Arterial Spin Labeled Perfusion Imaging in Acute Ischemic Stroke

Introduction

Arterial spin labeling (ASL) is an MRI method that enables the measurement of tissue perfusion without the use of exogenous contrast agents, by magnetically tagging the water in inflowing blood.¹ It was first proposed in the early 1990’s, and since then, has been primarily a research technique. Recent advances in pulse sequence design and the more widespread availability of higher field (i.e., 3 T) MRI scanners, both of which improve the signal-to-noise ratio (SNR) of the technique dramatically, has resulted in increased adoption in the clinical neuroimaging community. All three major MRI vendors now support some form of ASL imaging as a product.

At first glance, it would seem that ASL is a particularly ill-suited technique to image acute ischemic stroke. The ASL signal is proportional roughly to cerebral blood flow (CBF), which is markedly decreased in the core of large vessel ischemic stroke. Furthermore, what CBF does remain is often supplied by circuitous collateral routes, leading to longer arterial arrival times. Since the magnetic label decays with the blood T1 (typically between 1.2 – 1.8 sec at clinical magnetic field strengths), the ASL signal will not accurately represent CBF in these regions under normal operating conditions. Finally, there is extensive research and clinical experience with another MRI perfusion method, bolus dynamic susceptibility contrast (DSC), which can be acquired relatively rapidly in the acute setting. Remarkably, despite these limitations, an increasing number of studies are showing that there is a place for ASL in the workup of acute ischemic stroke with MRI.

In this review, I will highlight recommended imaging parameters for ASL stroke studies, discuss typical image quality and artifacts, review the recent clinical literature of ASL in stroke (particularly vis a vis the more routine and accepted method of DSC), mention limitations, and finally discuss the future of this promising non-contrast stroke perfusion technique.

ASL Technique: Background, Implementations, and Recommendations for Acute Stroke

ASL sequences differ from one another based on alterations of the tagging, in-flow, and the readout segments of the sequence. The goal of the tagging segment is to label the inflowing blood water; ideally, the goal is to label as much blood as possible as near to the brain as possible. A recent consensus white paper (Alsop et al., submitted manuscript, 2013) suggests the use of pseudocontinuous labeling, with a labeling time (TL) of 1500–1800 ms; longer TL (between 3000–4000 ms) is more efficient from a signal-to-noise ratio (SNR) per unit time perspective, though as of yet, there is little clinical experience with such sequences.

The requirements for the inflow segment are to allow time (called the post-label delay [PLD]) for the labeled water to traverse the proximal vasculature, reach the capillary bed, and exchange with tissue water. This same white paper suggests a PLD of 2000 ms as a default for adult clinical patients (Alsop et al., submitted manuscript, 2013). While longer PLD would be advantageous in acute stroke to account properly for late arriving or collateral flow, this incurs an SNR penalty as the magnetic label is decaying during this period. In our experience, a TL of 1500 ms and a PLD of 2000 ms is adequate for clinical triage of the vast majority of acute ischemic stroke patients, and has the advantage that one does not need to alter the protocol for acute stroke patients specifically. Furthermore, this TL-PLD combination can highlight late arriving collateral flow, since the label will still be present in the feeding arteries, an effect known as “arterial transit artifact”² (ATA) (Figure 1). However, if one knows that a patient is presenting for an acute stroke workup and is more interested in quantitative CBF rather than the collateral visualization, it might make sense to alter the routine protocol and perform the ASL study with a long label/long delay paradigm (e.g., TL/PLD 3000/3000 ms). Another promising approach is the use of multi-delay ASL, in which a fixed TL is coupled with images collected at multiple PLD’s,^3–5 allowing for dynamic assessment of inflow; with proper kinetic modeling, it may be possible to obtain a more accurate CBF assessment, as well as make an independent measurement of arrival time, which may itself be a valuable measurement.

Figure 1.

Images of acute stroke patients with (top) good and (bottom) poor collaterals, acquired at 3 T, following co-registration and normalization. Note the serpiginous high signal on the ASL maps (arrows) in the affected region in the patient with good collaterals, which is absent in the patient with poor collaterals. DSC-based hemodynamic maps demonstrate near normal CBF in the affected regions with good collaterals, despite the presence of Tmax delays suggesting critically hypoperfused tissue.

The goal of the imaging segment is to acquire a high SNR, high-resolution image of the tissue with minimal distortions and artifacts. Historically, 2-dimensional (2D) echo-planar imaging (EPI) single-shot imaging was used for this purpose. However, while multi-shot EPI approaches can mitigate some of the distortion problems at the expense of SNR, increased SNR has been demonstrated using 3D methods, particularly 3D fast spin-echo imaging using a stack-of-spirals approach,⁶ which is recommended by the previously described white paper. This also has the advantage of significantly reduced distortions, enabling imaging of challenging regions such as the inferior frontal and temporal lobes as well as the posterior fossa.

Typical Imaging Findings and Artifacts in Acute Stroke

As expected, patients with acute ischemic stroke typically present with a perfusion deficit (i.e., low to absent ASL CBF signal) in the affected region (Figure 1). Depending on the precise parameters for TL and PLD, arterial transit artifact may be visible at the periphery of the stroke, representing labeled blood that has not yet reached the capillary bed at the time of imaging. Reperfusion of acute stroke lesions is also common, either spontaneously or following successful intravenous or endovascular treatment. In these cases, increased CBF is often identified within the affected tissue (Figure 2). This phenomenon is known as “luxury perfusion.” In most cases, it is a good prognostic factor in patients undergoing stroke MRI, correlating with lack of subsequent infarct growth.⁷ It is occasionally difficult to distinguish between high ASL signal related to delayed arrival time and that related to parenchymal hyperperfusion. One helpful way to tell is if you identify low ASL signal in the territory distal to the brighter signal, it is probably ATA; if not, it is probably hyperperfusion.⁸

Figure 2.

61 year-old man with new left-sided weakness, imaged at 1.5 T. There is increased conspicuity of hyperperfusion with ASL compared with DSC imaging. Subtle increased relative CBF and reduced Tmax are noted in the same region on the DSC images. Multiple studies have shown that the presence of hyperperfusion following acute stroke is a good prognostic feature, with smaller infarct growth and better clinical outcomes.

Another artifact to be aware of is overall slow flow to the brain, which is sometimes seen in older patients or those with poor cardiac output. In this case, using standard ASL parameters, ASL signal is only seen in the large arterial structures. This is an extreme case of ATA, and makes interpretation of focal CBF deficits impossible. This has been called the “ASL borderzone” sign and is more prevalent at 1.5 T.⁹ In these cases, if there is time to repeat the study, often ASL acquisition with longer labeling and longer PLD can mitigate this problem and allow better visualization of the focal perfusion deficit.

Clinical ASL Stroke Studies

The first study of ASL in stroke was performed by Siewert et al.,¹⁰ demonstrating that ASL was capable of detecting perfusion alterations in stroke patients, and that these were largely concordant with gadolinium-based methods. In 2000, Chalela et al. performed the first significant clinical study on the use of ASL in acute stroke.¹¹ They demonstrated the feasibility of using ASL in this patient population, found the expected CBF decreases in the affected regions, and showed that CBF deficits correlated with the NIHSS. They also identified the frequent presence of ATA, and noted that it was related to better clinical outcomes, presaging future approaches that use ASL to map collateral blood flow.

During the ensuing decade, fewer studies were performed, mainly due to the challenges of implementing the available ASL sequences on clinical scanners. However, there has been a resurgence in the use of ASL generally, and in stroke specifically, in the current decade. This has been driven in part by better ASL sequence methods, as discussed above, some of which became widely available as vendor “works-in-progress,” and by the increased availability of 3 T imaging. In 2012, three large studies^12–14 were published comparing ASL perfusion to DSC, a gadolinium-based technique that was the standard for measuring cerebral perfusion. These studies had largely concordant findings, suggesting that ASL could be used in place of DSC without any change in interpretation or subsequent clinical management. To understand them, it is important to understand the landscape of acute stroke MRI at the time.

The dominant paradigm in stroke MRI is the diffusion-perfusion mismatch with a similar mismatch approach also used with CT perfusion.^{15, 16} While a somewhat simplistic framework,¹⁷ the basic idea is that regions with reduced diffusion, typically below a certain critical threshold, represent irreversibly infarcted tissue. Regions with critically low perfusion but normal diffusion represent tissue that is functionally compromised but not yet beyond the point of no return, provided that the tissue can be reperfused in an efficacious manner. This diffusion-negative, perfusion-positive tissue has been termed the ischemic penumbra, and is thought to be the target of any acute interventions. While measuring the diffusion lesion is relatively straightforward, the problem is really the perfusion images that are acquired during rapid passage of the gadolinium bolus. To produce meaningful information, they require postprocessing in which various assumptions are made and which varies widely between different sites and scanners. Several large clinical trials have suggested that blood arrival delay (measured as the time-to-peak [TTP] of the gadolinium concentration or normalized time to the maximum of the residue function [Tmax]) can provide a reliable measure of critically hypoperfused brain tissue.^18–20 Therefore, these three ASL stroke studies sought to compare the ASL findings to the DSC findings with regard to diffusion-perfusion mismatch stratification and to address the question of whether ASL could be substituted for DSC. This might be considered desirable, since the lack of contrast would enable the use of the mismatch paradigm in patients who cannot receive gadolinium contrast. Also, since the post-processing of ASL data is relatively simple compared with DSC, this might improve reliability and transferability between different sites and scanners.

Zaharchuk et al.¹² showed that clinical ASL was feasible at 1.5T using a pseudocontinuous labeling method with optimized readout. DSC was considered the gold-standard perfusion method for classifying patients into clinically relevant subgroups (i.e., those with a mismatch, no mismatch, or reperfusion). They found good inter-reader reliability, but found that exact agreement was only seen in about 60% of cases; this was higher (74%) for distinguishing the key categories of mismatch versus either matched or reperfused regions. In cases in which there were discrepancies, ASL tended to overestimate the perfusion deficit, likely due to the longer arrival times of the blood in affected regions.

Bokkers et al.¹³ performed a larger study of ASL in consecutive acute stroke patients at 3 T evaluating the conspicuity of perfusion lesions using both ASL and DSC. One interesting observation they made was that about 20% of patients in their series could not receive gadolinium due to reduced renal function, highlighting a key advantage of noncontrast ASL perfusion. In patients who could also receive DSC studies, they found that when a perfusion deficit was present on DSC, it was also seen on 82% of ASL images. The cases in which ASL was deemed normal tended to be the smaller lesions, such as small cortical and subcortical strokes. They also found that interpretable perfusion studies were achieved at similar rates with both methods, and that image quality was similar. They found agreement between DSC and ASL for significant mismatch was 88%; this higher agreement may have been due to the use of higher field strength. Overall, their study highlighted that particularly for the large territory strokes that are the greatest source of morbidity, ASL is a very reasonable alternative to DSC, and furthermore, it enabled evaluation of diffusion-perfusion mismatch in patients with contraindications to gadolinium.

Finally, Wang et al.¹⁴ demonstrated similar findings, focusing on consistency of ASL and DSC results as well as image quality. They found good concordance with multiple DSC biomarkers, including Tmax, mean transit time (MTT), and relative CBF. They also highlighted that ASL had particular value in visualizing hyperperfusion compared with DSC, a finding confirmed in a subsequent study by Bivard et al.,⁷ who also showed that ASL hyperperfusion at 24 hrs was associated with better clinical outcomes in acute stroke.

Since these studies, other studies have looked at whether more quantitative approaches can be applied to ASL. This is important, since quantitative image analysis may improve mismatch estimations compared with visual inspection. Also, ASL is inherently a quantitative technique, such as that absolute measures of CBF may provide more information than relative measures. Bivard et al.²¹ showed that the best relative threshold for ASL CBF maps to identify critical hypoperfusion was about 40% of mean contralateral CBF. Using this threshold, they showed high correlation between mismatch lesion volume defined by either DSC or ASL, though again they confirmed that ASL tended to overestimate the size of perfusion lesions due to late arriving flow. A similar study by Niibo et al.,²² using a quantitative CBF threshold of 20 ml/100 g/min, found that the ASL mismatch had 100% agreement with a mismatch based on DSC mean transit time >= 10 sec.

While ischemic stroke is primarily a disease of older adults, it does occur in children; because of its rarity, making the correct diagnosis can be challenging. ASL is a promising technique in this patient population, given their overall higher baseline CBF and shorter arterial arrival times. Chen et al.²³ describe their initial experience evaluating pediatric ischemic stroke with ASL, demonstrating good concordance of the location and character of the perfusion deficits with clinical and diffusion-weighted imaging (DWI) findings. Particularly interesting is the possibility that ASL might be sensitive to pre-clinical ischemic episodes in children with mitochondrial disease. Ikawa et al.²⁴ showed abnormally high ASL signal in a patient with myopathy, encephalopathy, lactic acidosis, and stroke-like episodes (MELAS) syndrome several months before the clinical appearance of a diffusion-positive stroke-like episode, suggesting it could be used for management. More recently, small studies have also been performed in peri-natal stroke,^{25, 26} a challenging group of patients, due to their small size, motion, and longer arterial arrival times. These studies have shown that ASL is feasible and can visual hypo- and hyperperfusion lesions. Peri-natal ASL avoids the difficulties of DSC in this patient group, but comes with its own set of challenges, given many factors that affect ASL CBF measurements, including effects of arrival time, hematocrit, and labeling efficiency.

Limitations of ASL in Acute Stroke

Despite the enthusiasm of the research community regarding ASL in acute stroke, several important limitations must be mentioned. Though it has many attractive properties, ASL is a lower SNR technique than DSC. There are several distinct “modes of failure” unique to ASL. One of these is an artifact that presents as overall reduced CBF to a vascular territory that is not due to true reduced flow, but rather to reduced labeling efficiency of the water in a specific artery at the labeling plane in the neck;²⁷ this artifact is specific to continuous and pseudocontinuous ASL, and while it occurs rarely, it can create confusion in acute stroke given that territorial infarcts might be expected to present with identical findings. Finally, motion artifacts affect ASL as they do all MRI sequences. More recent ASL sequences are background suppressed, and this affords a certain degree of insensitivity to motion. Several methods exist to address motion artifacts in ASL. One is to simply remove the raw images that have motion before signal averaging;²⁸ this is feasible, but ultimately reduces the SNR of the final perfusion images. Another approach is the use of prospective motion correction,²⁹ in which navigator images track the head in real-time, and are used to follow the head motion during image collection.

Future of Acute Stroke ASL Imaging

What will be the role of ASL in acute stroke in the next several years? Given the flexibility of the technique, it is hard to predict, but some areas of active research at the moment are the use of ASL to assess collateral blood flow and obtain truly quantitative CBF in stroke patients.

Several recent studies focusing on the mismatch hypothesis have suggested that different imaging profiles can be identified that predict lesion growth and patient outcome following early reperfusion in acute stroke.^18–20 However, the recent failure of several high-profile clinical stroke trials makes it clear that better patient selection for endovascular therapy is critical.^30–32 An increasing number of studies suggest that collateral flow may play a central role in sustaining tissue viability and reducing complications.^33–36 ASL has features that make it an ideal MRI-based modality to assess collaterals: it is non-invasive and combines features of perfusion and angiography, based on its sensitivity to delayed arterial arrival. Recently, several groups have evaluated the predictive ability of ASL to identify the regional pattern of collaterals.^37–39 Chng et al. and Wu et al. demonstrated good agreement of pulsed ASL sequences in patients with arterial steno-occlusive disease,^{37, 39} while Zaharchuk et al.³⁸ showed similar findings with pseudocontinuous ASL in Moyamoya disease patients. Figure 1 shows an example of the difference between good and poor collaterals on ASL imaging in two patients with acute ischemic stroke.

Ultimately, the most importance factor for tissue survival is probably not the presence of collaterals per se, but rather how effective they are at delivering CBF to the ischemic territory. While ASL can measure quantitative CBF relatively easily in normal subjects and those with mild arterial arrival delays, quantitation is very challenging in stroke patients. Several approaches show promise in this regard. The first is multi-delay ASL, which acquires images at multiple PLD’s, in order to disentangle the effects of CBF and arrival time; unfortunately, this comes at the price of reduced image quality, though newer methods, such as Hadamard encoding,⁴⁰ reduce the associated penalties. Another is the use of velocity-selective ASL (VS-ASL), a method that performs the labeling step in the imaged voxel itself, thereby reducing transit time effects.^{41, 42} VS-ASL has yet to be applied to an acute stroke study, and again has limitations in terms of SNR.

Conclusions

Despite the inherent challenges of ASL in cerebrovascular disease, it is beginning to enter the mainstream of clinical MRI stroke studies, showing many similarities to DSC perfusion imaging. Perhaps the most interesting cases are those in which differences between the techniques are seen, either due to collateral flow or reperfusion, that suggest that ASL may have additive value in stroke patients. Further sequence development is required and is ongoing to increase imaging speed and image quality.