W = 1357.5; p = 0.396), absolute motion (W = 1028; p = 0.150), relative motion (W = 1054; p = 0.205), and total outliers (W = 1145; p = 0.524). MMSE scores were significantly different between groups (W = 1621; p = 0.003), with MCI exhibiting lower scores associated with cognitive impairment. These statistical analyses and outliers are reported in Table 1.

For WMH volumes normalized by intracranial volume, we did not find any significant difference between the CN and MCI groups (t = 0.454; p = 0.651; d = 0.091) (Figure 1). Additionally, no significant differences in the normalized penumbra volumes were observed between groups (t = 0.729; p = 0.468; d = −0.156).

Figure 1.

Violin plots showing normalized white matter hyperintensity (WMH) volume (expressed as a percentage of total white matter volume) in cognitively normal (CN) and mild cognitive impairment (MCI) groups. A linear model with age and sex as covariates showed no significant difference between groups (t = 0.454, p = 0.651, Cohen's d = 0.091).

Figure 2 shows the mean values for all dMRI metrics in WMH, penumbra, and NAWM ROIs in both the CN and MCI groups. Within the WMH and penumbra ROIs, no significant differences were observed between CN and MCI groups for any dMRI metrics. Similarly, no differences were observed between groups for the NAWM ROIs across all dMRI metrics (p > 0.05 and small effect-size for all comparisons).

Figure 2.

Violin plots of mean diffusion MRI metrics within normal-appearing white matter (NAWM), white matter hyperintensities (WMH), and the WMH penumbra (WMHP) for cognitively normal (CN) and mild cognitive impairment (MCI) groups. No significant differences were observed between CN and MCI groups for any dMRI metric within corresponding tissue types (linear model controlling for age and sex). However, significant differences were observed within each group between tissue types (linear mixed-effects model with age and sex as covariates), with t-values reported for pairwise comparisons between NAWM, WMH, and WMHP. *p < 0.05; ***p < 0.001.

However, there were significant differences in all dMRI metrics between the WMH and NAWM ROIs within each group. WMHs were associated with higher f-index and lower FW-FA. Lower DKI metrics (i.e., MKT, KFA, MK, AK, and RK) were also observed in WMHs relative to NAWM. Higher MSD and lower MSK were observed in WMHs across both groups. Notably, higher F values and t-values (from Tukey comparisons) were observed in the CN group relative to the MCI group for all dMRI metrics.

In the WMH penumbra, the f-index, MKT, MK, RK, MSK, and MSD metrics showed intermediate values between WMH and NAWM. For these measures, significant differences were observed between the penumbra and both WMH and NAWM ROIs (all p < 0.05), except for RK, which did not differ between the penumbra and NAWM. On the other hand, FA was significantly higher in the penumbra compared to both WMH and NAWM ROIs (p < 0.001). KFA was also significantly higher in the penumbra than in WMHs (p < 0.001) but did not differ significantly from NAWM. For all metrics except AK, the magnitude of difference between the penumbra and NAWM was smaller than between the penumbra and WMH ROI, as indicated by lower t-statistics.

No significant correlation was found between MMSE scores and WMH normalized volume when analyzing the combined CN and MCI groups (t = −0.045, p = 0.963, ρ = −0.005), nor within the MCI group alone (t = −0.399, p = 0.692, ρ = −0.0622).

Figures 3 and 4 show the correlations between diffusion-related metrics within WMH and MMSE scores for all subjects (Figure 3) and MCI subjects only (Figure 4). For all subjects, no significant correlations were observed for FW-DTI or DKI metrics. For the MCI group only, statistically positive correlations (p < 0.05) were found for MKT and RK with MMSE scores, such that lower kurtosis values were associated with lower MMSE scores.

Figure 3.

Correlations between cognitive performance and diffusion MRI metrics white matter hyperintensities (WMH) for all subjects. Scatterplots show Mini-Mental State Examination (MMSE) scores versus mean diffusion metrics, including free-water index (fw index), free-water corrected fractional anisotropy (fw-FA), mean kurtosis tensor (MKT), kurtosis fractional anisotropy (KFA), mean kurtosis (MK), radial kurtosis (RK), axial kurtosis (AK), mean signal diffusion (MSD), and mean signal kurtosis (MSK). Linear regression lines with 95% confidence intervals (shaded) are shown. For each plot, the regression slope, p-value (p), and Spearman's coefficient (ρ) are reported. No significant correlations were observed across all subjects.

Figure 4.

Correlations between cognitive performance and diffusion MRI metrics white matter hyperintensities (WMH) in the mild cognitive impairment (MCI) group only. Scatterplots show Mini-Mental State Examination (MMSE) scores versus mean diffusion metrics, including free-water index (fw index), free-water corrected fractional anisotropy (fw-FA), mean kurtosis tensor (MKT), kurtosis fractional anisotropy (KFA), mean kurtosis (MK), radial kurtosis (RK), axial kurtosis (AK), mean signal diffusion (MSD), and mean signal kurtosis (MSK). Linear regression lines with 95% confidence intervals (shaded) are shown. For each plot, the regression slope, p-value (p), and Spearman's coefficient (ρ) are reported. * p < 0.05.

Discussion

This study investigated the relationship between WMHs and the associated penumbra, diffusion metrics from FW-DTI and DKI, and MMSE scores in CN and MCI cohorts. These complementary metrics provide valuable insights into the role of microstructural damage associated with WMH and NAWM ROIs.

While the advanced dMRI metrics utilized in this study are highly sensitive to microstructural changes,^33,43 it is important to distinguish between sensitivity to microstructural changes and the direct identification of biological mechanisms. For example, DKI provides critical insights into tissue integrity by revealing non-Gaussian diffusion, which indicates deviations from simple water diffusion patterns typically seen in healthy tissue.⁴⁴ However, DKI and FW-DTI techniques offer indirect measures of tissue complexity and do not provide direct histological evidence of the underlying cellular processes (e.g., demyelination, gliosis, or axonal damage) without complementary pathological data.⁴⁵ The sensitivity of these methods is demonstrated by their ability to detect significant differences in diffusion metrics between NAWM and WMHs, as supported by effect sizes. Despite the challenges of interpreting these changes mechanistically, lower kurtosis values observed in WMHs align with reduced microstructural complexity, a hallmark of neurodegenerative processes.^28,46

Advanced dMRI techniques were utilized to overcome the known limitations of DTI. FW-DTI corrects for PVEs and may provide more accurate DTI metrics, as well as the FW index; these metrics may be more directly indicative of neuroinflammatory responses or axonal degradation processes.⁴¹ In general, microstructural damage manifests as reduced FA values and higher FW values.²³ Kurtosis captures non-Gaussian characteristics of water diffusion; accordingly, the DKI model extends the DTI to model more complex tissue microstructure. Lower kurtosis values indicate less microstructural complexity, which may be associated with neurodegeneration.³¹ The related MSDKI model offers several advantages, such as simplifying data processing and analysis and reducing sensitivity to noise,⁴² which is particularly beneficial at high b-value ranges. By averaging signals across directions, the MSKI model produces the scalar MSD and MSK metrics.³³ However, this directional averaging also results in the loss of orientation information, which limits insight into fiber orientation and directional coherence in white matter. By combining DKI and MSDKI, we leverage the strengths of both models, retaining directional information when needed while enhancing robustness to noise. The integration of FW-DTI further offers complementary markers of neuroinflammation and tissue degeneration. Together, this multiparametric dMRI approach provides a more comprehensive characterization of white matter microstructure.

Robust differences in dMRI metrics between WMHs and NAWM were observed across both CN and MCI groups. These results suggest that WMHs, regardless of cohort's cognitive status, are associated with altered white matter microstructural integrity. The pathological changes associated with WMHs, including demyelination, axonal damage, microvascular dysfunction, inflammation, and tissue edema,⁴⁷ may result in higher FW and lower FA values. The consistency between MSD findings and the FW index supports this interpretation, indicating a coherent pattern of white matter pathology in WMHs. Additionally, reduced DKI values are broadly consistent with reduced microstructural complexity in WMHs.⁴⁸ These findings highlight the importance of accounting for WMHs in dMRI studies, particularly those in aging and cognitive decline.

In addition to changes in WMHs, we observed subtle microstructural changes in the WMH penumbra, revealing further vulnerabilities within areas typically considered NAWM.^19,49 These findings suggest that the WMH penumbra represents a distinct tissue region with intermediate microstructural characteristics, reflecting a gradient of microstructural changes extending outward from visible lesions. While the penumbra more closely resembled NAWM across most metrics, AK was aligned more closely to WMH; as reduced AK may indicate axonal damage,⁵⁰ this finding warrants further investigation. Interestingly, FA was paradoxically increased in the penumbra, which may reflect a compensatory response to the structural damage associated with WMHs or reparative mechanisms leading to localized changes in fiber organization or glial responses. Alternatively, increased FA could indicate tissue vulnerability, representing an early stage of microstructural compromise before overt lesion formation. Future studies investigating the longitudinal progression of these microstructural changes could clarify whether increased FA in the penumbra predicts subsequent WMH expansion or reflects ongoing tissue remodeling.

The absence of significant differences in both WMH volume and mean diffusion values in WMH ROIs between the CN and MCI cohorts may undermine the hypothesis of a direct link between WMH severity and the onset of cognitive decline. It is critical to acknowledge that WMHs manifest across diverse locations within the white matter, each with inherently distinct diffusion characteristics. Consequently, the absence of differential diffusion metrics within WMHs across the two groups might be anticipated, considering the inherent variability in baseline diffusion values across different white matter regions. Additionally, WMHs could be further subtyped into deep WMHs and periventricular WMHs, which have been shown to result from different pathophysiological processes.^51–53

Similar to WMHs, no significant differences in NAWM were observed between CN and MCI cohorts. However, previous studies have revealed subtle yet significant microstructural changes in white matter in MCI groups using dMRI techniques. For example, we previously assessed changes using a voxel-based approach in the same sample, finding significant clusters with higher FW and lower KFA values in the MCI group.³³ Similarly, other studies have found higher FW values in MCI and AD cohorts compared to CN controls.^23,54 Ji et al. showed that FW changes in NAWM, but not WMHs, contribute to dementia severity.⁵⁴ Dumont et al. found that accounting for both brain atrophy and WMHs produced FW values that distinguished between CN, MCI, and AD cohorts.³² Previous studies have also shown that the DKI model can differentiate between individuals with AD, MCI, and those with normal cognition.^31,55,56 In the present study, these differences are likely diluted due to using a NAWM mask across all white matter except the WMHs.

Overall, striking differences in dMRI metrics were observed between WMHs and NAWM, but no between-group dMRI differences were observed in either WMHs or NAWM. Interestingly, from the Tukey comparisons, higher t-values (a measure of the disparity between mean values relative to variability) were observed in the CN group relative to the MCI group for all dMRI metrics. This suggests greater variations in microstructural integrity between WMHs and NAWM in the CN group. In contrast, microstructural damage is present and similar in WMHs and NAWM in patients with MCI. These microstructural similarities between WMHs and NAWM in MCI cohorts indicate that the NAWM is already vulnerable to subtle microstructural changes that may be associated with cognitive decline.

In this study, we did not observe any significant correlation between MMSE scores and WMH normalized volume, either when analyzing the combined CN and MCI groups or within the MCI group alone. This finding suggests that overall WMH volume may not directly relate to global cognitive performance as measured by the MMSE in this cohort. Our results are consistent with previous reports indicating that WMH volume alone may not fully capture the complexity of cognitive decline, as cognitive impairment is likely influenced by a combination of lesion location, microstructural integrity, and other neurodegenerative processes beyond total lesion burden.^8,57

When examining diffusion-related metrics within WMH areas, we did not find any significant associations with MMSE scores in the combined CN and MCI analysis (Figure 3). However, when restricting the analysis to the MCI group only (Figure 4), significant positive correlations were identified between MMSE scores and several kurtosis-based DKI metrics, specifically MKT and RK. These results suggest that within individuals already exhibiting cognitive impairment, lower kurtosis values may be associated with poorer cognitive performance.⁵⁸ In contrast, no significant correlations were found for FW-DTI, MSDKI, KFA, MK, or AK, suggesting that not all diffusion metrics are equally sensitive to cognitive changes within WMH regions. This pattern may indicate that microstructural changes reflected by kurtosis metrics become more closely linked to cognitive function at later stages of impairment, while similar associations are not apparent when including cognitively unimpaired participants.

There are several limitations of note in the present study. Although DKI metrics are considered less susceptible to PVEs than DTI metrics,⁵⁹ they may still be influenced by PVEs. Another limitation is in the use of an automated toolbox to segment WMHs, which did not allow us to delineate between WMH sub-types (e.g., periventricular vs. deep WMHs). As such, we assessed only total WMH volume (and penumbra) without subregional stratification, which may obscure changes in specific WMH subtypes that have distinct pathological or clinical significance. Future studies could investigate advanced dMRI metrics in anatomically refined ROIs to yield more detailed insight into the microstructural variability within different WMH sub-types. Another limitation is that cardiovascular risk factors (reported in Table 1) were not included as a covariates. These factors are well-established confounders in cerebral small vessel disease research and may influence both the burden of WM hyperintensities and cognitive outcomes;⁶⁰ however, CRS did not differ between the CN and MCI cohorts in this study. Future studies with more comprehensive vascular profiling could help clarify potential confounding effects of cardiovascular risk factors on WMH microstructure and cognition. In addition, the relatively small sample size (CN = 55; MCI = 45) may have limited statistical power to detect subtle effects. Finally, longitudinal studies examining dMRI metrics across WMH subregions and penumbra may provide additional insight into the progression of white matter changes.

In conclusion, this study demonstrates that WMHs are associated with altered microstructural characteristics compared to NAWM in both CN and MCI cohorts. These findings underscore the importance of accounting for WMHs when assessing dMRI biomarkers in aging and cognitive decline. While no significant group differences were observed between the CN and MCI cohorts in either WMH volume or dMRI metrics across NAWM, WMH, and WMH penumbra regions, within-group analyses revealed significant microstructural differences between these regions. Post-hoc comparisons showed consistent differences between WMH and both NAWM and penumbra for all metrics, and between NAWM and penumbra for most metrics except KFA and RK. Although no statistically significant differences were found between groups, subtle, non-significant trends suggest possible vulnerabilities within the NAWM of the MCI group. Correlations between dMRI metrics and cognition were not observed in the combined CN and MCI group, but in the MCI group alone, lower kurtosis values were associated with poorer cognitive performance for select metrics. Overall, our results highlight the value of advanced dMRI biomarkers for characterizing white matter changes across tissue types in aging and cognitive decline.

Footnotes

Acknowledgements

Data collection and sharing for this project was funded by the Alzheimer's Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer's Association; Alzheimer's Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer's Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for NeuroImaging at the University of Southern California.

ORCID iDs

Maurizio Bergamino

Megan R Nelson

Elizabeth Keeling

Ashley M Stokes

Ethical considerations

The present study was conducted in accordance with the Helsinki Declaration and was approved at all ADNI sites by local Institutional Review Boards.

Consent to participate

All participants gave written, informed consent as part of the ADNI protocols prior to participation at the participating institutions.

Author contribution(s)

Maurizio Bergamino: Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Resources; Software; Supervision; Visualization; Writing – original draft.

Megan R Nelson: Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Software; Writing – original draft.

Elizabeth Keeling: Data curation; Investigation; Visualization; Writing – review & editing.

Ashley M Stokes: Conceptualization; Formal analysis; Project administration; Resources; Supervision; Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Barrow Neurological Foundation, Sam & Peggy Grossman Family Foundation, Samuel P. Mandell Foundation, and NIH/NIA (P30AG072980).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

All data are accessible without charge from the ADNI database ().

All software utilized in this study is readily accessible on the respective software websites. All in-house codes are available upon request.

References

Alber

Alladi

Bae

H-J

, et al. White matter hyperintensities in vascular contributions to cognitive impairment and dementia (VCID): knowledge gaps and opportunities. Alzheimers Dement (N Y) 2019; 5: 107–117.

Wardlaw

Valdés Hernández

Muñoz-Maniega

. What are white matter hyperintensities made of? Relevance to vascular cognitive impairment. J Am Heart Assoc 2015; 4: 001140.

Marstrand

Garde

Rostrup

, et al. Cerebral perfusion and cerebrovascular reactivity are reduced in white matter hyperintensities. Stroke 2002; 33: 972–976.

Stewart

Stringer

Shi

, et al. Associations between white matter hyperintensity burden, cerebral blood flow and transit time in small vessel disease: an updated meta-analysis. Front Neurol 2021; 12: 647848.

van Dalen

Mutsaerts

HJMM

Nederveen

, et al. White matter hyperintensity volume and cerebral perfusion in older individuals with hypertension using arterial spin-labeling. AJNR Am J Neuroradiol 2016; 37: 1824–1830.

De Guio

Vignaud

Chabriat

, et al. Different types of white matter hyperintensities in CADASIL: insights from 7-tesla MRI. J Cereb Blood Flow Metab 2018; 38: 1654–1663.

Jochems

ACC

Muñoz Maniega

Clancy

, et al. Magnetic resonance imaging tissue signatures associated with white matter changes due to sporadic cerebral small vessel disease indicate that white matter hyperintensities can regress. J Am Heart Assoc 2024; 13: e032259.

Debette

Markus

. The clinical importance of white matter hyperintensities on brain magnetic resonance imaging: systematic review and meta-analysis. Br Med J 2010; 341: c3666.

Zheng

Wang

Chen

, et al. Analysis of risk factors for white matter hyperintensity in older adults without stroke. Brain Sci 2023; 13: 835.

10.

Zhao

, et al. Volume of white matter hyperintensities increases with blood pressure in patients with hypertension. J Int Med Res 2019; 47: 3681–3689.

11.

Dadar

Mahmoud

Zhernovaia

, et al. White matter hyperintensity distribution differences in aging and neurodegenerative disease cohorts. Neuroimage Clin 2022; 36: 103204.

12.

Kim

Hong

Jeon

. Cerebral small vessel disease and Alzheimer’s disease: a review. Front Neurol 2020; 11: 927.

13.

Guo

Shi

. White matter hyperintensities volume and cognition: a meta-analysis. Front Aging Neurosci 2022; 14: 949763.

14.

Power

, et al. Association of white matter microstructural integrity with cognition and dementia. Neurobiol Aging 2019; 83: 63–72.

15.

Le Bihan

. Diffusion MRI: what water tells us about the brain. EMBO Mol Med 2014; 6: 569–573.

16.

Mueller

Lim

Hemmy

, et al. Diffusion MRI and its role in neuropsychology. Neuropsychol Rev 2015; 25: 250–271.

17.

Maniega

Valdés Hernández

Clayden

, et al. White matter hyperintensities and normal-appearing white matter integrity in the aging brain. Neurobiol Aging 2015; 36: 909–918.

18.

Promjunyakul

N-O

Dodge

Lahna

, et al. Baseline NAWM structural integrity and CBF predict periventricular WMH expansion over time. Neurology 2018; 90: e2119–e2126.

19.

Promjunyakul

N-O

Lahna

Kaye

, et al. Comparison of cerebral blood flow and structural penumbras in relation to white matter hyperintensities: a multi-modal magnetic resonance imaging study. J Cereb Blood Flow Metab 2016; 36: 1528–1536.

20.

Kern

Zagzoug

Gottesman

, et al. Diffusion tensor free water MRI predicts progression of FLAIR white matter hyperintensities after ischemic stroke. Front Neurol 2023; 14: 1172031.

21.

Maillard

Carmichael

Fletcher

, et al. Coevolution of white matter hyperintensities and cognition in the elderly. Neurology 2012; 79: 442–448.

22.

Pasternak

Sochen

Gur

, et al. Free water elimination and mapping from diffusion MRI. Magn Reson Med 2009; 62: 717–730.

23.

Bergamino

Walsh

Stokes

. Free-water diffusion tensor imaging improves the accuracy and sensitivity of white matter analysis in Alzheimer’s disease. Sci Rep 2021; 11: 6990.

24.

Archer

Moore

Shashikumar

, et al. Free-water metrics in medial temporal lobe white matter tract projections relate to longitudinal cognitive decline. Neurobiol Aging 2020; 94: 15–23.

25.

Ofori

DeKosky

Febo

, et al. Free-water imaging of the hippocampus is a sensitive marker of Alzheimer’s disease. Neuroimage Clin 2019; 24: 101985.

26.

Bergamino

Keeling

McElvogue

, et al. White matter microstructure analysis in subjective memory complaints and cognitive impairment: insights from diffusion kurtosis imaging and free-water DTI. J Alzheimers Dis 2024; 98: 863–884.

27.

Jensen

Helpern

Ramani

, et al. Diffusional kurtosis imaging: the quantification of non-Gaussian water diffusion by means of magnetic resonance imaging. Magn Reson Med 2005; 53: 1432–1440.

28.

Jensen

Helpern

. MRI Quantification of non-Gaussian water diffusion by kurtosis analysis. NMR Biomed 2010; 23: 698–710.

29.

Arab

Wojna-Pelczar

Khairnar

, et al. Principles of diffusion kurtosis imaging and its role in early diagnosis of neurodegenerative disorders. Brain Res Bull 2018; 139: 91–98.

30.

Schumacher

Ray

Hamilton

, et al. Free water imaging of the cholinergic system in dementia with Lewy bodies and Alzheimer’s disease. Alzheimers Dement 2023; 19: 4549–4563.

31.

Chu

Yan

, et al. Comparison of brain microstructure alterations on diffusion kurtosis imaging among Alzheimer’s disease, mild cognitive impairment, and cognitively normal individuals. Front Aging Neurosci 2022; 14: 919143.

32.

Dumont

Roy

Jodoin

P-M

, et al. Free water in white matter differentiates MCI and AD from control subjects. Front Aging Neurosci 2019; 11: 270.

33.

Nelson

Keeling

Stokes

, et al. Exploring white matter microstructural alterations in mild cognitive impairment: a multimodal diffusion MRI investigation utilizing diffusion kurtosis and free-water imaging. Front Neurosci 2024; 18: 1440653.

34.

Petersen

Aisen

Beckett

, et al. Alzheimer’s disease neuroimaging initiative (ADNI): clinical characterization. Neurology 2010; 74: 201–209.

35.

Biesbroek

Coenen

DeCarli

, et al. Amyloid pathology and vascular risk are associated with distinct patterns of cerebral white matter hyperintensities: a multicenter study in 3132 memory clinic patients. Alzheimers Dement 2024; 20: 2980–2989.

36.

Tournier

J-D

Smith

Raffelt

, et al. MRtrix3: a fast, flexible and open software framework for medical image processing and visualisation. Neuroimage 2019; 202: 116137.

37.

Jenkinson

Beckmann

Behrens

TEJ

, et al. FSL. Neuroimage 2012; 62: 782–790.

38.

Veraart

Novikov

Christiaens

, et al. Denoising of diffusion MRI using random matrix theory. Neuroimage 2016; 142: 394–406.

39.

Andersson

JLR

Sotiropoulos

. An integrated approach to correction for off-resonance effects and subject movement in diffusion MR imaging. Neuroimage 2016; 125: 1063–1078.

40.

Avants

Epstein

Grossman

, et al. Symmetric diffeomorphic image registration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain. Med Image Anal 2008; 12: 26–41.

41.

Hoy

Koay

Kecskemeti

, et al. Optimization of a free water elimination two-compartment model for diffusion tensor imaging. Neuroimage 2014; 103: 323–333.

42.

Henriques

Correia

Marrale

, et al. Diffusional kurtosis imaging in the diffusion imaging in Python project. Front Hum Neurosci 2021; 15: 675433.

43.

Afzali

Pieciak

Newman

, et al. The sensitivity of diffusion MRI to microstructural properties and experimental factors. J Neurosci Methods 2021; 347: 108951.

44.

Steven

Zhuo

Melhem

. Diffusion kurtosis imaging: an emerging technique for evaluating the microstructural environment of the brain. AJR Am J Roentgenol 2014; 202: W26–W33.

45.

Zhuo

Proctor

, et al. Diffusion kurtosis as an in vivo imaging marker for reactive astrogliosis in traumatic brain injury. Neuroimage 2012; 59: 467–477.

46.

Falangola

Jensen

Babb

, et al. Age-related non-Gaussian diffusion patterns in the prefrontal brain. J Magn Reson Imaging 2008; 28: 1345–1350.

47.

Kutzelnigg

Lucchinetti

Stadelmann

, et al. Cortical demyelination and diffuse white matter injury in multiple sclerosis. Brain 2005; 128: 2705–2712.

48.

Liu

, et al. Correlation study between the microstructural abnormalities of medial prefrontal cortex and white matter hyperintensities with mild cognitive impairment patients: a diffusion kurtosis imaging study. Psychiatry Res Neuroimaging 2025; 348: 111958.

49.

Mayer

Nägele

Petersen

, et al. Free-water diffusion MRI detects structural alterations surrounding white matter hyperintensities in the early stage of cerebral small vessel disease. J Cereb Blood Flow Metab 2022; 42: 1707–1718.

50.

Zhu

Zhuo

Qin

, et al. Performances of diffusion kurtosis imaging and diffusion tensor imaging in detecting white matter abnormality in schizophrenia. Neuroimage Clin 2015; 7: 170–176.

51.

Griffanti

Jenkinson

Suri

, et al. Classification and characterization of periventricular and deep white matter hyperintensities on MRI: a study in older adults. Neuroimage 2018; 170: 174–181.

52.

Spilt

Goekoop

Westendorp

RGJ

, et al. Not all age-related white matter hyperintensities are the same: a magnetization transfer imaging study. AJNR Am J Neuroradiol 2006; 27: 1964–1968.

53.

Cai

Sun

Chen

, et al. Different mechanisms in periventricular and deep white matter hyperintensities in old subjects. Front Aging Neurosci 2022; 14: 940538.

54.

Pasternak

Liu

, et al. Distinct white matter microstructural abnormalities and extracellular water increases relate to cognitive impairment in Alzheimer’s disease with and without cerebrovascular disease. Alzheimers Res Ther 2017; 9: 63.

55.

Struyfs

Van Hecke

Veraart

, et al.

Diffusion kurtosis imaging: a possible MRI biomarker for AD diagnosis?

J Alzheimers Dis 2015; 48: 937–948.

56.

Sun

Jiang

, et al. Diffusion kurtosis imaging in evaluating the mild cognitive impairment of occupational aluminum workers. Acad Radiol 2023; 30: 2225–2233.

57.

Prins

Scheltens

. White matter hyperintensities, cognitive impairment and dementia: an update. Nat Rev Neurol 2015; 11: 157–165.

58.

Gong

N-J

Wong

C-S

Chan

C-C

, et al. Correlations between microstructural alterations and severity of cognitive deficiency in Alzheimer’s disease and mild cognitive impairment: a diffusional kurtosis imaging study. Magn Reson Imaging 2013; 31: 688–694.

59.

Yang

Jensen

, et al. Effect of cerebral spinal fluid suppression for diffusional kurtosis imaging. J Magn Reson Imaging 2013; 37: 365–371.

60.

Wang

Chen

, et al. Risk factors of cerebral small vessel disease: a systematic review and meta-analysis. Medicine (Baltimore) 2021; 100: e28229.