Alterations in cerebral distal vascular features and effect on cognition in a high cardiovascular risk population: a prospective longitudinal study

Abstract

Background:

Alterations in cerebral vasculature are instrumental in affecting cognition. Current studies mainly focus on proximal large arteries and small vessels, while disregarding morphology and blood flow of the arteries between them (medium-to-large arteries).

Methods:

In this prospective study, two types of non-contrast enhanced magnetic resonance angiography (NCE-MRA) techniques, simultaneous non-contrast angiography and intraplaque hemorrhage (SNAP) and 3D Time-of-flight (TOF), were used to measure vascular morphologic features in medium-to-large intracranial arteries. Grey matter (GM) tissue level perfusion was assessed with arterial spin labeling (ASL) MRI. Twenty-seven subjects at high cardiovascular risk underwent baseline and 12-month follow-up MRI to compare the relationship between morphological features measured by NCE MRA, GM CBF by ASL MRI, and cognitive function measured by the Montreal Cognitive Assessment (MoCA).

Results:

Changes in both global medium-to-large arteries and posterior cerebral (PCA) distal artery length and branch numbers, measured on SNAP MRA, were significantly associated with alterations in MoCA scores (P<0.01), after adjusting for clinical confounding factors, total brain volume, and total white matter lesion (WML) volume. There were no associations between MoCA scores and vascular features on TOF MRA or ASL GM CBF.

Conclusions:

Alterations in vascular features of distal medium-to-large arteries may be more sensitive for detecting potential changes in cognition than cerebral blood flow alterations at the parenchymal level captured by perfusion ASL. Hemodynamic information from distal medium-to-large arteries provides an additional tool to advance understanding of the vascular contributions to cognitive function.

Introduction

Mounting evidence shows that alterations in cerebral vasculature are a contributor to declines in cognitive function, from mild cognitive impairment to dementia (1). Existing studies indicate that such alterations of cerebral vasculature may occur in both micro-and macro-vascular beds, such as rarefaction of microvessels, disruption of blood-brain-barrier, hypoperfusion, increased blood pulsatility, and decreased compliance of proximal large arteries (2,3). However, to date, most of these factors have not been incorporated into clinical use due to a lack of effective measurement techniques and large cohort prospective validation studies.

A recent study by Smith et al. suggests that features of cerebral vasculature, extracted from routinely applicable non-contrast enhanced magnetic resonance angiography (NCE-MRA), can potentially serve as an imaging biomarker for cognitive impairment and provide additional insights over other blood flow measurements (4). More specifically, artery length and branch number on either time-of-flight (TOF) or simultaneous non-contrast enhanced angiography and intraplaque hemorrhage (SNAP) (5), widely used MRA techniques, were significantly associated with Montreal Cognitive Assessment (MoCA) scores (6). In addition, the strength of associations from NCE-MRA are higher than other cerebral blood flow measurements, such as arterial spin labeling (ASL) (7), which is clinically used for measuring parenchyma perfusion level blood flow. The Smith et al. study demonstrates that focusing on different parts of cerebral vasculature may provide better markers to study these relationships. For example, vascular morphological features of distal arteries (i.e., the arteries distal to the A1-A2 connection point, M1-M2 connection point, or P1-P2 connection point) from MRA quantification (8) that cannot be assessed either by proximal flow dynamic features of phase contrast (PC) MRI (9) or capillary blood flow features of ASL. However, due to the cross-sectional nature of this study, it lacked the ability to reveal the prospective associations between changes in features of the cerebral vasculature and change in cognitive function. The predictive value of the NCE-MRA vascular features for changes in cognition needs to be explored and tested.

Hence, the purpose of this study was to test the hypothesis that alterations in cerebral distal medium-to-large arteries on NCE-MRA are potential imaging markers to track cognitive alterations, as determined by changes in a global cognition test named MoCA.

Materials and methods

Participants and cognitive assessment

The local Institutional Review Board approved this prospective study and all participants gave written informed consent prior to enrollment. We recruited 29 subjects as a sub-cohort of participants in an ongoing carotid atherosclerosis study (4). Among them, 27 subjects successfully underwent two scans within a 12-month interval, which is the population of the current study. Inclusion criteria were: at least one carotid artery with>15% stenosis; asymptomatic regarding carotid disease; age ≥18; and no contradictions to MR imaging. The exclusion criteria included severe chronic illness or chronic disability that will limit life expectancy and may lead to incomplete study procedures; history of neck radiation therapy, or with prior bilateral carotid endarterectomy or stenting; pregnant subjects; subjects with systemic inflammatory disease, or atrial fibrillation. The MoCA test is a 30-point cognitive screening measure that includes assessments of orientation, short-term memory, spatial construction, language, attention, and concentration.

MR Imaging Protocol

All MR studies were conducted on a 3.0-T Philips Ingenia CX scanner (Philips, Best, The Netherlands) with a 32-channel array head coil. We used 3D TOF and SNAP to visualize the intracranial arteries. 3D pseudo-continuous ASL (balanced pCASL, RF duration 0.48 ms, RF interval 1.21 ms, RF flip angle 27.81 degrees, maximum gradient strength 5 mT/m, mean gradient strength 0.36 mT/m) was performed to obtain parenchymal blood flow. Detailed imaging parameters for 3D TOF, 3D SNAP, and 3D pCASL are summarized in Table 1. Additional sequences included a conventional 2D T2-w fluid-attenuated inversion recovery (FLAIR) sequence to detect WML, and an emerging multi-contrast sequence named iSNAP (0.8 mm isotropic) (10) to acquire 3D whole-brain T1-w brain anatomical images.

Table 1.

MR protocol and scan parameters.

	3D TOF	SNAP	ASL perfusion
FOV, mm3	190 × 180 × 105	180 × 180 × 70	240 × 240 × 125
Voxel size, mm3	0.5 × 0.5 × 1	0.8 × 0.8 × 0.8	3 × 3 × 5
TR/TE, ms	20/3.5	10/5.6	4220/12
Flip angle, degree	18	11/5	90
Special parameters	Slab number: 6	IRTR: 1814 ms; TI: 500 ms	Repetition times: 8 Labeling duration: 1800 ms PLD: 2000 ms
Readout	SPGR	SPGR	GRASE
Acquisition time, min	6.5 min	3.5 min	5 min

Intracranial vascular feature analysis

The overall intracranial analysis workflow for TOF and SNAP is shown in Fig 1A. IntraCranial artery feature extraction (iCafe) is a custom-made, semi-automatic intracranial vascular map construction software (11), which represents each visible artery on MRA as a varying diameter tube. iCafe has been validated to generate highly reproducible vascular features such as total distal artery length (ICC=0.97, 95% CI: 0.93–0.99) and total branch number (ICC=0.92, 95% CI: 0.83–0.97). With the help of a deep learning-based open curve active contour algorithm (12) for automated artery segment tracing, a reviewer with 12 years of experience in neuroradiology examined each case and then made manual corrections, if needed. Three features were extracted in this study: artery length, branch number, and tortuosity, which is the ratio between artery length and the Euclidean distance of the two terminal points of an individual segment. The above features can be further divided into distal features and proximal features. Arteries distal to the A1-A2, M1-M2, and P1-P2 connection points were defined as distal arteries. With labeled arteries, territorial vascular features are also available, which are anterior artery territory (ACA), middle artery territory (MCA), and posterior artery territory (PCA).

Figure 1. The overall workflow of image post-processing for TOF, SNAP, and ASL perfusion-weighted image.

(A) Workflow in iCafe based method: SNAP is first transformed to a bright blood image and then a pre-trained segmentation model is used for vessel segmentation. A novel deep open snake tracker model is used to trace the vessels for both TOF and bright blood SNAP. The final step is to label different arteries for calculating territorial vascular features. (B) Workflow for CBF post-processing: grey matter (GM) cerebral blood flow (CBF) is segmented from whole-brain CBF using the T1-w image generated from the iSNAP sequence. Then, territorial CBFs (anterior communicating artery, ACA; middle cerebral artery, MCA; and posterior cerebral artery, PCA) are calculated.

ASL CBF Calculation

The overall ASL processing workflow is shown in Fig 1B. We used the recommended implementation of the hemodynamics model (13), and only measured grey matter (GM) CBF, since it has been shown to be more reliable compared to white matter (WM) CBF. An MPRAGE-like T1-w image was first acquired by the iSNAP sequence, then GM was segmented and co-registered with pCASL images to obtain pure GM CBF. The segmentation and registration processes were both conducted on the SPM platform (https://www.fil.ion.ucl.ac.uk/spm).

Brain volume measurement and WML segmentation

Brain volume was automatically measured by summing the volume of GM and WM segmented by SPM. WMLs were segmented based on FLAIR images and iSNAP T1-w images using a lesion segmentation tool (14) and SPM. A radiologist with over 10 years of experience manually corrected the segmentation results using FSL eyes (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FSLeyes), which then automatically calculated the total WML volume.

Statistical Analysis

All statistical analyses were performed by a statistician with more than 10 years of experience using the statistical computing language R (version 3.6.1; R Foundation for Statistical Computing, Vienna, Austria).

Demographic data were analyzed by first testing normality using the Lilliefors test. We applied univariable regression between cerebral vascular features (artery length, branch number, tortuosity) and changes in MoCA scores. Vascular feature changes were considered as independent variables and the change in MoCA scores as a dependent variable. Then, partial linear regression models were used to study the associations between intracranial vascular feature changes and changes in MoCA scores. Specifically, we first obtained the associations by adjusting total brain volume, total WML volume, and age (Model 1), due to these three factors being considered potential factors related to alterations in cognition. Then, additional clinical risk factors (hypertension, diabetes, smoking, obesity) were added as confounding factors, with the aim of further exploring the interactions between varied factors (Model 2). Lastly, Model 1 and Model 2 were further applied to study the relationships between baseline vascular features and changes in MoCA scores. All the significant levels were corrected using the Holm-Bonferroni method (15).

Results

Demographic and Clinical Data

Twenty-seven out of the twenty-nine enrolled subjects had both baseline and 12-month follow-up MoCA tests and MR imaging scans. The subjects’ demographic characteristics are shown in Table 2. For the MoCA scores, according to the MoCA classification criterion (www.mocatest.org), a score below 18 is considered the cut off MCI from dementia, and a score below 26 is considered the cut off to separate subjects with normal cognitive function from those with MCI. Baseline MoCA scores placed 11 subjects in the MCI category. However, only 8 subjects placed in the MCI category and 1 subject moved from MCI to dementia in the follow-up MoCA scores. 10 subjects had decreased scores at follow-up, while 11 subjects had increased scores at follow-up, and 6 subjects scores remained unchanged. Among the 21 subjects with altered MoCA scores, 6 subjects had either an increased or decreased score by over 3 points. The p-value [P = 0.5] of two-timepoint MoCA paired t-test was not significant.

Table 2. Clinical characteristics of the enrolled subjects (N = 27).

Systolic blood pressure, diastolic blood pressure, hypertension was not available for the one same subject.

Variable	Mean ± SD or N (%)
Age	72.5 ± 10.2
Male	19 (70)
BMI (kg/m2)	28.0 ± 3.7
Hypertension	17 (63)
Systolic blood pressure (mm Hg)	145.7 ± 12.5
Diastolic blood pressure (mm Hg)	80.6 ± 10.0
Use of antihypertensive drug	17 (63)
Diabetes	1 (4)
Smoking	13 (48)

BMI: body mass index.

Univariable regression for associations between intracranial vascular feature change and MoCA score changes

After labeling results for each artery, the following SNAP and TOF intracranial vascular feature analyses were further divided into proximal and distal parts. A univariable regression model was first used.

Artery length: Fig 2 shows the associations between SNAP total artery length change, TOF total artery length change, and changes in MoCA scores. We observed a statistically significant association between SNAP total artery length change and changes in MoCA scores: [r = 0.55, P = 0.003]. However, no significant associations were noted between TOF total artery length change and changes in MoCA scores: [r = 0.2, P = 0.321]. For the proximal artery analysis, there were no significant associations between SNAP proximal artery length change and changes in MoCA scores: [r = −0.02, P = 0.900], as well as TOF proximal artery length change and changes in MoCA scores: [r = −0.01, P = 0.969]. For the distal artery analysis, we observed a significant association between SNAP distal artery length change and changes in MoCA scores: [r = 0.55, P = 0.003], but no significant association between TOF distal artery length change and changes in MoCA scores: [r = 0.19, P = 0.344]. The above results indicate that SNAP artery length can detect changes that track with changes in MoCA scores, while TOF cannot detect such changes. Moreover, comparing the proximal and distal analysis results, the intracranial vascular changes that tracked with changes in MoCA scores were mainly from change in distal arteries. Branch number: Fig 2 also shows the associations between SNAP total artery branch number change, TOF total artery branch number change, and changes in MoCA scores. We observed significant association between SNAP total artery branch number change and changes in MoCA scores: [r = 0.71, P < 0.001], while no significant association was seen between TOF total artery branch number change and changes in MoCA scores: [r = 0.03, P = 0.874]. Similarly, only SNAP distal artery branch number change was significantly associated with changes in MoCA scores: [r = 0.72, P < 0.001], while no significant associations were found either in SNAP proximal artery branch number nor TOF distal or proximal artery branch numbers. GM CBF: Fig 2 demonstrates no significant association between ASL GM CBF change and changes in MoCA scores: [r = 0.1, P = 0.619]. Territorial analysis: Significant associations were observed between SNAP PCA artery length [r = 0.69, P < 0.001], SNAP PCA branch number [r= 0.77, P < 0.001] and changes in MoCA scores. No significant associations were found between other SNAP territories [artery length change: ACA: r = 0.31, P = 0.114; MCA: r = 0.39, P = 0.046; branch number change: ACA: r = 0.36, P = 0.063; MCA: r = 0.45, P = 0.018] or all territories of TOF [artery length change: ACA: r = 0.04, P = 0.842; MCA: r = 0.19, P = 0.354; PCA: r = 0.23, P = 0.247; branch number change: ACA: r = −0.06, P = 0.781; MCA: r = 0.01, P = 0.945; PCA: r = 0.00, P = 0.998]. Also, no significant associations between ASL GM CBF [ACA: r = 0.16, P = 0.413; MCA: r = 0.19, P = 0.354; PCA: r = 0.08, P = 0.708] and changes in MoCA scores. Holm-Bonferroni correction was applied to all above results.

Figure 2. Univariable regression associations between the percentage of total, proximal, and territorial distal SNAP artery length change, TOF artery length change, ASL GM CBF change, and changes in MoCA scores (L represents length, B represents branch number).

A significant positive association was found between the percentage of total SNAP artery length change and changes in MoCA scores in the whole population (r = 0.55, P = 0.003), while the percentage of total TOF artery length change and ASL GM CBF change were not significantly associated with changes in MoCA scores. No significant associations were found between the percentage of SNAP and TOF proximal artery length changes and changes in MoCA scores in the whole population. A significant association between the percentage of SNAP distal artery length change and changes in MoCA scores (r = 0.55, P = 0.003) was obtained, while no significant association was seen between TOF distal artery length change and changes in MoCA scores. A significant positive association between the percentage of SNAP PCA branch number change and MoCA score change was obtained (r = 0.7, P <0.01 and still significant after Bonferroni-Holm correction), while the percentage of other SNAP territory artery branch numbers and all territories of TOF artery branch number changes were not significantly associated with changes in MoCA scores.

Partial regression for associations between intracranial vascular features and changes in MoCA scores

Confounding factors including demographic, clinical variables, brain volume, and WML volume were added to further explore the interactions between different confounding factors and changes in MoCA scores. Due to the results shown in the previous section, that artery length and branch number in distal arteries were the significant associations, partial regression models were based on vascular features extracted from distal arteries. After applying confounding factors to the partial regression model, the SNAP distal artery length change [model 1: β = 0.7, P = 0.002; model 2: β = 0.8, P = 0.003], as well as the distal branch number change [model 1: β = 0.8, P < 0.001; model 2: β = 1.0, P < 0.001], still showed significant associations in two models, as shown in Table 3. However, all the baseline vascular features were not associated with MoCA score changes, as shown in Supplementary Table 1. We also investigated the relationship between territorial vascular features and cognition. As shown in Table 4, the SNAP distal artery length change [model 1: β = 0.7, P < 0.001; model 2: β = 0.8, P = 0.002] and distal branch number change [model 1: β = 0.8, P < 0.001; model 2: β = 0.8, P < 0.001] in PCA were significantly associated with changes in MoCA scores, while artery length and branch number change in ACA and MCA territories were not associated with changes in MoCA scores after using Holm-Bonferroni correction.

Table 3.

Linear regression summary between different vascular features and MoCA score change in different sequences.

	Model 1			Model 2
Predictors	β	(95% CI)	P-value	β	(95% CI)	P-value
SNAP, artery length change	0.7	0.3 – 1.1	0.002	0.8	0.3 – 1.3	0.003
TOF, artery length change	0.2	−0.3 – 0.7	0.371	0.3	−0.4 – 0.9	0.389
ASL, GM CBF change	0.2	−0.2 – 0.7	0.336	0.5	−0.2 – 1.2	0.343
SNAP, branch number change	0.8	0.5 – 1.2	<0.001	1.0	0.6 – 1.3	<0.001
TOF, branch number change	0.0	−0.5 – 0.6	0.930	0.0	−0.6 – 0.6	0.958
SNAP, tortuosity change	−0.2	−0.8 – 0.4	0.510	−0.4	−1.1 – 0.4	0.321
TOF, tortuosity change	−0.3	−0.8 – 0.1	0.174	−0.3	−0.9 – 0.3	0.271

^*Holm-Bonferroni correction was used to counteract the problem of multiple comparison. The bold P-values are still significant after correction.

Model 1 = adjusted for age and baseline total white matter lesion (WML) volume and total brain volume

Model 2 = Model 1 + other risk factors (hypertension + diabetes + smoking + obese)

Table 4.

Territory linear regression summary between different vascular features and MoCA score change in different sequences.

	Model 1									Model 2
	ACA			MCA			PCA			ACA			MCA			PCA
	β	(95% CI)	P-value*	β	(95% CI)	P-value	β	(95% CI)	P-value	β	(95% CI)	P-value	β	(95% CI)	P-value	β	(95% CI)	P-value
SNAP, artery	0.2	−0.2 – 0.8	0.201	0.6	0.1 – 1.0	0.012	0.7	0.4 – 1.1	<0.001	0.5	−0.2–1.1	0.130	0.7	0.2–1.1	0.012	0.8	0.3–1.2	0.002
TOF, artery	0.0	−0.5 – 0.5	0.983	0.2	−0.3 −0.7	0.378	0.3	−0.2 – 0.8	0.265	−0.1	−0.7–0.6	0.870	0.2	−0.3–0.8	0.389	0.3	−0.3–1.0	0.257
SNAP, branch	0.3	−0.2 – 0.8	0.168	0.5	0.1 – 1.0	0.0158	0.8	0.4 – 1.0	<0.001	0.4	−0.1–1.0	0.120	0.8	0.3–1.3	0.003	0.8	0.5–1.2	<0.001
TOF, branch	−0.2	−0.7 – 0.3	0.451	0.0	−0.5 – 0.5	0.983	0.1	−0.4 – 0.6	0.753	−0.3	−0.9–0.4	0.390	0.0	−0.6–0.6	0.929	0.1	−0.6–0.7	0.859
SNAP, tortuosity	0.0	−0.6 – 0.7	0.858	−0.3	−0.8 – 0.3	0.339	−0.2	−0.7 – 0.3	0.466	0.0	−0.7–0.8	0.925	−0.3	−0.9–0.4	0.370	−0.5	−1.2–0.2	0.129
TOF, tortuosity	−0.4	−0.9 – 0.2	0.134	−0.1	−0.6 – 0.4	0.708	0.2	−0.3 – 0.7	0.464	−0.5	−1.1–0.1	0.109	−0.1	−0.7–0.5	0.783	0.2	−0.4–0.8	0.466
ASL, GMCBF	0.3	−0.2 – 0.7	0.424	0.3	−0.2 – 0.7	0.216	0.2	−0.3 – 0.6	0.491	0.5	−0.2–1.1	0.122	0.5	−0.1–1.2	0.116	0.4	−0.4–1.2	0.333

^*Holm-Bonferroni correction was used to counteract the problem of multiple comparison. The bold P-values are still significant after correction.

Representative subjects

Three representative subjects with cognitive decline were selected to better present this trend (Fig 3). Artery length and branch number changes in SNAP tracked the change in MoCA scores, while the artery length and branch number of TOF, as well as the GM CBF change measured by ASL, did not consistently track with the change in MoCA scores. Six subjects with an increment of their MoCA scores by three points or greater are shown in Supplementary Fig 1, where vascular feature changes in SNAP track with the change in MoCA scores.

Figure 3. 3D TOF, 3D SNAP, and ASL perfusion-weighted images of three representative subjects.

Blue arrows point out the inconsistent change between SNAP and TOF, ASL GM CBF, while yellow arrows point out the consistent change in SNAP, TOF, and ASL GM CBF. The first subject had a decreased MoCA score from 23 to 18, with consistent artery length reduction in the posterior cerebral artery (PCA) and middle cerebral artery (MCA) territories in TOF and SNAP, along with a global reduction in GM CBF. The second subject had a decreased MoCA score from 29 to 26, with artery length reduction in MCA and PCA territories in SNAP but no obvious artery length change in TOF, and GM CBF decrease in PCA territory. The third subject had a decreased MoCA score from 29 to 22, with obvious artery length reduction in the MCA and PCA territories in SNAP but no obvious change in TOF and GM CBF.

Discussion

In this prospective study, we demonstrated that alterations in cerebral blood flow may be associated with alterations in cognition, specifically SNAP-captured alterations in distal artery length and distal artery branch number were found to be associated with changes in MoCA scores after adjusting for confounding factors. In addition, distal artery length and artery branch number in the PCA and MCA territories also tended to have significant associations with changes in MoCA scores before Holm-Bonferroni correction. The vascular feature changes of the PCA territory were significantly associated with changes in MoCA scores even after the Holm-Bonferroni correction. However, features of TOF and GM, CBF, and ASL did not show significant associations, which may suggest that SNAP is more sensitive than TOF and ASL for capturing changes in vascular features. Our results indicate that alterations in cerebral distal medium-to-large artery length and branch numbers may be associated with alterations in cognition in people at high cardiovascular risk.

Studies have shown that MoCA is more sensitive to cognitive changes in both cognitively impaired as well as healthy elderly subjects compared to MMSE, which is also a widely used cognitive screening test (6). Additionally, MoCA has higher sensitivity than MMSE in detecting MCI (16), which is more suitable for the population in our study. Along with good test-retest reliability for MoCA (0.92), the cognitive fluctuation observed in this study can be considered as a valid change. Although cognitive decline is more common in people at high cardiovascular risk, epidemiological research showed that 16%–50% of patients with MCI may return to normal cognition without targeted interventions (17–18). In our study population, 22% (6/27) of patients had a MoCA increment greater than or equal to three points, while 41% (11/27) of patients had increased MoCA scores, which is in line with the former studies.

Although the pathways of blood flow affecting cognition are still unclear, fluctuations in cognition been demonstrated to be related to vascular health (3, 15). In our study, the blood flow improvement can be observed from SNAP images among all patients with increases in MoCA scores of greater than or equal to three points (Supplementary Fig 1). In other studies, impaired cerebral hemodynamics has been shown to lead to cortical thinning, which may be related to cognition (19–20), and is a process that can be reversed when cerebral hemodynamics are restored (21–22). Restoring the luminal diameter via carotid endarterectomy has been shown to lead to clinical improvement in both cognitively normal and MCI patients, demonstrated by their multidimensional neuropsychological and behavioral assessments, with the improvement significantly stronger in MCI groups in verbal memory and attention scores after a 3-month follow-up test. Sixty percent of MCI subjects returned to normal cognitive function after such an intervention (15).

Alterations in vascular features (artery length, branch number, tortuosity) do not measure alterations in blood flow directly but may be indicators of the factors affecting luminal signal and visibility of distal medium-sized arteries on NCE-MRA. Those features are affected by blood flow, vessel wall condition, and mechanism of imaging technique, where the latter two factors are not likely to change abruptly in a 12-month period. More specifically, the artery length and branch numbers reflect the rarefaction of the cerebral vasculature, while tortuosity reflects the morphological structure of the vessels that may be related to vascular health. SNAP and TOF both measure intracranial arteries, however, their different mechanisms for imaging blood flow may explain the difference in their measurements’ associations to cognition change. The baseline study (4) demonstrated that the intracranial artery length and number of branches of distal arteries on NCE-MRA were found to be positively associated with MoCA scores after adjusting for clinical covariates, brain volume, and WML volume. However, only the artery length and number of branches of TOF were still significant after further adjusting blood flow measured by ASL and PC MRI. Although only the change of artery length and number of branches of distal arteries on SNAP were significantly associated with MoCA score changes, both the baseline and the follow-up results demonstrated the value of vascular morphological features to act as imaging biomarkers for cognitive function. The reason that only vascular morphological changes on SNAP can track the cognitive function changes may be that SNAP is more sensitive to slow blood flow in distal arteries. Previous studies showed that blood flow velocity plays a critical role in lumen signal intensity for both TOF and SNAP, however, it is more sensitive in SNAP than TOF (23–25). SNAP will be affected by the transit time of blood spin in distal arteries, and thus will be more sensitive to blood flow with low velocity in distal arteries (25).

Even though there is an association between distal medium-to-large vascular feature changes and changes in MoCA scores, no association between GM, CBF, and changes in MoCA scores were found. ASL perfusion cannot measure blood flow in distal medium-to-large arteries directly; however, it can measure cerebral blood flow at the parenchymal level, where the blood flow comes from distal medium-to-large arteries. Currently, the association between CBF and cognition remains unclear and controversial. Many studies demonstrated that reduction in global cerebral perfusion can lead to brain dysfunction and cognitive decline in both cognitively intact and cognitively impaired populations (3,26–28). However, it has been shown that GM CBF has significantly positive associations with MoCA scores in a univariable linear regression model but becomes non-significant after adjusting for clinical confounding factors and WML volume size (4). In addition, a large population cognition study found that there was no association between global CBF and cognition in patients with hemodynamic dysfunction along the heart-brain axis (29). Other results from the Mattsson et al. study showed that there was no significant regional CBF difference, except for posterior cingulate cortex precuneus, between controls and early or late MCI (30). As far as we can find, few studies have investigated whether alterations in CBF are associated with alterations in cognitive test scores. The mean GM CBF value (35.6 ml/100g/min) in this study is lower compared to previous reported values in an aged population (31). Our population had a higher mean age and high cardiovascular risk, which is believed to have lower perfusion level compared to normal aging subjects. While there is a possibility of label failure in cases of high carotid stenosis, the underestimation of GM CBF will be consistent between the two time points and will not affect the main findings of the current study. Based on our findings that CBF changes were not associated with changes in MoCA scores, we believe that GM CBF is not sensitive enough to track alterations in cognition in the population of the current study.

Regarding the territory-based analysis, PCA and MCA exhibit significant associations with changes in MoCA scores, and the PCA association persists even after the Bonferroni-Holm correction, while alterations in other territory features measured with TOF and ASL did not show associations with alteration in cognition. We hypothesize that the alterations in the PCA territory features were significantly associated with MoCA score changes because the blood flow alterations in the PCA territory (especially P2 segments) can be captured by SNAP due to its high sensitivity to slow flow, and this slow flow may affect the blood supply to specific regions in the posterior territory. A large population 4D flow study showed that the mean blood flow in the PCA proximal and distal segments had the smallest value compared to other territories (9). More importantly, they showed that the mean blood flow values for normal older middle-aged people, MCI, and those with Alzheimer’s disease decrease along with level of impairment, which may support our SNAP findings.

Among the strengths of the current prospective study, we have compared the alteration in cerebral blood flow in medium-to-large arteries and end-organ vasculature and alteration in cognition using TOF, SNAP, and ASL MRI simultaneously in a population at high cardiovascular risk. Distal medium-to-large artery analysis in this study bridges the gap between proximal large arteries and small vessels and capillary beds. We have advanced the current knowledge on the associations between various parts of vasculature and cognition. Changes in distal arteries are upstream to capillaries and it is possible that changes in distal arteries by iCafe measurements may become detectable before CBF changes in the capillaries. One possible interpretation for the relationship between the distal arteries and capillaries is that distal artery feature changes, due to global cerebral blood flow changes, will cause future hypoperfusion. Specifically, although reduction in perfusion levels will lead to cerebral hypoxia, the alterations in other factors, such as increase in oxygen extraction fraction (OEF), will help maintain CBF in a normal range until reaching the upper limit of the OEF, especially in people with vascular disease (32–33). During this period, the vascular features in distal arteries may be better surrogates to reflect cognition than CBF. Moreover, unlike post-stroke dementia or other cognition studies related to stroke, the current study explored the possibility of whether alterations in cerebral vasculature have already affected cognition before a stroke took place, or if it is a parallel pathway with stroke acting on cognition.

Limitations

The current study has some limitations. First, the sample size was limited. More accurate results can be expected from the partial linear regression model for a larger number of subjects. However, as a pioneer and preliminary study, it opens this field of research by studying the relationship between cerebral blood flow and cognitive function. Second,7 Tesla imaging is superior in delineating small vessels, compared to the 3 Tesla imaging in the current study. Imaging techniques regarding visualization of distal arteries need to be further improved. Third, subjects in this study all have carotid atherosclerosis, which may limit the generalization of the current findings. Additionally, the current one-year follow-up test may be subject to a practice effect on the result of alteration in cognition (34). Additional measurements related to blood flow velocity may add information to current iCafe measures.

Conclusion

The total distal artery length change and total distal artery branch number changes observed in SNAP were significantly associated with changes in MoCA scores in a 12-month follow-up study, while change in cerebral vascular features in TOF as well as GM CBF changes were not associated with changes in MoCA scores. Regarding vascular territory analysis, the distal artery length and branch numbers in the PCA territory were significantly associated with MoCA score changes, while no territory change was associated with MoCA score changes in measurements from TOF and GM CBF. The above relationships were independent of age, total brain volume, total WML volume, hypertension, diabetes, and smoking, suggesting that blood flow information in distal medium-to-large arteries may provide additional information for aiding the understanding of cognition and cerebrovascular health, with SNAP showcasing its potential capability to capture additional blood flow information.

Abbreviation Table:

ACA

Anterior cerebral artery

ASL

Arterial spin labeling

BA

Basilar artery

GM CBF

Grey matter cerebral blood flow

ICA

Intracranial cerebral artery

MCA

Middle cerebral artery

MCI

Mild cognitive impaired

MoCA

The Montreal Cognitive Assessment

MPRAGE

Magnetization-prepared rapid acquisition with gradient echo

NCE-MRA

Non-contrast enhanced magnetic resonance angiography

PC

Phase-contrast

PCA

Posterior cerebral artery

pCASL

pseudo-continuous ASL

SNAP

Simultaneous non-contrast angiography and intraplaque hemorrhage

TOF

Time-of-flight

WML

White matter lesion