Higher intracranial arterial pulsatility is associated with presumed imaging markers of the glymphatic system: an explorative study

Background: Arterial pulsation has been suggested as a key driver of paravascular cerebrospinal fluid flow, which is the foundation of glymphatic clearance. However, whether intracranial arterial pulsatility is associated with glymphatic markers in humans has not yet been studied. Methods: Seventy-three community participants were enrolled in the study. 4D phase-contrast magnetic resonance imaging (MRI) was used to quantify the hemodynamic parameters including flow pulsatility index (PI flow ) and area pulsatility index (PI area ) from 13 major intracerebral arterial segments. Three presumed neuroimaging markers of the glymphatic system were measured: including dilation of perivascular space (PVS), diffusivity along the perivascular space (ALPS), and volume fraction of free water (FW) in white matter. We explored the relationships between PI area , PI flow , and the presumed glymphatic markers, controlling for related covariates. Results: PI flow in the internal carotid artery (ICA) C2 segment (OR, 1.05; 95% CI, 1.01-1.10, per 0.01 increase in PI) and C4 segment (OR, 1.05; 95% CI, 1.01-1.09) was positively associated with the dilation of basal ganglia PVS, and PI flow in the ICA C4 segment (OR, 1.06, 95% CI, 1.02-1.10) was correlated with the dilation of PVS in the white matter. ALPS was associated with PI flow in the basilar artery (β = -0.273, p = 0.046) and PI area in the ICA C2 (β = -0.239, p = 0.041) and C7 segments (β = -0.238, p = 0.037). Conclusions: Intracranial arterial pulsatility was associated with presumed neuroimaging markers of the glymphatic system, but the results were not consistent across different markers. Further studies are warranted to confirm these findings.


Introduction
The glymphatic system is a vital fluid-clearance pathway in the brain.Subarachnoid cerebrospinal fluid (CSF) flows into the brain parenchyma along the periarterial space and into the interstitium through the Aquaporin-4 (AQP4) water channels, subsequently exchanges with interstitial fluid (ISF) and drains out of the brain along the perivenous space (Iliff et al., 2012).Dysfunction of the glymphatic system may be associated with a variety of neurological diseases, such as Alzheimer's disease, Parkinson's disease, and idiopathic normal pressure hydrocephalus (Ringstad et al., 2017;Rasmussen et al., 2018, Ringstad et al., 2018;Harrison et al., 2020).
Arterial pulsation has been proposed as the main driving force of glymphatic fluid transport.Early studies found that CSF tracers injected into the subarachnoid space gathered around arterioles, and the rapid perivascular influx of CSF tracers could be prevented by diminishing the pulsations of the cerebral arteries through ligation of the brachiocephalic artery (Rennels et al., 1985).Iliff et al. (2013) used in vivo two-photon microscopy to visualize vascular wall movement and directly measure the vascular pulsatility at different levels of the cerebrovascular tree.Moreover, they experimentally reduced the wall pulsatility of penetrating arteries by unilateral ligation of the internal carotid artery and found that the movement of CSF tracer into and through the cortex was slowed.On the contrary, CSF tracer influx into the brain parenchyma was significantly increased when elevating the pulsatility along penetrating arteries.Mestre et al. (2018) quantified CSF flow velocities in PVSs of living mice and reported that CSF flow was pulsatile and matched the speed of the arterial wall motion related to the cardiac cycle.Furthermore, the backflow in the PVS increased when the arterial pulsation was dampened due to elevated blood pressure.Although mounting evidence based on animal experiments demonstrates the role of arterial pulsation in glymphatic clearance, the extensive difference between humans and mice emphasizes the importance of validation in the human brain (Hodge et al., 2019).Understanding the relationship between arterial pulsation and glymphatic clearance in humans may offer insights into the treatment of related brain disorders.
In recent years, several magnetic resonance imaging (MRI) techniques have become widely used to study glymphatic-related changes in humans.Intrathecal administration of the gadolinium-based contrast agent and tracking its clearance from the brain using dynamic contrastenhanced (DCE) MRI is considered a gold-standard method for measuring glymphatic clearance (Ringstad et al., 2017(Ringstad et al., , 2018)).Nonetheless, due to its invasive nature, several other imaging markers have been proposed.First, dilation of perivascular spaces (PVSs), the main conduit of glymphatic flow, has been proposed as a marker of an impaired glymphatic system.Dilation of PVS has been related to aging, hypertension, and a variety of neurodegenerative disorders (Zhu et al., 2010;Shi et al., 2020;van den Kerkhof et al., 2023).Second, to measure fluid transport in the perivenous space along the deep medullary veins (DMVs), Taoka et al. proposed a method called diffusion tensor image analysis along the perivascular space (DTI-ALPS).Although still under debate, the ALPS index showed a good correlation to glymphatic clearance measured by DCE-MRI and has been found to decrease in many neurological disorders (Zhang et al., 2021).Third, the volume fraction of free water (FW) in the white matter may reflect water stagnation related to impaired glymphatic clearance.Indeed, increased water diffusivity has been reported in mice with AQP4 water channel deletion, suggesting enlargement and fluid stagnation in the interstitial space (Gomolka et al., 2023).While their direct link with glymphatic function still needs further validation, these metrics demonstrated great potential to serve as in vivo markers.
In this study, we aimed to test whether intracerebral arterial pulsatility is associated with these presumed glymphatic markers in humans.Recently, 4D phase-contrast MRI (4D flow) with three-directional encoding has been developed as an effective method for visualization and quantification of hemodynamics in blood vessels.It can simultaneously image a wide range of blood vessels while maintaining accuracy.Here, we estimated the arterial pulsatility index (PI) in 13 segments of intracranial arteries using 4D flow and explored the relationship between PIs and the three presumed glymphatic markers.

Clinical data
The demographic information and vascular risk factors were collected, including age, sex, hypertension, hyperlipidemia, diabetes, and history of smoking.Vascular risk factors (VRF) were recorded as a status (yes or no) and defined according to the self-reported medical history or medical treatment.The status of each vascular risk factor is marked as 0 or 1, and all of them are summarized into a total score, ranging from 0 to 4 (Jochems et al., 2020).The Mini-Mental State Examination (MMSE) was used to assess the global cognitive condition of the participants.

Image acquisition
All the MRI sequences were performed on a United Imaging uMR790 3.0T scanner (Shanghai, China) using a 32-channel head coil.T1 weighted images were acquired with a 3D fast spoiled gradient-echo sequence, the parameters were TR, 6.7 ms, TE, 2.9 ms, flip angle, 9 • , field of view, 256 × 240 mm 2 , voxel size, 1 × 1 × 1 mm 3 .T2 weighted images were acquired with a MATRIX (modulated flip angle technique in refocused imaging with extended echo train) sequence, the parameters were TR, 3000 ms, TE, 407.4 ms, echo train length, 180, field of view, 256 × 240 mm 2 , voxel size, 0.8 × 0.8 × 0.8 mm 3 .T2 weighted FLAIR images were acquired with inversion recovery MATRIX sequence, the parameters were TR, 6500 ms, TE, 434.52 ms, TI, 1938ms, echo train length, 220, field of view, 256 × 220 mm 2 , voxel size, 1 × 1 × 1 mm 3 .Diffusion tensor imaging was acquired with multiple b values (500 s/ mm 2 , 16 diffusion directions; 1000 s/mm 2 , 32 directions; 2000 s/mm 2 , 32 directions) using a spin-echo planar imaging sequence, the parameters were TR, 8682 ms, TE, 75.8 ms, field of view, 224 × 224 mm 2 , voxel size, 2 × 2 × 2 mm 3 , 70 slices.The phase encoding direction was posterior-anterior.Two additional volumes were acquired without diffusion weighting (b, 0 s/mm 2 ) and another b0 image with an opposite phase encoding direction was acquired to correct echo planar imaging (EPI) distortions.All 4D phase-contrast (PC) MRI data were acquired by using a 4D flow sequence based on gradient recalled echo sequence with added flow encoding.The scan with the following imaging parameters: imaging volume, 224 × 170 × 60 mm 3 , velocity encoding (venc), 100 cm/s, voxel size, 1 × 1 × 1.5 mm 3 , resampled size= 1 × 1 × 0.75 mm 3 , TR, 29.3 ms, TE, 4.09 ms, flip angle, 8 • , bandwidth, 350 Hz/pixel, scan time, 10 min.A pulse oximeter was put onto the subject's finger during the exam to collect the cardiac trigger, in order to reconstruct images retrospectively.After completion of the acquisition, magnitude data providing the anatomic information and three flow images representing the velocities Vx, Vy, and Vz respectively, which were reconstructed into 20 frames per cardiac cycle, were generated.

Pulsatility analysis
Pulsatility analysis from 4D flow data was conducted in the cvi42 software (https://www.circlecvi.com/) and the steps included data cropping, preprocessing, vessel segmentation, and hemodynamic measurement.The static tissue and vessel mask were automatically identified and adjusted manually to make the static tissue mask distributed across the images evenly.The preprocessing consisted of offset correction, and phase anti-aliasing was performed.In order to exclude the hemodynamic effect of the physiologic vascular tortuosity, 13 arterial segments running straightly were selected: internal carotid artery (ICA) C2, C4, and C7 segments (left and right), middle cerebral artery (MCA) proximal M1 segment (left and right), anterior cerebral artery (ACA) A1 segment (left and right), posterior cerebral artery (PCA) P1 segment (left and right), and basilar artery (BA) middle segment.Then, hemodynamic measurement planes were manually placed orthogonal to the vessel orientation (Fig. 1A).Subsequently, the contour of the vessel crosssection would be drawn and propagated automatically, and the pulsatile vessel waveform would be extracted instantaneously (Fig. 1B).Check the edge of the contour in every time frame and adjust it manually if necessary.
Here, the area PI and flow PI were measured to represent the arterial pulsatility.The area PI was calculated for each vessel segmentation: PI area, (Amax -Amin) / Amean; A, cross-sectional area of the vessel.The flow PI was calculated using the Gosling equation: PI flow, (Qmax-Qmin) / Qmean; Q, flow rate; the Qmax and Qmin were defined within the cardiac cycle.The bilateral PIs were averaged to obtain the mean PI in all predefined vessels except the BA middle segment.The lateral PI represented the mean PI in the participants with arterial hypoplasia or with a vessel diameter that was too small to be measured.Considering the small lumen area of downstream arteries, PI area measured using the current imaging resolution may not well-reflect pulsatile changes, so we only included PI area acquired from C2, C4, C7, and BA in the following analyses.

PVS visual rating
The dilated PVS (dPVS) was assessed according to the STandards for ReportIng Vascular changes on nEuroimaging (STRIVE) on T2 images (Wardlaw et al., 2013).It was defined as a round, oval, or linear lesion with a maximum diameter < 3 mm and has a CSF-like signal (hyperintense on T2), perpendicular to the brain surface and parallel to perforating vessels.We estimated the severity based on the number of dPVS with a rating scale of 0 to 4 in basal ganglia (BG) and centrum semiovale (CSO) separately: 0, no PVS, 1 <= 10 PVS, 2, 11 to 20 PVS, 3, 21 to 40 PVS, and 4 >= 40 PVS (Potter et al., 2015).Dilated PVSs were counted in the slice with the highest number in one cerebral hemisphere by two postgraduate students (LX, 5-year experience in medical imaging; LL, 4-year experience in radiology), trained together and blinded to clinical information, evaluated PVS score.Cohen's kappa was used to assess the consistency between the results from the two raters.Disagreements were solved by discussion.

DTI-ALPS calculation
DTI processing was performed in FSL 6.0 (https://fsl.fmrib.ox.ac.uk/fsl).The preprocessing steps included skull stripping, denoising, removing Gibbs artifact, EPI distortion correction, and eddy current correction.The b0 and b1000 images were used to fit an ellipsoid model using DTIFIT in FSL.Then, all the generated results including the fractional anisotropy (FA) map and mean diffusivity (MD) map were normalized to the MNI space using linear registration.The DTI-ALPS index was calculated based on the assumption that the major difference between the x-axis diffusivity in both areas (projection fiber area, Dxproj and association fiber area, Dxassoc) and the diffusivity (Dyproj, Dzassoc) which is perpendicular to them has originated from the existence of the movement water molecules in the perivascular space (Taoka et al., 2017).A cross region of interest (ROI) containing 5 voxels (40mm 3 ) was placed on the projection fiber and association fiber respectively in bilateral cerebral hemispheres according to previous literature, and the diffusivities in the directions of the x-axis, y-axis, and z-axis were obtained.The DTI-ALPS index was calculated using the following formula: DTI-ALPS index, mean (Dxproj, Dxassoc) / mean (Dyproj, Dzassoc).The ALPS indices in the bilateral hemispheres were calculated respectively and the mean of them was used in further analyses.
Additionally, to exclude possible influences of the white matter microstructure on ALPS, we extracted FA and MD values in the projection and association areas using the same ALPS ROIs.The average of the left and right projection and association FA/MD values were then used as covariates for the statistical analyses.

FW calculation
FW map was calculated using the script from the MarkVCID project (https://markvcid.partners.org/)based on diffusion images with b = 0 or 1000s /mm 2 .Briefly, the signal was fitted to a two-compartment model in each voxel, including a FW compartment (isotropic tensor) and a non-FW tissue compartment (anisotropic tensor) (Pasternak et al., 2009;Hoy et al., 2014).A FW map represents the fractional volume (ranging from 0 to 1) of the FW compartment.Finally, the generated FW maps were registered to structural images through b0 images and mean white matter FW was extracted.

Brain segmentation
The brain volume estimation was performed using CAT12 toolbox (https://neuro-jena.github.io/cat/).T1 images were segmented into gray matter, white matter, and CSF.The intracranial volume (ICV) was calculated as the sum of white matter, gray matter, and CSF.

WMH segmentation
White matter hyperintensity (WMH) segmentation was performed using BIANCA (Brain Intensity AbNormality Classification Algorithm, https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/BIANCA), an automated and supervised method for WMH detection based on the nearest neighbor (k-NN) algorithm (Griffanti et al., 2016;Huang et al., 2021a,b).BIANCA has been trained to segment WMH in the local dataset in our previous study.We performed visual assessments of the segmentation results and manually corrected the masks when necessary.The total WMH volume was further normalized by ICV to remove the effect of head size.

Statistical analysis
All statistical analyses were conducted in IBM SPSS 26.First, univariate analyses were performed to investigate the relationship between three potential glymphatic markers and each predictor respectively.Then, the associations between PIs and each covariate were also investigated.Pearson's correlation was used to investigate the association between two continuous variables.For the categorical variable sex, ALPS, FW, and PIs in different arterial segments were compared using two-sample t-tests separately and PVS scores were compared using the Chi-square test.The difference of ALPS/FW/PIs among groups with varying VRF total scores was assessed through analysis of variance (ANOVA), followed by post-hoc analysis using Bonferroni correction.Kendall's tau-b correlation was used to investigate the relationship between two ordinal variables (ordinal variable and continuous variable), such as PVS score and VRF total score.
Second, we performed ordinal logistic regression to investigate the association between PIs and PVS scores in various arterial segments.Both univariate ordinal regression (model 1) and multivariable ordinal regression were applied, in which PVS score was set as the dependent variable.The multivariable ordinal regression was performed twice.One was adjusted for age, sex, and ICV (model 2).The other was adjusted for age, sex, ICV, and VRF total score (model 3).Results were presented as odds ratios (ORs) with 95 % confidence intervals (95 % CIs).Here the ICV was included as a covariate because previous studies demonstrated its association with PVS scores or volumes (Huang et al., 2021;Zeng et al., 2022).
Last, univariate (model 1) and multiple linear regression were used to investigate the relationship between PIs, DTI-ALPS, and FW.The DTI-ALPS and FW were set as the dependent variable in turn.In multiple regression analyses, we controlled age, sex, and ICV (model 2).Considering the possible influence of WMH on the measurements of FW and ALPS, the normalized WMH volume was also included as a confounding factor in model 2. Furthermore, to remove possible influences of white matter microstructural properties on ALPS assessment, the mean FA and MD values from the ALPS ROIs were also included in model 2 during the ALPS analysis.Then, multiple regression models were additionally adjusted for VRF total score (model 3).
In all regression analyses, standardized beta was used to reflect the predictive power of each variable.For multiple regression analyses, collinearity was examined to avoid biased fitting using the variable inflation factor (<10).For each model, we performed false discovery rate (FDR) corrections on the analyses of 13 vessel segments to control false positives.Nonetheless, considering the explorative nature of the current study.We kept all statistical results in the paper.The p-value for statistical significance was set at 0.05, 2-tailed.

Participants characteristics
A total of 73 participants (mean age (SD), 63.2 (6.9), female/male, 40/33) were included in the analyses.The demographics and clinical characteristics were summarized in Table 1.Among the participants, 33 (45.2 %) had hypertension, 20 (27.4 %) had hyperlipemia, 13 (17.8%) had diabetes, and 20 (27.4 %) had a history of smoking.In general, the vascular burden among the participants was mild.The median of WMH volume was 0.58 ml.The hemodynamics parameters including PI area , PI flow in all arterial segments, and the presumed glymphatic markers (BG PVS score, CSO PVS score, DTI-ALPS, and FW), were presented in Table 2.In total, 72 PIs of ICA C2, BA, and ACA A1segment, 73 PIs of ICA C4, C7, and MCA M1 segments, and 71 PIs of PCA segment were analyzed in the research due to anatomic variations or technical problems.There were good agreements between the two observers in the assessments of BG-PVS (Kappa, 0.639) and CSO-PVS (Kappa, 0.654) scores.

Associations between the three glymphatic markers and covariates
PVS score and covariates.Age was associated with BG-PVS score (r, 0.262, p, 0.005), but showed no relationship with CSO-PVS score.There were no differences in PVS scores between females and males.ICV and VRF total score showed no relationship with PVS scores.

Associations between PIs and covariates
Age was significantly associated with increased PI flow in all arterial segments (eFig.1 in the Supplementary Material e.g.ICA C2: r, 0.611, p < 0.001), but showed no relationship with PI area in all arterial segments.Compared to males, females had lower PI flow in the PCA P1 segment (0.776 vs 0.872, p, 0.045).There was no significant differences of PI area in all arterial segments between females and males.ICV had positive relationships with PI area in the ICA C7 segment (r, 0.313, p, 0.007), and showed no significant association with PI flow in all selected segments.Then, no significant association was found between PIs and VRF total scores.The detailed results were shown in the eTable 1-3 in the Supplementary Material.

Association between PI flow and glymphatic markers
The results of ordinal logistic regression analyses, focusing on the relationship between flow pulsatility and PVS scores, were shown in Table 3. (1) PI flow and BG-PVS.In univariate regression analyses, PI flow in all arterial segments except for those in the PCA P1 segment and the BA segment was positively associated with BG-PVS score.After controlling for age, sex, and ICV, only PI flow in the ICA C2 segment (Model 2, OR, 1.05; 95 % CI, 1.01-1.10;per 0.01 increase in PI, same for the other analyses) and C4 segment (Model 2, OR, 1.05; 95 % CI, 1.01-1.09)was associated with BG-PVS score respectively.The associations remained significant with the VRF total score added for correction.(2) PI flow and CSO -PVS.Higher PI flow in the ICA C4/C7, ACA A1, PCA P1, and the BA segment was significantly correlated with higher CSO-PVS score in univariate regression analyses.However, only the association between the PI flow in the ICA C4 segment and CSO-PVS score remained significant while adjusting for age, sex, ICV, and VRF total score (Model 3, OR, 1.06, 95 % CI, 1.02-1.10).
PI flow and ALPS index.PI flow in the BA segment was associated with ALPS (Table 4. β, -0.274, p, 0.019), and the association remained significant when additionally adjusting for age, sex, normalized WMH volume, VRF total score, and mean FA and MD values in ROIs of ALPS measurement.(Table 4. β,p,0.046).PI flow in other arterial segments was not associated with ALPS.
PI flow and FW.Higher PI flow in all arterial segments was significantly associated with FW in univariate models (Table 4).After correcting for age, sex, and normalized WMH volume, only PI flow in the MCA M1 segment was positively associated with FW (Table 4. β, 0.199, p, 0.049), and PI flow in the ICA C2 segment had a trend of correlating with FW (Table 4. β, 0.174, p, 0.065).All the correlations disappeared when additionally adjusting for VRF total score.

Association between PI area and glymphatic markers PI area and PVS scores.
There was no significant association between PI area in each arterial segment and PVS scores in either univariate or multivariate regression analyses (Table 5).

Discussion
This study explored the relationship between intracranial arterial pulsatility and three presumed glymphatic markers (PVS, ALPS, and FW) in humans.PI flow was found to be strongly associated with PVS score and FW in almost all vessel segments, and many associations survived after controlling for the covariates.ALPS was associated with PI flow in the BA segment and PI area in the C2 and C7 segments.These results provided in vivo evidence for understanding the association between intracranial arterial pulsatility and variations in the glymphatic system in humans.
We used 4D flow to measure vessel PIs at major intracranial arteries.PIs in the selected arterial segments were at a level similar to the results of other studies, whether obtained by 2D PC-MRI, Transcranial Doppler ultrasound (TCD), or 4D flow (Mok et al., 2012;Meckel et al., 2013;Wahlin et al., 2013;Zarrinkoob et al., 2016;Holmgren et al., 2020;Kneihsl et al., 2020).The cerebral arterial PI has been demonstrated as a measure sensitive to both central arterial stiffness and downstream microvascular resistance (Shi et al., 2018;Wardlaw et al., 2019).Under the influence of aging and vascular risk factors, the aorta and peripheral vessels may become thicker, stiffer, and less flexible, leading to the transfer of excessive pulsatile energy to the cerebral circulation (Mitchell et al., 2004(Mitchell et al., , 2011;;Fico et al., 2022).Consistent with previous studies, we found robust positive associations between age and intracranial arterial PIs (Mitchell et al., 2011;Zarrinkoob et al., 2016).
We found that higher PI flow was associated with PVS dilation in both BG and CSO regions.Several previous studies have investigated similar research questions using different techniques and yielded inconsistent results.One study used TCD to obtain the PI along the MCA M1 segment based on lacunar stroke patients, finding no significant relationship with dilated PVS in BG region (Nam et al., 2020).Using 2D PC-MRI, van den Kerkhof et al. enrolled 45 subjects to measure velocity PI both in the small lenticulostriate arteries and MCA and showed that PIs in the selected segments were not associated with dilated BG or CSO PVS (van den Kerkhof et al., 2023).Interestingly, Shi et al. reported that higher flow PI in the venous sinuses, but not intracranial arteries, was associated with higher BG-PVS score, and there was no correlation between PIs and CSO-PVS (Shi et al., 2020).The discrepancy might be because the study included patients with lacunar stroke.They reported a higher PI flow in the ICA (mean: 1.27) than our results (0.81), suggesting that the ceiling effect might have influenced the analyses.In a recent study, Vickner et al. used 4D flow MRI to investigate the potential temporal ordering among CSVD markers and PIs in ICA and cerebral distal arteries.In this study, cerebral arterial PIs were not associated with SVD markers at baseline and could not predict changes in PVS (Vikner et al., 2022).Comparatively, we both utilized the 4D-flow method and included community subjects, but there are a few differences in technical details, including PI measurements, PVS assessments, and CSVD burden.These may have caused discrepancies in the results.Currently, only one study found that velocity PI in the MCA (assessed using TCD) was related to a higher BG-PVS number in patients with mild burden of CSVD (Nam et al., 2020).
Some indirect evidence may provide support for our findings.In the Vanderbilt Memory and Aging Project (n, 327), higher aortic stiffness was related to greater BG-PVS volume and count (Bown et al., 2023).A similar relationship was also found in a cohort of 782 hypertensive individuals (Riba-Llena et al., 2018).One study (n, 80) assessed cerebrovascular reactivity (CVR), which is reversely related to vessel stiffness, and found a negative relationship between CVR and BG-PVS count in cognitively normal older adults.This indicates that higher vessel stiffness is associated with higher BG-PVS count (Libecap et al., 2022).In general, it seems that the association between pulsatility and PVS dilation was more robust in a relatively healthy participant cohort.Indeed, PVS dilation is affected by a variety of reasons including aging, vascular disease, Alzheimer's disease, etc.In community subjects, the interferences from disease pathologies are relatively mild, and we utilized a more advanced method that allows us to measure PIs in multiple vessel segments simultaneously and avoids physiology-induced variations in  PIs measurements.These are probable reasons that we associations between PIs and PVS in both BG and CSO regions.Nonetheless, validation in larger cohorts is needed due to the complexity of this issue.PI flow in all vessel segments was associated with FW.To the best of our knowledge, no previous studies have reported an association between intracranial arterial pulsatility and brain FW.One related study in community subjects (n, 2422) found that hypertension was associated with higher carotid-femoral pulse wave velocity (CFPWV), an index reflecting vessel stiffness, and CFPWV was positively associated with brain FW content (Maillard et al., 2017).This finding aligns with our results, suggesting that aortic stiffness would allow high pulsatility transferred to downstream intracranial vessels and change fluid dynamics in the brain.Nonetheless, many associations disappeared after adjusting for age, sex, and normalized WMH volume.The observed effect size appears to be small, indicating that validation in larger samples might be necessary.Notably, besides the effect of the glymphatic system, other potential mechanisms, such as inflammation and blood-brain barrier disruption (Pasternak et al., 2015;Altendahl et al., 2020;Hillmer et al., 2023), have been suggested to relate to increased FW.Nonetheless, considering that our participants were normal community people, the impact of these pathologies should be mild.
ALPS was associated with PI flow in the BA segment and PI area in the C2 and C7 segments.ALPS has been extensively used in recent studies to evaluate the glymphatic function and was found reduced in many neurological disorders.On the other hand, increased pulsatility in BA has been demonstrated associated with neurological deterioration after stroke (Yoo et al., 2022).Considering that BA has little anatomical relationship with deep medullary veins in the parietal lobe where we measure the ALPS index, it is possible that PI changes in the BA are a reflection of global vascular pathology.PI area in the ICA segments, on the other hand, may indeed contribute to fluid motion in downstream PVS.To be noted, here PIs were measured from large arteries, but ALPS is supposed to reflect fluid motion in the peri-venous spaces (Taoka et al., 2017).While cerebral arterial and venous pulsatility are expected to be correlated to a certain extent, the pulsatile blood flow tends to be dampened as it flows to downstream vessels (Belz, 1995;Zarrinkoob et al., 2016).Interestingly, it is possible to measure fluid flow in the peri-arterial space with some new diffusion imaging methods (Harrison et al., 2018;Wen et al., 2022).Validation using these methods may help answer this question better.
We did not find stable associations between PI area and the presumed glymphatic markers.While there were some occasional findings, the statistical significance was marginal.This might be attributed to the technical limitations of 4D flow on clinical 3T scanners.To achieve simultaneous flow quantification across a large volume, we cannot set the spatial resolution too high, given the necessity to maintain an adequate signal-to-noise ratio and a reasonable scan duration.On the other hand, the intracranial arteries have relatively small vessel diameters.Therefore, the accuracy of PI area estimation is compromised.Previously, one study used 2D PC-MRI with reconstructed spatial resolution 0.25 × 0.25 × 3 mm 3 to estimate area pulsatility in healthy subjects (mean age: 46.2).They reported higher area pulsatility in similar vessel segments than our results.The PI area in the BA was 0.13~0.27and the PI area in the ICA was between 0.11~0.3,higher than the current results (van Tuijl et al., 2020).Nonetheless, there are also several limitations of 2D PC-MRI, such as the need for multiple scans and precise placement of acquisition planes.A recent study showed that area PIs measured from 4D flow using a spatial resolution of 0.78×0.78×0.8mm 3 were well-correlated to 2D PC-MRI, suggesting the potential of this new technique (van Hespen et al., 2022).
Strengths of this study include its novel research design and the use of advanced multi-model imaging methods.The 4D flow imaging method allowed us to find consistent associations across different vessel segments.We covered three mainstream markers presumably reflecting both structural and functional changes in the glymphatic system.Based on this design, we found interesting associations that would help improve our understanding of the association between vascular pulsatility and variations in the glymphatic system.A few limitations exist, though.First, we could not image penetrating arteries as their diameters are only a few hundred micrometers.It has been proven as a difficult task even on ultra-high field MRI.We presumed that there was a structural-functional continuity between large and small vessels in the human body, which has been implied in some previous studies (Nichols, 1990;Fico et al., 2022;van den Kerkhof et al., 2023a).Nonetheless, we must admit that PIs in large arteries cannot fully reflect the situations in penetrating arteries.Secondly, due to a similar imaging resolution issue, the PI area values were lower than in some literature reports.The negative association between PI area and glymphatic markers still needs to be confirmed in future studies.Third, while our aim was to understand the relationship between vascular pulsatility and glymphatic function, employing the gold-standard intrathecal injection DCE-MRI method to assess glymphatic function in healthy controls is nearly unfeasible.Therefore, we chose the three most widely used markers to represent glymphatic-related changes.It should be noted that the pathophysiological underpinnings of these markers are still undergoing validation.Caution should be exercised when interpreting the results.

Conclusions
In summary, we demonstrated that intracranial arterial pulsatility was associated with presumed neuroimaging markers of the glymphatic system, providing important in vivo evidence of the role of cerebral arterial pulsation in glymphatic activity.

Declaration of competing interest
None.

Fig. 1 .
Fig. 1. (A) thirteen selected arterial segments with measurement planes placed perpendicular to the vessel orientations.(B) pulsatile flow waveforms of selected vessels through the cardiac cycle from one participant.Abbreviations: ICA: internal carotid artery; MCA: middle cerebral artery; ACA: anterior cerebral artery; PCA: posterior cerebral artery; BA: basilar artery.

Table 2
Summary of flow pulsatility indices, area pulsatility indices, and presumed glymphatic markers.

Table 3
Association between flow pulsatility and PVS score.

Table 4
Association between flow pulsatility and ALPS, FW. multivariable linear regression model, adjusted for age, sex, and normalized WMH volume.In ALPS analysis, the mean FA and MD values in ROIs were included as covariates additionally.Model 3: multivariable linear regression model, adjusted for age, sex, normalized WMH volume, and VRF total score.In ALPS analysis, the mean FA and MD values in ROIs were included as covariates additionally.PI area ALPS index.After adjusting for age, sex, normalized WMH volume, VRF total score, and mean FA and MD values in ROIs, higher PI area in the ICA C2 segment (Table 6, β, -0.239, p, 0.041) and ICA C7 segment (Table 6, β, -0.238, p, 0.037) were correlated with lower ALPS.PI area and FW.PI area in the ICA C7 (Table 6, β, 0.231, p, 0.049) was positively associated with FW in the univariate model.However, while adjusting for age, sex, normalized WMH volume, and VRF total score, the correlation disappeared.

Table 5
Association between area pulsatility and PVS score.

Table 6
Association between area pulsatility and ALPS, FW.