Periprocedural Arterial Spin Labeling and Dynamic Susceptibility Contrast Perfusion in Detection of Cerebral Blood Flow in Patients With Acute Ischemic Syndrome
Background and Purpose—To compare the diagnostic performance of arterial spin-labeling (ASL) and dynamic susceptibility contrast (DSC) perfusion in detecting cerebral blood flow (CBF) changes before and after endovascular recanalization in acute ischemic syndrome.
Methods—The inclusion criteria for this retrospective study were patients with acute ischemic syndrome who underwent endovascular recanalization and acquisition of both ASL and DSC before and after revascularization. ASL-CBF and multiparametric DSC maps were evaluated for image quality, location, and type of perfusion abnormality. Relative CBF (rCBF) was calculated in the infarction core and hypoperfused areas using coregistered ASL and DSC. Core and hypoperfused rCBF were used for paired pretreatment and posttreatment comparisons. Interobserver and intermodality agreement were evaluated by κ test, and t test was calculated for ASL and DSC rCBF values.
Results—Twenty-five patients met our inclusion criteria. Five studies were rated nondiagnostic, resulting in 45 pairs of DSC–ASL available for comparison. ASL and DSC agreed on type and location of the perfusion abnormality in 71% and 80% of cases, respectively. The image quality of ASL was lower than DSC, resulting in interobserver variability for the type (κ=0.45) and location (κ=0.56) of perfusion abnormality. ASL was unable to show any type of perfusion abnormality in 11% of patients. In successfully recanalized patients, hyperperfusion (rCBF >1) was detected in 100% on DSC and 47% on ASL.
Conclusions—ASL is less sensitive than DSC for detecting rCBF changes in patients with acute ischemic syndrome, particularly with respect to hyperperfusion after successful recanalization.
- acute stroke
- brain ischemia
- magnetic resonance
- magnetic resonance perfusion
Several clinical trials have suggested that patients with a mismatch between their infarction volume and the volume of hypoperfused tissue may respond favorably to revascularization therapies.1–3 The identification of a perfusion–diffusion mismatch, although still controversial, has been the driving force behind the growth of magnetic resonance (MR) perfusion imaging.1 Both dynamic susceptibility contrast (DSC) and arterial spin-labeling (ASL) have been used for evaluation of cerebral perfusion in patients with stroke,4,5 each with different strengths and limitations.6–8
Recent technical advances have significantly improved the quality and acquisition speed of ASL techniques.9,10 In addition, recent concerns about gadolinium-induced nephrogenic systemic fibrosis in patients with extremely poor renal function11 have resulted in renewed interest in ASL for the clinical evaluation of cerebral perfusion status in patients with acute ischemic syndrome (AIS). Several recent studies also have shown that ASL can detect cerebral blood flow (CBF) alterations in the setting of acute stroke, many with concordant comparative analysis with DSC.7,12,13 In this study, we selected a unique population of patients with AIS and evaluated the performance of ASL for the detection of CBF changes, before and after invasive recanalization, by performing a qualitative and quantitative comparative analysis to DSC imaging performed concurrently.
Our Health Insurance Portability and Accountability Act-compliant retrospective study was performed with the approval of The University of California at Los Angeles (UCLA) Institutional Review Boards. Electronic medical records of patients with suspected AIS who presented to a single comprehensive stroke center from September 2010 to March 2012 were reviewed. Twenty-five consecutive patients who met our inclusion criteria were included in this study. Inclusion criteria included the following:
(1) acute onset of neurological deficits meeting the standard eligibility criteria for invasive recanalization, including intra-arterial tissue plasminogen activator (<6 hours) or clot retrieval (<9 hours);14–16 for carotid dissection patients (n=4), recanalization was attempted regardless of the timing criteria
(2) periprocedural magnetic resonance imaging (MRI) perfusion imaging acquired; and (3)
absence of previous intracranial hemorrhage, brain surgery, or large territorial infarction.
The number of patients achieving recanalization, the type of recanalization using postprocedure angiography and the thrombolysis in cerebral infarction (TICI) scoring method,17 the baseline National Institutes of Health Stroke Scale (NIHSS) scores, the median time from last known well to first MRI, the median time from first MRI to groin puncture, the median time from first MRI to second posttreatment MRI, and the type of scanner used (3.0 or 1.5 T) were documented for each patient.
All patients underwent MRI on either a 1.5-T (Siemens Avanto; Erlangen, Germany) or 3.0-T (Siemens Trio; Erlangen, Germany) MR system. The imaging protocol included diffusion-weighted imaging, fluid attenuation inversion recovery imaging, gradient recalled echo, MR angiography, and DSC and ASL perfusion-weighted imaging. DSC images were acquired using a gradient-echo echoplanar imaging sequence with repetition time of 1.9 seconds (2.5 seconds for 1.5 T), echo time of 30 ms (45 ms for 1.5 T), field of view of 22 cm, and matrix size of 128×128, resulting in 26 slices with 5-mm slice thickness. A generalized autocalibrating partially parallel acquisition (GRAPPA) factor of 2 was used for parallel acquisition resulting in a 2-minute scan time. During dynamic acquisition, a single dose of 0.1 mmol/kg of gadolinium contrast agent (Magnevist [Bayer Health Care] or Multihance [Bracco Diagnostics]) was injected at a rate of 5 mL/s. ASL perfusion scans were acquired using a pseudocontinuous arterial spin labeling (PCASL) pulse sequence with background suppression using a 3-dimensional gradient-echo and spin-echo readout, a 2-second postlabeling delay, and a labeling pulse duration of 1.5 seconds, resulting in 30 pairs of tag and control images consisting of 26 total image slices, each 5 mm thick. A matrix size of 64×64, field of view of 22 cm, GRAPPA acceleration factor of 2, repetition time of 4 seconds, and echo time of 22 ms was used. A similar acquisition scheme has been used in other published quantitative PCASL investigations.9,10,12
ASL data analysis was performed in-house using the Interactive Data Language software program. ASL images were motion-corrected, pairwise-subtracted between the tag and control images, and averaged to generate a mean difference image, ΔM. Quantitative CBF (CBFASL) maps were calculated as follows9,18:(1)
where R1a=0.72 s−1 (0.61 s−1 at 1.5 T) is the longitudinal relaxation rate of blood, M0 is the equilibrium magnetization of brain tissue, α=0.8 is the tagging efficiency, τ=1.5 seconds is the duration of the labeling pulse, w=2 seconds is the postlabeling delay time, and λ=0.9 g/mL is blood/tissue water partition coefficient. Note that this equation assumes that the labeled blood spins remain primarily in the vasculature rather than exchanging freely with tissue water, an assumption that is justified in patients with stroke with prolonged arterial transit times.19
DSC images were processed using a commercially available Food and Drug Administration–approved software (PerfScape/NeuroScape; Olea Medical SAS). DSC analysis consisted of the following steps: (1) truncation of the first 5 time points in the DSC time series because the MR signal does not reach steady-state before this time; (2) calculation of prebolus signal intensity on a voxel-wise basis; and (3) conversion of truncated DSC time series to a concentration-time curve based on the T2* relaxivity of the contrast agent. The arterial input function was selected automatically by the perfusion software, and the relative CBF (rCBF) was calculated using a block-circulant singular-value decomposition technique.20 Parametric maps of CBF, time to peak, and Tmax were exported from the software for subsequent analysis.
Quantitative Image Evaluation
Diffusion-weighted imaging, DSC, and ASL images for each patient were registered to a high-resolution (1.0-mm isotropic), T1-weighted brain atlas (MNI152; Montreal Neurological Institute) using a 12-degree-of-freedom transformation with a mutual information cost function. This was followed by visual inspection to ensure adequate alignment. To maintain consistency, we used quantitative values described in the literature to define the infarction core, defined as apparent diffusion coefficient value <550×10−6 mm2/s.21,22 Likewise, regions of hypoperfusion were defined as Tmax >4 seconds.23 These regions also were aligned with ASL and DSC images in Montreal Neurological Institute (MNI) space through the same transformation matrices used for the diffusion-weighted imaging and ASL images. Subsequently, regions of interest were placed over the infarction core and hypoperfused area to extract the corresponding ASL and DSC-CBF values. Regions of interest then were mirrored to the contralateral hemisphere and the ratio of rCBF within the infarction core and hypoperfusion regions with respect to contralateral normal-appearing tissue were calculated using both ASL and DSC measurement techniques. Manual restriction of the regions of interests was applied when necessary. The same process was repeated on the posttreatment scans using the fixed regions of interests in MNI space to ensure evaluation of rCBF values in precisely the same regions in both pretreatment and posttreatment scans.
Qualitative Image Evaluation
The ASL-CBF maps and multiparametric perfusion maps from DSC studies were evaluated on a commercially available image viewer. The observers could adjust image contrast, select a viewing color scheme, and select size of the images. Two experienced neuroradiologists blinded to treatment and clinical information reviewed the ASL and DSC perfusion maps independently and in separate reading sessions. A 3-scale imaging score was used to evaluate the image quality with regard to susceptibility-mediated distortion at tissue interfaces, noise, motion, and delineation of major structures such as the ventricles, thalami, basal ganglia, brain stem, and posterior fossa, with scoring as follows: (1) poor image quality, not interpretable; (2) fair diagnostic image quality, some distortion and noise, limits detail delineation of major structures; and (3) good image quality, no to minimal distortion with detailed delineation of all structures. In addition to image quality, the observers were asked to note the presence of hypoperfusion and hyperperfusion, or both, when present. Perfusion abnormality was identified from a collective analysis of the time to peak, Tmax, and CBF maps. A perfusion deficit was defined as an area with visually perceptible increased time to peak and Tmax and decreased CBF on DSC maps, and with decreased perfusion signal on ASL when compared with the surrounding brain tissue and with the homologous contralateral hemisphere. Any hyperperfused region was defined as an area of visually perceptible decreased time to peak and Tmax and increased CBF on DSC with increased perfusion signal on ASL. The observers were asked to localize the site of the perfusion abnormality into 2 possible categories: (1) basal ganglia involved and (2) more distal middle cerebral artery (MCA) (cortical and white matter without basal ganglia involvement). These data were used to determine interobserver agreement. The observers also noted the regions of delayed arterial arrival seen as surrounding serpiginous high signal on ASL images when present. Finally, and in a different reading session, any discrepancy in the qualitative scores of perfusion deficit with regard to the type and location of perfusion abnormality between the readers was resolved by consensus agreement. These scores were used to perform comparative analysis between DSC and ASL.
Statistical analysis was performed using MedCalc (version 12.2.1; MedCalc Software). The Wilcoxon signed-rank test was used to compare the mean ratings of ASL and DSC perfusion maps. The weighted kappa test was used to evaluate the interobserver and intermodality agreements, t test was calculated for the quantitative values of rCBF on both the ASL and DSC perfusion maps, and significance level was defined as P<0.05 (2-sided).
A total of 25 patients (15 men) with a mean age of 63.5 years (range, 19–82) met our inclusion criteria. All patients had evidence of large arterial steno-occlusive disease, including carotid-T occlusion (n=6), M1 occlusion (n=12), M2 occlusion (n=3), and occlusive carotid dissection (n=4). The therapeutic procedures for these patients included mechanical clot retrieval (n=16), intraarterial tissue plasminogen activator (n=5), and stent placement for carotid dissection (n=4). A total of 17 patients (68%) achieved adequate recanalization as defined by a TICI score ≥2a and as confirmed by postprocedural angiography, including TICI 3 (n=6; 24%), TICI 2b (n=8; 32%), and TICI 2a (n=4; 16%).
NIHSS scores at baseline ranged from 3 to 27, with a median of 16. The median time from last known well to first MRI was 5 hours (range, 1–8.9 hours). Median time from first MRI to groin puncture was 60.4 minutes (range, 25–303 minutes), and the median time from first MRI to second posttreatment MRI was 7 hours (range, 2.3–11.4 hours).
Five studies (10%) were deemed nondiagnostic, including 3 ASL (6%) studies with motion and significant ASL border zone artifacts and 2 DSC (4%) examinations with susceptibility and motion artifacts, leaving a total of 45 pairs of ASL-DSC studies for both qualitative and quantitative analysis. The overall image quality scores for DSC examinations were 3 and 2 to 3 (median±range), without a significant difference between the observers (P=0.9). The image quality scores for ASL studies were 2 and 2 to 3 (median±range) for both observers, without a significant difference between the observers (P=0.4). The lower image quality of ASL has resulted in some interobserver variability (κ=0.6) in comparison with DSC (κ=0.95). Qualitative analysis with interobserver assessment for DSC and ASL examinations is detailed in Tables I and II in the online-only Data Supplement.
Using consensus scores between 2 observers, a direct intermodality comparison was performed between the ASL and DSC for image quality, type, and location of perfusion abnormality (Table 1). The image quality of ASL, although diagnostic, was rated significantly lower than DSC for both pretreatment (P=0.04) and posttreatment (P=0.005) groups. The median of image quality scores was significantly higher for DSC (P=0.001). All DSC studies in the pretreatment and posttreatment scans were able to detect a perfusion abnormality.
All recanalized patients (n=17) demonstrated evidence of luxury perfusion whether pure hyperperfusion (n=7) or a combination of both hypoperfusion and hyperperfusion (n=10). All patients demonstrating persistent hypoperfusion on posttreatment imaging were nonrecanalized, whereas 2 additional nonrecanalized patients demonstrated a combination of hypoperfusion and hyperperfusion. ASL and DSC agreed on type of perfusion abnormality present in 32 of 45 cases (71%), with overall intermodality agreement of κ=0.458 (Table 2). Among disagreements, 5 cases (11%) with the hypoperfusion and hyperperfusion pattern identified by DSC appeared as only hypoperfusion on ASL imaging. Further, ASL imaging did not show any type of perfusion abnormality in an additional 5 cases (11%) that were evaluated as abnormal by DSC (Figure 1).
In terms of location of perfusion abnormality, ASL agreed with DSC in identifying the location of the perfusion abnormality in 31 of 40 patients (68%), including 23 patients with involvement of basal ganglia and 8 without. However, ASL failed to demonstrate the presence of a confirmed perfusion abnormality in the BG in 9 patients (20%; Figure 2), resulting in an intermodality agreement of only κ=0.556. In another 5 patients in whom ASL was unable to detect a perfusion abnormality, the distribution of perfusion deficits involved the basal ganglia in 4 cases and no basal ganglia involvement in the other. The delayed arterial transient effect was noted in 10 of 17 (58%) successfully recanalized patients. Twenty-three studies were performed on a 3.0-T scanner and 22 were performed on a 1.5-T MR scanner.
There was no significant difference between the image quality of DSC studies acquired at 3 T vs 1.5 T (P=0.4). The image quality of ASL images acquired on 3 T was slightly higher than images acquired on 1.5 T, although it did not reach statistical significance (P=0.09).
The quantitative rCBF values in pretreatment and posttreatment groups using both ASL and DSC are summarized in Table 3.
Both ASL and DSC demonstrated decreased rCBF without a statistically significant difference between modalities within the infraction core (0.48±0.31 DSC vs 0.52±0.35 ASL; P=0.64) and hypoperfused area (0.76±0.47 DSC vs 0.57±0.27 ASL; P=0.09). Hyperperfusion (rCBF >1) was detected by DSC in 3 cases within the hypoperfused area and by ASL in 3 cases (1 within the infarction core and the others within the hypoperfused area). An important observation in the pretreatment group was the ability of DSC to demonstrate a significant difference in rCBF within the penumbral hypoperfused area as compared with the infarction core (P=0.0001). This difference did not reach statistical significance for ASL (P=0.54; Table 3).
After recanalization, both ASL and DSC demonstrated overall increases in rCBF (Table 3). In nonrecanalized patients, there was no significant increase in rCBF values between pretreatment and posttreatment groups in the infarction core or hypoperfused area (Table 3). Hyperperfusion (rCBF >1) was detected in only 1 nonrecanalized patient, within the infarction core, by both ASL and DSC. In recanalized patients, there was a significant increase in rCBF values between pretreatment and posttreatment scans by both ASL and DSC in the infarction core and hypoperfused area (Table 3). Similar to pretreatment studies, ASL was unable to demonstrate a significant difference in rCBF within the hypoperfused area in comparison with the infarction core (P=0.83), whereas this difference was significant on DSC (P=0.01). Although the difference in quantitative rCBF within the infarction core for DSC and ASL was not significant (P=0.29) in recanalized patients, increased rCBF measured within the hypoperfused area was significantly higher by DSC vs ASL (P=0.01). This can be explained by the fact that hypoperfused penumbral areas in all recanalized cases (n=17) demonstrated hyperperfusion (rCB >1) on posttreatment DSC compared with only 8 cases (47%) on ASL. Posttreatment hyperperfusion also was detected within the infarction core in 9 cases by DSC compared with 6 cases by ASL.
Both qualitative and quantitative changes of CBF were generally concordant using ASL and DSC. Expected changes in rCBF in recanalized patients were detected by both modalities with moderate agreement. There are, however, some important limitations for ASL centered around the areas of image quality, sensitivity for the detection of hyperperfusion, and demonstration of rCBF changes before and after intervention.
The image quality of ASL is significantly lower when compared with DSC, resulting in higher interobserver variability in terms of determination of the type and location of the perfusion abnormality. We hypothesize that the lower image quality of ASL is related to known and inherent limitations of ASL, including its lower signal-to-noise ratio and use of low-imaging matrix. For optimization of ASL signal-to-noise ratio, we incorporated many recently suggested strategies, including the use pseudocontinuous labeling pulse sequence, background suppression, and optimized image readout,24 resulting in only modest gains vs DSC. One major handicap of ASL is the current practical limitation to a matrix size of 64×64 mm required for short acquisition time, which generally results in voxel sizes 4-times larger than those of DSC.
Second major finding of our results suggests that ASL in its current state is less sensitive in the detection of hyperperfusion than DSC. Although overall ASL agreed with DSC in the detection of the type of perfusion abnormality in 71% of cases, ASL failed to show any type of perfusion abnormality in 11% of patients. In addition, in 11% who exhibited both hypoperfusion and hyperperfusion on DSC imaging, ASL showed only hypoperfusion. Quantitative analysis of the CBF in successfully recanalized patients showed significantly lower rCBF values in reperfused area in ASL and approximately half of cases demonstrated hyperperfusion (rCBF >1) on ASL in comparison with 100% on DSC. This was somewhat discrepant with the result of recent study by Wang et al12 that showed a more promising comparison between the ASL and DSC-CBF values in patients with acute stroke and, in particular, higher sensitivity for detection of hyperperfusion using ASL. Some differences between their results and ours could be attributed to several factors. In the Wang et al12 study, the described hyperperfusion pattern seems to be a predominantly heterogeneous pattern of CBF in patients with acute ischemic stroke, in which hyperperfusion is thought to be the result of spontaneous recanalization. In contrast, the main focus of our study is mechanically revascularized patients. We believe that the pathophysiology and cerebral hemodynamics in spontaneous recanalization (which frequently involves smaller distal arteries) are different from those that occur after the recanalization of large main arteries such as the internal carotid and proximal MCAs. The general kinetic model for accurate assessment of CBF using ASL is based on the assumption that there is complete exchange of labeled blood and tissue spins. However, this assumption may not hold in the latter group, in which the rapid blood flow rate and luxury perfusion occurring after recanalization of major arteries can affect the performance of ASL.25 Although the sensitivity to arterial transit delay in ASL perfusion MR can be minimized with a long postlabeling delay, the gain is compromised by the T1 relaxation of blood during this period. Although in this study a postlabeling delay of 2 seconds was used as a trade-off between maintaining sufficient image quality while allowing adequate delay to detect cerebral perfusion, this may not be the optimal delay for detection of hyperperfusion. The use of different postlabeling delay times in the pretreatment scan when the artery is occluded vs the posttreatment scan after arterial recanalization may be required to accurately evaluate the rCBF using ASL methods. An alternate approach is to obtain images using multiple postlabeling delays, thus effectively mapping the inflow of label into the tissue26 to determine actual arrival times, and then to fit calculated CBF to kinetic models.27 In the Wang et al12 study, absolute values of CBF were calculated for both ASL and DSC. We chose to use the more conventional approach of using the rCBF values because the absolute quantification of CBF on DSC can be influenced by many confounding factors which can be difficult to control.28,29 This is likely a significant source of discrepancy between their result and ours. As with any other quantitative analysis, differences in results can be introduced through the use of different postprocessing techniques and operators. The perfusion scale and algorithm used in their study are different from ours, which explains some of the discrepancies.
Finally, the third major finding of our results suggests that ASL in its current state fails to show differences in rCBF within the infarction core vs hypoperfused regions in both pretreatment and posttreatment groups. Again, this can be explained by high dependency of ASL to arterial arrival time, which results in relative insensitivity to detect differences in rCBF in regions with variable transit delay when the transient delay is greater than postlabeling delay. In other words, because the tagged blood does not reach the capillary bed by the postlabeling delay time (2 seconds) in either the penumbral tissue (defined by Tmax >4 seconds) or within the infarction core (where the Tmax can be >10 seconds), ASL is unable to detect the differences in rCBF. New techniques, such as velocity-selective ASL,30 which is theoretically insensitive to arrival time, may help to address some of these limitations.
This study has several limitations, including a relatively small sample size drawn from a single institution possibly introducing a size selection bias, the sequential rather than simultaneous performance of DSC and ASL imaging possibly introducing an artifactual performance difference between both techniques that is attributable to temporal rather than performance factors, and retrospective study design, possibly introducing unknown patient selection bias. In addition, DSC was assumed to be the comparative standard, possibly introducing a modality-specific bias. These limitations were mitigated by confirmation of ischemia in every patient enrolled by catheter angiography and follow-up imaging, standardization of the optimized image acquisition parameters between the ASL and DSC imaging protocols, making DSC and ASL scanning immediately sequential, making group rather than individual comparisons, and maximizing sample size in this difficult and critically ill patient population.
ASL performs with moderate agreement with DSC in the evaluation of CBF changes in patients with AIS before and after invasive recanalization. Qualitative and quantitative differences exist between 2 modalities regarding detection of hyperperfusion in successfully recanalized patients, with ASL being less sensitive for detection of rCBF changes in patients with AIS in particular. Knowledge of these limitations is helpful for clinicians to interpret these studies with caution and highlights the need for further research and optimization of the ASL technique before it can be incorporated as the major perfusion imaging modality in the acute setting.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.112.672956/-/DC1.
- Received August 31, 2012.
- Revision received November 12, 2012.
- Accepted November 19, 2012.
- © 2013 American Heart Association, Inc.
- Albers GW,
- Thijs VN,
- Wechsler L,
- Kemp S,
- Schlaug G,
- Skalabrin E,
- et al
- Hacke W,
- Albers G,
- Al-Rawi Y,
- Bogousslavsky J,
- Davalos A,
- Eliasziw M,
- et al
- Wang DJ,
- Alger JR,
- Qiao JX,
- Hao Q,
- Hou S,
- Fiaz R,
- et al
- Bokkers RP,
- Hernandez DA,
- Merino JG,
- Mirasol RV,
- van Osch MJ,
- Hendrikse J,
- et al
- 15.↵The penumbra pivotal stroke trial: Safety and effectiveness of a new generation of mechanical devices for clot removal in intracranial large vessel occlusive disease. Stroke. 2009;40:2761–2768.
- 16.↵The interventional management of stroke (IMS) II study. Stroke. 2007;38:2127–2135.
- Higashida RT,
- Furlan AJ,
- Roberts H,
- Tomsick T,
- Connors B,
- Barr J,
- et al
- Chalela JA,
- Alsop DC,
- Gonzalez-Atavales JB,
- Maldjian JA,
- Kasner SE,
- Detre JA
- Schaefer PW,
- Hassankhani A,
- Putman C,
- Sorensen AG,
- Schwamm L,
- Koroshetz W,
- et al