Imaging of cerebrovascular reserve and oxygenation in Moyamoya disease

Abstract

This study aimed to determine whether measurements of cerebrovascular reserve and oxygenation, assessed with spin relaxation rate R2′, yield similar information about pathology in pre-operative Moyamoya disease patients, and to assess whether R2′ is a better measure of oxygenation than other proposed markers, such as R2* and R2. Twenty-five pre-operative Moyamoya disease patients were scanned at 3.0T with acetazolamide challenge. Cerebral blood flow mapping with multi-delay arterial spin labeling, and R2*, R2, and R2′ mapping with Gradient-Echo Sampling of Free Induction Decay and Echo were performed. No baseline cerebral blood flow difference was found between angiographically abnormal and normal regions (49 ± 12 vs. 48 ± 11 mL/100 g/min, p = 0.44). However, baseline R2′ differed between these regions (3.2 ± 0.7 vs. 2.9 ± 0.6 s⁻¹, p < 0.001), indicating reduced oxygenation in abnormal regions. Cerebrovascular reserve was lower in angiographically abnormal regions (21 ± 38 vs. 41 ± 26%, p = 0.001). All regions showed trend toward significantly improved oxygenation post-acetazolamide. Regions with poorer cerebrovascular reserve had lower baseline oxygenation (Kendall's τ = −0.24, p = 0.003). A number of angiographically abnormal regions demonstrated preserved cerebrovascular reserve, likely due to the presence of collaterals. Finally, of the concurrently measured relaxation rates, R2′ was superior for oxygenation assessment.

Introduction

In cerebrovascular diseases, perfusion and metabolic function of the brain may be affected concurrently.¹ These changes can be assessed through MR-based measurements of cerebrovascular reserve (CVR) and oxygenation. CVR, the brain's capacity to respond to vasodilatory challenges, provides information beyond baseline perfusion. Indeed, reduced CVR is strongly associated with increased risk of stroke and transient ischemic attack² and has been shown to be a predictor of poor outcome after extracranial–intracranial (EC–IC) bypass.³ Oxygenation, on the other hand, directly assesses tissue function and has shown promise as a biomarker in conditions including brain tumors⁴ and Alzheimer's disease.⁵ Both CVR and oxygenation have shown value in ischemic stroke.^6–8 Combined measurements of CVR and oxygenation may characterize different stages of autoregulatory failure due to hypoperfusion.¹

CVR can be mapped with MRI techniques such as arterial spin labeling (ASL)^9,10 and blood oxygen-level dependent (BOLD) imaging. The former technique quantifies cerebral blood flow (CBF), while the latter measures the combined effect of blood volume, flow, and oxygenation. Acetazolamide injection is a common vasodilatory challenge and causes a robust (30%–60%) CBF increase in normal brain regions,¹¹ enabling CVR mapping. Another common vasodilatory challenge is the hypercapnic challenge, where change in end-tidal partial pressure of carbon dioxide allows for CVR to be normalized across patients. Brain oxygenation also may be mapped with MRI. One general approach, termed quantitative blood oxygen-level dependent (qBOLD) imaging,^12,13 uses measurement of proton transverse relaxation rate R2′ (=1/T2′, the relaxation difference between rates R2* = 1/T2* and R2 = 1/T2), a measurement of local susceptibility from paramagnetic deoxyhemoglobin, to reflect tissue oxygenation. R2 and R2* are also considered to be sensitive to oxygenation.¹⁴ If CBF is augmented while tissue metabolism is stable, tissue oxygenation increases.¹⁵ Tissue oxygenation increase, when unaccompanied by change in blood volume, reduces the amount of paramagnetic deoxyhemoglobin in the imaging voxel, thereby reducing R2′ and R2*.

Moyamoya disease patients provide an ideal cohort for studying pathological changes in CVR and oxygenation. This disease is characterized by progressive stenosis or occlusion of major cerebral arteries.¹⁶ Due to impaired brain blood supply, patients are at increased risk of ischemic stroke, transient ischemic attack, and hemorrhage. Patients may demonstrate reduced or even negative CBF augmentation (a phenomenon known as “steal”, indicating severe disease) in affected regions during the acetazolamide challenge. However, some impacted regions also develop collateral vasculature, which can help compensate for baseline perfusion deficits and may improve outcome,¹⁷ even in the presence of severe angiographic abnormality.

In this study, we aimed to characterize both CVR (with ASL pre- and post-acetazolamide) and oxygenation (with R2′) in pre-operative Moyamoya disease patients. We also compared R2′ oxygenation measurements to concurrently measured R2* and R2, to assess which relaxation rate should be preferred for oxygenation measurement.

Materials and methods

Patient population

Patients were eligible for this prospective study if they had nonsurgically treated Moyamoya disease and were able to undergo the protocol below. Written informed consent was obtained and the study was HIPAA compliant with approval from the Stanford University Institutional Review Board. Twenty-five patients (age: mean 38 years, range 19–70 years; female: N = 20, mean 38 years, range 19–52 years; male: N = 5, mean 39 years, range 21–70 years) were included in this study between October 2014 and August 2015.

MRI and acetazolamide challenge protocol

All imaging was performed at 3.0T (MR750, GE Healthcare, Waukesha, WI, USA). Noncontrast 3D time-of-flight MR angiography was performed to image the Circle of Willis (TOF MRA: TR/TE 21/2.3 ms, voxel size 0.5 × 0.7 × 1.2 mm³, 48 mm craniocaudal coverage). T1-weighted anatomical images were acquired for structural segmentation, using a 3D inversion recovery-prepared fast spoiled gradient-recalled sequence (TR/TE/TI 9.2/3.7/400 ms, flip angle 13°, voxel size 0.9 × 0.9 × 1.2 mm³).

Acetazolamide (1 g IV) was administered to all patients. Prior to and at least 10 min after injection, a set of two sequences was repeated (Figure 1). The first was multi-delay pseudocontinuous ASL (3D stack-of-spirals fast spin-echo (FSE) readout, TR/TE 6484–6518/23.3–25.1 ms, label time 2000 ms, 5 post-label delays [PLDs] equally spaced between 700 and 3000 ms, voxel size 3.4 × 3.4 × 4 mm³ with 36 axial slices for whole-brain coverage). This sequence produces CBF maps that account for the potential differences in arterial arrival time that can occur in cerebrovascular disease due to slow flow or collaterals.^18–20 Oxygenation imaging was performed using a massively multi-echo 2D Gradient-Echo Sampling of Free Induction Decay and Echo sequence (GESFIDE: 2D Cartesian, GeneRalized Autocalibrating Partially Parallel Acquisitions [GRAPPA] acceleration factor 2, TR 2000 ms, 42 echoes with TE 3–130 ms, spin echo at 100 ms, voxel size 1.9 × 1.9 × 1.5 mm³ and inter-slice gap 1mm, 14 slices).²¹ High spatial resolution images were acquired to minimize intra-voxel dephasing due to macroscopic gradients. Slice prescription covered a 34 mm-thick slab, with the inferior-most slice positioned through the superior aspect of the lateral ventricles.

Figure 1.

Imaging protocol. Anatomical imaging sequences included MRA and T1-weighted IR-SPGR sequence. There was always at least 10 min between repetitions of the GESFIDE and multi-delay ASL sequences, which allowed for the effect of acetazolamide to stabilize.

Data pre-processing and analysis

After acquisition, images were processed in a MATLAB (Mathworks, Natick, MA, USA) pipeline, which utilizes FSL (Oxford University, Oxford, UK), SPM8 (University College London, London, UK), and custom in-house developed code. To avoid introducing artifacts, GESFIDE images were analyzed in their native space, without co-registration or filtering. Transverse relaxation rate maps were calculated using mono-exponential fitting,²¹ producing multiple transverse relaxation rates concurrently (R2*, R2, and R2′ [=R2*−R2]). Quantitative CBF maps were automatically generated from the ASL images as shown in Dai et al.,²² and then co-registered to the GESFIDE images.

ROI definitions

Following brain extraction, the T1-weighted anatomical images were co-registered to the GESFIDE images and segmented into gray matter, white matter, and cerebrospinal fluid. Prior infarcts visible on T1-weighted imaging as hypointensity were automatically excluded by the FSL segmentation algorithm. Data were analyzed in annular, mixed-cortical ROIs corresponding to canonical vascular territories (anterior cerebral artery (ACA), middle cerebral artery (MCA), and posterior cerebral artery (PCA)), as shown in Figure 2. ROIs were defined as 12 annular radial segments at two levels, each with thickness 14 mm, with exclusion of the superior-most and inferior-most 1.5 mm slices.

Figure 2.

Definition of regions of interest (ROIs) for analysis of perfusion and oxygenation. (a) illustrates the sagittal view of the imaging volume of the GESFIDE sequence (bold white box) and the multi-delay ASL sequence (dashed box). Analysis was only performed on data from the GESFIDE imaging volume. ROIs were defined by summing adjacent slices to form two slabs, with center of the slabs illustrated at the levels of the cyan and yellow slices shown in (a). (b) illustrates an example case where the right MCA is occluded and the right ACA is stenosed, as indicated by the red arrows. At each level, six radial mixed-cortical segments were defined symmetrically to the midline, as shown in (c) and (d), shown superimposed on the T1-weighted anatomical images. Red segments correspond to angiographically abnormal ROIs, while green segments are angiographically normal ROIs.

Definition of tissue status

Each ROI was defined as “normal” or “abnormal” angiographically and functionally. The primary definition was based on MR angiography. ROIs approximating the territories of occluded or stenotic supratentorial cerebral arteries (ACA, MCA, and PCA) were classified as “abnormal,” while the rest were classified as “normal.” Classification was performed by a neuroradiologist with over 10 years of experience (author GZ) in a blinded manner without knowledge of CBF or oxygenation measurements. To better understand relationships between angiographic abnormalities, oxygenation, and CVR, we performed a secondary analysis in which normal and abnormal region classification was based on that region's CVR. Functionally abnormal ROIs were defined as those with less than 10 ml/100 g/min of absolute CBF increase, and/or less than 10% CBF increase.²³

Statistical analysis

Measurements were compared between normal and abnormal regions, as well as pre- and post-acetazolamide, using the Wilcoxon rank sum test and Wilcoxon signed rank test, respectively. The two definitions of tissue status (angiographic and functional) were tested for agreement using the unweighted Cohen's kappa statistic. Expecting collaterals to cause some angiographically abnormal regions to have normal CVR, but not the reverse, we tested for asymmetry using McNemar's test. To test which relaxation rate was most sensitive to perfusion, we assessed the correlation between ΔCBF and each of the three concurrently measured relaxation rates (R2′, R2*, and R2). To evaluate correlation between variables without assuming Gaussian distribution, we examined the overall trend among data quartiles using the Jonckheere-Terpstra trend test and assessed correlation across all data points using Kendall's tau coefficient. All statistical analyses were performed using Stata Release 14.1 (StataCorp LP, College Station, TX, USA). Due to the inherent dependence between multiple ROIs from the same subject, clustering within patient was adjusted for in all statistical analysis, using the cluster-adjusted “somersd” procedure for the Wilcoxon tests and tau correlations, and by cluster-based bootstrap estimation for the kappa statistic and McNemar tests. Potential intrinsic differences in tissue composition and vasculature, and therefore perfusion and relaxation, were accounted for in all statistical analysis by stratifying ROIs by location—anterior (ACA and MCA territories) and posterior (PCA territories). Significance level was set to 0.05.

Results

Patient cohort disease extent

ASL-based CBF and R2′-based oxygenation imaging was performed successfully pre- and post-acetazolamide in all patients. As is typical for Moyamoya disease, most (but not all) had unaffected posterior circulation, with similar numbers of angiographically abnormal ROIs in each hemisphere. A total of 14 patients had bilateral disease, 6 had left-sided unilateral disease, and 5 had right-sided unilateral disease. Out of 300 ROIs in total, 49% were classified as angiographically abnormal and 35% as functionally abnormal.

Perfusion and CVR

Baseline CBF did not differentiate between angiographically normal and abnormal regions (48 ± 11 vs. 49 ± 12 mL/100 g/min, p = 0.44). However, there was significantly higher CBF augmentation (ΔCBF: 19 ± 14 vs. 9 ± 18 mL/100 g/min, p = 0.001), reflecting better CVR in the angiographically normal regions. More details are shown in Table 1. Histogram analysis (Figure 3) shows that abnormal regions demonstrated highly heterogeneous response to the acetazolamide challenge, with some ROIs augmenting robustly while others experienced reduced augmentation or steal.

Table 1.

Summary of perfusion and oxygenation measurements (mean ± standard deviation), comparing angiographically normal and abnormal regions.

	Angiographic ROI definition
	Normal (N = 154)	Abnormal (N = 146)	Abnormal–normal difference	Abnormal vs. normal
Pre-acetazolamide CBF (mL/100 g/min)	48 ± 11	49 ± 12	3% ± 34%	NS: p = 0.75
Post-acetazolamide CBF (mL/100 g/min)	67 ± 15	58 ± 20	−12 ± 38%	p = 0.03
ΔCBF (mL/100 g/min)	19 ± 14	9 ± 18	−51 ± 123%	p = 0.002
ΔCBF (%)	41 ± 26	21 ± 38	−49 ± 112%	p = 0.002
Pre-acetazolamide CBF vs. Post-acetazolamide CBF	p<0.001	p = 0.02	NA	NA
Pre-acetazolamide R2′ (s−1)	2.9 ± 0.6	3.2 ± 0.7	13 ± 31%	p = 0.01
Post-acetazolamide R2′ (s−1)	2.8 ± 0.6	3.2 ± 0.7	14 ± 33%	p = 0.02
ΔR2′ (s−1)	−0.1 ± 0.4	−0.1 ± 0.4	−25 ± 540%	NS: p = 0.91
ΔR2′ (%)	−2.5 ± 13.8	−1.9 ± 12.0	−34 ± 719%	NS: p = 0.91
Pre-acetazolamide R2′ vs. Post-acetazolamide R2′	NS: p = 0.14	NS: p = 0.13	NA	NA

Figure 3.

Histograms and fitted normal distributions of pre- and post-acetazolamide measurements of CBF in all ROIs for all subjects, in angiographically normal and abnormal ROIs. CBF measurements were consistently augmented in normal ROIs. However, in addition to reduced average augmentation, abnormal ROIs displayed heterogeneous behavior. Some ROIs (black arrow) demonstrated cerebrovascular steal, while other ROIs (white arrow) augmented robustly. The former indicated very poor CVR, while the latter had preserved CVR, likely as a result of adequate collaterals.

Oxygenation

Baseline R2′ was 13% higher (corresponding to lower oxygenation) in angiographically abnormal ROIs compared with normal ROIs (3.2 ± 0.7 vs. 2.9 ± 0.6 s⁻¹, p < 0.001). ΔR2′ (the change in R2′ after acetazolamide) decreased in both regions, which would be consistent with improved oxygenation, but this change was not statistically significant (p = 0.14, p = 0.13 in normal and abnormal regions, respectively). ΔR2′ also did not significantly differ between angiographically normal and abnormal regions (normal: −0.1 ± 0.4 s⁻¹ or −2.5 ± 13.8%; abnormal: −0.1 ± 0.4 s⁻¹ or −1.9 ± 12.0%; p = 0.70). Figure 4 illustrates a case where ROIs with poor CVR had higher baseline R2′ measurements than neighboring ROIs with good CVR. This was a general finding of the study, where a significant correlation between the baseline R2′ and CBF augmentation was found (Kendall's τ = −0.24, p = 0.003) (Figure 5), demonstrating that areas with poor CVR tended to have reduced oxygenation at baseline. Further analysis by quartile of ΔCBF also showed a linear relationship in quartile centroids (Figure 5), and the Jonckheere-Terpstra test observed a significant (p < 0.001) downward trend in R2′ as ΔCBF increased.

Figure 4.

Example CBF, CVR, and R2′ imaging in a patient (F, 47y) with bilateral Moyamoya disease. (a) This subject has severe bilateral disease—only the right ACA and left PCA territories were considered angiographically normal; b) regional mean baseline CBF shows no clear correlation with angiographic status, even recording high CBF in regions considered to be chronically hypoperfused; c) percentage CBF change after acetazolamide indicates several territories (indicated by arrows) with very low values (<10 mL/100 g/min or < 10%) and therefore poor CVR; d) regions with low or negative CBF augmentation are defined as functionally abnormal; e) mean baseline R2′ indicates that more severe the CVR defect, the more elevated are the R2′ values (arrows), indicating lower tissue oxygenation.

Figure 5.

Baseline R2′ and acetazolamide-induced ΔCBF measurements are negative correlated, both among the (a) individual ROI measurements and in the (b) ΔCBF quartile means. The quartiles are shown in different colors on (a) for illustrative purposes. Since higher R2′ represents poorer oxygenation, these relationships show that regions with poor CVR tend to have lower baseline oxygenation.

Impact of defining tissue status based on angiographic versus functional criteria

Agreement between the angiographic and functional definitions was 67% overall and was higher in the posterior territories (82%) than in the anterior territories (60%), as would be expected based on the typical distribution of disease in Moyamoya. Unweighted Cohen's kappa statistic demonstrated significant (p < 0.001) but moderate (κ = 0.33 with 95% confidence interval 0.19 to 0.50) agreement between the two definitions of tissue status. Although there was asymmetry between the two types of discordant pairs (i.e. where the definitions disagreed), McNemar's test did not show significant asymmetry (p = 0.15). Nevertheless, the trend of a larger group of angiographically abnormal but functionally normal ROIs is consistent with the presence of adequate collaterals (Table 2 and Supplemental Table 1).

Table 2.

Comparison between MRA-based angiographic and CVR-based functional definitions of ROI status.

		Angiographic definition
		Normal	Abnormal	Total
Functional Definition	Normal	125 (42%)a	70 (23%)b	195 (65%)
	Abnormal	29 (10%)b	76 (25%)a	105 (35%)
	Total	154 (51%)	146 (49%)	300 (100%)

^aModerate agreement between definitions: κ = 0.33, with ^binsignificant (p = 0.15) McNemar's test for asymmetry.

No baseline CBF differences were observed between normal and abnormal regions based on MR angiographic anatomy. If the ROIs were defined functionally based on CVR, a greater and significant (p = 0.01) baseline CBF difference was observed between functionally normal and abnormal regions (Supplemental Table 2). Baseline R2′ still reflects functional tissue status (abnormal vs. normal: 3.3 ± 0.7 vs. 2.9 ± 0.6 s⁻¹, p = 0.02).

Concurrent R2′, R2* and R2 Measurements

Comparing angiographically abnormal with normal ROIs, R2′ was 13 ± 31% (p < 0.001) higher at baseline and 14 ± 33% (p < 0.001) higher after acetazolamide (Table 1), but neither R2 nor R2* differentiated between angiographically abnormal and normal tissues, either at baseline or post-acetazolamide. All three relaxation rates experienced net reduction in all regions, though the changes were only significant for ΔR2* and ΔR2 in angiographically normal regions (ΔR2*: −0.3 ± 0.6 s⁻¹ or −1.3 ± 3.1%, p = 0.02; ΔR2: −0.2 ± 0.4 s⁻¹ or −1.0 ± 2.6%, p = 0.049). Again, all observed trends were negative, suggesting improved oxygenation after acetazolamide. Full results are tabulated in Supplemental Table 3. Finally, correlation analysis of each of R2′, R2*, and R2 versus CBF revealed that R2′ had the strongest correlation and the highest statistical significance (R2′: Kendall's τ = −0.28 and p < 0.001; R2*: τ = −0.17 and p = 0.02; R2: τ = 0.02 and p = 0.79). When data are broken down by three groupings (angiographic status of ROI, presence of acetazolamide, and ROI position), the negative correlation between R2′ and CBF is more evident in the absence of data noise introduced by these factors. This is visualized in Supplemental Figure S1.

Discussion

Perfusion and CVR versus oxygenation mapping for Moyamoya disease

We demonstrated that baseline CBF and R2′ measurements yield different information in pre-operative Moyamoya patients. Specifically, there was no difference in baseline CBF between angiographically normal and abnormal regions. This finding was in agreement with some studies^3,24 but not others.^25,26 Disagreement may have arisen due to differences in imaging modalities and sequences, ROI definition, and possibly the different severity of disease in different patient cohorts. Acetazolamide injection allowed for the assessment of CVR, which did differ between angiographically normal and abnormal regions overall, though some angiographically abnormal regions had good CVR and vice versa. In contrast, R2′ could distinguish between normal and abnormal regions without an acetazolamide challenge, with a 13% increase in R2′ in abnormal regions, reflecting poorer baseline oxygenation. Moreover, reduced oxygenation (higher R2′) correlated with poor CVR (lower ΔCBF) (Figure 5). Although correlation does not indicate equivalence, our data demonstrates the potential for oxygenation mapping to supplement or even replace the acetazolamide challenge paradigm in Moyamoya disease. This could reduce the length and complexity of the MRI scan, enable scanning in sulfa allergic patients, and avoid potential side effects in sensitive individuals.

Our study focused on GESFIDE measurement of oxygenation through the relaxation rate R2′, which has been suggested by qBOLD modeling to be most sensitive to oxygenation changes.¹² A biophysical model¹² indicates that tissue R2′ is influenced by a combination of deoxygenated blood volume (DBV) and tissue oxygenation, which have competing effects on R2′ during an acetazolamide challenge. As a result of increasing CBF, increasing blood volume will increase R2′, while increasing tissue oxygenation (because of increased blood supply without a commensurate increase in tissue utilization¹⁵) will reduce R2′. Our observation, that there is no significant acetazolamide-induced ΔR2′ in all regions, is consistent with the presence of these competing factors. However, all measurements, independent of ROI definition, demonstrated trends toward improved oxygenation following acetazolamide, suggesting that the oxygenation effects likely outweigh any blood volume changes.

Finally, simultaneous acquisition of oxygenation and perfusion information provide more data in areas with true perfusion deficits, allowing the progression and severity of chronic ischemia to be examined within the framework proposed by Powers.¹

Concurrent R2′, R2*, and R2 measurements

Because all three transverse relaxation rates, R2′, R2*, and R2, are considered to be sensitive to oxygenation,¹⁴ and given that all three are simultaneously measured with GESFIDE, we examined which of these would be most appropriate for oxygenation measurements. While we observed pre-acetazolamide R2′ trends consistent with physiology, pre-acetazolamide R2* and R2 did not distinguish between normal and abnormal regions at all. Furthermore, modeling R2 is particularly complicated. Aside from the same competing factors^27,28 influencing blood R2, and the small blood volume limiting SNR in measurements of even large blood R2 changes, baseline tissue R2 can also be affected by structural changes associated with pathology. Thus, even measurable differences in tissue R2 in pathologies cannot be attributed to physiology. R2*, being the sum of R2′ and R2, suffers from all the problems outlined so far, especially blood volume effects.²⁹ Finally, our statistical analysis also showed that R2′ is the only relaxation rate that was significantly sensitive to perfusion. Indeed, as shown by correlation analysis and as illustrated in Supplemental Figure 1, R2′ and CBF displayed consistently negative correlation, in spite of variations in the slopes and offsets of regression lines caused by other factors. Conventional wisdom states that hypoperfusion drives tissue oxygenation reduction, especially in a chronic cerebrovascular disease. Therefore, due to the insensitivity of R2* and R2 to angiographic abnormality and perfusion, we can conclude that R2′ should be the preferred transverse relaxation rate for measuring oxygenation in cerebrovascular disease.

Implications of mismatches between tissue status definitions

The large number of mismatches between angiographic and functional definitions of tissue status (Table 2) indicates the importance of CVR mapping in addition to angiographic assessment. The greater number of regions found to be angiographically abnormal while functionally normal suggests that regions with occluded or stenotic primary vasculature could retain some degree of augmentation, likely due to adequate collateral formation.

Comparison to literature

Single-delay ASL has been used to measure CBF and CVR in Moyamoya patients in previous studies.^30,31 Our results suggest that multi-delay ASL can also be used in this setting. Given its ability to measure and correct for arterial arrival times, multi-delay ASL may have superior performance in patients with cerebrovascular disease.³² In agreement with literature, we also found that CVR measured via the acetazolamide challenge revealed important information about tissue status and potential presence of collaterals.^10,33

Our study also supports the use of R2′ as a metric for oxygenation, as it was the only rate to significantly differentiate between normal and abnormal regions. Baseline R2′ values measured here were similar to those from studies using GESFIDE in normal subjects.^13,34,35 There is very little data regarding transverse relaxation changes in cerebrovascular disease patients, and most of it is in acute stroke. Using separate gradient and fast spin echo measurements, Geisler et al.³⁶ found R2′ increases of around 20% in diffusion-positive stroke lesions at 1.5T in patients scanned within the first 6 h of symptoms. Siemonsen et al.³⁷ found similar changes; while they presented their data only in graphical form, it appears that R2′ of the infarcted regions was approximately 20%–30% higher than normal white matter. Our study presents the first R2′ measurements under conditions of chronic ischemia. Given the controversy surrounding EC-IC bypass in these patients^38,39 having an oxygenation metric may be helpful for patient selection.

Limitations

Our study suffered several limitations. First, the ROIs only roughly approximated vascular territories. Given the presence of chronic collateralization, they probably do not correspond directly with the feeding arteries and may have spanned territories of both normal and abnormal cerebral arteries. In addition, the imaging volume determined by the GESFIDE sequence had limited coverage of posterior circulation. The imprecise definition of ROIs may have contributed to disagreement between the two definitions of tissue status. Second, the perfusion-based functional definition of tissue status used thresholds used in prior literature²³ without consideration of the range of CVR due to normal physiological variation.^40,41 Thus, some ROIs may have been miscategorized. Third, the GESFIDE imaging sequence and its current implementation suffered from low SNR,²¹ which made any individual analysis of results very difficult; for this reason, only group level comparisons are reported. Our study is a proof of concept and demonstrates the need for continued research on robust oxygenation mapping techniques. Further studies must be performed to find the best imaging sequence and parameters. Fourth, in regions with extremely slow blood flow, even multi-delay ASL can still suffer from arterial transit artifacts, where labeled spins remain in the arteries even at the longest PLD. BOLD imaging may be used instead, though it would not produce quantitative measurement of ΔCBF. Fifth, the confounding effect of DBV on R2′ could not be explored using the protocol in this study. There is no imaging method that can isolate and measure DBV. Measuring cerebral blood volume (CBV) as a surrogate for DBV with a contrast agent at baseline would interfere with subsequent GESFIDE and ASL sequences and could not be performed in a challenge paradigm. A noncontrast method such as Vascular-Space-Occupancy (VASO)⁴² could be considered, though it would only allow for relative measurements in the absence of contrast. Lastly, we did not measure individual patients' hematocrit, which would be expected to affect R2′.

Conclusion and future work

In summary, this study has demonstrated significant results indicating that oxygenation measurements may be able to provide similar diagnostic information as CVR measurement obtained through a perfusion challenge using acetazolamide. We report the first estimates of R2′ changes in chronic ischemia and demonstrate that R2′ is the superior relaxation rate for oxygenation mapping compared to R2* and R2. Future studies should concentrate on the relationship of CVR and R2′ changes with patient outcome measures and upon improved sequences for R2′ mapping.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: NIH 1R01-NS66506, P41-EB015891

Declaration of conflicting interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: WWN, TC, JR, ZZ, and MEM—no conflict of interest. GZ—related: grant from National Institutes of Health (NIH); unrelated: consultancy for GE Healthcare, and grants/grants pending from NIH.

Authors' contributions

WWN, TC, JR, ZZ, MEM, and GZ significantly contributed to study design, protocol development, methodology development, data analysis, interpretation, and writing and revision of the manuscript. WWN, TC, and ZZ implemented the MRI sequences. WWN and GZ performed participant enrolment and data collection. WWN and JR designed and performed all statistical analysis, interpreted the results, and wrote sections of the manuscript related to statistical analysis.

Supplementary material

Supplementary material for this paper can be found at http://jcbfm.sagepub.com/content/by/supplemental-data