Collateral blood flow plays a critical role in supporting the cerebral circulation in the setting of acute and chronic cerebral ischemia.3,23,24
In response to a decrease in local perfusion pressure, there is recruitment of flow from both circle of Willis and leptomeningeal anastomoses, which compensate for the lack of anterograde flow.3
Patients with the same vascular occlusion may have significantly different outcomes based on their ability to recruit collateral pathways to restore flow to the ischemic region during the minutes and hours after an acute event.25,26
The eventual failure of collateral pathways is thought to lead to infarct growth and is ultimately responsible for the decreasing efficacy of stroke therapy with time.27
Timely knowledge about the status of collaterals affects the decision-making process regarding acute therapy in individual patients with ischemic stroke.23,27
For example, Kucinski et al25
demonstrated that the best predictor of favorable outcome in the retrospective series of patients undergoing intra-arterial thrombolysis was the presence of good collateral flow as judged by the initial DSA study. For all these reasons, having the ability to assess the location and amount of collateral perfusion using a noninvasive test such as MRI would be desirable.
Given its sensitivity to arrival time and its ability to quantify CBF, ASL combines features of both angiography and perfusion. Previous studies have shown good correlation with gold standard CBF imaging of gray matter in healthy subjects,28
but it is likely that it underestimates CBF in regions with delayed arterial arrival times. This is because the label decays with the blood T1, which is on the same order as arterial arrival delays in patients with Moyamoya disease.29
However, this drawback for quantitation may be turned to advantage for visualizing collaterals. With ASL, late-arriving flow appears as serpiginous high ASL signal within cortical vessels, which has been termed ATA.11,30
ATA was seen frequently in a small group of acute ischemic stroke patients and was associated with tissue survival and improved clinical outcome.31
Also, patients with chronic hypoperfusion and ATA had poor cerebrovascular reserve in response to acetazolamide.11
ATA is dependent on several sequence parameters, particularly the labeling time and the PLD. Only a single PLD was used in this study; ASL sequences with a range of PLD times exist and can be helpful for quantifying CBF32–34
but require longer imaging times for equivalent signal-to-noise ratio. There is some evidence that choosing a single PLD in a range that is highly sensitive to delay can increase the sensitivity for identifying pathology.12,35
This study shows that an ASL sequence with a single moderately long PLD can identify regions with collateral flow and can differentiate between poor and robust collateral flow, as determined using a DSA-based collateral grading scale. In particular, the agreement between consensus ASL and DSA scores for the distinction of normal perfusion versus collateral flow was quite good. Also, there was no evidence of a systematic bias in the ASL scoring compared with the consensus DSA. These findings are consistent with a previous study36
that used a high-field (3-T) multiple PLD ASL method that also used perfusion territory imaging, an ASL method that can separate flow contributions from different cerebral arteries. These investigators examined a population of patients with a variety of cerebrovascular disease, some of whom had Moyamoya disease, and found a κ
value of 0.72 for distinguishing collateral from antegrade perfusion, similar to that in the current study (κ
=0.65). In addition to validating these general results, the current study also examines how well different readers agree on both the DSA and ASL grading scales, showing that agreement is higher with the ASL method. Furthermore, we suggest that high-field imaging, perfusion territory imaging, and multiple PLD ASL imaging are not required to identify collateral flow with a similar degree of accuracy. This is important because these modifications to the ASL experiment required additional imaging time (10 minutes for the ASL and 4 minutes for the 3-dimensional time-of-flight angiogram that is required to plan the ASL sequence). The current study demonstrates good performance using a faster ASL imaging protocol (6 minutes), making it more feasible in the clinical setting.
We also undertook understanding the relationship between collateral scores and CBF. Previous reports have suggested that identification of collaterals on DSA was not a good predictor of the adequacy of cerebral perfusion based on oxygen extraction fraction measurements using positron emission tomography.37,38
Our findings show that gold standard Xe CT-based CBF increases as collateral score increases for both DSA-based and ASL-based consensus collateral scores. This effect was more pronounced with the ASL-based scores than the DSA-based scores (Supplemental Table II
). Interestingly, the highest correlation was for Xe CT CBF based on ASL collateral score. ASL CBF also increased with increasing collateral score but is unreliable in the presence of known long arterial arrival delays. This likely explains why the ASL CBF measurements are highest in regions rated to have robust collaterals (score = 2), because this flow is destined for more distal slices. Bolus perfusion-weighted imaging, using either CT or MRI, also can be used to estimate CBF, cerebral blood volume, mean transit time, and normalized bolus arrival time and could be used for this application. However, it is challenging to acquire quantitative CBF measurements using such nondiffusible intravascular tracers, particularly in patients with Moyamoya disease.29,39
may mitigate problems associated with pure delays, but not with dispersion.41
We performed perfusion-weighted imaging measurements in a subset of these patients, but we do not present these data because the acquisition parameters were not standardized, and the imaging planes did not completely cover the regions of interest interrogated in the remainder of the study. To our knowledge, few methods using bolus perfusion-weighted imaging to distinguish normal from collateral flow have been reported,42
and we hope to explore this more thoroughly in a prospectively recruited study. This ASL method may be particularly amenable for use in patients who cannot receive gadolinium-containing contrast agents or who may need multiple studies (either in a single or consecutive sessions) to assess either cerebrovascular reserve or effects of treatment.
The limitations of this study include the difficulties of applying the DSA collateral grading system,7
initially developed for acute ischemic stroke as part of the PROACT 2 study, to a chronic cerebrovascular disease such as Moyamoya disease. Also, perfusion territory imaging using vessel-selective ASL was not used,36,43,44
which may further help identify regions fed via collateral pathways. Such a method would allow one to distinguish slow antegrade flow attributable to a high-grade stenosis from leptomeningeal or other retrograde-type collateral flow, which was not possible with the current study. Historically, such methods have required additional scan time, but newer techniques have been described that achieve vessel selectivity without the scan time penalties.45
It is probable that such methods would prove even more accurate. Also, we do not report quantitative CBF using the ASL method. Given the long arterial arrival delays in Moyamoya patients,29
it is likely that some degree of CBF underestimation occurs in the regions perfused by collaterals. Finally, we recognize that the use of a consensus reading from 2 separate readers for each modality allows only a limited evaluation of the variability related to multiple readers, which is reported in the .