The differential relationship between white matter cerebrovascular perfusion and microstructure between normal cognition and mild cognitive impairment

Abstract

Background

Mild cognitive impairment (MCI) has been related to impairment in cerebrovascular perfusion and microstructural magnetic resonance imaging (MRI) parameters. However, the relationship between cerebrovascular perfusion and microstructure remains understudied, especially within normal-appearing white matter (NAWM).

Objective

This study aimed to investigate the differential relationship between white matter cerebrovascular perfusion and microstructure between normal cognition (NC) and MCI using multiple MRI measurements within NAWM.

Methods

We examined differences between 24 MCI and 55 NC participants in NAWM for cerebrovascular perfusion and microstructural measures, including cerebrovascular reactivity (CVR), cerebral blood flow, arterial transit time (ATT), and intracellular volume fraction (ICVF), isotropic volume fraction, and orientation dispersion index in models adjusted for vascular risk factors including hypertension, glycemic status, and the presence of APOE ε4 allele.

Results

In the voxel-wise analyses within NAWM, participants with MCI exhibited lower CVR, prolonged ATT, and lower ICVF. Additionally, ATT and ICVF demonstrated a negative interrelationship in voxel clusters affected by MCI.

Conclusions

Impaired ATT and ICVF in NAWM appeared to be closely interlinked in MCI, while CVR served as an independent imaging biomarker. Further investigation into the intrinsic link between ATT and ICVF in NAWM is warranted.

Keywords

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

Introduction

Associations between cerebrovascular perfusion imaging biomarkers and cognitive status have been demonstrated in previous studies. Impairments in cerebral blood flow (CBF) have shown the potential contribution of CBF alterations to cognitive decline.^1,2 Arterial spin labeling (ASL) magnetic resonance imaging (MRI) provides various cerebrovascular perfusion imaging biomarkers, such as CBF and arterial transit time (ATT). ATT represents the time spent for the ASL bolus to travel from the labeling plane to brain locations of interest and hence forms a map of the transit times. Although ATT is not solely a measure of brain perfusion, as it can be affected by the diameter and tortuosity of the feeding vessels, prolonged ATT may represent impaired perfusion at rest.^3–6 Specifically, prolonged ATT has been observed in individuals with Alzheimer's disease (AD) or mild cognitive impairment (MCI).^7,8

Recent studies suggest that cerebrovascular reactivity (CVR) in response to hypercapnia may be a more sensitive biomarker of perfusion abnormalities than resting CBF.^9,10 CVR is an indicator of the compensatory dilatory capacity of CBF in response to a vasoactive stimulus, such as a hypercapnic challenge. We have shown that decreased CVR may imply poor hemodynamics and is associated with cognitive decline.⁹ Perfusion abnormalities may be particularly important for white matter (WM) even though it has lower perfusion than gray matter (GM) due to its longer ATT.¹¹ This is because periventricular WM receives blood primarily from the watershed areas of the brain, making it especially susceptible to impaired perfusion.^12,13

Few studies have investigated WM perfusion in relation to cognitive status or cognitive performance. However, most studies have focused on white matter hyperintensities (WMHs) because WMH lesions have very low perfusion.^10,14 Although some studies have examined the perfusion of normal-appearing white matter (NAWM) in relation to cognitive impairment and its comparison with WMH,^15,16 direct investigations of NAWM perfusion specifically in MCI and cognitively normal (NC) individuals remain limited. However, understanding impairments in NAWM is important, as they may indicate vascular or microstructural contributions to AD in cognitively impaired individuals and could also serve as precursors of WMH.

Neurite orientation dispersion and density imaging (NODDI) was introduced to estimate the microstructural complexity of axons and dendrites using multi-shell diffusion acquisition.¹⁷ Intracellular volume fraction (ICVF) in NODDI represents axonal density in WM, also known as neurite density index. Isotropic volume fraction (ISOVF) reflects the proportion of free water, such as cerebrospinal fluid or edema, while the orientation dispersion index (ODI) quantifies the variability in neurite orientation, capturing the complexity of fiber organization. Lower ICVF, higher ISOVF, and ODI changes have been observed in MCI in previous studies.^18,19

We hypothesized that a significant correlation exists between impaired cerebrovascular perfusion and the deterioration of NAWM microstructure in cases of MCI. This presumption is underpinned by the higher vulnerability of oligodendrocytes, responsible for the formation of myelin sheaths, to impaired perfusion in comparison to neurons and astrocytes.^20,21 Consequently, in this study, it is hypothesized that impaired cerebrovascular perfusion in WM is related to NAWM structural damage. Furthermore, existing animal studies suggest that cerebral hypo-perfusion induces the loss of oligodendrocytes and WM damage.^22,23 To investigate this, we conducted voxel-wise analyses using multiple MRI measurements. The cerebrovascular perfusion parameters included CBF, ATT, ASL-based CVR ( $C V R_{C B F}$ ), and blood oxygen level dependent (BOLD)-based CVR ( $C V R_{B O L D}$ ). ICVF, ISOVF, and ODI were employed as a WM microstructural measurement.

Methods

This study involves human subjects. The study protocol was approved by the Wake Forest University Health Sciences Institutional Review Board Office. All participants provided their written informed consent to participate in this study.

Study participants

The Wake Forest Alzheimer's Disease Research Center focuses on the contribution of metabolic and vascular factors to AD expression. The study targets middle-aged and older adults (aged 55 and older) who are at a higher risk of AD. Participants for the clinical Core were recruited from a variety of sources, including the Kulynych Memory Assessment Clinic, local neurology, geriatric psychiatry, and primary care practices, as well as through collaborations with community leaders, advertisements, and community presentations. Among these Clinical Core participants, 79 individuals enrolled in additional MRI sequences including T1-weighted, T2-fluid attenuated inversion recovery (FLAIR), dynamic single-post labeling delay (PLD) pseudo-continuous ASL (PCASL), multi-PLD PCASL, and NODDI. The exclusion criteria for the Clinical Core have been described previously.²⁴ Additionally, patients with severe neurodegenerative diseases such as stroke, Parkinson's disease, AD, multiple sclerosis, or those with a recent severe head injury involving loss of consciousness for more than 30 min within the past year, or with permanent neurologic sequelae, were excluded from this study.

Cognitive status

Cognitive status was adjudicated at baseline on all available participants; 24 MCI participants were diagnosed according to National Institute of Aging/Alzheimer's Association criteria.²⁵ Adjudication was determined by consensus from an expert panel comprised of investigators with extensive experience in assessing cognitive status. The panel included neuropsychologists, neurologists, and geriatricians.

Hypertension and impaired glycemic status

Brachial blood pressure was measured to define hypertensive status for each participant using a DINAMAP automated blood pressure device (GE Healthcare, Chicago, IL) in a seated position in a quiet and dark room after a 5-min rest. 45 hypertensive participants were defined as systolic blood pressure (SBP) $\geq$ 130 mmHg, diastolic blood pressure (DBP) $\geq$ 80 mmHg, and/or the current use of antihypertensive medications. The antihypertensive medications were described in a previous study.²⁶ Oral glucose tolerance tests (OGTT) were performed on study participants without diabetes at study entry. Blood glucose was analyzed using the HemoCue whole blood glucose analyzer (HemoCue, Angelholm, Sweden). For those with a diagnosis of diabetes or who refused OGTT, hemoglobin A1c was measured. 35 participants who were prediabetic or diabetic with type 2 diabetes mellitus were considered as impaired glycemic status, which was defined by glucose $\geq$ 140 mg/dL at 120 min of OGTT or hemoglobin A1c $\geq$ 5.7%.²⁶

Apolipoprotein E (APOE) ε4

Apolipoprotein E (APOE) genotype was determined by TaqMan (Thermo Fisher Scientific, Waltham, MA) using single nucleotide polymorphisms.^24,26 APOE ε4 participants were classified by the presence of one or more APOE ε4 alleles. Notably, no participants in this study cohort carried both APOE ε2 and APOE ε4 alleles.

MRI acquisition

Participants underwent MRI scans, including T1-weighted, T2-FLAIR, single-PLD PCASL, multi-PLD PCASL, and NODDI. Experiments were performed on a 3T Siemens Skyra MRI scanner with a 32-channel head coil (Siemens Healthineers, Erlangen, Germany). T1-weighted structural images with full brain coverage were acquired using a three-dimensional volumetric magnetization prepared rapid gradient echo sequence with resolution of 1 × 1 × 1 mm (TR = 2300 ms; TE = 2.98 ms; TI = 900 ms; flip angle = 9 $\circ$ ; field of view (FOV) = 240 × 256). T2-FLAIR images were acquired using a three-dimensional inversion recovery gradient echo sequence with resolution of 1 × 1 × 1 mm (TR = 5000 ms; TE = 383 ms; TI = 1800 ms).

Dynamic single-PLD PCASL images were obtained during resting and computer-controlled $C O_{2}$ gas challenge (2D echo planar imaging (EPI) without background suppression; TR = 4000 ms; TE = 25 ms; matrix = 64 × 64 × 28; flip angle = 90°; FOV = 21 cm × 21 cm; slice thickness = 5 mm; labeling duration = 1.8 s; PLD = 1.2 s, resolution = 3 × 3 × 4 mm). Dynamic single-PLD PCASL provided baseline and hypercapnic CBF. Note that baseline and hypercapnic CBF images were used solely to calculate $C V R_{C B F}$ . Baseline and hypercapnic BOLD signals were also acquired from dynamic single-PLD PCASL by averaging tag and control images. The averaging was performed based on signal magnitude to avoid perfusion signal weighing. Baseline and hypercapnic BOLD images were also used to compute $C V R_{B O L D}$ .

Multi-PLD PCASL images were acquired with a total of eight PLDs with varying bolus. The first three PLDs were 100 ms with different PCASL labeling durations (500, 1100, and 1700 ms, respectively). The other 5 PLDs ranged from 600 to 3000 ms in increments of 600 ms, with PCASL labeling duration of 1800 ms. Each TI had minimum TR (1700, 2300, 2900, 3500, 4100, 4700, 5300, and 5900 ms for the 8 TIs, respectively). In PCASL, TI is the summation of labeling duration and PLD. A 2D single-shot EPI acquisition without background suppression was used to cover the whole brain (matrix size = 56 × 70 × 36, resolution = 3 × 3 × 4 mm). Flow crusher gradients on all three orthogonal axes were applied immediately after an excitation RF pulse to remove flowing signals faster than 10 cm/s. Multi-PLD PCASL provided ATT and CBF.

Diffusion MRI was acquired with the following parameters: 2 mm isotropic voxel; TR = 3500 ms; TE = 106 ms; FA = 90; 9 b0 images; 30 diffusion-weighing directions at b-value = 711 s/mm² and 60 directions at b-value = 2855 s/mm².

Hypercapnic challenge

Inducing a hypercapnic challenge is one of the most robust methods for obtaining CVR. CVR is usually calculated as the percent difference between baseline CBF and hypercapnic CBF. The dynamic single-PLD PCASL acquisition time was 5 min comprising 2 min of the baseline scan, 2 min of the hypercapnic period, and 1 min of recovery back to baseline. The hypercapnic challenge was induced by breathing in $C O_{2}$ -enriched air controlled by an MRI-compatible, computer-controlled gas delivery system (RespirAct, Thornhill Medical, North York, Canada). Hypercapnia was induced by setting the end-tidal $C O_{2}$ ( $P_{E T} C O_{2}$ ) target to 10 mmHg higher than the resting level, while maintaining the end-tidal oxygen levels. The mean resting $P_{E T} C O_{2}$ level for all participants was at 39.64 mmHg (SD of 3.80, ranging from 27.78 to 48.08 mmHg). The mean of hypercapnic $P_{E T} C O_{2}$ level for all participants was at 49.47 mmHg (SD of 4.01, ranging from 37.27 to 58.26 mmHg). A dynamic single-PLD PCASL scan was acquired, and CBF and BOLD changes were recorded during rest and hypercapnia. The baseline and hypercapnic CBF maps were created after averaging the volumes from each respective time period (Figure 1). The CBF quantification process is further described in section 2.8. Additionally, the changes in BOLD during rest and hypercapnia were observed, and the baseline and hypercapnic BOLD maps were created in the same manner (Figure 1).

Figure 1.

Cerebrovascular reactivity measurements using a hypercapnic paradigm. Dynamic PCASL scan allowed simultaneous $C V R_{C B F}$ and $C V R_{B O L D}$ mapping. Per-voxel based CVR images were created. CVR: cerebrovascular reactivity.

Image processing: T1-weighted image

The processing of the structural T1-weighted image comprised file conversion into the Neuroimaging Informatics Technology Initiative (NIFTI) format, normalization to the Montreal Neurological Institute (MNI) template, and tissue segmentation. Image normalization and segmentation were performed using the SPM12 and CAT12 tools.²⁷ The normalization and segmentation processes also provided tissue probability maps (TPMs) for each participant based on TPM templates of SPM12.²⁷ The segmentation process included skull stripping and tissue segmentations for WM. Whole brain GM and WM masks were created for each participant based on the respective TPMs, where GM or WM probabilities were greater than 0.5.

Image processing: single-PLD PCASL, multi-PLD PCASL, ICVF, ISOVF, and ODI

The image processing of the single-PLD PCASL and multi-PLD PCASL images consisted of converting to NIFTI, head motion correction, subtraction between tag and control images, tissue segmentation, partial volume correction (PVC), and normalization. For multi-PLD PCASL, susceptibility distortion was also corrected after NIFTI conversion. Two M0 images were acquired using opposing phase-encoding directions such as left to right and right to left, which were used to determine the susceptibility-induced off-resonance field distortion using the FMRIB Software Library (FSL) topup.²⁸ The spm_realign function from SPM12 was used for head motion correction for both single- and multi-PLD PCASL images.²⁷ The subtraction between tag and control images for both single-PLD PCASL and multi-PLD PCASL provided 4D perfusion-weighted images (PWIs) for each participant. The 4D PWI from the single-PLD PCASL comprised a series of 3D PWIs, while the 4D PWI from the multi-PLD PCASL consisted of a set of 3D PWIs with varying PLDs.

The 3D PWIs of the single-PLD PCASL were evaluated using an ASL kinetic model and then converted into absolute CBF units (ml/100 g/min) for each participant. The cerebrospinal fluid (CSF) signal ( $M_{0 C S F}$ ), estimated from the CSF region of the initial volume of the PCASL scans, was converted to $M_{0 B l o o d}$ using assumed $T_{2} *$ relaxations of blood and CSF to serve as a reference value during CBF quantification.^29–31 The baseline and hypercapnic CBF were derived by averaging the CBF images from their respective time windows, such as the baseline and hypercapnic challenge. BOLD images were created by averaging tag and control images of the head motion-corrected single-PLD PCASL. The baseline and hypercapnic BOLD were also derived by averaging the BOLD images from their respective time windows. The CBF and BOLD images were further processed with tissue segmentation, and PVC. A kernel-based PVC algorithm was applied to each image modality.³² Partial volume corrected ATT and CBF maps were estimated from the 4D PWI of the multi-PLD PCASL using a fast and robust deep-learning method based on our previous study.³

Diffusion MRI images collected with reversed phase-encode blips were corrected for susceptibility-induced off-resonance field distortion using FSL topup,²⁸ and for distortion induced by eddy currents and head motion using FSL eddy.³³ The corrected images were then smoothed with a 3D 2-mm Gaussian kernel. ICVF, ISOVF, and ODI images were estimated using accelerated microstructure imaging via convex optimization from corrected and smoothed NODDI data.³⁴

The normalization process for all images involved aligning with T1-weighted images and then standardizing to the reference image: the MNI template, with resolution of 1.5 $\times$ 1.5 $\times$ 1.5 mm. Co-registration of all images to the T1-weighted images was performed using a rigid body transformation implemented in SPM12.²⁷ For NODDI parameters, the B0 image was co-registered to each subject's T1-weighted image, and the transformation matrix derived from this process was applied to the ICVF, ISOVF, and ODI images. For normalization to MNI space, the deformation field obtained from the spatial normalization of each subject's T1-weighted image to the MNI template was applied. Note that the 1.5 mm isotropic MNI template was provided from CAT12 toolbox of SPM12.²⁷

CVR calculations

Both $C V R_{C B F}$ and $C V R_{B O L D}$ were calculated as the voxel-wise relative difference between baseline and hypercapnic challenge and divided by the change of $P_{E T} C O_{2}$ in mmHg using Equation 1 and 2, respectively.

C V R_{C B F} = 100 \times \frac{\frac{(C B F_{H y p e r c a p n i a} - C B F_{b a s e l i n e})}{C B F_{b a s e l i n e}}}{Δ P_{E T} C O_{2}}

(1)

C V R_{B O L D} = 100 \times \frac{\frac{(B O L D_{H y p e r c a p n i a} - B O L D_{b a s e l i n e})}{B O L D_{b a s e l i n e}}}{Δ P_{E T} C O_{2}}

(2)

WMH segmentation and WMH fractions

The T2-FLAIR images were co-registered to the T1-weighted images using SPM12.²⁷ In this study, WMH segmentation was achieved using a U-Net enhanced with multi-scale highlighting of foregrounds (HF), a method developed in 2021.³⁵ This U-Net with multi-scale HF method provided a WMH mask of each participant in their native space. Since the created WMH masks were probability maps, they were thresholded at a value of 0.5. Each participant's WMH mask was then normalized to the MNI template. It is important to note that each participant's NAWM mask was created by excluding the WMH region from their WM mask. Additionally, a WMH fraction for each participant was calculated based on the ratio between the skull-stripped whole brain volume and their respective WMH volumes.

Experiments and statistical analyses

Participant demographics were compared between MCI and NC using chi-square tests and t-tests. All analyses on image data were performed on 1.5 mm isotropic MNI space including statistical tests and calculations for minimum voxel clusters. For each perfusion and microstructural imaging metric, separate statistical analyses were performed: a global analysis in the whole-brain NAWM and a voxel-wise NAWM analysis. Each participant's NAWM mask was applied to each image modality to exclude WMH regions from the analyses. For the global NAWM analyses, a mean value of each image metric was calculated for each participant. Separate multiple linear regression analyses were conducted to investigate the relationship between each imaging parameter (CBF, ATT, $C V R_{C B F}$ , $C V R_{B O L D}$ , ICVF, ISOVF, or ODI), and cognitive status, adjusted for covariates: age, sex, years of education, and vascular risk factors such as hypertension status, impaired glycemic status, and the presence of APOE ε4 allele. For each regression model, the raw mean value of CBF, ATT, $C V R_{C B F}$ , $C V R_{B O L D}$ , ICVF, ISOVF, or ODI within NAWM served as the outcome variable, while cognitive status, age, sex, years of education, and vascular risk factors were the predictors. Voxel-wise analyses were performed for each imaging parameter using the same manner. Raw voxel values were used as the outcome variable in these analyses. In the voxel-wise analysis, since the WMH regions were defined individually for each participant, the sample size for each voxel of the brain was not always identical. Notably, voxel-wise analysis was performed only when data from over 60 subjects were available, focusing exclusively on NAWM regions.

To explore the inter-relationship between cerebrovascular perfusion and microstructural parameters, we conducted voxel-wise linear regression analyses between perfusion and microstructural parameters separately in MCI and NC groups. These analyses were also adjusted for covariates, including age, sex, years of education, hypertension status, impaired glycemic status, and presence of APOE ε4 allele. Notably, we analyzed only imaging parameters that showed a statistically significant difference between MCI and NC when analyzed individually.

Additional multiple linear regression analyses were performed to determine whether WMH fraction varied by cognitive status, adjusted for the same covariates and vascular risk factors including hypertension status, impaired glycemic status, and the presence of APOE ε4 allele. In these analyses, the log-transformed WMH fractions were used as the outcome variables while cognitive status, age, sex, years of education, and vascular risk factors were the predictors.

The statistical analyses were conducted in MATLAB R2023a with an alpha = 0.05 for all analyses. The minimum cluster sizes of the voxel-wise analyses for each image modality were corrected for multiple comparisons. A Gaussian-shaped autocorrelation function with full width at half maximum was used with family-wise error (FWE)-corrected cluster threshold of p < 0.05. Specifically, 3dFWHMx and 3dClustSim from analysis of functional neuroimages (AFNI) (α=0.05, p < 0.05, nearest neighbor = 3, bi-sided) were used to avoid inflated false positives.³⁶

Results

Demographics and cognitive status

The characteristics of the study population are summarized in Table 1. We observed differences between MCI and NC in age, and years of education. Although the cognitive score was not used in this study, a difference in the Mini-Mental State Examination score between MCI and NC was also observed (Table 1). Additionally, Supplemental Table 1 provides participant demographics with partially available positron emission tomography (PET)-determined amyloid- $β$ deposition status to further characterize the study participants. The PET acquisitions and amyloid- $β$ status assessment procedures were described in a previous study.³⁷ Amyloid- $β$ positivity was classified by trained visual raters.³⁷

Table 1.

Characteristics of study population.

Characteristics	ALL n = 79		MCI n = 24		NC n = 55		p	T-stat/χ²
Characteristics	n/mean	%/SD	n/mean	%/SD	n/mean	%/SD	p	T-stat/χ²
Female, n	61	77%	16	67%	45	81%	0.236	1.404
Age, y	68.7	7.2	72.5	6.6	67.0	6.7	0.002	2.015
Years of education, y	15.6	2.4	14.5	2.3	16.1	2.3	0.005	2.015
APOE $ε$ 4 allele, n	26	33%	9	37.5%	17	31%	0.754	0.098
Hypertension, n	45	57%	17	71%	28	51%	0.162	1.954
Impaired glycemic status, n	35	44%	15	62.5%	20	36%	0.057	3.627
MMSE, score	28.4	1.5	27.5	1.7	28.8	1.2	0.004	−3.117
APOE $ε 2$ allele, n	7	8.9%	2	8.3%	5	9.1%	1	0

p-values were obtained from Chi-square test and t-test for categorical variables and quantitative variables, respectively, for statistical test between MCI and NC.

MCI: Mild cognitive impairment; MMSE: Mini-Mental State Examination.

Cerebral blood flow

In the global NAWM analysis, a non-significant trend of lower CBF ( $β$ = −4.76 ml/100 g/min, p-value = 0.057) was observed in hypertensive participants compared to non-hypertensive participants, while no significant difference was found difference between MCI and NC (Supplemental Table 2). In the voxel-wise analysis, hypertension showed a statistically significant association with lower CBF (Supplemental Figure 1). The covariate-adjusted mean CBF values for the non-hypertensive and hypertensive groups in the voxel-wise analysis were 45.22 ml/100 g/min (95% confidence interval (CI): 41.72 to 48.71 ml/100 g/min) and 36.68 ml/100 g/min (95% CI: 33.89 to 39.47 ml/100 g/min) within the voxel clusters, respectively (Supplemental Figure 1). However, no statistical relationship was found between MCI and CBF in the voxel-wise analyses.

Arterial transit time

Participants with MCI exhibited a non-significant trend toward prolonged ATT at the global NAWM level ( $β$ = 109.79 ms, p-value = 0.056) (Supplemental Table 2). In the voxel-wise analysis, voxel clusters associated with MCI exhibited a statistically significant prolongation of ATT (Figure 2). The voxel clusters were globally distributed within the WM. In contrast, significantly shorter ATT was observed in participants with APOE ε4 allele in the voxel-wise analysis. The covariates-adjusted mean ATT values for participants without and with APOE ε4 allele groups were 1859.16 ms (95% CI: 1785.35 to 1932.97 ms) and 1594.45 ms (95% CI: 1477.86 to 1711.04 ms) within the voxel clusters, respectively (Supplemental Figure 2).

Figure 2.

(A) The voxel clusters that showed statistically significant difference in ATT between NC and MCI participants, adjusted for covariates: age, sex, years of education, hypertension status, impaired glycemic status, and APOE ε4 allele (MNI Z coordinate 53, 56, 59, and 62 are shown from left to right). The color bar indicates the covariates-adjusted t-statistics. The minimum cluster size of ATT analysis was 1790 voxels. (B) The violin plots illustrate the covariate-adjusted ATT of each cognition group within the voxel clusters. The horizontal lines and white dots represent the mean and median values, respectively. ATT: arterial transit time; MCI: Mild cognitive impairment; MNI: Montreal Neurological Institute.

Cerebrovascular reactivity

In the global NAWM level analyses, neither $C V R_{C B F}$ nor $C V R_{B O L D}$ showed a statistically significant relationship with MCI or other vascular risk factors (Supplemental Table 2). In addition, voxel-wise analyses demonstrated no significant statistical relationships between $C V R_{C B F}$ and any vascular risk factors or MCI. In contrast, $C V R_{B O L D}$ revealed two voxel clusters with statistically significant relationships with MCI (Figure 3). The participants with MCI showed lower $C V R_{B O L D}$ within these clusters. The majority of these voxel clusters were located in the left anterior corona radiata, left superior corona radiata, right superior corona radiata, and left posterior corona radiata, based on the Johns Hopkins University (JHU) WM atlas.

Figure 3.

(A) The voxel clusters that showed statistically significant difference in $C V R_{B O L D}$ between NC and MCI participants, adjusted for covariates: age, sex, years of education, hypertension status, impaired glycemic status, and APOE ε4 allele (MNI Z coordinate 63, 66, 69, and 72 are shown from left to right). The color bar indicates the covariates-adjusted t-statistics. The minimum cluster size of $C V R_{B O L D}$ analysis was 1301 voxels. (B) The violin plots illustrate the covariate-adjusted $C V R_{B O L D}$ of each cognition group within the voxel clusters. The horizontal lines and white dots represent the mean and median values, respectively. MCI: Mild cognitive impairment; MNI: Montreal Neurological Institute.

Microstructural parameters

In the global NAWM, participants with MCI showed a non-significant trend toward lower ICVF values ( $β$ = −0.012, p-value = 0.066), while no differences were observed between groups with respect to vascular risk factors (Supplemental Table 2). The voxel-wise analysis identified voxel clusters with statistically significantly lower ICVF values in participants with MCI (Figure 4). These voxel clusters were globally located within the WM. However, no differences in ISOVF or ODI were observed between groups, either in the global NAWM or voxel-wise analyses, with respect to MCI or vascular risk factors.

Figure 4.

(A) The voxel clusters that showed statistically significant difference in ICVF between NC and MCI participants, adjusted for covariates: age, sex, years of education, hypertension status, impaired glycemic status, and APOE ε4 allele (MNI Z coordinate 54, 58, 62, and 66 are shown from left to right). The color bar indicates the covariates-adjusted t-statistics. The minimum cluster size ICVF analysis was 1983 voxels. (B) The violin plots illustrate the covariate-adjusted ICVF for each cognition group within the voxel clusters. The horizontal lines and white dots represent the mean and median values, respectively. ICVF: intracellular volume fraction; MCI: Mild cognitive impairment; MNI: Montreal Neurological Institute.

Associations between cerebrovascular perfusion parameters and ICVF

Voxel-wise analyses of NAWM identified significant associations between MCI and three parameters: ATT, $C V R_{B O L D}$ , and ICVF. Among these, ATT exhibited a substantially greater voxel overlap with ICVF compared to $C V R_{B O L D}$ in participants with MCI. Specifically, the overlap between ATT and ICVF comprised 3204 voxels, whereas the overlap between $C V R_{B O L D}$ and ICVF was limited to 320 voxels. The 3204 overlapping voxels accounted for 17.16% and 19.26% of the impaired ATT and ICVF voxels in NAWM within the MCI group, respectively. Additionally, the overlap between ATT and $C V R_{B O L D}$ included 534 voxels. The overlapping voxels between ATT and ICVF formed three distinct clusters, as illustrated in Figure 5.

Figure 5.

(A) Three overlapping voxel clusters between ATT and ICVF in right-posterior-inferior (RPI) orientations: (A) cluster #1 in blue (MNI coordinate X = 43, Y = 107, Z = 60), (B) cluster #2 in green (MNI coordinate X = 35, Y = 62, Z = 53), and (C) cluster #3 in red (MNI coordinate X = 77, Y = 91, Z = 65). ATT: arterial transit time; MNI: Montreal Neurological Institute.

Voxel-wise linear regression analyses between perfusion and microstructural parameters identified a voxel cluster demonstrating a significant negative association between ATT and ICVF in participants with MCI (Figure 6). Notably, this cluster was located in close proximity to the one formed by the overlapping ICVF-ATT voxels (Figure 5C). The majority of these voxels were concentrated in the left anterior and left superior corona radiata. In contrast, no other voxel-wise linear regression analyses revealed statistically significant associations between perfusion and microstructural parameters in either the MCI or NC groups.

Figure 6.

(A) The voxel clusters that showed statistically significant difference between ICVF and ATT in MCI participants, adjusted for covariates: age, sex, years of education, hypertension status, impaired glycemic status, and APOE ε4 allele (MNI coordinate X = 77, Y = 95, Z = 65). The minimum cluster size of ICVF-ATT analysis was 845 voxels. (B) Illustration of the relationship between ICVF and ATT within the voxel cluster shown in (A). ATT: arterial transit time; ICVF: intracellular volume fraction; MCI: Mild cognitive impairment; MNI: Montreal Neurological Institute.

White matter hyperintensities in MCI

The covariate-adjusted multiple linear regression analyses showed no statistically significant difference in WMH fractions between MCI and NC in this study cohort. WMH fractions had p-values greater than 0.1. There were no vascular risk factors showing statistically significant differences in WMH fractions.

Discussion

In this analysis of CBF, ATT, CVR, and white matter microstructure in well-characterized NC and MCI participants, the results in the NAWM demonstrated that MCI was associated with delayed arterial transit by ATT, lower reactivity to hypercapnia by $C V R_{B O L D}$ , and lower microstructural integrity by ICVF. On the other hand, no statistically significant relationship was found between MCI and $C V R_{C B F}$ , CBF, ISOVF, or ODI. Previous studies have shown consistent results, independently demonstrating a correlation between cognitive status and $C V R_{B O L D}$ , ATT, and ICVF.^7,10,38,39 However, most studies on $C V R_{B O L D}$ and ATT have specifically focused on either gray matter or WMH; as such, brain perfusion in NAWM is understudied, especially in MCI. Moreover, while some studies have investigated the relationship between brain perfusion and microstructural changes, they have primarily focused on CBF and diffusion tensor imaging or diffusional kurtosis imaging measures.^40,41 In contrast, our study not only investigated the independent associations between MCI and cerebrovascular perfusion or microstructure in NAWM, but also examined the relationship between cerebrovascular perfusion parameters from ASL and the microstructural parameter from NODDI.

Among the investigated NODDI parameters, no statistical relationships were found between ISOVF or ODI and MCI, despite previous studies reporting increased ISOVF and ODI changes in MCI.^18,19 Consequently, ICVF may serve as a more sensitive marker for MCI in our study cohort. Lower CBF is typically observed in cases of impaired cognitive performance or MCI.^1,42 However, no differences in CBF were observed in this study cohort. This discrepancy could be due to prolonged ATT. Longer ATT reduces CBF signal, which potentially reduces CBF sensitivity in WM. Even though the implementation of multiple PLDs increased the accuracy and sensitivity of CBF quantification, the prolonged ATT in this cohort might have decreased CBF sensitivity. This is because delayed ATT could lead to underestimated CBF values.⁴³ Furthermore, several studies have shown that ATT can be a more sensitive parameter than CBF for physiological correlations.^7,44,45 It is important to note that some vascular risk factors, such as hypertension and the presence of APOE ε4 allele showed statistically significant association with some cerebrovascular perfusion parameters. Lower CBF was observed in the participants with hypertension. This observation was consistent with previous findings.²⁶ The shortened ATT in participants with APOE ε4 allele needs further investigation, as the direct relationship between the APOE ε4 allele and ATT remains understudied. Although a few studies have explored the association between the APOE ε4 allele and ATT, none have analyzed WM perfusion.^46,47 Shortened ATT in participants with the APOE ε4 allele may represent increased vulnerability to vessel narrowing, stiffening, or both. The absence of observed CBF changes in the presence of the APOE ε4 allele in this relatively healthy cohort may suggest potential compensatory mechanisms to maintain perfusion under cerebrovascular influences. However, this should not be interpreted as evidence of such compensatory mechanisms without further study.

There are several limitations in this study. We utilized a single echo scheme for single-PLD PCASL and BOLD acquisitions to cover a larger brain area in a shorter acquisition time. Typically, ASL scans employ a shorter TE. However, to minimize the impact of BOLD signal fluctuations on the ASL, we strategically collected baseline and hypercapnic scans during the plateau phase, reviewing the acquired data. In addition, the selected PLD of 1.2 s in single-PLD PCASL might not be long enough for WM quantification given that ATT can be 2000 ms or longer in deep WM.⁴⁸ This short PLD limits the assessment of ASL-based $C V R_{C B F}$ of WM. For instance, the mean ATT within the WM was 1899 ms in our study cohort, and 71% of WM voxels exhibited ATT values exceeding the effective PLD. The effective PLD, which accounts for the delay between slice acquisition, ranges from 1.2 s for the bottom slice to 2.145 s for the top slice. In WM voxels where the ATT exceeds the effective PLD, single-PLD PCASL may lead to underestimation of cerebral blood flow and yield suboptimal results, as the labeled blood may not have fully arrived by the time of image acquisition. To capture over 80% of WM voxels with ATT longer than the effective PLD, a PLD of approximately 1886 ms would be required for 2D acquisition. Moreover, due to the brief time window of hypercapnic challenge, we used single-PLD PCASL to provide baseline CBF and hypercapnic CBF for the calculation of $C V R_{C B F}$ . In this study, $C V R_{B O L D}$ showed better sensitivity than $C V R_{C B F}$ in WM in the participants with MCI, which may be attributable to the technical limitations associated with $C V R_{C B F}$ from single-PLD of 1.2 s. In addition, this study was conducted with a relatively small cohort of only 79 subjects with smaller groups of participants for each vascular risk factor. This limitation confines the generalizability of the findings, making them potentially specific to this study cohort. Finally, other factors beyond the included vascular risk factors may contribute to the delayed ATT observed in our MCI cohort. For instance, increased feeding vessel tortuosity or diameter could also prolong ATT.^5,6 However, due to the limited data available, we were unable to explore this in greater depth. Future studies could consider utilizing magnetic resonance angiography to investigate this potential factor further.

The intrinsic physiological correlations between the reported MRI parameters and MCI still need to be explored. In this study, although a direct correlation between cerebrovascular perfusion and WM microstructural integrity was not established, the findings elucidated a notable association between prolonged ATT and microstructural deterioration in MCI. It is pertinent to note that while ATT does not directly quantify perfusion, it offers valuable insights into cerebrovascular dynamics, potentially linking altered blood flow characteristics to changes in WM microstructural integrity. This study thereby contributes to the insightful understanding of the relationship between cerebrovascular perfusion, as indirectly inferred through ATT, and WM microstructural integrity, specifically through ICVF.

In conclusion, this study demonstrated that markers of impaired WM perfusion and microstructure (lower $C V R_{B O L D}$ , longer ATT and lower ICVF) were observed in NAWM of participants with MCI. This study also showed a negative association between ATT and ICVF in the reported voxel cluster (Figure 6). Thus, the findings suggest that $C V R_{B O L D}$ , ATT, and ICVF may serve as more sensitive imaging biomarkers than $C V R_{C B F}$ , CBF, ISOVF, and ODI in NAWM, specifically in revealing the association between cerebrovascular perfusion or microstructure and MCI. Moreover, prolonged ATT and impaired microstructural axonal density appear to be closely interlinked, while $C V R_{B O L D}$ served as an independent imaging biomarker. These observations warrant further research into the inherent nature of physiological connection between ATT and axonal density of WM in the brain.

Supplemental Material

sj-docx-1-alz-10.1177_13872877251360740 - Supplemental material for The differential relationship between white matter cerebrovascular perfusion and microstructure between normal cognition and mild cognitive impairment

Supplemental material, sj-docx-1-alz-10.1177_13872877251360740 for The differential relationship between white matter cerebrovascular perfusion and microstructure between normal cognition and mild cognitive impairment by Donghoon Kim, Sarah Yoon, Timothy M. Hughes, Yu-Chien Wu, Danielle Harvey, Megan E Lipford, Samuel N Lockhart, Suzanne Craft, Laura D Baker, Christopher T Whitlow, Stephanie E Okonmah-Obazee, Christina E Hugenschmidt, Matthew Bobinski and Youngkyoo Jung in Journal of Alzheimer's Disease

Footnotes

ORCID iDs

Donghoon Kim

Sarah Yoon

Timothy M. Hughes

Yu-Chien Wu

Danielle Harvey

Megan E Lipford

Samuel N Lockhart

Suzanne Craft

Laura D Baker

Christopher T Whitlow

Christina E Hugenschmidt

Matthew Bobinski

Youngkyoo Jung

Ethical considerations

This study was approved by the Institutional Review Boards at Wake Forest School of Medicine.

Consent to participate

Informed written consent was obtained from all participants.

Author contribution(s)

Donghoon Kim: Formal analysis; Investigation; Methodology; Writing – original draft; Writing – review & editing.

Sarah Yoon: Data curation; Investigation.

Timothy M Hughes: Investigation; Validation; Writing – review & editing.

Yu-Chien Wu: Validation; Writing – review & editing.

Danielle Harvey: Validation; Writing – review & editing.

Megan E Lipford: Methodology; Writing – review & editing.

Samuel N Lockhart: Data curation; Writing – review & editing.

Suzanne Craft: Funding acquisition; Investigation; Resources.

Laura D Baker: Investigation; Writing – review & editing.

Christopher T Whitlow: Resources.

Stephanie E Okonmah-Obazee: Data curation.

Christina E Hugenschmidt: Investigation; Writing – review & editing.

Matthew Bobinski: Resources.

Youngkyoo Jung: Conceptualization; Investigation; Methodology; Supervision; Validation; Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the Wake Forest Alzheimer's Disease Research Center (NIH P30AG072947) and an NIH Research grant (RF1NS110043).

Declaration of conflicting interests

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Suzanne Craft is an Editorial Board Member of this journal but was not involved in the peer-review process of this article nor had access to any information regarding its peer-review. Danielle Harvey has consulted for NervGen Pharma Corp. Christopher T. Whitlow has consulted for Biogen and Genentech.

Data availability statement

The data and code that support the findings of this study are available on request from the senior author.

Supplemental material

Supplemental material for this article is available online.

References

Bracko

Njiru

Swallow

, et al. Increasing cerebral blood flow improves cognition into late stages in Alzheimer’s disease mice. J Cereb Blood Flow Metab 2020; 40: 1441–1452.

Iturria-Medina

Sotero

Toussaint

, et al. Early role of vascular dysregulation on late-onset Alzheimer’s disease based on multifactorial data-driven analysis. Nature Commun 2016; 7: 11934.

Kim

Lipford

, et al. Parametric cerebral blood flow and arterial transit time mapping using a 3D convolutional neural network. Magn Reson Med 2023; 90: 583–595.

Johnston

Maldjian

, et al. Multi-TI arterial spin labeling MRI with variable TR and bolus duration for cerebral blood flow and arterial transit time mapping. IEEE Trans Med Imaging 2015; 34: 1392–1402.

Dai

Fong

Jones

, et al. Effects of arterial transit delay on cerebral blood flow quantification using arterial spin labeling in an elderly cohort. J Magn Reson Imaging 2017; 45: 472–481.

Farkas

Luiten

. Cerebral microvascular pathology in aging and Alzheimer's disease. Prog Neurobiol 2001; 64: 575–611.

Suo

Lei

Cheng

, et al. Multidelay multiparametric arterial spin labeling perfusion MRI and mild cognitive impairment in early stage Parkinson's disease. Hum Brain Mapp 2019; 40: 1317–1327.

Sun

Wang

, et al. Potential diagnostic applications of multi-delay arterial spin labeling in early Alzheimer’s disease: the Chinese imaging, biomarkers, and lifestyle study. Front Neurosci 2022; 16: 934471.

Kim

Hughes

Lipford

, et al. Relationship between cerebrovascular reactivity and cognition among people with risk of cognitive decline. Front Physiology 2021; 12: 645342.

10.

Blair

Thrippleton

Shi

, et al. Intracranial hemodynamic relationships in patients with cerebral small vessel disease. Neurology 2020; 94: e2258–e2269.

11.

Van Gelderen

De Zwart

Duyn

. Pittfalls of MRI measurement of white matter perfusion based on arterial spin labeling. Magn Reson Med 2008; 59: 788–795.

12.

Wang

Guo

, et al. Reduced perfusion in normal-appearing white matter in mild to moderate hypertension as revealed by 3D pseudocontinuous arterial spin labeling. J Magn Reson Imaging 2016; 43: 635–643.

13.

Lin

Wang

Lan

, et al. Multiple factors involved in the pathogenesis of white matter lesions. BioMed Res Int 2017; 2017: 9372050.

14.

Sam

Crawley

Conklin

, et al. Development of white matter hyperintensity is preceded by reduced cerebrovascular reactivity. Ann Neurol 2016; 80: 277–285.

15.

Meng

Yang

Jin

. Research progress on MRI for white matter hyperintensity of presumed vascular origin and cognitive impairment. Front Neurol 2022; 13: 865920.

16.

Makedonov

Black

MacIntosh

. Cerebral small vessel disease in aging and Alzheimer's disease: a comparative study using MRI and SPECT. Eur J Neurol 2013; 20: 243–250.

17.

Zhang

Schneider

Wheeler-Kingshott

, et al. NODDI: practical in vivo neurite orientation dispersion and density imaging of the human brain. Neuroimage 2012; 61: 1000–1016.

18.

Shrestha

Sun

, et al. Microstructural white matter alterations in mild cognitive impairment and Alzheimer’s disease: study based on neurite orientation dispersion and density imaging (NODDI). Clin Neuroradiol 2020; 30: 569–579.

19.

Wen

Mustafi

, et al. White matter alterations in early-stage Alzheimer's disease: a tract-specific study. Alzheimers Dement (Amst) 2019; 11: 576–587.

20.

Juurlink

. Response of glial cells to ischemia: roles of reactive oxygen species and glutathione. Neurosci Biobehav Rev 1997; 21: 151–166.

21.

Petty

Wettstein

. White matter ischaemia. Brain Res Rev 1999; 31: 58–64.

22.

Farkas

Donka

de Vos

, et al. Experimental cerebral hypoperfusion induces white matter injury and microglial activation in the rat brain. Acta Neuropathol 2004; 108: 57–64.

23.

Kurumatani

Kudo

Ikura

, et al. White matter changes in the gerbil brain under chronic cerebral hypoperfusion. Stroke 1998; 29: 1058–1062.

24.

Coffin

Suerken

Bateman

, et al. Vascular and microstructural markers of cognitive pathology. Alzheimers Dement 2022; 14: e12332.

25.

Albert

DeKosky

Dickson

, et al. The diagnosis of mild cognitive impairment due to Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement 2011; 7: 270–279.

26.

Hughes

Lockhart

Suerken

, et al. Hypertensive aspects of cardiometabolic disorders are associated with lower brain microstructure, perfusion, and cognition. J Alzheimers Dis 2022; 90: 1589–1599.

27.

Penny

Friston

Ashburner

, et al. Statistical parametric mapping: the analysis of functional brain images. London: Elsevier, 2011.

28.

Andersson

Skare

Ashburner

. How to correct susceptibility distortions in spin-echo echo-planar images: application to diffusion tensor imaging. Neuroimage 2003; 20: 870–888.

29.

Wong

. Quantifying CBF with pulsed ASL: technical and pulse sequence factors. J Magn Reson Imaging 2005; 22: 727–731.

30.

Çavuşoğlu

Pfeuffer

Uğurbil

, et al. Comparison of pulsed arterial spin labeling encoding schemes and absolute perfusion quantification. Magn Reson Imaging 2009; 27: 1039–1045.

31.

Buxton

Frank

Wong

, et al. A general kinetic model for quantitative perfusion imaging with arterial spin labeling. Magn Reson Med 1998; 40: 383–396.

32.

Asllani

Borogovac

Brown

. Regression algorithm correcting for partial volume effects in arterial spin labeling MRI. Magn Reson Med 2008; 60: 1362–1371.

33.

Andersson

Sotiropoulos

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

34.

Daducci

Canales-Rodríguez

Zhang

, et al. Accelerated microstructure imaging via convex optimization (AMICO) from diffusion MRI data. Neuroimage 2015; 105: 32–44.

35.

Park

Hong

Duffy

, et al. White matter hyperintensities segmentation using the ensemble U-net with multi-scale highlighting foregrounds. Neuroimage 2021; 237: 118140.

36.

Cox

Chen

Glen

, et al. FMRI Clustering in AFNI: false-positive rates redux. Brain Connect 2017; 7: 152–171.

37.

Grayson

Short

Lee

, et al. T cell exhaustion is associated with cognitive status and amyloid accumulation in Alzheimer’s disease. Sci Rep 2023; 13: 15779.

38.

Cantin

Villien

Moreaud

, et al. Impaired cerebral vasoreactivity to CO2 in Alzheimer's disease using BOLD fMRI. Neuroimage 2011; 58: 579–587.

39.

Vogt

Hunt

Adluru

, et al. Cortical microstructural alterations in mild cognitive impairment and Alzheimer’s disease dementia. Cereb Cortex 2020; 30: 2948–2960.

40.

Niu

Guo

Chang

, et al. The correlation between changes in gray matter microstructure and cerebral blood flow in Alzheimer’s disease. Front Aging Neurosci 2023; 15: 1205838.

41.

Lacalle-Aurioles

Navas-Sánchez

Aleman-Gomez

, et al. The disconnection hypothesis in Alzheimer’s disease studied through multimodal magnetic resonance imaging: structural, perfusion, and diffusion tensor imaging. J Alzheimers Dis 2016; 50: 1051–1064.

42.

Binnewijzend

Kuijer

Benedictus

, et al. Cerebral blood flow measured with 3D pseudocontinuous arterial spin-labeling MR imaging in Alzheimer disease and mild cognitive impairment: a marker for disease severity. Radiology 2013; 267: 221–230.

43.

Fan

Jahanian

Holdsworth

, et al. Comparison of cerebral blood flow measurement with [15O]-water positron emission tomography and arterial spin labeling magnetic resonance imaging: a systematic review. J Cereb Blood Flow Metab 2016; 36: 842–861.

44.

Van Dalen

Mutsaerts

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.

45.

Mutsaerts

Van Dalen

Heijtel

, et al. Cerebral perfusion measurements in elderly with hypertension using arterial spin labeling. PloS One 2015; 10: e0133717.

46.

Dounavi

Mak

Swann

, et al. Differential association of cerebral blood flow and anisocytosis in APOE ε4 carriers at midlife. J Cereb Blood Flow Metab 2023; 43: 1672–1684.

47.

Damestani

Jacoby

Yadav

, et al. Associations between age, sex, APOE genotype, and regional vascular physiology in typically aging adults. Neuroimage 2023; 275: 120167.

48.

Alsop

Detre

Golay

, et al. Recommended implementation of arterial spin-labeled perfusion MRI for clinical applications: a consensus of the ISMRM perfusion study group and the European consortium for ASL in dementia. Magn Reson Med 2015; 73: 102–116.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.59 MB