Normative Cerebral Hemodynamics in Middle-aged and Older Adults Using 4D Flow MRI: Initial Analysis of Vascular Aging

Abstract

Background

Characterizing cerebrovascular hemodynamics in older adults is important for identifying disease and understanding normal neurovascular aging. Four-dimensional (4D) flow MRI allows for a comprehensive assessment of cerebral hemodynamics in a single acquisition.

Purpose

To establish reference intracranial blood flow and pulsatility index values in a large cross-sectional sample of middle-aged (45–65 years) and older (>65 years) adults and characterize the effect of age and sex on blood flow and pulsatility.

Materials and Methods

In this retrospective study, patients aged 45–93 years (cognitively unimpaired) underwent cranial 4D flow MRI between March 2010 and March 2020. Blood flow rates and pulsatility indexes from 13 major arteries and four venous sinuses and total cerebral blood flow were collected. Intraobserver and interobserver reproducibility of flow and pulsatility measures was assessed in 30 patients. Descriptive statistics (mean ± SD) of blood flow and pulsatility were tabulated for the entire group and by age and sex. Multiple linear regression and linear mixed-effects models were used to assess the effect of age and sex on total cerebral blood flow and vessel-specific flow and pulsatility, respectively.

Results

There were 759 patients (mean age, 65 years ± 8 [SD]; 506 female patients) analyzed. For intra- and interobserver reproducibility, median intraclass correlation coefficients were greater than 0.90 for flow and pulsatility measures across all vessels. Regression coefficients β ± standard error from multiple linear regression showed a 4 mL/min decrease in total cerebral blood flow each year (age β = −3.94 mL/min per year ± 0.44; P < .001). Mixed effects showed a 1 mL/min average annual decrease in blood flow (age β = −0.95 mL/min per year ± 0.16; P < .001) and 0.01 arbitrary unit (au) average annual increase in pulsatility over all vessels (age β = 0.011 au per year ± 0.001; P < .001). No evidence of sex differences was observed for flow (β = −1.60 mL/min per male patient ± 1.77; P = .37), but pulsatility was higher in female patients (sex β = −0.018 au per male patient ± 0.008; P = .02).

Conclusion

Normal reference values for blood flow and pulsatility obtained using four-dimensional flow MRI showed correlations with age.

© RSNA, 2023

Supplemental material is available for this article.

See also the editorial by Steinman in this issue.

Summary

Normative cerebrovascular flow rates and pulsatility indexes were obtained in a large sample of middle-aged and older adults using four-dimensional flow MRI and both showed strong correlations with age.

Key Results

• ■ Intracranial hemodynamics were obtained in 759 cognitively unimpaired middle-aged and older adults using four-dimensional flow MRI.

• ■ Blood flow rates and pulsatility indexes were established in 17 major vessel segments for the entire group, as well as by age and sex.

• ■ Blood flow (β = −0.95 mL/min per year; P < .001) and pulsatility (β = 0.01 arbitrary unit per year; P < .001) measures were strongly correlated with age. There was no evidence of sex differences in blood flow (P = .37); however, pulsatility was higher in female patients (P = .02).

Introduction

Vascular aging is known to occur with loss of vascular compliance and decreased cerebral blood flow (1). This is particularly relevant to cerebrovascular and neurologic health. Cardiovascular disease risk factors are associated with vascular dementia, and Alzheimer disease and related dementias (ADRDs) (2,3). It is essential to study the relationship between normal vascular aging, cerebrovascular disease, and neurodegeneration as the aging population increases. In 2022, the national cost of ADRD care was estimated to be around $321 billion (4). Noninvasive reproducible quantitative vascular measures are necessary to understand normal vascular aging and assess cerebrovascular disease. It is therefore imperative to first establish baseline or reference vascular measures in large normative populations of older adults to better understand what is “normal” and to disentangle the contributions of cerebrovascular disease in neurodegeneration and ADRD.

Two commonly used markers to assess flow and loss of vascular compliance are blood flow rate and pulsatility index. Flow rates describe the volume of blood passing through a specific vessel per unit of time and are sensitive to a wide range of vascular disorders, including stroke and stenoses. Pulsatility index reflects arterial compliance and downstream vascular resistance (5,6) and has been associated with cerebral small vessel disease (7) and vascular cognitive impairment (8). Flow rates and pulsatility indexes can be used to study the effects of progressive hemodynamic and vessel wall changes associated with normal vascular aging and cerebrovascular disease (9). Such changes can increase risk of age-related vascular diseases and can lead to increased cerebral pulsatility that can cause microvasculature and parenchymal damage (10). Furthermore, studies have demonstrated decreased blood flow, increased pulsatility, and increased arterial stiffness in individuals with Alzheimer disease compared with age-matched control patients (11–13). Four-dimensional (4D) flow MRI is a validated method to noninvasively quantify cerebral hemodynamics (14,15). Four-dimensional flow MRI can depict blood velocities within the entire brain with submillimeter resolution, which makes it well-suited for comprehensive, large-scale studies.

The primary aim of our study is to apply 4D flow MRI to measure resting intracranial blood flow rates and pulsatility indexes in the main cerebral arteries and venous sinuses in a large cognitively unimpaired cohort of 759 middle-aged (45–65 years) and older (>65 years) adults to establish normative values across a wide age spectrum. As a secondary aim, we investigate the associations of age and sex with intracranial blood flow and pulsatility in this cross-sectional sample.

Materials and Methods

Patients and Study Design

This single-center, retrospective, Health Insurance Portability and Accountability Act–compliant study was approved by the local institutional review board. Written informed consent was obtained for all patients. S.C.J. is a consultant for Eisai, Merck, and Roche Diagnostics. H.A.R. has previously consulted for GE Healthcare, Bracco Diagnostics, Bayer, and Guerbet. The remaining authors, who were not employees of or consultants for these companies, had control of inclusion of data and information that might present a conflict of interest.

Patients were selected as a convenience sample from the Wisconsin Alzheimer’s Disease Research Center and Wisconsin Registry for Alzheimer’s Prevention cohorts undergoing 4D flow MRI between March 2010 and March 2020. These longitudinal cohorts consist of individuals enriched with parental history of Alzheimer disease. All patients underwent neuropsychologic examinations to determine cognitive status, obtained with consensus reviews on the basis of criteria developed by the National Institute on Aging and Alzheimer’s Association (16,17). A subset of these patients (31%; 390 of 1248) underwent Pittsburgh compound B PET to determine amyloid-β burden. To examine age-related changes, those with Alzheimer disease pathologic results (based on a previously established Pittsburgh compound B threshold [18]) or cognitive impairment were excluded (Fig 1). Additionally, 4D flow MRI data sets with poor image or cardiac gating quality were excluded.

Figure 1:

Patient and four-dimensional (4D) flow MRI data set selection process. ADRC = Alzheimer’s Disease Research Center, DVR = distribution volume ratio, 4D = four-dimensional, NIA-AA = National Institute on Aging–Alzheimer’s Association, PiB = Pittsburgh compound B, WRAP = Wisconsin Registry for Alzheimer’s Prevention. * From the 1248 total patients, Pittsburgh compound B PET was only performed in a subset of 390 patients (31%).

Previous studies (11–13,19–24) have used subsets of these cognitively unimpaired patients as control patients to compare blood flow and pulsatility measures with patients who have Alzheimer disease. However, these studies were smaller and/or assessed a limited number of vessel segments. In this study, we reprocessed all cognitively unimpaired patients from the Wisconsin Alzheimer’s Disease Research Center and Wisconsin Registry for Alzheimer’s Prevention cohorts and analyzed flow and pulsatility in all major arterial and venous vessel segments.

MRI Acquisition

We acquired 4D flow data using three separate 3-T clinical systems (GE Healthcare): Discovery MR750 with a 32-channel head coil (n = 565) or an eight-channel head coil (n = 46), Signa PET/MRI with an eight-channel coil (n = 8), or Signa Premier using a 48-channel coil (n = 140). A noncontrast-enhanced, radially undersampled phase-contrast vastly undersampled isotropic projection reconstruction (25,26) 4D flow sequence was used with the following parameters: repetition time msec/echo time msec, 7.7/2.6; flip angle, 8°; isotropic resolution, 0.69 mm; image volume, 22 × 22 × 10 cm³; 11 000 projections; velocity encoding, 80 cm/sec; four- (n = 322) or five-point (n = 437) velocity encoding scheme; and scan time, 5.6 minutes or 7.1 minutes (depending on encoding scheme). Velocity, magnitude, and complex difference data were reconstructed into 20 cardiac time frames evenly spaced over the cardiac cycle on a patient-specific basis using retrospective pulse oximeter gating.

Image Processing

An interactive 4D flow processing tool (Fig 2A) was developed by using software (Matlab v2020b; Mathworks) to facilitate rapid, robust, and reproducible postprocessing for large cohort analysis (27,28). The initial portion of the processing tool is completely automated and involves global thresholding, centerline skeletonization, orthogonal cut-plane generation, in-plane k-means clustering segmentation, and hemodynamic parameter extraction at each centerline point. To reduce noise, quantitative values were spatially averaged over five neighboring centerline points. After initial processing, data were loaded into an interactive three-dimensional interface. A user then selected vessel segments from a three-dimensional view, which updated a separate control window in real-time displaying flow curves, two-dimensional cut-plane images, and segmentation contours. Reporting functions then saved vessel-specific hemodynamics of interest. This analysis tool is publicly available (https://github.com/uwmri/QVT).

Figure 2:

A contrast-unenhanced four-dimensional (4D) flow MRI data set in a 67-year-old female patient without cognitive impairment. (A) Semiautomated cranial 4D flow MRI postprocessing platform. The parameter tool (left) is an interactive three-dimensional interface with color-coded vessel centerlines overlayed on a semitransparent angiogram. Individual vessel segments can be selected for hemodynamic analysis with this tool. The control window (middle) shows local (in-plane) cross-sectional images of magnitude, velocity, complex difference, and segmentation contours, and flow profiles. This information is updated in real time when a vessel is selected from the parameter tool. Blood flow profiles over the cardiac cycle are displayed for the current centerline (black line) and neighboring centerlines (red and blue lines). The built-in visual tool (right) can display three-dimensional velocity vector arrows or “glyphs” within the segmented angiogram. Axial, sagittal, and coronal two-dimensional conventional magnitude sections, simultaneously obtained from the 4D flow MRI examination, are shown. Velocity arrows (image inset in visual tool) are color-coded by velocity magnitude and point in the direction of blood flow. (B) Four-dimensional flow MRI angiograms show vessel locations for hemodynamic measurements performed in this study. Coronal, axial, and sagittal maximum intensity projections (MIPs) from the three-dimensional complex difference data set, which is used for vessel segmentation during postprocessing, are shown. ACA = anterior cerebral artery, BA = basilar artery, ICA = internal carotid artery (C1 or C3 segment), L = left, MCA = middle cerebral artery, PCA = posterior cerebral artery, R = right, SSS = superior sagittal sinus, STR = straight sinus, TS = transverse sinus, VA = vertebral artery.

Pulsatility indexes and cycle-averaged volumetric flow rates were obtained in 17 major vessel segments: cervical internal carotid arteries (ICAs), cavernous ICAs, vertebral arteries, basilar artery, anterior cerebral arteries, middle cerebral arteries, posterior cerebral arteries, straight sinus, superior sagittal sinus, and transverse sinuses. Bilateral vessels (listed as plural) were assessed individually. Pulsatility index was defined as the difference between peak systolic and minimum diastolic flow rates, normalized by the mean flow over the cardiac cycle. Total cerebral blood flow was computed as the sum of flow in the cervical ICAs and basilar artery. Using standardized vessel measurement locations (Fig 2B, Table S1), flow analysis was independently performed by two observers: observer 1 (A.P., with 4 years of radiology experience) and observer 2 (G.S.R., with 5 years of flow postprocessing experience). Observers 1 and 2 were assigned 457 and 302 data sets, respectively, and were blinded to subsequent hemodynamic analyses.

Data sets were inspected for velocity aliasing and, if found, automated 4D Laplacian unwrapping (29) was performed. The quality of in-plane segmentation was evaluated for all vessels and the initial automated segmentation was manually adjusted if inadequate. Vessel segments not visualized (eg, anatomic variants) were reported. Total processing times for each data set were recorded.

Statistical Analysis

Statistical analysis was performed by using software (R version 2.20; Foundation for Statistical Computing) by observer 2. For this study, blood flow and pulsatility were primary outcome measures with age and sex as the main predictors of interest. ComBat data harmonization was applied to account for potential biases from different scanner, coil, and velocity encoding scheme configurations using empirical Bayes framework, as previously reported (30).

To test intra- and interobserver reproducibility of blood flow and pulsatility measurements, observer 2 reanalyzed 30 random data sets and observer 1 analyzed these same data sets 1 month after initial analysis to reduce recall bias. Bland-Altman plots were obtained to visually inspect agreement between observers. Bias, limits of agreement, Pearson correlation, and intraclass correlation coefficients were reported. Internal consistency was assessed using conservation of flow (ie, sum of inflow should equal total outflow). Correlation plots and Pearson correlation coefficients (r values) were obtained in the posterior (basilar and vertebral arteries) and anterior (ICA, anterior cerebral artery, and middle cerebral arteries) circulations. Correlation matrices for flow and pulsatility measures were also obtained.

Descriptive statistics of flow and pulsatility (mean ± SD) for all vessels were reported for the entire sample, by age (half-decade) and by sex. Histograms were used as a data visualization tool to understand distributions of outcome variables. Violin plots were used to view flow and pulsatility distributions, quartiles, and outliers (detected using the interquartile range method). A two-sample t test was used to investigate age differences between sexes. P values less than .05 indicated statistical significance. Multiple linear regression analysis was used to assess the relationship between total cerebral blood flow and age and sex. An adjusted coefficient of determination (R²) was calculated for this model. Two linear mixed-effects models were used to understand the overall associations between age and sex with (1) blood flow and (2) pulsatility across all vessel segments using random intercepts and age slopes at the vessel level with crossed random effects at the patient level. Model performance was assessed using Akaike information criterion by iteratively adding random effects, then fixed effects, to our model. For fixed effects, confidence intervals were computed with likelihood profiles and P values estimated with Satterthwaite method. For the mixed-effects model, age was mean-centered to improve convergence and pulsatility indexes were log-transformed to address positive skewness. For all regression models, assumptions of linearity, normality of residuals, heteroscedasticity, and multicollinearity were checked using regression diagnostics obtained from the Performance package in R (Foundation for Statistical Computing).

Results

Patient Characteristics and Image Processing

A total of 1248 4D flow MRI examinations from individual patients were identified. Abnormal neuropsychologic examinations and/or patients with global Pittsburgh compound B distributed volume ratio greater than 1.19 excluded 230 and 109 patients, respectively. An additional 150 patients were excluded because of poor image quality or poor cardiac gating, and included were 759 patients who were cognitively unimpaired with 4D flow MRI scans suitable for analysis (mean age, 65 years ± 8 [SD]; age range, 45–93 years; 506 female patients, 253 male patients). Inclusion and exclusion criteria are outlined in Figure 1. Patient demographics for the included sample are shown in Table 1. Postprocessing and manual vessel selection took around 5 minutes per patient. In 354 total segments (from 215 patients), velocity aliasing was observed, primarily in the middle cerebral arteries and anterior cerebral arteries. Phase unwrapping successfully corrected aliasing in 280 vessels. Poor segmentation quality was observed in 70 vessel segments (62 patients) and 66 segments were successfully manually segmented. A total of 452 vessel segments (333 patients) were not visualized because of vessel hypoplasia or anatomic variants, observed primarily in the anterior cerebral arteries, vertebral arteries, and transverse sinuses. A more detailed distribution of aliasing, poor segmentation, and nonvisualization by vessel segment is in Table S2.

Table 1:

Patient Demographics

Intraobserver Agreement

Table S3 shows demographics for the 30 patients selected for reproducibility analysis. Pearson and intraclass correlation coefficients were above 0.90 for all flow measures and 15 of 17 pulsatility measures. Median intraclass correlation coefficient of flow and pulsatility measures across all vessels was 1.00 and 0.96, respectively. Bland-Altman plots show both flow and pulsatility measures in an example vessel (left cervical ICA) in Figure S1. Intraobserver agreement and correlation metrics for all vessels are provided in Table S4. Mean bias was 0.27 mL/min (mean limits of agreement, −6.15 to 6.69 mL/min). Maximum flow bias was 1.37 mL/min in the left middle cerebral artery and maximum pulsatility bias was 0.08 arbitrary units (au) in the right anterior cerebral artery.

Interobserver Agreement

Pearson and intraclass correlation coefficients were above 0.90 for all flow measures and 15 of 17 pulsatility measures. Median intraclass correlation coefficient flow and pulsatility measures across all vessels were 0.99 and 0.95, respectively. Bland-Altman plots in Figure S1 also show interobserver agreement, demonstrating similar biases compared with the intraobserver analysis. Interobserver agreement and correlation metrics for all vessels are in Table S5. Mean bias was 0.78 mL/min (mean limits of agreement, −10.9 to 12.5 mL/min). Maximum flow bias was 5.04 mL/min in the left cavernous ICA and 0.07 au for pulsatility in the right anterior cerebral artery.

Internal Consistency and Correlations

Figure S2 shows correlation plots (conservation of flow assessment) for the anterior and posterior cerebral circulations. For the anterior circulation, a strong correlation was observed between the sum of flow in the middle and anterior cerebral arteries and flow in the cavernous ICA (r = 0.84; P < .001). For the posterior circulation, a strong correlation was observed between the sum of flow in the vertebral and basilar arteries (r = 0.95; P < .001).

More broadly, correlation of flow between all 17 vessel segments showed positive correlations in 129 of 136 total observations. Similarly, pulsatility was positively correlated in 132 of 136 observations. Correlation matrices are shown in Figure S3.

Descriptive Statistics

Median and interquartile ranges and flow and pulsatility distributions across all vessels are shown as violin plots in Figure 3. Visually, flow values were mostly normally distributed, aside from the left and right transverse sinuses, which demonstrated skewed distributions. For bilateral segments, flow was qualitatively similar between the left and right sides, aside from the transverse sinuses, where the right transverse sinus was typically dominant. Distributions of raw pulsatility indexes were right skewed but appeared mostly normal after log transformation. A higher number of outliers were observed for pulsatility indexes, relative to flow rates.

Figure 3:

Violin plots. (A) Blood flow rates for all vessel segments, including total cerebral blood flow (TCBF), color coded by vascular region. Kernel density estimations for all vessels follow normal distributions, except for left transverse sinus (LTS) and right transverse sinus (RTS), which were asymmetric. (B) Pulsatility index for each vessel segment. Distributions were skewed right in most vessels and bimodal in some vessels (eg, cavernous internal carotid artery [ICA] and vertebral artery [VA]). Pulsatility index tended to increase in smaller and more distal vessel segments. There were outliers, which may be from velocity noise. ACA = anterior cerebral artery, BA = basilar artery, L = left, MCA = middle cerebral artery, PCA = posterior cerebral artery, R = right, SSS = superior sagittal sinus, STR = straight sinus, TS = transverse sinus.

Mean ± SD of flow rates (Table 2) and pulsatility indexes (Table 3) for all vessels were tabulated for the entire sample, by age group and by sex. There was no evidence of a difference in mean age between male patients (mean age, 65 years ± 8) and female patients (mean age, 65 years ± 8; P = .64). It was qualitatively observed that flow tended to decrease and pulsatility tended to increase across increasing age groups.

Table 2:

Volumetric Blood Flow Rates for Each Vessel Tabulated for the Entire Sample by Age and by Sex

Table 3:

Pulsatility Indexes for Each Vessel Tabulated for the Entire Sample by Age and by Sex

Linear Regression

Multiple linear regression analysis of total cerebral blood flow with age and sex as explanatory variables is shown in Figure 4. For regression coefficients (β value) ± standard error, total cerebral blood flow decreases by approximately 4 mL/min per year with age (β = −3.94 mL/min per year ± 0.44; model R² = 0.10; P < .001). However, total cerebral blood flow was not different between sexes (β = −7.50 mL/min per male patient ± 7.14; P = .29).

Figure 4:

Multiple linear regression plot of total cerebral blood flow with age and sex as predictors. There is no evidence of a difference (P = .29) in total cerebral blood flow between male (blue) and female (red) patients. There was a negative correlation (model R² = 0.10; P < .001) between total cerebral blood flow and age. The regression equation suggests a 3.94 mL/min decrease in total cerebral blood flow per year of age, consistent with findings from other studies. The R² value of 0.10 suggests a high degree of individual variance not accounted for by age and sex.

Table 4 lists estimates for fixed and random effects for both flow and pulsatility models. Lowest Akaike information criterion was observed for the full pulsatility model (Table 4) with crossed random effects and age and sex as fixed effects. However, for the flow model, lower Akaike information criterion was observed without sex as a fixed effect (Akaike information criterion difference, −1). Despite a lower Akaike information criterion, sex was included to assess these effects. Figures 5 and 6 show random effects and predicted trajectories across age for each vessel. Fixed effects showed that blood flow for all vessels decreased around 1 mL/min per year of age (β = −0.95 mL/min per year ± 0.16; P < .001), with no evidence of sex differences (β = −1.60 mL/min per male patient ± 1.77; P = .37). Inversely, fixed effects showed that log-transformed pulsatility for all vessels increases around 0.01 au for each year of age (β = 0.011 au per year ± 0.001; P < .001) and that log-transformed pulsatility is decreased in men (β = −0.018 au per male patient ± 0.008; P = .02). Coefficient estimates of vessel-level random effects are shown in Table S6.

Table 4:

Results of Linear Mixed-Effect Models Assessing Effect of Age and Sex on Flow and Pulsatility

Figure 5:

Results from linear mixed-effects models for blood flow over all vessel segments. Dot plots of random effects for both intercept and age at the vessel level are shown (left side). Mixed-effects regression plots show a clear relationship between flow over age for all color-coded vessel segments over each vascular region (right side). ACA = anterior cerebral artery, BA = basilar artery, ICA = internal carotid artery (C1 or C3 segment), L = left, MCA = middle cerebral artery, PCA = posterior cerebral artery, PI = pulsatility index, R = right, SSS = superior sagittal sinus, STR = straight sinus, TS = transverse sinus, VA = vertebral artery.

Figure 6:

Results from linear mixed-effects models for pulsatility over vessel segments. Dot plots of random effects by vessel are shown (left side). To account for right-skewed distributions of pulsatility index, pulsatility values were log transformed. The mixed-effects regression plots (right side) show that pulsatility tends to increase over age for all vessel segments. PI = pulsatility index, ICA = internal carotid artery (C1 or C3 segment), MCA = middle cerebral artery, ACA = anterior cerebral artery, PCA = posterior cerebral artery, BA = basilar artery, VA = vertebral artery, SSS = superior sagittal sinus, STR = straight sinus, TS = transverse sinus, L = left, R = right.

Discussion

Reference values and ranges for cerebrovascular hemodynamics are needed to better understand cerebrovascular function and disease but are lacking. Our study investigated vessel-specific measures of blood flow and pulsatility in a large normative study sample of 759 cognitively unimpaired middle-aged (45–65 years) and older (>65 years) adults using four-dimensional flow MRI. Descriptive statistics of blood flow rates and pulsatility indexes were tabulated for the entire sample, by age, and by sex in 13 arterial and four venous vessel segments. Excellent intra- and interobserver agreement of flow and pulsatility measures were observed across all vessel segments (all median intraclass correlation coefficients, >0.90). Similarly, conservation of flow correlation analysis demonstrated high internal flow consistency (all r > 0.80). Multiple linear regression demonstrated an overall total cerebral blood flow decrease of around 4 mL/min per year (model R² = 0.10; P < .001), and linear mixed-effects models showed a strong age effect on blood flow (main effect, −0.95 mL/min per year; P < .001) and pulsatility (0.011 au per year; P < .001) with sex differences in pulsatility (−0.018 au per male patient; P = .02).

Reference flow and pulsatility values have been reported in similar populations using two-dimensional phase-contrast MRI (31–34) and transcranial Doppler US (35–38). Relative to other studies, pulsatility indexes were consistent in the larger arteries but were relatively increased in our study in distal arteries (eg, posterior cerebral arteries [34]). However, flow measures and their relationship with age are consistent with other studies. For example, Buijs et al (31) reported an average total cerebral blood flow of 536 mL/min ± 99 (n = 65) between age 60 and 69 years using two-dimensional phase-contrast MRI, whereas our study reported 538 mL/min ± 94 (n = 365) using two-dimensional phase-contrast MRI. Similarly, Scheel et al (35) reported an average total cerebral blood flow of 603 mL/min ± 106 (n = 30) between age 60 and 85 years using US. However, two-dimensional phase-contrast MRI requires tedious plane prescription and is limited to only vessels imaged in-plane. Similarly, US is limited by spatial resolution, field of view, operator dependence, and narrow acoustic bone windows. These factors may greatly reduce the quality of obtained hemodynamic measures and limit the number of vessels analyzed. Four-dimensional flow MRI is advantageous because it allows for retrospective analysis of any vessel of interest in a single volumetric acquisition. Vessel-specific hemodynamics can therefore be obtained across the entire vascular network and can show specific flow distributions that may be differentially altered in some vascular diseases. Furthermore, the effect of vessel structure (eg, anatomic variants and vessel tortuosity) on hemodynamics could be explored using 4D flow MRI, which captures both morphologic structure and hemodynamics simultaneously. Our study extends previous two-dimensional phase-contrast and US studies by measuring both flow and pulsatility in all major intracranial arteries and sinuses while also benefiting from a large sample size. However, 4D flow MRI reduces spatial and temporal resolution, which can restrict analysis in small vessels and may reduce accuracy of time-resolved measurements. However, advances in acquisition (eg, phase-contrast vastly undersampled isotropic projection reconstruction) and reconstruction schemes have allowed for improved resolutions in reasonable scan times.

Multiple linear regression and linear mixed-effects models demonstrate clear decreases in blood flow with increasing age, with an overall total cerebral blood flow decrease of 3.94 mL/min per year, aligning well with other studies (31,32,35). The physiologic basis behind this occurrence is not clear, but several theories (39) suggest that decreased brain parenchyma and metabolism may have a role. It has been hypothesized that reduced cerebral blood flow may partially be the result of age-related central arterial stiffening, increasing the transmission of high-energy pulsatile pressure waveforms into the microcirculation, which leads to microvascular remodeling, increased vascular resistance, and decreased cerebral blood flow (40). This hypothesis is supported by our results and those in other studies (34,41), showing increased cerebral pulsatility index over age. Whereas pulsatility index does reflect cerebral vascular resistance, it is also dependent on the interplay between arterial pressure, arterial compliance, and cerebral perfusion pressure (5). These factors are associated with age-related arterial stiffening and cerebrovascular disease (1). Sex differences in flow have been reported (42) but they were not observed here. Pulsatility index, however, was higher in female patients. Although these results have been observed previously (43), the physiologic mechanisms are unclear but may be related to postmenopausal vascular changes (44).

Our study had limitations. First, we did not exclude patients with vascular disease (eg, hypertension, diabetes, and smoking history) and may not have represented normal hemodynamics of healthy adults. Second, our study sample was not a population-based sample but a convenience sample that had been enriched for risk of Alzheimer disease (ie, higher prevalence of APOE4 genotype and parental history of dementia); therefore, there may have been some patients with presymptomatic Alzheimer disease who may have showed incipient hemodynamic changes. Furthermore, there was a lack of diversity in our sample (mostly female and White patients). These factors may have limited applicability to the general population. Third, whereas postprocessing was mostly automated, analysis required manual selection of vessel segments (~5 minutes per patient) and may have limited clinical translation. Finally, whereas hemodynamics were spatially smoothed, flow waveforms were not temporally filtered. This may have reduced accuracy of pulsatility indexes, particularly in smaller vessels that were prone to the effects of noise (6).

In conclusion, normal reference values for blood flow and pulsatility obtained by using four-dimensional flow MRI showed correlations with age. Four-dimensional flow MRI provided a noninvasive assessment of cerebral luminal blood flow and pulsatility with reasonable imaging times. These reference values are an important first step in defining normative cerebral hemodynamics. Ultimately, the large amount of hemodynamic data obtained, with patient biometric data, can be used to better understand how cardiovascular risk factors and dementia risk factors (eg, APOE4 and white matter hyperintensities) are related to normal vascular aging. Future studies could apply methods such as “robust norms” or Gaussian mixture modeling to detect vascular disease subpopulations, apply atlas-based or deep learning methods to completely automate postprocessing, and use temporal filtering to improve pulsatility index calculations.