The Effects of Intracranial Stenosis on Cerebral Perfusion and Cognitive Performance

Abstract

Background:

Intracranial stenosis (ICS) may contribute to cognitive dysfunction by decreased cerebral blood flow (CBF) which can be measured quantitatively by arterial spin labelling (ASL). Interpretation of CBF measurements with ASL, however, becomes difficult in patients with vascular disease due to prolonged arterial transit time (ATT). Recently, spatial coefficient of variation (sCoV) of ASL signal has been proposed that approximates ATT and utilized as a proxy marker for assessment of hemodynamic status of cerebral circulation.

Objective:

We investigate the association of ICS with CBF and sCoV parameters and its eventual effects on cognition in a memory clinic population.

Methods:

We included 381 patients (mean age = 72.3±7.9 years, women = 53.7%) who underwent 3T MRI and detailed neuropsychological assessment. ICS was defined as≥50% stenosis in any intracranial vessel on 3D Time-of-Flight MR Angiography. Gray matter sCoV and CBF were obtained from 2D EPI pseudo-continuous ASL images.

Results:

ICS was present in 58 (15.2%) patients. Patients with ICS had higher gray matter sCoV and lower CBF. The association with sCoV remained statistically significant after correction for cardiovascular risk factors. Moreover, ICS was associated with worse performance on visuoconstruction, which attenuated with higher sCoV. Mediation analysis showed that there was an indirect effect of ICS on visuoconstruction via sCoV.

Conclusion:

These findings suggest that compromised CBF as detected by higher sCoV is related to cognitive impairment among individuals diagnosed with ICS. We also showed that sCoV partially mediates the link between ICS and cognition. Therefore, sCoV may provide valuable hemodynamic information in patients with vascular disease.

Keywords

Arterial spin labeling cerebral perfusion coefficient of variation cognition intracranial stenosis

INTRODUCTION

The pathogenesis of Alzheimer’s disease and its prodromal stages were traditionally attributed to ove-rproduction and deposition of amyloid-β in cerebral vessels and parenchyma [1]. Growing body of evidence suggests that cerebral hypoperfusion resulting from vascular disease may also lead to neuronal death and hence may serve as useful biomarkers of cognitive decline [1, 2].

In vivo assessment of cerebral blood flow (CBF) via arterial spin labelling (ASL) is a promising non-invasive tool for early detection of ‘vascular causes’ to cognitive impairment [3, 4]. However, CBF measurement with ASL is affected by the arterial transit time (ATT), the time it takes the labeled blood to travel from the labeling plane to an imaged voxel [5]. Because of the rapid T1-decay of the labeled blood, ASL timing parameter, i.e., the post-labeling delay (PLD), are a tradeoff between signal-to-noise ratio and arrival of labeled blood in the distal br-ain tissue [6]. As such, the interpretation of the ASL perfusion measurement becomes unreliable in patients with large vessel disease. Recently, the spa-tial coefficient of variation (sCoV) is proposed to distinguish between the complete arrival of the labeled blood-identified by low sCoV due to a homogeneous label distribution within gray matter, and label presence within the proximal vessels-with high sCoV from vascular patterns in combination with relatively low ASL signal in distal regions [5]. As it has been suggested previously, the ability of the extracranial and intracranial cerebral vasculature to deliver labelled blood to the parenchyma, as captured by sCoV, could be a proxy parameter of global vascular health and be indicative of vascular pathology. This is further confirmed by its strong correlation with ATT as well as with several indicators of cerebrovascular pathology, e.g., white matter hyperintensity volume [7, 8], age [7], sex [7], and cortical microinfarcts [9]. However, the utility of sCoV as a proxy marker for assessment of the hemodynamic status of cerebral circulation needs further investigation.

We investigated the association between ICS and perfusion, quantified by two ASL parameters (CBF and sCoV) and their effects on cognitive performance in a memory-clinic population. We further determined whether decreased perfusion is the key link between ICS and cognition.

MATERIALS AND METHODS

Study population

This study was conducted as part of an ongoing prospective memory-clinic study which recruits pa-tients from National University Hospital in Singapore. Three diagnostic categories at baseline were considered eligible for study inclusion [10]: 1) No cognitive impairment (NCI): individuals who had no objective cognitive impairment on neuropsychological tests, or functional loss; 2) Cognitive impairment no dementia (CIND) patients who were impaired in at least one cognitive domain on a neuropsychological test battery without loss of daily functions; and 3) Dementia was diagnosed according to Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition (DSM-IV) criteria.

Recruitment took place from August 12, 2010 to September 14, 2017. From the total of 582 patients, 19 had no MRI scans, 26 had no magnetic resonance angiography (MRA), and 29 lacked an ASL sequence giving a sample of 508 for ASL image analysis. All study participants underwent clinical and physical assessments, MRI scanning, and cognitive testing on the same day.

Ethics approval for this study was obtained from National Healthcare Group Domain-Specific Review Board. Written informed consent was obtained, in the preferred language of the participants, by bilingual study coordinators prior to recruitment.

Neuroimaging

MRI was performed on a 3T Siemens Magnetom Trio Tim scanner, using a 32-channel head coil, at the Clinical Imaging Research Centre of the National University of Singapore. A three-dimensional (3D) T1-weighted image was acquired with the following parameters: 1.0×1.0×1.0 mm³ voxels; repetition time (TR) 2300 ms; echo time (TE) 1.9 ms; inversion time (TI) = 900 ms; flip angle = 9°; matrix 256×256. The parameters of the three-dimensional time-of-flight MRA images (3D TOF MRA) were: TR = 22 ms, TE = 3.4 ms, flip angle = 20°, 192 mm field of view (FoV), 218×256 acquisition matrix, slice thi-ckness = 0.80 mm, and an acquisition time of 6 min and 28 s. The pseudo-continuous arterial spin labeling (pCASL) with a 2D gradient-echo echo planar imaging readout was used with the following parameters: voxel size = 3 x 3 x 5 mm³, 24 slices, labeling duration = 1656 ms, initial post-labeling delay = 1500 ms, slice readout time = 49.94 ms, leading to a PLD range of 1500–2649 ms across all slices or a mean PLD of 2074 ms, TR/TE=4000/9 ms, and generalized auto-calibrating partially parallel acquisitions (GRAPPA) factor = 3. Two ASL volumes of 23 control-label pairs each were acquired with a 1-h interval and were concatenated into one ASL time series to decrease physiological fluctuations.

Intracranial stenosis on MRA

ICS was defined based on the criteria published previously in our memory clinic population [11]. Briefly, ICS was defined as arterial narrowing exc-eeding 50% of the luminal diameter in any of the following intracranial arteries: vertebral (segment V4), basilar, internal carotids (ICA), posterior cerebral (PCA) (segments P1 and P2), posterior communicating, middle cerebral (MCA) (segments M1 and M2), and anterior cerebral arteries (ACA) (segments A1 and A2), as assessed on 3D TOF MRA images. We also assessed if the vessels were stenosed on one side (unilateral) or both sides (bilateral). The vessels were first visually assessed on the reconstructed coronal images. The final decision was based on the MRA source images. One rater (SH), who was bli-nded to clinical data, graded each participant’s MRA. The previously established intra-rater reliability was excellent (intraclass correlation coefficient, ICC = 0.88). We also carefully examined the presence and origins of all branches of the vertebrobasilar system and the anterior cerebral arteries where normal variations are common. The diameters of the anterior and posterior circulation were examined on the axial plane as well as on reconstructed coronal images to distinguish the normal variations from stenoses. Typically, an abrupt or tapered occluded lumen and an associated increase in external diameter were main features used to distinguish ICS from normal variations. The diameter of the branches was measured 0.5 cm distal to their origins. A difference in the diameter of the distal vertebral arteries of more than 1 mm was accepted as the criterion for determining left or right vertebral dominance (10% had left vertebral dominance whereas 3% had right vertebral dominance). If the difference was smaller than or equal to 1 mm, it was accepted as codominant (87% of the cases). Vessels visualized as continuous segments of at least 0.8 mm in diameter were considered as hypoplastic. In one case the left vertebral artery and in three cases the right vertebral artery ended as the posterior inferior cerebellar artery. The anterior circle was complete in 85% of the cases where as it was incomplete in 15% of the cases out of which 10.8% had absent anterior communicating artery and 4.2% had hypoplastic or absent A1 segment. Normal variations of MCA and ICA were not found in any case.

Other MRI markers

Cortical infarcts were identified as focal lesions involving cortical gray matter with hyperintense rim on FLAIR images, with a center similar to cereb-rospinal fluid intensity, loss of tissue of variable magnitude, as well as prominent adjacent sulci and ipsilateral ventricular enlargement. Cerebral microbleeds were defined as focal rounded hypointense lesions on susceptibility-weighted images with a blooming effect, and were graded using the Brain Observer Microbleed Scale [12]. Cortical microinfarcts were defined as cortical lesions of 5 mm or less in diameter, perpendicular to the cortical surface, hypointense on T1-weighted images, hyperintense on T2-weighted images and hyperintense or isointense on FLAIR images [13].

ASL image processing

ASL post-processing was performed with the Ex-ploreASL software based on SPM and Matlab (Ma-thWorks, MA) [14, 15]. Image processing included motion correction and rigid-body registration of the CBF map to a gray matter map and normalized into a common space using Diffeomorphic Anatomical Registration analysis using Exponentiated Lie algebra (DARTEL) [16]. From the CBF map, two ASL parameters were acquired: CBF and sCoV. CBF reflects perfusion in mL blood/100 g tissue/min and was calculated in the gray and white matter regions of interest (ROIs). The ROIs were defined as voxels that contained more than 70% of gray matter and white matter, respectively, as assessed on segmented T1-weighted images. For white matter ROI, an outer layer of 6 mm was additionally removed to exclude contamination by gray matter CBF [17]. sCoV, which was shown to serve as a proxy of ATT, was defined as the standard deviation (SD) of the CBF/mean CBF within the region of interest [7]. The CBF and sCoV analyses were obtained from anterior and posterior flow territories, which were defined as territories corresponding to the anterior and middle cerebral artery, and posterior cerebral artery, respectively, as obtained from the atlas [18]. The absolute asymmetry index (AI) was calculated for CBF and sCoV separately using the following formula: $AI = \frac{ASL (left) - ASL (right)}{0.5 (ASL (left) + ASL (right))}$ (1)

Both left- and right-sided lateralization will lead to higher AI, with zero value representing a symmetrical CBF or sCoV. Quality assessment of the ASL scans was performed by two independent raters (DF and HM) based on visual assessment. Scans were classified into 1 of the 4 categories: 1) Unusable - scans that were incomplete had labeling errors or severe motion artifacts (n = 127), 2) Angiogram - scans with predominantly vascular contrast, and no or minimal tissue perfusion (n = 54), 3) Acceptable - scans that had minor artifacts and reasonable tissue perfusion (n = 161), or 4) Good (n = 166). Patients with unusable scans were excluded from all analyses. For the sCoV analysis, a total of 381 patients were used, including patients with angiogram, acceptable, and good scans. For the CBF analysis, the 54 patients who had an angiogram-like ASL scan were excluded from analysis, resulting in a study population of 327. An-giogram-like scans showed limited information about CBF and thus need to be excluded from the CBF analysis. On the other hand, the amount of vascular artifacts and the ratio of vascular artifacts to perfusion still contain information about the arterial transit time and can be thus used in the sCoV analysis.

Neuropsychological test battery

Trained research psychologists administered brief cognitive tests: the Mini-Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MoCA) and a formal detailed neuropsychological test battery that has been locally validated in Singapore. This battery assesses seven domains, five of which are non-memory domains.

The non-memory domains tested were: Executive function: Frontal Assessment Battery and Maze Task; Attention: Digit Span, Visual Memory Span, and Auditory Detection; Language: 15-item Boston Naming Test and Verbal Fluency; Visuomotor speed: Symbol Digit Modality Test, Digit Cancellation; and Visuoconstruction: Weschler Memory Scale-Revised (WMS-R) Visual Reproduction Copy task, Clock Drawing, and Weschler Adult Intelligence Scale-Re-vised (WAIS-R) subtest of Block Design.

The memory domains tested were: Verbal memory: Word List Recall and Story Recall; and Visual memory: Picture Recall and WMS-R Visual Reproduction.

The assessment was done in the participant’s habitual language and was completed in approximately an hour. For each participant, raw scores from each individual test within a domain were first transformed to standardized Z-scores using the mean and standard deviation [SD] of that test in this cohort. A higher Z-score reflected a better performance on that test. Subsequently, for each participant a mean Z-score for each domain was calculated by averaging the Z-scores of all the individual tests within that domain. These mean Z-scores of each domain were then standardized using the mean and SD of that domain-specific mean Z-score. Finally, composite Z-score reflecting global cognitive functioning was calculated by averaging the seven domain-specific mean Z-scores, which were also standardized using the corresponding mean and SD [19].

Covariates assessment

A detailed questionnaire was collected from all patients to obtain information on age, sex, race, education, and smoking history. Previous medical history of hypertension, hyperlipidemia, and diabetes mellitus together with the medications were also noted and subsequently verified by medical records. Physical examination included height, weight, and blood pressure. Systolic and diastolic blood pressures were measured using a digital automatic blood pressure monitor after the patient rested for five minutes. Blood pressure was measured twice, five minutes ap-art. The mean of two readings was considered as the relevant blood pressure. Hypertension was defined as previous diagnosis of hypertension by a physician or use of antihypertensive medication. Diabetes mellitus was defined as a previous diagnosis of diabetes by a physician or use of anti-diabetic medication. Hyperlipidemia was defined as a previous diagnosis of hyperlipidemia by a physician or use of lipid lowering medication. Smoking was categorized into ever (past and current smokers) and never smokers.

Statistical analyses

Differences in the characteristics between patients with and without ICS were performed using chi-squ-ared test for categorical variables and Mann-Whitney U-test for continuous variables. The association between ICS (presence/absence) and ASL parameters (CBF and sCoV) as well as asymmetry index were determined using linear regression models. We also examined the association of ICS in the anterior (ACA and MCA) and posterior circulations (vertebral, basilar, PCA and posterior communicating) and internal carotid arteries with ASL parameters. We again used regression analyses to determine associ-ation between ICS in anterior, middle, posterior and internal carotid arteries and sCoV and CBF in the respective arterial territories. Similarly, linear regression analyses were constructed between ICS and cognition. In all regression models, ICS was treated as the exposure and ASL parameters, asymmetry in-dex and cognition were treated as outcomes. All the models were initially adjusted for age and sex (model I), and the cognition analysis was further adjusted for education. Both models were additionally adjusted for smoking, systolic blood pressure, hypertension, hyperlipidemia, diabetes and diagnosis (model II). The final model was further adjusted for cortical infarcts and additionally for cerebral microbleeds and cortical microinfarcts in the cognition analysis (model III) to determine the independent effects of ICS on cognition. We performed mediation analysis after examining which cognitive domain was significantly associated with ICS. Mediation analysis was conducted between ICS and cognitive domain using ASL parameters as mediators. Bootstrapping was applied (1000 times) and significance was based on the upper and lower bootstrapped 95% CI. For mediation analysis, we controlled for potential confounding effects from age, sex, education, smoking, systolic blood pressure, hypertension, hyperlipidemia, diabetes, diagnosis, cortical infarcts, cortical microinfarcts, and cerebral microbleeds. All statistical analyses were performed using SPSS software (v.25.0, IBM, USA) and mediation analyses were conducted using PROCESS macro.

RESULTS

A total of 58 (15.2%) patients had ICS. Of these, 40 (68.9%) patients had single artery stenosis and 18 (31.3%) had multiple artery stenosis (≥2 arteries). The frequency of stenosis in different arteries among patients were: ICA: 15 (25%), MCA: 15 (25%), ACA: 5 (8.6%), PCA: 5 (8.6%), vertebral: 30 (51.7%), and basilar: 6 (10.3%). Unilateral stenosis (left sided) was present in 17 patients (4.5%), right sided unilateral stenosis was present in 22 (5.8%) and 19 (4.9%) patients had bilateral stenoses. Patients who were excluded (n = 201, based on no MRI (19 patients), no MRA (26 patients), no ASL sequence (29 patients), and unusable ASL scans (127 patients), were older, less educated, had a higher prevalence of hypertension and more likely to consumed antihypertensives and antihyperlipidemics. Moreover, the excluded group was more likely to smoke, more cognitively impaired (global and domain specific cognitive performance) and had a diagnosis of dementia compared to included patients (n = 381) (Table 1).

Table 1

Characteristics of included and excluded participants

Variables	Included (N = 381)	Excluded (N = 201)^*	p
Age, y, mean (SD)	72.1 (7.9)	74.6 (7.7)	0.001
Men, no. (%)	166 (46)	77 (43.5)	0.587
Education, y, mean (SD)	7.45 (5.03)	5.75 (4.77)	<0.001
Hypertension, no. (%)	246 (68.1)	137 (77.4)	0.026
Diabetes, no. (%)	123 (34.1)	66 (37.3)	0.463
Hyperlipidemia, no. (%)	266 (73.7)	134 (75.7)	0.614
Systolic blood pressure,	142.9 (18.8)	145.9 (20.1)	0.096
mmHg, mean (SD)
Antihypertensives, no. (%)	217 (56.9)	134 (66.6)	0.010
Antihyperlipidemics, no. (%)	241 (63.3)	132 (65.6)	0.020
Antidiabetic drugs, no. (%)	104 (27.3)	64 (31.8)	0.009
Smoking, ever, no. (%)	81 (21.3)	53 (26.4)	0.034
Cortical infarcts, no. (%)	46 (12)	30 (14.9)	0.036
Cerebral microbleeds, no. (%)	196 (51.4)	68 (33.8)	0.824
Cortical microinfarcts, no. (%)	55 (14.4)	30 (14.9)	0.048
Global cognition, mean (SD)	0.24 (0.93)	–0.48 (0.96)	<0.001
Executive function, mean (SD)	0.21 (0.91)	–0.43 (1.02)	<0.001
Attention, mean (SD)	0.21 (0.91)	–0.42 (1.04)	<0.001
Language, mean (SD)	0.22 (0.90)	–0.43 (1.04)	<0.001
Verbal memory, mean (SD)	0.20 (0.99)	–0.40 (0.88)	<0.001
Visual memory, mean (SD)	0.19 (0.98)	–0.39 (0.91)	<0.001
Visuoconstruction, mean (SD)	0.24 (0.92)	–0.47 (0.99)	<0.001
Visuomotor speed, mean (SD)	0.23 (0.96)	–0.46 (0.91)	<0.001
Diagnosis
NCI, no. (%)	95 (24.9)	21 (10.4)	<0.001
CIND, no. (%)	170 (44.6)	71 (35.3)
Dementia, no. (%)	116 (30.4)	109 (54.2)

SD, standard deviation; no, number; mmHg, millimeter of mercury. ^*The excluded group included 19 subjects with no MRI, 26 with no MRA, 29 with no ASL sequence and 127 with unusable ASL scans.

Table 2 shows the comparison between patients with and without ICS. Men had higher burden of ICS compared to women (18.7%, p = 0.01). Patients with ICS were likely to be hypertensive (20.5% higher), diabetic (18.3%), hyperlipidemic (16.1%), and had higher systolic blood pressure. Moreover, patients with ICS were more likely to consume antidiabetics, were smokers, had higher prevalence of cortical infarcts on the scans and had a higher sCoV (19.4%) and reduced CBF (18.1%) than patients without ICS (p < 0.013 for all). Patients with ICS also had lower scores in the visuoconstruction domain compared to patients without ICS (p = 0.028) and were likely to be diagnosed with dementia (p≤0.001).

Table 2

Characteristics of the study population (n = 381)

Variables	ICS (Presence) (N = 58)	ICS (Absence) (N = 323)	p
Age, y, mean (SD)	72.6 (9.3)	71.9 (7.7)	0.609
Men, no. (%)	34 (61.8)	132 (43.1)	0.010
Education, y, mean (SD)	7 (5.28)	7.53 (4.98)	0.473
Hypertension, no. (%)	47 (85.5)	199 (65)	0.003
Diabetes, no. (%)	27 (49.1)	96 (31.4)	0.011
Hyperlipidemia, no. (%)	48 (87.3)	218 (71.2)	0.013
Systolic blood pressure,	148.3 (18.2)	141.9 (18.8)	0.024
mmHg, mean (SD)
Antihypertensives, no. (%)	38 (65.5)	179 (55.4)	0.153
Antihyperlipidemics, no. (%)	41 (70.7)	200 (61.9)	0.202
Antidiabetic drugs, no. (%)	23 (39.7)	81 (25.1)	0.022
Smoking, ever, no. (%)	18 (34)	63 (21.2)	0.043
Cortical infarcts, no. (%)	17 (29.3)	29 (9)	<0.001
Cerebral microbleeds, no. (%)	38 (65.5)	158 (49.1)	0.021
Cortical microinfarcts, no. (%)	16 (38.1)	39 (23.9)	0.065
sCoV of gray matter, median (IQR)	0.86 (0.46)	0.72 (0.28)	<0.001
CBF of gray matter, median (IQR)	40.9 (18.0)	48.3 (20.5)	0.003
Global cognition, mean (SD)	–0.18 (0.93)	0.03 (1.01)	0.128
Executive function, mean (SD)	–0.17 (1.02)	0.03 (0.99)	0.151
Attention, mean (SD)	–0.38 (1.02)	0.01 (0.99)	0.752
Language, mean (SD)	–0.19 (0.81)	0.03 (1.03)	0.114
Verbal memory, mean (SD)	–0.11 (0.84)	0.02 (1.02)	0.347
Visual memory, mean (SD)	–0.16 (0.87)	0.03 (1.01)	0.183
Visuoconstruction, mean (SD)	–0.26 (1.02)	0.04 (0.98)	0.028
Visuomotor speed, mean (SD)	–0.21 (0.97)	0.04 (1.00)	0.092
Diagnosis
NCI, no. (%)	10 (17.2)	85 (26.3)	<0.001
CIND, no. (%)	27 (46.5)	143 (44.2)
Dementia, no. (%)	21 (36.2)	95 (29.4)

ICS, intracranial stenosis; SD, standard deviation; no, number; mmHg, millimeter of mercury; IQR, interquartile range; sCoV, spatial coefficient of variation; CBF, cerebral blood flow.

Figure 1 shows higher sCoV in patients with ICS (Fig. 1A) and low sCoV in patients without ICS (Fig. 1B). Figure 1C shows the decreased CBF in patients with ICS compared to those without ICS (Fig. 1D). Table 3 shows the association between ICS and sCoV and CBF. The presence of ICS was associated with higher sCoV (mean difference: 0.15, 95% CI: 0.05 –0.24, p = 0.003) and reduced CBF (mean difference: –0.12, 95% CI: –0.22; –0.01, p =0.034) in gray matter in age- and sex-adjusted models. After further adjustment for the covariates such as smoking, systolic blood pressure, hypertension, hyperlipidemia, diabetes and diagnosis (Model II) as well as for cortical infarcts (Model III), the associations of ICS with sCoV were slightly attenuated whereas for CBF, it no longer exceeded the threshold for statistical significance. The effect sizes of associations between ICS and CBF in deep white matter were comparable to that of gray matter CBF but non-significant for all the models.

Fig. 1

Spatial coefficient of variation and cerebral blood flow in patients with and with intracranial stenosis. Low (A) and high (B) spatial coefficient of variation in patient without and with ICS of the right internal carotid artery, respectively. Although the contralateral MCA territory (green oval) also clearly shows vascular pattern/signs of delayed label arrival, the stenosed MCA territory (red oval) shows a stronger spatial heterogeneity with proximal hyperintense vessels (top rows) combined with a distal hypointense tissue region (bottom rows). These asymmetries cancel out on the population level (C and D) due to the heterogeneity of the vascular anatomy and of the laterality of the ICS. However, patients with ICS show on average lower cerebral blood flow and higher signal heterogeneity (D) compared to patients without ICS (C). ICS, intracranial stenosis; MCA, middle cerebral artery.

Table 3

Association of intracranial stenosis with spatial coefficient of variation (n = 381) and cerebral blood flow (n = 327) with standardized mean difference and 95% confidence interval

Presence of stenosis	sCoV of gray matter Mean difference (95% CI)	CBF of gray matter Mean difference (95% CI)	CBF of deep white matter Mean difference (95% CI)
Model I	0.15 (0.05; 0.24),	–0.12 (–0.22; –0.01),	–0.08 (–0.19; 0.03),
	p = 0.003	p = 0.034	p = 0.157
Model II	0.13 (0.03; 0.23),	–0.07 (–0.17; 0.02),	–0.10 (–0.29; 0.09),
	p = 0.011	p = 0.129	p = 0.303
Model III	0.11 (0.01; 0.22),	–0.07 (–0.17; 0.03),	–0.11 (–0.30; 0.08),
	p = 0.031	p = 0.150	p = 0.262

sCoV, spatial coefficient of variation; CBF, cerebral blood flow; CI, confidence interval. Model I: age and sex. Model II: Model I + smoking, systolic blood pressure, hypertension, hyperlipidemia and diabetes, diagnosis. Model III: Model II+cortical infarcts.

When analysis was performed separately for stenosis in anterior and posterior circulation as well as internal carotid artery with ASL parameters (data not shown), no association was observed between stenosis in anterior and posterior circulation and internal carotid artery with sCoV. A borderline significant association was observed between posterior circulation stenosis and CBF in age and sex adjusted models (mean difference: –0.12, 95% CI: –0.24; –0.01, p = 0.041) which became attenuated after controlling for cardiovascular risk factors, diagnosis and cortical infarcts (mean difference: –0.12, 95% CI: –0.24; 0.00, p = 0.059).

Presence of ICS was associated with the asymmetry index representing lateralization of sCoV (mean difference: 0.12, 95% CI: 0.02; 0.21, p = 0.014) in the hemisphere ipsilateral to that of the stenosis in age and sex adjusted models (Table 4). These associations became attenuated after correcting for smoking, systolic blood pressure, hypertension, hyperlipidemia, diabetes, and diagnosis. Finally, after correcting for cortical strokes, these associations remained in the similar direction albeit non-significant. Presence of ICS was also associated with asymmetry index of CBF but only in age and sex adjusted models. Region specific analyses showed that the stenosis in anterior cerebral artery was associated with higher sCoV and reduced CBF in anterior cerebral territory (mean difference for sCoV: 0.40, 95% CI: 0.03; 0.77, p = 0.033 and mean difference for CBF: –0.39, 95% CI: –0.76; –0.03, p = 0.034) in age, sex and cardiovascular risk factors adjusted models (data not shown). Further adjustment for diagnosis and cortical infarcts attenuate these associations.

Table 4

Association of intracranial stenosis with asymmetry index with standardized mean difference and 95% confidence interval

Presence of stenosis	Asymmetry index of sCoV Mean difference (95% CI)	Asymmetry index of CBF Mean difference (95% CI)
Model I	0.12 (0.02; 0.21),	–0.09 (–0.19; –0.00),
	p = 0.014	p = 0.047
Model II	0.16 (0.00; 0.31),	–0.06 (–0.16; 0.04),
	p = 0.049	p = 0.203
Model III	0.15 (–0.02; 0.30),	–0.07 (–0.17; 0.03),
	p = 0.075	p = 0.198

sCoV, spatial coefficient of variation; CBF, cerebral blood flow. Model I: age and sex. Model II: Model I + smoking, systolic blood pressure, hypertension, hyperlipidemia, and diabetes, diagnosis. Model III: Model II+cortical infarcts.

The presence of ICS was associated with worse performance on visuoconstruction in age, sex and education adjusted models (Table 5). The negative effects of ICS on other cognitive domains were also observed albeit non-significant. The association with visuoconstruction remained significant after correc-ting for smoking, systolic blood pressure, hyperten-sion, hyperlipidemia, diabetes, diagnosis, cortical infarcts, cerebral microbleeds, and cortical microinfarcts. Further adjustment for sCoV attenuated the association between ICS and performance on visuoconstruction domain. Mediation analysis showed that there was an indirect effect of ICS on visuoconstruction via sCoV: indirect effect –0.049, boot strapped (95% CI: –0.108, –0.003) (partial mediation effect) (Fig. 2).

Table 5

Association of intracranial stenosis with cognition with standardized mean difference and 95% Confidence Interval

Stenosis	Global cognition	Cognitive domains

		Executive function	Attention	Language	Visual memory	Verbal memory	Visuomotor speed	Visuoconstruction
	Mean difference (95% CI)	Mean difference (95% CI)	Mean difference (95% CI)	Mean difference (95% CI)	Mean difference (95% CI)	Mean difference (95% CI)	Mean difference (95% CI)	Mean difference (95% CI)
Model I	–0.08 (–0.16; 0.00),	–0.07 (–0.16; 0.02),	–0.02 (–0.11; 0.06),	–0.07 (–0.16; 0.02),	–0.06 (–0.15; 0.03),	–0.04 (–0.12; 0.05),	–0.09 (–0.16; –0.01),	–0.12 (–0.19; –0.03),
	p = 0.059	p = 0.118	p = 0.592	p = 0.147	p = 0.210	p = 0.429	p = 0.029	p = 0.006
Model II	–0.01 (–0.29; 0.26),	0.03 (–0.28; 0.35), p	0.18 (–0.14; 0.49),	0.01 (–0.31; 0.33)	–0.04 (–0.34; 0.26),	0.05 (–0.23; 0.33),	–0.04 (–0.29; 0.23),	–0.16 (–0.15; –0.01),
	p = 0.918	p = 0.840	p = 0.264	p = 0.969	p = 0.792	p = 0.715	p = 0.787	p = 0.035
Model III	–0.02 (–0.29; 0.25),	–0.02 (–0.32; 0.28),	0.15 (–0.16; 0.47),	0.01 (–0.31; 0.33),	–0.05 (–0.35; 0.26),	0.04 (–0.24; 0.32),	–0.01 (–0.27; 0.24),	–0.15 (–0.44; 0.14),
	p = 0.889	p = 0.901	p = 0.345	p = 0.946	p = 0.761	p = 0.796	p = 0.913	p = 0.313

sCoV, spatial coefficient of variation; CI, confidence interval. Model I: Age, sex and education. Model II: Model I + smoking, systolic blood pressure, hypertension, hyperlipidemia, diabetes, diagnosis, cortical infarcts, cortical microinfarcts, and cerebral microbleeds. Model III: Model II+sCoV.

Fig. 2

Mediation analysis between intracranial stenosis and visuoconstruction. Spatial coefficient of variation (sCoV) mediated the association between intracranial stenosis and visuoconstruction. Partial mediation was observed between intracranial stenosis and visuoconstruction domain after adjusting for age, sex, education, hypertension, hyperlipidemia, diabetes mellitus, diagnosis, cortical infarcts, cortical microinfarcts, cerebral microbleeds, and sCoV. ‘a’ denotes the effect of intracranial stenosis (exposure) on spatial coefficient of variation (mediator), ‘b’ denotes the effect of spatial coefficient of variation (mediator) on visuoconstruction (outcome), ‘c’ denotes the total effect of intracranial stenosis on visuoconstruction; ‘c” denotes the direct effect of intracranial stenosis on visuoconstruction excluding effects from spatial coefficient of variation (mediator); and ‘c-c” denotes the indirect effect of intracranial stenosis on visuoconstruction (mediation effect). ^* p < 0.05, ^** p < 0.01, ^*** p < 0.001.

DISCUSSION

We showed a clear association between ICS, hemodynamic parameters, and cognitive functioning, and showed the utility of ASL imaging and the sCoV parameter to quantify these hemodynamic changes. The presence of ICS was associated with higher sCoV in gray matter independent of age, sex, cardiovascular risk factors, and diagnosis. Unilateral stenosis was associated with the asymmetry index representing the lateralization of CBF and sCoV. Most importantly, sCoV partially mediated the association between ICS and visuoconstruction.

Previously, limited data has shown that brain hemispheres with symptomatic ICS had prolonged mean transit time and increased cerebral blood volume while CBF was not significantly different when compared with hemispheres with asymptomatic ICS. It was suggested that prolonged mean transit time indicated hemodynamic disturbances that increases the risk of vascular damage in the brain [20]. Further data has shown that the sCoV associated more strongly with ATT than CBF [5 , 21]. A recent study has shown that sCoV and ATT were positively correlated with positron emission tomography (PET) mean transit time and negatively correlated with PET CBF. These results signify that slow flow due to reduction of cerebral perfusion pressure results in incomplete arrival of labeled blood and prolonged mean transit time and reduced CBF as measured by PET, which causes the appearance of vascular artifacts and increases the sCoV of pCASL CBF images [21]. However, as ATT scans are not usually performed in clinical practice because of scanning time limitations, it remains a challenge in elderly population. Therefore, the use of the sCoV as a proxy ATT parameter may have a value in clinical ASL studies. One additional explanation for the association between ICS and sCoV could be that sCoV has a higher reliability and power than the CBF measurements in an elderly population with compromised macro-vasculature and thus, is useful for detection of more subtle hemodynamic changes. Indeed, the CBF measurements require the label to be present in tissue compartment whereas for sCoV, the presence of label in imaging voxel is sufficient irrespective of tissue or vascular compartment. Interestingly, it has been shown that ICS may cause a certain translesional pressure drop which cause CBF in the downstream territory to be normal, but the mean arterial transit time could be prolonged due to the flow-limiting effects of the ste-notic lesion [22]. However, the current study could only offer a single cross-sectional snapshot of the correlations between these factors, which in vivo could be far more complicated and dynamic.

The association of ICS with asymmetry index in this study showed lateralization of sCoV to the side of the stenosis. The advantage of sCoV in detecting the side of stenosis is attributed to the relatively high vascular signal present in the ipsilateral hemisphere, which could be due to longer arrival time of the la-beled blood or vascular reconfiguration over time suggesting sCoV as a marker of collateral perfusion in patients with large to medium vessel disease. Thus, sCoV may provide important information regarding the extent to which the hemodynamic status is com-promised and how this contributes to cognitive imp-airment. To what extent our findings can be explained by longer ATT or local vascular reconfiguration cannot be differentiated with these data. Similarly, the association observed between ACA and higher sCoV and lower CBF might be a chance finding and should be interpreted with caution as the number of cases with ACA were few (n = 5).

A few studies have shown that the atherosclerosis in cervicocerebral arteries was related to lower performance in visuospatial and verbal cognitive do-mains [23]. Moreover, patients with asymptomatic carotid stenosis improved significantly with respect to their neuropsychological assessments after prophylactic carotid surgery, including improvement in attention, visual memory, and psychomotor speed [24]. The present study extends the previous findings by demonstrating that patients with ICS were more likely to have poorer cognitive functioning, in par-ticular tasks assessing cortical functions, i.e., visuoconstruction. However, these associations became attenuated when adjusting for sCoV leading to the possibility that this association is partially mediated by vascular insufficiency. This was further confirmed by the mediation analysis where we showed that ICS influenced visuoconstruction performance via sCoV. Interestingly, tasks that are known to be affected by vascular damage in subcortical regions (i.e., attention and visuomotor speed) were relatively less affected. It has been reported before that cognitive deficit in patients with carotid stenosis may occur as a result of diffuse ischemic damage; however, hypoperfusion was not directly measured in these studies [25]. The current findings thus, show the importance of measuring the vascular insufficiency, e.g., by ASL sCoV as proxy measure in addition to ASL CBF, providing insights into the underlying mechanisms of ICS and cognitive impairment in elderly with compromised macrovasculature. Moreover, ASL sCoV might be a useful marker of vascular insufficiency in Asian population where prevalence of ICS is much higher co-mpared to Caucasian population (40–50% versus 8–10%). Our current findings contribute to the literature by reporting how ASL could be a helpful non-invasive biomarker to monitor the occurrence of ICS for initiating preventive therapy. Our study has the following limitations. First, cross-sectional design of our study does not allow studying the temporal rel-ationship between the presence of ICS, ASL par-ameters, and cognition. Second, cerebral perfusion can be affected by collaterals, leading to vascular reconfiguration and increased arterial transit time that can be visible as regions of ASL hyperintensities [26]. With a single post label delay ASL image, it is difficult to differentiate these different sources of ASL hyperintensities and thus was not possible in the current study. Third, MRA may overestimate the measured degree of stenosis due to blood flow turbulence or presence of the normal variations, which was mitigated by confirming stenosis in the vessel lumen on source images. Fourth, our definition of ICS was relatively lenient, including a range of moderate (>50%) to severe stenosis (>80%) and complete occlusion. This may complicate the interpretation of our findings, as different stenosis grades are likely to give rise to different perfusion deficits. Since both subgroups were relatively small, i.e., six and eight subjects, respectively, we were unable to perform separate analyses on the subgroups. Fifth, angiograms with intra-arterial contrast are considered to be the gold standard to diagnose occlusive disease of the intracerebral arteries; however, it would not have been feasible or ethical to perform such invasive investigations with contrast safety issues in the memory clinic populations. Sixth, we did not have information on other chronic diseases such as depression, thyroid dysfunction or chronic kidney disease, which may affect our associations with cognitive performance. Finally, as our study was performed in a memory clinic population, it limits the generalizability to population-based setting. The strengths of the study include extensive neuropsychological assessment and availability of 3T MRA neuroimaging to grade stenosis in individuals. In the mediation analysis, we used the bootstrap test, which is superior to the originally proposed Sobel test. Moreover, we found that both the indirect path and the total effect were significant, and that the direct effect and indirect path had the same sign leading to the clear conclusion of sCoV being a partial mediator. Finally, in this study, we are interested in exploring how ICS relates to ASL parameters including the novel sCoV and whether it can help understand the mechanism for cognitive impairment in memory clinic patients. This will also help us understand a novel way of analyzing and interpreting the ASL data in memory clinic patients including AD, rather than on incremental improvement of the currently know CBF effects.

In conclusion, ICS is associated with higher sCoV in gray matter. sCoV partially mediated the association between ICS and visuoconstruction. Thus, sCoV may provide hemodynamic information in patients with large to medium vessel disease and might be more robust compared to CBF, mainly because it can also be used in patients with cerebrovascular disease. The current findings need further verification in larger, longitudinal studies with repeated imaging exams to delineate the dynamic evolution in the stenotic lesion, collateral circulation and the perfusion status, and clinical follow-up to correlate with these imaging markers.

Footnotes

ACKNOWLEDGMENTS

This study is funded by the Singapore National Medical Research Council (grants NMRC/CG/NUHS/2010 and NMRC/CG/013/2013).

Authors’ disclosures available online ().

References

Jack

Jr. , Knopman

, Jagust

, Shaw

, Aisen

, Weiner

, Petersen

, Trojanowski

(2010) Hypothetical model of dynamic biomarkers of the Alzheimer’s pathological cascade. Lancet Neurol 9, 119–128.

Wierenga

, Hays

, Zlatar

(2014) Cerebral blood flow measured by arterial spin labeling MRI as a preclinical marker of Alzheimer’s disease. J Alzheimers Dis 42, S411–419.

Yoshiura

, Hiwatashi

, Noguchi

, Yamashita

, Ohyagi

, Monji

, Nagao

, Kamano

, Togao

, Honda

(2009) Arterial spin labelling at 3-T MR imaging for detection of individuals with Alzheimer’s disease. Eur Radiol 19, 2819–2825.

Iturria-Medina

, Sotero

, Toussaint

, Mateos-Pérez

, Evans

(2016) Early role of vascular dysregulation on late-onset Alzheimer’s disease based on multifactorial data-driven analysis. Nat Commun 7, 11934.

Mutsaerts

, van Dalen

, Heijtel

, Groot

, Majoie

, Petersen

, Richard

, Nederveen

(2015) Cerebral perfusion measurements in elderly with hypertension using arterial spin labeling. PloS One 10, e0133717.

Alsop

, Detre

, Golay

, Günther

, Hendrikse

, Hernandez-Garcia

, Lu

, MacIntosh

, Parkes

, Smits

, van Osch

, Wang

, Wong

, Zaharchuk

(2015) 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 73, 102–116.

Mutsaerts

, Petr

, Václavů

, van Dalen

, Robertson

, Caan

, Masellis

, Nederveen

, Richard

, MacIntosh

(2017) The spatial coefficient of variation in arterial spin labeling cerebral blood flow images. J Cereb Blood Flow Metab 37, 3184–3192.

Ingala

, Petr

, Pålhaugen

, Gajdos

, Fladby

, Selnes

, Barkhof

, Bjornerud

, Rektorova

, Mutsaerts

HJMM

(2018) Investigating arterial spin labeling as a large vessel correlate of SVD, AD and PD. Alzheimers Dement 14(Suppl 7), P456.

Ferro

, Mutsaerts

, Hilal

, Kuijf

, Petersen

, Petr

, van Veluw

, Venketasubramanian

, Boon Yeow

, Biessels

, Chen

(2019) Cortical microinfarcts in memory clinic patients are associated with reduced cerebral perfusion. J Cereb Blood Flow Metab 40, 1869–1878.

10.

Hilal

, Chai

, Veluw

, Shaik

, Ikram

, Venketasubramanian

, Richards

, Biessels

, Chen

(2016) Association between subclinical cardiac biomarkers and clinically manifest cardiac diseases with cortical cerebral microinfarcts. JAMA Neurol 74, 403–410.

11.

Hilal

, Xu

, Ikram

, Vrooman

, Venketasubramanian

, Chen

(2017) Intracranial stenosis in cognitive impairment and dementia. J Cereb Blood Flow Metab 37, 2262–2269.

12.

Cordonnier

, Potter

, Jackson

, Doubal

, Keir

, Sudlow

, Wardlaw

, Al-Shahi Salman

(2009) improving interrater agreement about brain microbleeds: Development of the Brain Observer MicroBleed Scale (BOMBS). Stroke 40, 94–99.

13.

Hilal

, Sikking

, Shaik

, Chan

, van Veluw

, Vrooman

, Cheng

, Sabanayagam

, Cheung

, Wong TY Venketasubramanian

, Biessels

, Chen

, Ikram

(2016) Cortical cerebral microinfarcts on 3T MRI: A novel marker of cerebrovascular disease. Neurology 87, 1583–1590.

14.

Mutsaerts

HJMM

, Petr

, Thomas

, De Vita

, Cash

, van Osch

MJP

, Golay

, Groot

PFC

, Ourselin

, van Swieten

, Laforce

Jr. , Tagliavini

, Borroni

, Galimberti

, Rowe

, Graff

, Pizzini

, Finger

, Sorbi

, Castelo Branco

, Rohrer

, Masellis

, MacIntosh

GENFI investigators (2018) Comparison of arterial spin labeling registration strategies in the multi-center GENetic frontotemporal dementia initiative (GENFI). J Magn Reson Imaging 47, 131–140.

15.

Mutsaerts

, Petr

, Groot

, Vandemaele

, Ingala

, Robertson

, Václavu

, Groote

, Kuijf

, Zelaya

, O’Daly

, Hilal

, Wink

, Kant

, Caan

MWA

, Morgan

, de Bresser

, Lysvik

, Schrantee

, Bjørnebekk

, Clement

, Shirzadi

, Kuijer

JPA

, Anazodo

, Pajkrt

, Richard

E BR

, Reneman

, Masellis

, Günther

, MacIntosh

, Achten

, Chappell

, van Osch

MJP

, Golay

, Thomas

, de Vita

, Bjørnerud

, Nederveen

, Hendrikse

, Asllani

, Barkhof

(2020) ExploreASL: An image processing pipeline for multi-center ASL perfusion MRI studies. Neuroimage 219, 117031.

16.

Ashburner

(2007) A fast diffeomorphic image registration algorithm. Neuroimage 38, 95–113.

17.

Mutsaerts

, Richard

, Heijtel

, van Osch

, Majoie

, Nederveen

(2013) Gray matter contamination in arterial spin labeling white matter perfusion measurements in patients with dementia. Neuroimage Clin 4, 139–144.

18.

Tatu

, Moulin

, Bogousslavsky

, Duvernoy

(1998) Arterial territories of the human brain: Cerebral hemispheres. Neurology 50, 1699–1708.

19.

Hilal

, Xin

, Ang

, Tan

, Venketasubramanian

, Niessen

, Vrooman

, Wong

, Chen

, Ikram

(2015) Risk factors and consequences of cortical thickness in an Asian population. Medicine 94, e852.

20.

Lan

, Leng

, Abrigo

, Ip

, Soo

, Leung

, Chen

, Wong

(2017) Cerebral perfusion difference between hemispheres with symptomatic and asymptomatic intracranial arterial stenosis. Int J Stroke 12, NP19–NP20.

21.

Ibaraki

, Nakamura

, Toyoshima

, Takahashi

, Matsubara

, Umetsu

, Pfeuffer

, Kuribayashi

, Kinoshita

(2019) Spatial coefficient of variation in pseudo-continuous arterial spin labeling cerebral blood flow images as a hemodynamic measure for cerebrovascular steno-occlusive disease: A comparative 15O positron emission tomography study. J Cereb Blood Flow Metab 39, 173–181.

22.

Lan

, Leng

, Ip

, Soo

, Abrigo

, Liu

, Fan

, Ma

, Ip

, Chan

, Mok

, Liebeskind

, Wong

, Leung

(2020) Sustaining cerebral perfusion in intracranial atherosclerotic stenosis: The roles of antegrade residual flow and leptomeningeal collateral flow. J Cereb Blood Flow Metab 40, 126–134.

23.

López-Olóriz

, López-Cancio

, Arenillas

, Hernández

, Jiménez

, Dorado

, Barrios

, Soriano-Raya

, Miralbell

, Cáceres

, Forés

, Pera

, Dávalos

, Mataró

(2013) Asymptomatic cervicocerebral atherosclerosis, intracranial vascular resistance and cognition: The AsIA-neuropsychology study. Atherosclerosis 230, 330–335.

24.

Bossema

, Brand

, Moll

, Ackerstaff

, de Haan

, van Doornen

(2006) Cognitive functions in carotid artery disease before endarterectomy. J Clin Exp Neuropsychol 28, 357–369.

25.

Mathiesen

, Waterloo

, Joakimsen

, Bakke

, Jacobsen

, Bonaa

(2004) Reduced neuropsychological test performance in asymptomatic carotid stenosis: The Tromso Study. Neurology 62, 695–701.

26.

Zaharchuk

, Do

, Marks

, Rosenberg

, Moseley

, Steinberg

(2011) Arterial spin-labeling MRI can identify the presence and intensity of collateral perfusion in patients with moyamoya disease. Stroke 42, 2485–2491.