RENDIMIENTOS

Clinical Investigative Study Reproducibility of ABC/2 Method to Determine Infarct Volume and Mismatch Percentage with CT Perfusion Kris F. French, MD, Julie K. Martinez, RN, Adam H. DeHavenon, MD, Natalie R. Weathered, MD, Matthew Grantz, MD, Shawn M. Smith, MD, Michael Wilder, MD, Ulrich A. Rassner, MD, John C. Kircher, PhD, L.Dana. Dewitt, MD, Jana J. Wold, MD, Robert E. Hoesch, MD, PhD From the Department of Neurology, University of Utah, Salt Lake City, UT (KFF, JKM, AHD, NRW, MG, LD, JJW); Department of Neuroradiology, University of Utah, Salt Lake City, UT (MW, UAR); Department of Educational Psychology, University of Utah, Salt Lake City, UT (JCK); Department of Neurocritical Care (SMS, REH).

ABSTRACT BACKGROUND

Our aim is to implement a simple, rapid, and reliable method using computed tomography perfusion imaging and clinical judgment to target patients for reperfusion therapy in the hyper-acute stroke setting. We introduce a novel formula (1–infarct volume [CBV]/penumbra volume [MTT] × 100%) to quantify mismatch percentage. METHODS

Twenty patients with anterior circulation strokes who underwent CT perfusion and received intravenous tissue plasminogen activator (IV tPA) were analyzed retrospectively. Nine blinded viewers determined volume of infarct and ischemic penumbra using the ABC/2 method and also the mismatch percentage. RESULTS

Interrater reliability using the volumetric formula (ABC/2) was very good (intraclass correlation [ICC] = .9440 and ICC = .8510) for hemodynamic parameters infarct (CBV) and penumbra (MTT). ICC coefficient using the mismatch formula (1–MTT/CBV × 100%) was good (ICC of .635). CONCLUSIONS

The ABC/2 method of volume estimation on CT perfusion is a reliable and efficient approach to determine infarct and penumbra volumes. The 1–CBV/MTT × 100% formula produces a mismatch percentage assisting providers in communicating the proportion of salvageable brain and guides therapy in the setting of patients with unclear time of onset with potentially salvageable tissue who can undergo mechanical retrieval or intraarterial thrombolytics.

Keywords: Stroke, computed tomography, thrombolytics. Acceptance: Received July 25, 2012, and in revised form October 12, 2012. Accepted for publication November 8, 2012. Correspondence: Address correspondence to to Kris French, M.D., University of Utah, Clinical Neurosciences Center, 175 N Medical Dr E, Salt Lake City, UT 84132. E-mail: Kris.french@hsc. utah.edu. Contributorship statement: All authors contributed to the work of this paper. Dr. Hoesch and Dr. Wold reviewed and supervised all work and the process of the study and generation of the manuscript. Dr. French is the corresponding and first author of the paper. Drs. DeHavenon, Weathered, Grantz, Wilder, Smith, and DeWitt were all involved in reviewing and editing the manuscript as well as involvement in blinded viewing of the data as described in the manuscript. Dr. Rassner was involved in the methods section especially with the neuroradiology description. Dr. Kircher was involved with the statistics generation. Sources of Funding: There were no sources of funding for this study. Disclosures: Kris French: None; Julie Martinez: None; Adam DeHavenon: None; Natalie Weathered: None; John Kircher: None; Ulrich Rassner: None; Matthew Grantz: None; Shawn Smith: None; Michael Wilder: None; L.D. Dewitt: None; Jana Wold: None; Robert Hoesch: None J Neuroimaging 2014;24:232-237. DOI: 10.1111/jon.12001

Introduction Brain imaging is an essential element in the approach to patients with hyperacute and acute ischemic stroke. Determination of infarct volume relative to the volume of tissue at risk (ischemic

232

Copyright

penumbra) can be an important step in planning thrombolysis, interventional therapy, or immediate poststroke care, such as induced or permissive hypertension.1 Several methods are available to measure volume on head imaging including visual

◦ 2013 by the American Society of Neuroimaging C

estimation, conventional manual planimetry, and observer calculations.1 Visual estimation is a gestalt approach to quickly determine the extent of involvement and the magnitude of possible mismatch on CT perfusion. However, it has been demonstrated that visual approximation can result in overestimation of infarct size and underestimation of mismatch, which may result in exclusion of patients for thrombolysis.2 Manual planimetric measurements are performed by an experienced radiologist by tracing the area of abnormal signal and estimating the volume of region involved.3 Although the planimetric method has been determined to be more accurate in determining infarct volume, it is time consuming and not practical in the hyperacute and acute settings.3 Observer calculations, such as the ABC/2 formula have been used with CT imaging in the estimation of stroke and intracerebral hemorrhage volumes.4 The ABC/2 calculation was introduced as a fast method for estimating intracerebral hemorrhage, in which it is used as a rapid assessment tool requiring one minute or less by the observer to calculate.2,5 This calculation is based on the formula for an ellipse and is generated by measuring the cross sectional area (A x B), multiplying the area by the number of slices (corrected for slice thickness), then, dividing the result by two (ABC/2). Because of good interrater reliability for hemorrhage volume, application of the ABC/2 formula has been useful in neurocritical care for prognosis and intervention guidance in ICH.2 The ABC/2 method has also been studied in both MRI and CT imaging for ischemic stroke.2,3 On MRI, the ABC/2 method has near perfect intraclass correlation (ICC; .99), but consistently overestimates the size of infarct by a mean of 7 mL compared to the planimetric method.3 Nonetheless, conventional planimetry and ABC/2 have been highly correlated in a number of stroke studies using CT, suggesting that the ABC/2 calculation could be a valid volumetric tool in acute ischemic stroke.4 Both CT and MRI play important roles in acute stroke and each modality has particular strengths and weaknesses. Current techniques to predict outcome in acute stroke patients are controversial and include the use of imaging studies such as CT and MRI to determine the relative volume of infarct to salvageable, but ischemic, brain tissue.6 Thrombolytic treatment of the culprit clot is the mainstay of reperfusion therapy in acute stroke. CT and MRI are potential imaging modalities for determining which patients may benefit more from thrombolysis. MRI has higher sensitivity for acutely infarcted brain and ischemic penumbra. However, contrary to previous teaching, not all regions of restricted diffusion reflect irreversibly damaged tissue and portions can sometimes be reversed with reperfusion.7 Despite increased sensitivity for ischemic tissue, acquisition of high quality MRI requires far more time than CT imaging and a number of contraindications to MRI, such as a pacemaker, preclude performance of MRI on all patients. Furthermore, not all medical centers house an MRI scanner and not all medical centers with MRI scanners are capable of performing MR perfusion to determine the regions of ischemic penumbra. CT has several advantages over the use of MR, especially in the setting of acute stroke during which emergent decisionmaking by the provider is paramount. CT is a more rapid and available imaging technique than MRI and is an option regard-

less of whether the patient has a pacemaker or other contraindication to MRI. Although all medical centers might not have CT perfusion, these technologies are more uniformly available than MRI.8 CT is the most common imaging performed in the acute stroke setting and is the only imaging study required before administration of IV tissue plasminogen activator (tPA). Because of the wide availability and relative ease of performance of CT compared to MRI, and the fact that nearly all patients who present with acute stroke undergo brain CT imaging, CT perfusion is frequently obtained at the time of standard brain CT imaging. It has been demonstrated that there is near perfect interobserver agreement using CT perfusion and it also improves confidence in diagnostic accuracy.9 Good agreement has been demonstrated between CT perfusion and MRI in determining lesion size and location.10 Occasionally, the ischemic region is located outside the scanning range of CT perfusion, and this represents one of its major disadvantages: with each bolus injection only one brain section can be evaluated.10 Furthermore, CT perfusion requires administration of IV contrast, which is contraindicated in renal failure and in patients with contrast allergy. CT perfusion also requires delivery of far more radiation than a standard brain CT.10 Because of the wide applicability of CT and CT perfusion in the setting of acute and hyperacute ischemic stroke, a method to reliably and easily communicate volume measurements on CT perfusion is needed. The purposes of this study are: (1) to determine the interrater reliability among viewers using the established ABC/2 calculation for cerebral blood volume (CBV; infarct volume) and mean transit time (MTT; penumbra volume); and (2) to introduce a novel formula (1–CBV/MTT × 100%) to quantify the mismatch percentage using CT perfusion.

Methods Patient Selection and Study Design After approval from the Institutional Review Board, we retrospectively reviewed clinical and imaging findings of patients admitted to the neurology service at a large academic hospital between June 2010 and June 2011. All consecutive acute ischemic stroke patients at the University of Utah who were greater than 18 years of age and received IV tPA were eligible for inclusion. Patients who did not undergo CT perfusion or had posterior circulation strokes were excluded. A search for appropriate patients was performed using the University of Utah Stroke Center patient database. This search revealed a total of 63 acute ischemic stroke patients who received IV tPA. A total of 38 patients in this group underwent CT perfusion. Twelve of these patients did not have documented evidence of acute ischemic stroke and an additional 4 patients had posterior circulation strokes. Therefore, a total of 22 patients were included in the final analysis (Fig 1). NIHSS scores were recorded at the time of admission to the hospital. Modifed Rankin scale was assessed at 3 months after stroke onset. Patient outcomes were obtained from clinic visits and chart review.

Image Analysis, Acquisition, and Processing CT perfusion was performed on all patients included in this study using a 64-slice Siemens Somatom Definition or Somatom

French et al: Reproducibility of ABC/2 Method

233

Fig 2. ABC/2 volumetric calculation. The image on the left demonstrates cerebral blood volume images and the image on the right demonstrates mean transit time.

Statistical Analysis Fig 1. Total acute ischemic strokes that received iv tPA from June 2010 to June 2011.

Definition AS with 8.4 cm coverage utilizing a 4-dimensional spiral technique. Siemens Syngo Neuro Volume Perfusion CT software utilized deconvolution method (central volume theorem), generating cerebral blood flow (CBF), CBV, time to drain (TTD), and MTT. The images were obtained from the University picture archiving and communication system (PACS). CBV and MTT hemodynamic maps were stored on flash drives deidentified of patient data. Infarct regions were measured using the CBV images and penumbra regions were measured on MTT images. An advanced image analysis tool was used to determine infarct and penumbra volumes. The MTT images were chosen for penumbra instead of CBF because MTT values of normal gray and white matter are not significantly different and tend to be a more reliable parameter than CBF.8

ABC/2 Measurements Volumetric calculations were based on prior data using the elliptical model of the ABC/2 volumetric formula.2,5 Nine blinded observers performed the calculations for the 22 patients. The nine blinded observers consisted of three neurovascular attendings, three neurovascular fellows, and three senior level neurology residents. No a priori education was provided for volumetric calculations besides a detailed instruction sheet. The MTT and CBV perfusion images were placed on flash memory drives along with instructions and data sheets for recording measurements. The nine observers selected the image (Fig 2) with the largest region of infarct or ischemia. Using measurement tools on the PACS software, the largest diameter in the x-direction (A) and the y-direction (B) were obtained. C was determined as the number of slices (1 cm thickness) with observable infarct or ischemia on the images. The lesion volumes for CBV and MTT were calculated using the ABC/2 formula for each image and reported in cubic centimeters. The mismatch percentage was then calculated utilizing the CBV and MTT volumes which were placed into the formula, 1–CBV/MTT × 100%.

234

Journal of Neuroimaging Vol 24 No 3 May/June 2014

Data analyses were performed using Statistical Package for the Social Sciences (SPSS), OriginPro 8.5, and SigmaPlot software. Interrater reliability using ICC coefficient with two-way ANOVA was evaluated on the CBV, MTT, and mismatch calculations between nine blinded observers. ICC was interpreted as follows: 0-.2 indicates poor agreement, .2-.4 indicates fair agreement, .4-.6 indicates moderate agreement, .6-.8 indicates strong agreement, and >.8 indicates almost perfect agreement. Kolmogorov-Smirnov testing at the .05 level was used to test datasets for normality. Datasets with a normal distribution were described using means and standard error of the mean. All other continuous datasets were described using medians with interquartile range (IQR).

Results Pooled patient demographics and comorbid conditions are outlined in Table 1. Individual patient demographics (n = 22) and outcomes are outlined in Table 2. Most patients (59%) were male and the mean age was 63 years. Only 13% (3/22 patients) of the patients had a prior diagnosis of atrial fibrillation. New or suspected diagnosis of atrial fibrillation was made in 6 more patients by the time of discharge, therefore, 9 patients overall had a diagnosis of atrial fibrillation. No patients had a past history of stroke. All patients received IV tPA and 27% (6/22 patients) required intraarterial thrombolysis or mechanical clot retrieval. Time to IV tPA ranged from 40 minutes to 220 minutes with mean of 132 ± 10 minutes. Intracerebral hemorrhage Table 1.

Patient Demographics, NIHSSS, and Comorbid Conditions Total Patients (n = 22)

Age (mean, SEM, years) Weight (mean, SEM, kg) NIHSS score (mean, SEM) Gender (male, % male) Diabetes Previous stroke Atrial fibrillation Hypertension Previous use of aspirin

63.4 ± 4.4 74.9 ± 3.9 12.5 ± 1.4 13 (59%) 7 (32%) 0 (0%) 9 (41%) 13 (59%) 7 (32%)

n = sample population; SEM = standard error of the mean; NIHSS = National Institute of Health Stroke Scale; kg = kilograms.

Table 2.

Patient

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

Individual Patient Outcomes and Hospital Stay Length (n = 22)

Age

Gender

Death from Stroke

83 72 51 85 52 56 65 53 73 53 67 62 19 86 79 73 75 26 22 86 56 88

M F F F M M M F M F M M M F M M F M F M F M

N Y N Y Y N N N N N N N Y N N Y N N N N N N

Time to Tpa (Min)

155 180 90 150 120 150 40 210 120 220 175 80 125 103 107 105 180 120 100 70 85 220

ICH

90-Day mRS Score

Length of Stay (days)

N Y N Y Y N N N N N N N Y N N N N N N N N N

1 6 1 6 6 1 1 0 1 0 2 0 6 0 0 6 2 1 2 2 1 1

2 1 2 6 9 3 4 2 5 5 8 2 6 3 2 6 9 6 5 6 4 4

Table 3 also presents the median and IQR for the mismatch calculation (1–CBV/MTT × 100%). The median mismatch was 81.5% (39.6-100%) for all nine observers with an ICC of .6385. The three neuro-vascular attendings generated an ICC for mismatch percentage of .9230, higher than the overall ICC for mismatch percentage. The three neurology fellows generated an ICC for mismatch percentage of .554, whereas the three neurology residents obtained an ICC of .4991 for mismatch percentage. In general, residents estimated the penumbra volume on MTT to be higher, leading to higher median mismatch (81.5%) compared to attendings (80.0%) and fellows (75.8%). Figure 3 represents scatter plots that present the range and distribution of the data for each of the 22 patients (x-axis). There are nine data points plotted for each patient representing the volume or mismatch percentage given by each of the nine observers. Figure 3(a) displays the infarct volume (CBV, y-axis) reported by each observer for each patient and illustrates the good agreement (ICC of .944) between the nine different raters. Figure 3(b) displays the penumbral volumes (MTT, y-axis) and generally good consistency as manifested by relatively tight clustering of the measurements (ICC of .8510). Figure 3(c) displays the mismatch percentages (y-axis) and shows more variability of data between observers for each patient (ICC of .6385) compared to the data for the prior two parameters (CBV and MTT).

Min = minutes; ICH = intracerebral hemorrhage; mRS = modified rankin score. Length of stay includes only acute hospitalization, not inpatient rehabilitations.

Discussion

occurred in 18% (4/22) and two of these patients underwent surgical decompression. Mortality was 23% (5/22 patients). The mean length of acute hospitalization (not including inpatient physical rehabilitation) was 4 ± .4 days. The mean 90-day modified Rankin score (mRS) was 2 ± .5. Table 3 displays the median volumes, IQR, and ICC using the ABC/2 calculation for the nine observers taken together and also divided by level of experience (attendings fellows, residents). Overall, median (IQR) values for CBV and MTT were 24.9 mL (0-112.9 mL) and 214.3 mL (0-585.3 mL), respectively. ICC between all nine observers was .944 for CBV and .851 for MTT volume.

The major findings of this study are: (1) infarct volume (CBV) and penumbra volume (MTT) can be reproducibly and reliably calculated using the ABC/2 formula on CT perfusion; (2) mismatch percentage can be calculated from these volumes using the novel formula 1–CBV/MTT × 100%; and (3) mismatch percentage calculation demonstrates good reliability among observers and could represent an intuitive tool for communicating percentage of brain at risk and not yet infracted. These calculations are simple and fast and communicate a quantitative and logical conclusion about the CT perfusion. A small mismatch percentage communicates a small area of potentially salvageable tissue whereas a large mismatch percentage corresponds to a large area of salvageable tissue. For example, a result of 25% tells the physician that there is not

Table 3.

Overall and Observer-Specific Interrater Reliability Coefficients (Single Measure ICC) CBV (95% CI) Median (IQR, mL)

Total (k = 9) Attendings (k = 3) Fellows (k = 3) Residents (k = 3)

24.9 (0-112.9) 28.0 (0-117.7) 26.5 (0-99.9) 24.4 (0-124.8)

ICC (95% CI)

.9440 (.9089-.9739) .9480 (.8854-.9732) .9280 (.8961-.9758) .9320 (.8957-.9757)

MTT (95% CI) Median (IQR, mL)

214.3 (0-585.3) 192.3 (0-538.3) 195.8 (0-583.1) 262.2 (0-598.7)

ICC (95% CI)

.8510 (.7959-.9366) .8810 (.7238-.9291) .8430 (.8058-.9579) .8430 (.7968-.9445)

Mismatch (95% CI) Median (IQR, %)

81.5 (39.6 - 100) 80.0 (34.7-100) 75.8 (30.2-100) 86.7 (56.2-100)

ICC (95% CI)

.6385 (.4866-.7925) .9230 (.8463-.9670) .5540 (.3748–.7963) .4991 (.2213-.7416)

CT = computed tomography; ICC = intraclass correlation coefficient; IQR = interquartile range; mL = milliliter; k = rater; CBV = cerebral blood volume; MTT = mean transit time; 95% CI = confidence interval. ICC interpreted as follows: 0-.2 indicates poor agreement, .2-.4 indicates fair agreement, .4-.6 indicates moderate agreement, .6-.8 indicates strong agreement, and > .8 indicates almost perfect agreement.

French et al: Reproducibility of ABC/2 Method

235

Fig 3. Three scatter plots (A-C) demonstrating the variability of

data derived for each patient. Y-axis is CBV data (A), MTT data (B), and mismatch percentage (C). The x-axis consists of the patent number (n = 22). A. Scatter plot demonstrating range of agreement on infarct volume (CBV) for each patient (ICC = .944). B. Scatter plot demonstrating range of agreement on penumbra volume (MTT) for each patient (ICC = .8510). C. Scatter plot demonstrating range of agreement on mismatch percentage for each patient (ICC = .6385).

236

Journal of Neuroimaging Vol 24 No 3 May/June 2014

much penumbra to be saved. Whereas, a 100% mismatch communicates that all of the tissue is potentially salvageable. These percentages seem instinctive. Further studies will determine whether mismatch percentage corresponds to a proportion of patients who racanalize and to long-term patient outcomes. We found that the overall interrater agreement between nine blinded observers demonstrated good ICC of .6385 for mismatch percentages. When separated out, the ICC for CBV and MTT among all nine raters was better at .944 and .851, respectively. The agreement between neurovascular attendings was higher (ICC = .923) than the agreement for either the neurology fellows (ICC = .554) or the neurology residents (ICC = .4991). However, these ICCs are still considered to demonstrate fairly good agreement. Much of the discrepancy between the interobserver reliability may have come from difficulty for less experienced physicians (fellows and residents) in locating the region of stroke or ischemia. The difference in agreement between attendings, fellows, and residents suggests that experience plays a role in the reliability of volumetric measurements. There is a clear improvement in agreement from the level of neurology resident to neurology fellow to neurovascular attending. Further education about imaging techniques and the volumetric procedures would likely increase the degree of agreement. With further education and, consequently, better interrater reliability, these volumetric calculations can be easily and dependably compared between observers to assist in decision making for acute and hyperacute stroke therapies. Currently, there is a paucity of standardized practices for imaging studies to direct thrombolytic therapy in the stroke patient. Penumbra-guided imaging has not yet been validated as a reliable method for targeting stroke patients for therapy because of variability in observer methods. The unpredictability of estimating “tissue at risk” can result in different treatment decisions and this has resulted in the need to evaluate methods to standardize penumbra imaging and quantification. The Stroke Imaging Repository Consortium was created to address issues such as clinical utility of the ischemic penumbra as well as other potential imaging techniques to be used as biomarkers.11 Validation of MRI as a biomarker for several criteria including lesion volume and diffusion-perfusion mismatch have been investigated and are potentials for application of neuro-imaging standardization in stroke.12 There are several limitations to our study. First, this study is a preliminary retrospective evaluation of interobserver reliability with a relatively small sample size including only 22 patients with acute stroke who received IV tPA. A large cohort would be required to evaluate the clinical accuracy of the method in all stroke patients. Moreover, three perfusion scans did not cover the entire area of stroke which is a known limitation of the use of CT perfusion. In addition, many hospitals do not have observers with as much experience in interpreting CT perfusion as vascular neurologists and this may lower the generalizable agreement between all observers. Consideration of evaluating both neuroradiologists as well as non-specialized physicians in further studies is warranted. CT perfusion parameters to characterize the ischemic penumbra that are selected during post-processing of data can influence the final

results including MTT and CBV.13 Automated postprocessing of venous outflow function and arterial input function increase interobserver reliability in the measurement of CBV and MTT and this is a potential area of improvement in stroke care.13 It is important to also recognize that radiation exposure is potentially harmful to patients and there are protocols that may decrease the total amount of radiation during CT perfusion scanning. There is also an overall lack of clinical therapeutic utility of the use of these quantitative measures. Future studies that will examine outcome should help to clarify this issue. Follow-up neuroimaging with MRI or CT and clinical outcome measures can help to determine the true predictive value of the initial CT perfusion scan in regard to reversibly and irreversibly damaged brain tissue. Future studies will address these outcomes. Utilizing the ABC/2 elliptical calculation with CT perfusion hemodynamic parameters for infarct (CBV) and penumbra (MTT) appears to be a simple, reproducible, and fast approach to determination of the size of brain at risk.2 Using the novel formula, 1–CBV/MTT × 100%, an estimation of difference between the irreversibly damaged and potentially reversibly ischemic tissue can be performed. Use of the 1–CBV/MTT × 100% formula gives a quick estimation of mismatch percentage that assists providers in communicating the proportion of salvageable brain and may guide therapy in the setting of patients with unclear time of stroke onset with potentially salvageable tissue who have the option of mechanical clot retrieval or intraarterial thrombolytics. These conclusions will form the foundation for the validation of CT perfusion as a potential imaging biomarker in stroke management. Future studies to address the clinical utility in patient outcomes can help to verify the accuracy of this quantitative mismatch method.

References 1. Schellinger PD, Fiebach JB, Hacke W. Imaging-based decision making in thrombolytic therapy for ischemic stroke: present status. Stroke 2003;34(2):575-583.

2. Sims JR, Gharai LR, Schaefer, et al. ABC/2 for rapid clinical estimate of infarct, perfusion, and mismatch volumes. Neurology 2009;72(24):2104-2110. 3. Pedraza S, Puig J, Blasco G, et al. Reliability of the ABC/2 method in determining acute infarct volume. J Neuroimaging 2012; 22(2):155-159. 4. Gomez-Marino R, Andre C, Novis SA. Volumetric determination of cerebral infarction in the acute phase using skull computed tomography without contrast: comparative study of 3 methods. Arq Neuropsiquiatr 2001;59(2-B):380-383. 5. Kothari RU, Brott T, Broderick JP, et al. The ABCs of measuring intracerebral hemorrhage volumes. Stroke 1996;27(8):13041305. 6. 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;40:27612768. 7. Wechsler LR. Imaging evaluation of acute ischemic stroke. Stroke 2011;42(1 Suppl):S12-S15. 8. Wintermark M, Flanders AE, Velthuis B, et al. Perfusion-CT assessment of infarct core and penumbra: receiver operating characteristic curve analysis in 130 patients suspected of acute hemispheric stroke. Stroke 2006;37:979-985. 9. Agarwal S, Jones PS, Alawneh JA, et al. Does perfusion computed tomography facilitate clinical decision making for thrombolysis in unselected acute patients with suspected ischaemic stroke? Cerebrovasc Dis 2011;32:227-233. 10. Schaefer PW, Barak ER, Kamalian S, et al. Quantitative assessment of core/penumbra mismatch in acute stroke: CT and MR perfusion imaging are strongly correlated when sufficient brain volume is imaged. Stroke 2008;39(11):2986-2992. 11. Leiva-Salinas C, Hom J, Warach S, Wintermark M. Stroke imaging research road map. Neuroimaging Clin N Am 2011;21(2): 239-245. 12. Pedraza S, Puig J, Blasco G, et al. Magnetic resonance imaging biomarkers of ischemic stroke: criteria for the validation of primary imaging biomarkers. Drug News Perspect 2009;22(8): 481-486. 13. Soares BP, Dankbaar JW, Bredno J, et al. Automated versus manual post-processing of perfusion-CT data in patients with acute cerebral ischemia: influence on interobserver variability. Neuroradiology 2009;51:445-451.

French et al: Reproducibility of ABC/2 Method

237