Regional hypoxic cerebral vasodilation facilitated by diameter changes primarily in anterior versus posterior circulation

Abstract

The inability to quantify cerebral blood flow and changes in macrocirculation cross-sectional area in all brain regions impedes robust insight into hypoxic cerebral blood flow control. We applied four-dimensional flow magnetic resonance imaging to quantify cerebral blood flow (ml • min⁻¹) and cross-sectional area (mm²) simultaneously in 11 arteries. In healthy adults, blood pressure, O₂ Saturation (SpO₂), and end-tidal CO₂ were measured at baseline and steady-state hypoxia (FiO₂ = 0.11). We investigated left and right: internal carotid, vertebral, middle, anterior, posterior cerebral arteries, and basilar artery. Hypoxia (SpO₂ = 80±2%) increased total cerebral blood flow from 621±38 to 742±50 ml • min⁻¹ (p < 0.05). Hypoxia increased cerebral blood flow, except in the right posterior cerebral arteries. Hypoxia increased cross-sectional area in the anterior arteries (left and right internal carotid arteries, left and right middle, p < 0.05; left and right anterior p = 0.08) but only the right vertebral artery of the posterior circulation. Nonetheless, relative cerebral blood flow distribution and vascular reactivity (Δ%cerebral blood flow • ΔSpO₂⁻¹) were not different between arteries. Collectively, moderate hypoxia: (1) increased cerebral blood flow, but relative distribution remains similar to normoxia, (2) evokes similar vascular reactivity between 11 arteries, and (3) increased cross-sectional area primarily in the anterior arteries. This study provides the first wide-ranging, quantitative, functional and structural data regarding intracranial arteries during hypoxia in humans, highlighting cerebral blood flow regulation of microcirculation and macrocirculation differs between anterior and posterior circulation.

Introduction

The human brain has a critical requirement for cerebral blood flow (CBF) to meet constantly high metabolic demands. The inability to maintain appropriate CBF is linked to many neurologic and cerebrovascular diseases, such that cerebrovascular health remains a substantial concern for the American population.^1,2 These diseases are commonly expressed heterogeneously across brain regions, such that, non-uniform vascular reactivity or CBF distribution may serve to prevent, or promote, pathologic progression.

With the prevalence of disorders like sleep apnea, as well as human visitation to high altitude, cerebrovascular responses to hypoxia have increased in relevance.^3,4 Evidence in both animal^5,6 and human studies^7–9 suggest the CBF response to hypoxia may be regionally specific, but conclusions are limited due to limited number of arteries studied, as well the use of Doppler methods which primarily provide blood velocity but not volumetric flow. For example, during moderate hypoxia (SpO₂ above 80% associated with sleep apnea), changes in blood velocity were observed in the cerebral arteries, while no changes in CBF were observed in extracerebral vessels. ⁷ Interestingly, the posterior cerebral circulation (vertebral artery) displayed greater reactivity compared to the anterior circulation (internal carotids and middle cerebral arteries) during severe hypoxia (SpO₂ = 70%), suggesting CBF responses to hypoxia are non-uniform. ⁷ However, recent discovery of middle cerebral artery dilation during hypoxic stress^10,11 has highlighted the difficulty in interpreting physiological significance of changes in blood velocity alone, without providing volumetric flow measures that incorporate vessel area changes. In fact, quantifying CBF in cerebral arteries is rarely accomplished due to poor ultrasound access to intracranial vessels. Taken together, comprehensive and quantitative examination of the conduit cerebral arterial structural and functional responses to hypoxia have not yet been attained in humans. Consequently, mechanistic insight into where and how CBF is regulated during systemic hypoxia has been hindered.

Therefore, the purpose of this experiment was to quantify total and regional CBF, cerebral arterial cross-sectional area (CSA), and CBF distribution during systemic hypoxia in conscious humans, to comprehensively test whether hypoxic cerebrovascular reactivity differs between brain regions. This investigation was possible due to recent advances in magnetic resonance imaging (MRI) methodology with a 4D MR Flow ¹² approach with high spatial resolution called Phase Contrast Vastly undersampled Isotropic Projection Reconstruction (PC VIPR).^13,14 This methodology allows non-invasive measurement of human cerebral vascular CSA and flow over a large imaging volume within a single acquisition, without use of a contrast agent.^15–22 Simultaneous measurement of volumetric flow and CSAs of 11 major cerebral arteries offers extensive new insight into whether hypoxic cerebral vasodilation varies between extracerebral versus cerebral, or anterior versus posterior arteries, in addition to whether macrocirculation contributes to change in vascular resistance that are observed.

Methods

Subjects and informed consent

Twelve healthy young volunteers (six females, 30 ± 2 years, Table 1) were enrolled in the study. Subjects were lean (BMI < 25 kg • m⁻²), normotensive, non-smoking, and not currently on any medications. Subjects were instructed to arrive on the study day fasted (≥ 4 h) and to refrain from caffeine, exercise, alcohol, over the counter medications, supplements, and nonsteroidal anti-inflammatory drugs for a minimum of 24 h prior to study day. Women were studied during the early follicular phase of the menstrual cycle (days 1–5) and were not pregnant (urine pregnancy test). Procedures were approved by the Institutional Review Board of the University of Wisconsin – Madison and conformed to the standards set by the Declaration of Helsinki. Written informed consent was obtained before study participation.

Table 1.

Subject characteristics.

Characteristic	Mean ± SE
Age (years)	30 ± 2
Height (cm)	174 ± 2
Mass (kg)	70 ± 3
Body mass index (kg • m−2)	23 ± 0.5
Waist circumference (cm)	81 ± 2
Hip circumference (cm)	99 ± 1
Fasting glucose (mg • dL−1)	70 ± 2
Total cholesterol (mg • dL−1)	155 ± 9
High-density lipoprotein (mg • dL−1)	62 ± 5
Low-density lipoprotein (mg • dL−1)	78 ± 14
Triglycerides (mg • dL−1)	83 ± 14
Systolic blood pressure (mmHg)	119 ± 2
Diastolic blood pressure (mmHg)	75 ± 2
Mean arterial pressure (mmHg)	90 ± 2

Screening visit

Before participation in the study, all subjects completed a laboratory visit to determine eligibility and safety before completing the experimental procedures. During this visit, height, weight, hip and waist circumference, fasting blood glucose and lipids, as well as blood pressure were measured.

Instrumentation

After the screening visit, the imaging study was completed on a clinical 3 T MRI system (Discovery MR750, GE Healthcare, Waukesha, WI, USA) using an eight-channel head coil. Heart rate (HR, pulse rate via pulse oximeter), blood pressure (MAP, by automated sphygmomanometry), arterial oxygen saturation (SpO₂, pulse oximeter), and breath-by-breath end-tidal CO₂ (P_ETCO₂) ⁷ were collected continuously throughout the imaging study and acquired with an MRI compatible monitor (Medrad Veris MR Vital Signs Patient Monitor, Bayer Healthcare, Whippany, NJ, USA). In addition, the subject wore a two-way non-rebreathing Hans-Rudolph valve (2630 series, Hans Rudolph Inc., Shawnee, KS) and nose clip to experimentally control gas concentrations of inhaled air.

Hypoxia

Initially, a 3-min scout scan was conducted to prescribe the 4D Flow MR imaging volume over the Circle of Willis. After the scout scan, subjects rested quietly while breathing room air for 2 min before the acquisition of baseline PC VIPR flow MRI scans. After normoxia baseline scans, subjects breathed a hypoxic gas mixture (11% O₂, and balance nitrogen). To maintain isocapnia, breath-by-breath P_ETCO₂ was monitored, and CO₂ (100% CO₂) was titrated into the gas mixture to maintain a steady level of baseline P_ETCO₂. Once a steady level of oxygen saturation and end-tidal CO₂ was reached (no change in % SpO₂ and P_ETCO₂ for > 1 min), typically within 5 min, CBF was measured with PC VIPR MRI.

Flow imaging

4D Flow MRI measurements were obtained with PC VIPR, which has been shown to provide accurate measurement of pulsatile blood flow in cerebral arteries. ¹⁵ The scan parameters used were: imaging volume = 22 × 22 × 22 cm³, (0.69 mm) ³ acquired isotropic resolution, scan time = 5 min 30 s, velocity encoding (Venc) = 100 cm/s, flip angle = 20°, TR/TE = 6.7/2.8 ms, 20 reconstructed cardiac time frames using retrospective cardiac gating and temporal view sharing.^17,23

Post-processing

All PC VIPR flow MRI processing was completed using an in-house tool developed in commercial software (Matlab, The Mathworks, Natick, MA, USA). This tool has been show to be a reliable, fast, and user-independent method for the assessment of intracranial arteries from 4D flow MRI.^16,17 With a centerline-processing scheme, a user can interactively select a vessel position and an underlying algorithm conducts the segmentation of a vessel plane that is perpendicular to the local vessel path and one voxel width thick (0.69 mm) to simultaneously provide the corresponding velocity, CSA, and flow measures.

Flow and CSA parameters were averaged over five consecutive cross sections to produce a 0.69 mm × 5 = 3.45 mm long segment in each artery of interest. This averaging of cross sections provides a more robust estimate of the flow parameters, ¹⁶ especially in the presence of complex flow near branches. Eleven arteries of interest were assessed for each subject as shown in Figure 1: left and right vertebral arteries (VA), the inferior basilar artery (BA), left and right internal carotid arteries (ICA), left and right middle (MCA), left and right anterior (ACA), and left and right posterior cerebral arteries (PCA). VA flow was measured 4–5 mm from the junction with the basilar artery, and basilar flow was assessed in the most inferior portion near this junction. PCA measurements were taken 4–5 mm from the junction with the BA. ICA measurements were performed in the straight portion of the ICA C4 segment. ²⁴ ACA (A1 segment) and MCA (M1 segment) measurements were performed 4–5 mm from their junction with the ICA. Total CBF was calculated as the sum of ICAs and BA as all the flow into the brain will have derived from these arteries (flow from VA feed the BA). ²⁵ Intracerebral blood flow was calculated as the sum of ACAs, MCAs, and PCAs. Therefore, ICAs, VAs, and BA were referred to as extracerebral arteries, and ACAs, MCAs, and PCAs were referred to as intracerebral arteries. Vascular reactivity was calculated as the Δ in blood flow relative to the Δ in SpO₂ from normoxia to hypoxia (Δ ml • min⁻¹ • Δ SpO₂⁻¹). Relative vascular reactivity was calculated as the %Δ in blood flow relative to the Δ in SpO₂ (Δ% ml • min⁻¹ • Δ SpO₂⁻¹). Flow distribution was calculated as the flow through the given artery divided by CBF, multiplied by 100%. ²⁵

Figure 1.

Four-dimensional MRI- (PC VIPR) derived angiograms from a representative subject without contrast agent. Blue squares indicate cut-planes for 3.45 mm long measurements within a given vessel. (a) Coronal view, (b) Axial view, (c) Sagittal view of intracranial arteries segmented from the PC VIPR angiogram. Angiograms with overlaid velocity tracings along measurement points (d) Coronal view, (e) Axial view, and (f) Sagittal view. Arteries examined: ICA: right (r) and left (l) internal carotid artery; VA: right (r) and left (l) vertebral artery; BA: basilar artery; MCA: right (r) and left (l) middle cerebral artery; ACA: right (r) and left (l) anterior cerebral artery; PCA: right (r) and left (l) posterior cerebral artery.

Statistical analysis

Statistical analysis was completed using SigmaPlot Version 12.0 (Systat Software, Inc.). Not all arterial responses passed in Shaprio–Wilk tests of normality. Therefore, non-parametric statistical tests were used as a more conservative statistical approach for all variables. Subject characteristics and hemodynamic values were analyzed using a Wilcoxon Signed Rank Test to determine differences between normoxia and hypoxia. Changes in CSA, diameter, and blood flow from baseline (Δ) were tested with One-Sample Wilcoxon Signed Rank Test to determine if the values are different than zero (zero would represent no change from baseline). Differences between reactivity of vessels were examined with Kruskal-Wallis One Way ANOVA. Multiple comparisons were examined with Tukey test. All data are presented as mean ± standard error, and significance was determined a priori at p < 0.05.

Results

Figure 1 illustrates representative images generated from the PC VIPR data. The centerline maps of the entire arterial circulation facilitated artery selection and quantification of key variables in all major cerebral arteries simultaneously. Figure 1(a) to (c) identifies the locations for five single voxel slices used for quantification of blood flow and vessel characteristics. All subjects had normal angiograms with only one presenting P1 hypoplasia and a fetal type left PCA, which is prevalent in 15%–32% of healthy humans. ²⁶ Additionally, one subject had no visible left VA, and a separate volunteer had no visible right VA. Therefore, statistical analysis of variables derived from each of those arteries was conducted with an n = 11.

Systemic hemodynamic responses are summarized in Table 2. Hypoxia increased HR and blood pressure (Table 2) (p < 0.05), and reduced arterial saturation from 98±1% to 80±2% (Table 2). P_ETCO₂ decreased slightly during hypoxia (∼1 mmHg, p = 0.03), despite monitoring P_ETCO₂ and titration of CO₂ into gas mixture to maintain isocapnia.

Table 2.

Hemodynamic parameters during normoxia (baseline) and hypoxia.

	Normoxia	Hypoxia
Heart rate (bpm)	66 ± 2	82 ± 2*
SpO2 (%)	98 ± 0.4	80 ± 2*
MAP (mmHg)	98 ± 3	101 ± 3*
Systolic BP (mmHg)	132 ± 4	133 ± 4
Diastolic BP (mmHg)	82 ± 3	85 ± 3*
PETCO2 (mmHg)	40 ± 0.6	39 ± 0.8*

Data are presented as mean ± SE. Hypoxia increased HR, MAP, DBP, and reduced SpO₂ and P_ET CO₂.

MAP: mean arterial pressure; BP: blood pressure.

*

Significantly different between hypoxia and normoxic baseline (P < 0.05).

Hypoxia increased total CBF from 626 ± 37 to 742 ± 50 ml • min⁻¹ representing a 19 ± 3% increase from baseline (p < 0.05). Similar increases in CBF were observed when calculating extracerebral (BA and both ICAs, Δ 120 ± 24 ml • min⁻¹) or intracerebral flow (2 MCAs + 2 ACA + 2 PCAs, Δ 113 ± 26 ml • min⁻¹). The intracerebral CBF was quantitatively lower than the total CBF due to contributions to smaller unexamined vessels.^16,25

Blood flow increased in all arteries examined, except the right PCA (Figure 2(a)). Interestingly, the relative distribution of flow throughout the brain was unaltered by hypoxia (Figure 2(b)). When expressed as relative increase in CBF, to account for differences in normoxic baseline CBF between different arteries, the increase in CBF was similar across all 11 arteries (Figure 2(c), p > 0.05).

Figure 2.

Cerebral blood flow quantified in 11 arteries supplying the brain. (a) absolute blood flow in each artery at baseline and hypoxia, (b) percent distribution of blood flow in each artery at baseline and hypoxia, and (c) percent change in blood flow from baseline. Data are mean ± SE, * indicates statistical significance from normoxia, p < 0.05. VA—r and VA—l n = 11.

To assess if hypoxic sensitivity differed between vessels, flow reactivity was calculated (Figure 3). In terms of absolute flow (Δ ml • min⁻¹ • Δ SpO₂⁻¹, Figure 3(a)), vessels of the posterior circulation (PCAs and VAs) were found to be significantly less reactive than the ICAs. However, when compared on a relative CBF basis, all vessels were found to have a similar vascular reactivity to the hypoxia ranging between 1% and 1.6%Δ ml • min⁻¹ • Δ SpO₂⁻¹ (Figure 3(b)).

Figure 3.

Hypoxic vascular reactivity in 11 arteries supplying the brain. (a) Absolute blood flow reactivity and (b) Relative blood flow reactivity. Data are means ± SE, #, significantly different than ICA—r,, significantly different than ICA—l, $, significantly different than MCA—r, p < 0.05. VA (r and l) n = 11.

CSA and arterial diameter increased in response to hypoxia in several arteries (Figure 4, Table 3). Most arteries associated with the anterior cerebral circulation dilated. Specifically, both right and left ICAs and MCAs were observed to dilate (∼15 ± 5% and 11 ± 4%, p < 0.05), and ACA vessels trended to increase (∼ Δ 8 ± 7%, p = 0.08). In contrast, CSA of most arteries of the posterior circulation (BA, PCA, VA) remained unchanged to hypoxia, with the exception of right VA dilation (Figure 4(b)).

Figure 4.

Hypoxia-induced change in cross-sectional area (CSA) from baseline. (a) Arteries of the anterior circulation and (b) Arteries of the posterior circulation. Data ± SE, * indicates Δ statistically significant p < 0.05. VA (r–l) n = 11.

Table 3.

Artery cross-sectional area and diameter at normoxia (baseline) and during systemic hypoxia.

		CSA (mm2)		Diameter (mm)
Artery		Normoxia	Hypoxia	Normoxia	Hypoxia
ICA	Right	13.5 ± 1.3	15.8 ± 1.6*	4.1 ± 0.2	4.4 ± 0.2*
	Left	13.0 ± 1.5	14.3 ± 1.7*	4.0 ± 0.2	4.2 ± 0.2*
MCA	Right	6.7 ± 0.4	7.5 ± 0.5*	2.9 ± 0.1	3.1 ± 0.1*
	Left	6.9 ± 0.4	7.5 ± 0.6*	2.9 ± 0.1	3.1 ± 0.1*
ACA	Right	5.2 ± 0.4	5.6 ± 0.4	2.6 ± 0.1	2.7 ± 0.1
	Left	4.9 ± 0.4	5.3 ± 0.7	2.5 ± 0.1	2.6 ± 0.2
VA	Right	5.6 ± 0.7	6.4 ± 1.0*	2.5 ± 0.3	2.7 ± 0.3*
	Left	6.4 ± 1.0	6.3 ± 1.0	2.4 ± 0.4	2.4 ± 0.3
BA		7.4 ± 0.7	7.3 ± 0.6	2.4 ± 0.2	2.3 ± 0.1
PCA	Right	3.8 ± 0.2	3.7 ± 0.2	4.2 ± 0.2	4.5 ± 0.3
	Left	2.2 ± 0.1	2.2 ± 0.1	2.3 ± 0.1	2.4 ± 0.1

Data are presented as mean ± SE. Anterior circulation is the top portion of table, and bottom portion of the table is the posterior circulation. ICA: internal carotid arteries; MCA: middle cerebral arteries; ACA: anterior cerebral arteries; VA: vertebral arteries; BA: basilar artery; PCA: posterior cerebral arteries; CSA: cross-sectional area.

*

Significantly different between hypoxia and normoxic baseline (p < 0.05).

Discussion

To our knowledge, this is the first study in humans to simultaneously quantify regional cerebrovascular functional responses to hypoxic exposure, in all medium to large cerebral arteries. The novel findings of this study were: (1) CBF increased in 10 of the 11 vessels studied, (2) CBF distribution in intracerebral arteries remained similar to normoxia; (3) relative vascular reactivity to hypoxia is similar between extracerebral and intracerebral arteries; and (4) anterior conduit arteries dilate in response to hypoxia more so than posterior arteries. Collectively, the results of this study show that regional blood flow responses and relative vascular reactivity to moderate hypoxia are fairly uniform throughout the cerebral circulation. However, CSA changes in arteries feeding anterior brain regions suggest regulation of arterial resistance differs between the anterior and posterior circulations.

CBF and distribution response to hypoxia

Absolute CBF during resting normoxia (626 ± 37 ml • min⁻¹) is similar to other human studies using ultrasound or MRI technologies.^17,27 By measuring total and regional CBF, our data allow the opportunity to better understand distribution of CBF under both normoxic and hypoxic conditions. The first observation is that during normoxia, the relative CBF (% of total) through the arteries examined was similar to those observed at rest using similar MRI scanning techniques. ²⁵ The absolute increases in CBF in response to hypoxia in each vessel varied widely (Figure 2(a)), with the greatest increases in ICA, since they provide CBF to several intracerebral arteries. Surprisingly, the relative CBF distribution was unchanged during hypoxia (Figure 2(b)), offering the first comprehensive data supporting that at moderate levels of hypoxia, each macrovascular region responds similarly to the reduction in arterial oxygen content.

Reactivity to hypoxia

Due to differences in severity of systemic isocapnic hypoxia in the literature, comparing reactivity (both total and vessel specific) appears to be the most appropriate (% Δ ml • min⁻¹ • Δ SpO₂⁻¹). The ∼1%–1.6% increase per % decrease in SpO₂ observed in the current study (Figure 3(b)) fits within the range of 0.5%–2.5% increase per decrease of SpO₂ reported in the literature using various CBF measurements and degrees of isocapnic hypoxia.^{7,10,11,28–34} Regionally, specific vascular hypoxic reactivity to SpO₂ ∼80% was quite similar across 11 arteries (Figure 3(b)). Thus, our work verifies previous studies suggesting hypoxic reactivity is similar in ICA and VA during mild to moderate hypoxia, however, do not support the concept of regional differences in hypoxic reactivity. ⁷ Importantly, new findings extend this observation to seven additional arteries, including intracerebral vessels that are difficult to access. Taken together, this study provides the first data to quantify CBF responses in MCA, ACA, PCA, and BA, which collectively indicate a largely uniform hypoxic vascular reactivity across all medium to large cerebral arteries.

Literature directly comparing regional CBF responses to hypoxia is sparse. In the most comprehensive prior study, Willie et al. ⁷ quantified blood flow responses in the left ICA and right VA (echo and Doppler ultrasound), as well as blood velocity responses in a MCA and a PCA (trans-cranial Doppler, TCD) to graded isocapnic hypoxia (∼SpO₂ 70%, 80%, 90%). Interestingly, Willie et al. found CBF was unchanged in the neck vessels at 80% SpO₂ despite increases in blood velocity in the PCA and MCA. In fact, CBF only increased in the extracerebral vessels at 70% SpO₂, ⁷ when the VA demonstrated 50% greater reactivity than ICA, suggesting flow increases to a greater extent in the VA supplying blood to the brainstem. ⁷

The present data are the first to use absolute measures of CBF to reflect true reactivity in a comprehensive approach across the cerebral artery circulation. Previous work in both animals and humans has provided indirect evidence that hypoxic vascular reactivity is similar between regions.^28,35,36 Physiologic insight has been limited in these studies due to lack of fully quantified CBF responses and/or a limited number of arteries studied. Despite these limitations, previous work largely argues regional reactivity is uniform. Specifically, our results confirm prior reports from Willie et al.^7,28 that VA and ICA exhibit similar reactivity at ∼80% hypoxia. Only at more severe hypoxia (70%) did the data suggest relative vascular reactivity to hypoxia is higher in VA, ²⁸ which to our knowledge are the only data from humans suggesting extracerebral vascular reactivity may differentiate at more severe desaturation.

The report of greater reactivity in VA versus ICA at SpO₂ = 70% suggests more severe hypoxia may uncover heterogeneous vascular reactivity or regional CBF distribution. Our choice of ∼80% SpO₂ was aimed toward applied relevance since most mountaineering activities and even clinically relevant sleep disorders are experienced above 80% SpO₂.^28,37 Other potential differences in reactivity may be due to differences in experimental protocols. It is plausible that brainstem blood flow is sensitive to duration (≥ 30 min) and/or severity of hypoxia.^7,27 The current protocol used less than 10 min exposure. Additionally, technical differences between MRI and ultrasound may have contributed to the small differences between studies.

A final consideration is the segment of artery examined differs between studies. Specifically, the imaging volume for PC VIPR scans was centered over the Circle of Willis, and therefore measurements of ICA and VA flow were performed superior (anatomically) to those made in previous studies using ultrasound.^7,30 However, this is unlikely to explain any differences in results, since these data add to two independent studies suggesting ICA and VA hypoxia reactivity is similar above a SpO₂ of 80%.^7,30 Thus, three reports suggest the majority of the segments of ICAs, and VAs exhibit similar hypoxic reactivity to moderate systemic hypoxia. Collectively, the most likely explanation for greater VA reactivity in the previous study ⁷ is more severe hemoglobin desaturation.

Taken together with the relative distribution of CBF during 80% hypoxia (Figure 2), our data strongly support the conclusion that during moderate hypoxia, the overarching, integrated hypoxic vascular reactivity does not differ between the posterior and anterior cerebral circulations, or between intracerebral and extracerebral arteries.

Contribution of CSA to CBF

Table 3 and Figure 4 summarize changes observed in CSA and diameter. These data support previous studies demonstrating changes in MCA,^10,11, ICA,^27,28 and VA ²⁷ diameter contribute to regulation of cerebrovascular resistance during hypoxia.^10,11 This study provides new data on intracranial artery CSA (2 × ACA, 2 × PCA, and BA), which provides novel insight into cerebrovascular control that was previously unattainable. These new data demonstrate arteries of the anterior circulation vasodilate (2 × ICA, 2 × MCA, trend for 2 × ACA, Figure 4(b)), whereas, posterior arteries (PCA, BA) do not dilate with the exception of the right VA (Figure 4(b)). These data have noteworthy implications for understanding hypoxic vascular control. First, studies which do not account for increases in arterial diameter ³⁸ may severely underestimate CBF responses, and therefore limit insight into the contribution of conduit arteries to CBF regulation. Second, despite similar integrative hypoxic vascular reactivity (Figure 3(b)) between anterior and posterior cerebral arteries, the anatomical sites where cerebrovascular resistance changes with hypoxia are not uniform throughout the brain (Figure 4). In other words, anterior and posterior arteries achieve increased CBF differently from one another. The absence of an increase in CSA in most of the posterior circulation (VAs, BA, PCAs) indicates downstream pial arteries are the primary sites of vascular control in response to 80% SpO₂ hypoxia. Conversely, CSA increases in ICAs, MCAs, and a strong trend for ACAs (Table 3) indicate that in addition to pial arteries, the macrocirculation plays an important role in hypoxic CBF regulation in the anterior circulation.

Experimental considerations

There are a few important limitations that should be noted. Despite the attempt to maintain isocapnia by titration of CO₂, we observed a slight decrease of ∼1 mmHg in P_ETCO₂. Therefore, it is possible that our values slightly underestimate the hypoxic response, as it is well documented that hypocapnia lowers hypoxic increases in CBF.^39,40 Blood pressure during screening visit indicated healthy levels, but blood pressure was higher during scanning procedures. We speculate the elevated blood pressure is due to the stress of being placed inside the MRI bore, with a mouthpiece and nose clip inside of a very confining head coil. Importantly, blood pressure did not change drastically with hypoxia, and expressing data relative to perfusion pressure (vascular conductance, data not shown) did not alter conclusions. Additionally, we did not examine greater severity of hypoxia, where regional differences in integrated hypoxic reactivity, or distribution of flow, may be uncovered. ⁷ Finally, we have no direct measures of arterial blood gases. We substituted measures of breath-by-breath P_ETCO₂ as surrogate measures of arterial CO₂ considering P_ETCO₂ has shown to be strongly related to direct measures of arterial CO₂ (r² = 0.98). ⁷

An additional consideration is the MRI sequence used in the current study, PC VIPR. PC VIPR is not a commercially available scanning sequence, however, it has been validated extensively for anatomical and hemodynamic assessment. In vivo validation of flow measures is challenging for the lack of a gold standard. Prior work with traditional manual placement of analysis planes and vessels contours includes: (1) vessel area comparisons with contrast-enhanced MRA in the abdomen, ⁴¹ (2) pressure measurements compared with catheter-based invasive measures in a canine neck aneurysm model, ⁴² a canine carotid artery stenosis model, ⁴³ and a porcine study of renal artery stenosis, ⁴⁴ (3) flow comparisons with 2D PC MRI in the chest, ⁴⁵ abdomen, ⁴⁶ and for intracranial vessels, ¹⁵ and (4) flow comparisons with TCD and 2D PC MRI for intracranial vessels. ²⁰ Good correlations were found for vessel area measurements across various diameters ⁴¹ as well as flow measures obtained with 2D PC and TCD with systematically lower peak flow for 2D and 4D PC MRI compared to TCD. ²⁰ Processing with the recently introduced automated processing procedure used here was validated in a cranial phantom and in vivo study and ensures full reproducibility by eliminating user input in the segmentation/contour processing. ¹⁶ Measures are established at five consecutive planes generated automatically perpendicular to the centerline across a 3.5 mm segment of each artery, to reduce the impact of any potential measurement errors. ¹⁷ While even minor patient motion can impact high spatial resolution scans such as this one, radial acquisitions have been shown to suffer much less benign artifacts in form of averaging effects rather than ghosting artifacts seen in traditional spinwarp imaging. ⁴⁷ This property has found widespread use in PROPELLER-type diffusion brain imaging ⁴⁸ that also uses motions sensitizing bipolar gradients like PC VIPR.

Conclusions

The purpose of this experiment was to quantify CBF, distribution of CBF, and arterial CSA, in order to comprehensively assess the existence of regional differences within the cerebrovascular network during systemic hypoxia in humans. The ability to use advanced MRI methodology to simultaneously quantify CSA and volumetric flow through 11 cerebral arteries adds new quantitative structural and functional insight into control of CBF during hypoxia. Our results indicate during moderate hypoxia: (1) CBF increases in 10 of 11 arteries, however, distribution of CBF remains similar to normoxia, (2) integrated hypoxic vascular reactivity was similar between arteries of anterior and posterior circulation as well as between extracerebral and intracerebral arteries, (3) increases in artery CSA highlight the potential for regulation of CBF at the level of macrocirculation, and (4) increases in CSA occur predominantly in the anterior circulation, suggesting a differential regulation of vascular resistance in anterior compared to posterior cerebral arteries. This study provides the first detailed, quantitative, functional, and structural data of intracranial vessels during hypoxia in humans. Considering human travel to altitudes and nocturnal desaturations during sleep apnea, regional differences in vascular control observed in this study hold important implications for deeper understanding of cerebrovascular health and disease.