| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Neuroradiology |
1 From the Department of Imaging, Lawson Health Research Institute, London, Ontario, Canada (B.D.M., T.Y.L.); Departments of Radiology (D.H.L., I.B.G., D.P.) and Clinical Neurosciences (V.B., R.K.C.), London Health Sciences Centre, London, Ontario, Canada; Departments of Radiology (A.J.F., R.I.A., S.S.) and Neurology (S.E.B., D.J.S.), Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada; Departments of Clinical Neurosciences (S.B.C., A.M.D.) and Radiology (M.G.), Foothills Medical Centre, Calgary, Alberta, Canada; and Department of Neuroscience, Ottawa Health Research Institute, Ottawa, Ontario, Canada (M.J.H.). Received April 19, 2007; revision requested June 13; revision received July 31; accepted August 28; final version accepted November 8. Supported in part by GE Healthcare and the Canadian Stroke Network. T.Y.L. is a consultant to GE Healthcare and is associated with Robarts Research Institute, which licenses software to GE Healthcare. Address correspondence to T.Y.L., Imaging Research Laboratories, Robarts Research Institute, 100 Perth Dr, London, ON, Canada N6A 5K8 (e-mail: tlee{at}imaging.robarts.ca).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
Materials and Methods: Ethics committee approval and informed consent were obtained. Thirty patients with stroke underwent unenhanced CT, CT angiography, and CT perfusion studies at admission. Additionally, CT angiography was performed 24 hours after the stroke, and an unenhanced CT study was performed 5–7 days after the stroke. Five patients were excluded; the remaining patients (10 men, 15 women; mean age, 70 years ± 13 [standard deviation]) were separated into those with recanalization (n = 16) and those without recanalization (n = 9) at 24 hours. For patients with recanalization, the final infarct was outlined on unenhanced CT images obtained 5–7 days after the event and was superimposed on coregistered maps from the CT perfusion study performed at admission. Ischemic white matter tissue (cerebral blood flow [CBF] < 14 mL/min/100 g) was identified at the admission CT perfusion study, and the penumbra was defined as the difference between the ischemic region and the infarct region.
Results: Infarct regions showed a matched decrease in CBF and cerebral blood volume (CBV) at admission, whereas penumbra regions showed a significant (P < .05) decrease in CBF but no change in CBV (P > .05) from contralateral values. A threshold CBF · CBV value of 8.14 was the most sensitive (95%, 20 of 21 regions) and specific (94%, 32 of 34 regions) parameter for differentiating between regions of ischemic white matter that recovered and regions of ischemic white matter that infarcted.
Conclusion: The product of CBF and CBV derived from CT perfusion data provided the best differentiation between regions of ischemic white matter that infarcted and regions of ischemic white matter that recovered 5–7 days after a stroke.
© RSNA, 2008
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
Despite the large number of imaging studies devoted to defining the ischemic penumbra and infarct core, one major limitation of many of these studies is the failure to distinguish between the gray and the white matter of the brain. Although these tissue compartments have distinctly different perfusion and metabolic requirements (3,4), many investigations have mixed both tissue types in deriving thresholds for infarction or, alternatively, focused only on thresholds for gray matter (5–8). In a recent study (9), we segmented gray matter of the brain from white matter and cerebrospinal fluid in computed tomographic (CT) perfusion imaging of patients with acute stroke and identified CT perfusion parameter(s) that best differentiated between ischemic gray matter that infarcted (infarct) and ischemic gray matter that recovered (penumbra). We showed that logistic regression analysis by using an interaction term (CBF · CBV) between cerebral blood flow (CBF) and blood volume (CBV) provided a sensitive and specific predictor of tissue viability.
Results of magnetic resonance (MR) imaging studies (10–13) suggest that perfusion and diffusion thresholds for infarction are different between gray and white matter and that the use of tissue-specific thresholds may improve the accuracy of tissue classification. We hypothesized that use of a combination of CT perfusion parameters involving CBF and CBV could enable accurate differentiation between white matter penumbra (operationally defined as tissue with CBF < 14 mL/min/100 g at admission that did not progress to infarction at CT performed without contrast material [unenhanced CT] 5–7 days after the event) and white matter infarct (defined as hypoattenuating areas on unenhanced CT images obtained 5–7 days after the event). Thus, the purpose of our study was to prospectively determine the parameters derived from CT perfusion data at admission that best differentiate ischemic white matter that recovers from that which infarcts, with the latter being defined at an unenhanced CT examination performed 5–7 days after the event.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
Participants
Ethics committee approval was obtained from all participating institutions. Informed consent was obtained from all patients or family members prior to the study. The 30 patients initially included in this study were the same as those described in a recent publication (9). All participants showed signs and/or symptoms of middle cerebral artery occlusion, presented to the emergency department within 7 hours of stroke onset, and were prospectively recruited for this study between November 2002 and January 2005. Exclusion criteria were as follows: evidence of brain stem infarct, intracranial hemorrhage, previous stroke with residual deficit, minor stroke symptoms (National Institutes of Health Stroke Scale [NIHSS] score, <4), clinically important hyperglycemia, impaired renal function and/or known allergy to contrast media; pregnancy, and age less than 18 years. Of the 30 patients, 25 contributed data and five were excluded for the following reasons: lack of ischemic signs at admission imaging and infarction at follow-up unenhanced CT (n = 2), large hemorrhage and edema requiring hemicraniectomy (n = 1), admission CT perfusion study performed at incorrect location (n = 1), and excessive movement resulting in unacceptable perfusion maps (n = 1). The remaining 25 patients were divided into those who experienced partial or full recanalization (n = 16, recanalized group) and those in whom occlusion persisted (n = 9, occluded group) at 24 hours, both as documented with CT angiography (Table 1, Fig 1). Patients were enrolled regardless of therapy, and results of the CT perfusion study performed at admission did not influence treatment decisions (patients and family members providing informed consent were aware of this).
|
|
For consistency, one author (B.D.M., with 4 years of experience with CT perfusion imaging) was responsible for generating functional maps by using CT perfusion software. The arterial input function and the venous output function, respectively, were obtained for each CT perfusion study as the time-attenuation curves in an anterior cerebral artery and the superior sagittal or transverse sinus. The CT perfusion software (CT Perfusion 3; GE Healthcare) used in this study deconvolved the arterial time-attenuation curve with the tissue time-attenuation curve from 2 x 2-pixel blocks to calculate quantitative CBF and CBV maps. The arterial input function was corrected for partial volume averaging by scaling its area relative to that of the venous output function, in which partial volume averaging effects are absent (14). This correction procedure is required for the calculation of quantitative CBF and CBV values with deconvolution; these values have been validated by using microspheres in animal studies and positron emission tomography (PET) and xenon CT in human patients (15–17). Additionally, a perfusion-weighted map was created by averaging cine CT perfusion images over the duration of the first pass of contrast material though the brain in the CT perfusion study; this served to accentuate the inherent attenuation difference (in Hounsfield units) by the additional CBV difference between gray and white matter.
Segmentation and Image Analysis
All postprocessing analysis of images was performed with custom software (IDL, version 5.6; RSI, Boulder, Colo) by one author (B.D.M.), who was blinded to the clinical outcomes. The perfusion-weighted map was used to segment white matter from gray matter on the basis of Hounsfield unit thresholds to produce a mask for white matter. This white matter region of interest (ROI) was then applied to the admission CBF and CBV maps, avoiding gray matter, cerebrospinal fluid, and skull bones. The admission perfusion-weighted, CBF, and CBV maps were automatically coregistered because they were derived from the same CT perfusion study.
For patients in the recanalized group, delayed unenhanced CT images were registered to baseline images by using custom software to adjust for in-plane rotational changes by minimizing the differences between the rotated delayed and baseline unenhanced CT images. The final infarct, as defined by areas of decreased attenuation on the registered 5–7-day unenhanced CT image, was outlined by an experienced neuroradiologist (D.H.L.), who was blinded to results from imaging performed at admission and other time points as well as to clinical outcomes. The final infarct region was then superimposed on the admission CBF map, and the white matter mask obtained from the perfusion-weighted image was applied to define ischemic (CBF < 14 mL/min/100 g) and oligemic or normal (CBF 
14 mL/min/100 g) white matter. The penumbra was operationally defined as the difference between the ischemic white matter region (not infarcted at the 5–7-day unenhanced CT study) and the infarct ROI. Accordingly, D.H.L. outlined ischemic white matter outside the infarct as the penumbra while blinded to imaging results at other time points and patient outcome. A value of 14 mL/min/100 g was chosen on the basis of results from previous literature and results in the occluded group that showed that white matter with CBF values below this threshold progressed to infarction when recanalization was not achieved (see Results). A mirror image of the ischemic region including both the infarct and the penumbra was created in the contralateral side of the brain to obtain CBF and CBV values. This process yielded a maximum of three distinct ROIs for each section of a patient study: the contralateral ROI, the penumbra ROI (with CBF < 14 mL/min/100 g and no infarction at 5–7-day unenhanced CT study), and the final infarct ROI (with hypoattenuation at the 5–7-day unenhanced CT study). Mean CBF, CBV, and CBF · CBV values and white matter region volume (in cubic centimeters) were calculated for all regions.
For patients in the occluded group, the volume of ischemic white matter on the admission CBF map (CBF < 14 mL/min/100 g) and the volume of the final infarct on the corresponding 5–7-day unenhanced CT sections were calculated. Infarcts occupying less than 1 cm3 of white matter were excluded from all analyses to minimize errors due to registration, segmentation, and partial volume averaging. All volumes described in this study, including final infarct volume, were limited to the 20-mm slab of tissue covered at the CT perfusion study. To reduce the influence of large vessels, pixels with CBF greater than 100 mL/min/100 g or CBV greater than 8 mL/100 g were excluded from further analysis for penumbra, infarct, and contralateral ROIs. Elimination of large vessels at dynamic contrast material–enhanced CT yields values that correlate well with those obtained at xenon CT and PET (15,18).
Statistical Analysis
Statistical analyses were performed by one author (B.D.M.); results were verified by coauthors. For all patients, mean and standard deviation of the NIHSS score at admission, time from stroke onset to imaging, and age were calculated. For patients in the occluded group, a Pearson correlation and paired t test were performed between the volume of ischemic (CBF < 14 mL/min/100 g) tissue on admission CT perfusion images and the final infarct volume on the corresponding 5–7-day unenhanced CT sections. Additional t tests were performed to assess differences between the total penumbra-plus-infarct volume in patients in the recanalized group and both the total ischemic volume and the total infarct volume in patients in the occluded group.
One-way analysis of variance was performed with the Tukey post-hoc test for significant differences in mean CBF, CBV, and CBF · CBV values between penumbra, final infarct, and contralateral ROIs for both the occluded and the recanalized groups. P < .05 was considered to indicate a significant difference for all comparisons. For patients in the recanalized group, mean CBF, CBV, and CBF · CBV values for all penumbra and infarct regions were entered as predictors into a logistic regression analysis to determine which parameter or combination of parameters provided the highest degree of sensitivity and specificity and greatest degree of separation between infarct and penumbra data sets. Sensitivity and specificity for infarction were calculated from the regression model by using each parameter independently. All statistical analyses were performed by using software (SPSS, version 13 for Windows; SPSS, Chicago, Ill).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
|
|
Results of logistic regression analysis performed by using the product of mean admission CBF and CBV, CBF · CBV, as a predictor in penumbra and final infarct regions showed that a threshold of 8.14 resulted in the highest sensitivity (95%; 20 of 21 regions) and specificity (94%; 32 of 34 regions) for infarction, with an overall accuracy of 95% (52 of 55 regions) of regions classified correctly (Fig 3).
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
Logistic regression analysis identified CBF · CBV as the best parameter for differentiating between final infarct and penumbra regions at a threshold of 8.14. The product of the two individual parameters resulted in less overlap between penumbra and infarct values and fewer data points between the 5% and 95% probability lines than either CBF or CBV alone. CBF · CBV provided a superior distinction between final infarct and penumbra than either CBF or CBV independently, while also achieving a maximal separation between the two groups. The choice of a hyperbolic discriminant, CBF · CBV, is suggested by logistic regression as the interaction term between the two main predictors CBF and CBF and reflects the pathophysiology of stroke: For increasingly ischemic CBF (<14 mL/min/100 g), an increasing CBV is required to maintain potential viability pending recanalization and vice versa. Another discriminant with a similar inverse relationship such as the linear discriminant (CBV = approximately 2 – CBF/8) performs equally well: It is 95% (20 of 21 regions) sensitive and 97% (33 of 34 regions) specific and thus 96% (53 of 55 regions) accurate.
A threshold of 14 mL/min/100 g was chosen in our study to define ischemic white matter at risk of infarction. The volume of ischemic white matter at the admission CT perfusion study correlated with final white matter infarct volume on the corresponding delayed unenhanced CT images for patients in the occluded group. This suggests that our CBF threshold for defining penumbra was appropriate because the entire ischemic area, by this definition, progressed to infarction when recanalization was not achieved. Additionally, this threshold is similar to values obtained in studies (13,19) in which MR perfusion imaging was used to define tissue at risk of infarction in white matter and is in agreement with results of previous studies in which a 34% reduction in CBF from contralateral values was used to define tissue at risk of infarction (5,16). Although an ischemic threshold of 14 mL/min/100 g was used in our study, mean values for some penumbra regions exceeded this value. This discrepancy arises because the penumbra region was manually outlined to encompass the entire region of perfusion abnormality, and there were pixels within this outlined area that had a CBF of greater than 14 mL/min/100 g, resulting in mean values greater than 14 mL/min/100 g.
CBF values in our study are similar to previously reported values for white matter in contralateral, penumbra, and infarcted tissue (13,19,20). However, in our study, we found that CBV values for infarcted regions on the 5–7-day unenhanced CT images showed a significant reduction from contralateral values, while CBV values for penumbra regions were not significantly different from contralateral values. These findings are in agreement with results of recent studies (10,20) of white matter involving CT perfusion imaging and MR perfusion imaging but differ from findings of another MR perfusion imaging study (13) that revealed increased CBV in both ischemic tissue that recovered and ischemic tissue that infarcted. One difference from our study is the fact that the latter study did not assess recanalization until 30 days, and it is possible that the elevated CBV values were derived from tissue that was viable at the time of measurement but that later infarcted. However, this does not explain the increased CBV for recovered ischemic tissue that was reported (13). The exclusion of pixels with a CBF of more than 100 mL/min/100 g or a CBV of more than 8 mL/100 g is unlikely to have a large effect on the CBV values for penumbras because large vessels are not prominent in white matter.
Infarcted white matter in patients in the occluded group had CBF and CBV values intermediate between values obtained for infarct and values obtained for penumbra in patients experiencing recanalization. Additionally, the volume of infarcted tissue on the 5–7-day unenhanced CT images for patients in the occluded group was not significantly different than the sum of the penumbra and infarct volumes in patients experiencing recanalization. This suggests that infarcted tissue on the 5–7-day unenhanced CT images in patients in the occluded group contained both viable and nonviable tissue at admission. Stroke severity according to the NIHSS score was not significantly different between the two groups of patients (P > .05); again, this suggests that without recanalization, penumbra white matter identified in this study progressed to infarction.
An important limitation of our study was the relatively small sample size of 25 patients with middle cerebral artery occlusion, from among which only 16 patients with recanalization of the occluded vessel provided data for the logistic regression analysis. Additionally, the thresholds in our study were derived by using an ROI analysis, which may result in an overestimation of the thresholds for infarction and may be applicable only to patients experiencing complete middle cerebral artery occlusion. Manual tracing of the infarcted region by using decreased attenuation (in Hounsfield units) on unenhanced CT images obtained 5–7 days after a stroke may not reflect the true final infarct because of edema and incomplete maturation of the infarct. Although there are limitations to this technique, other methods for defining the infarct also suffer from similar problems, including atrophy of the brain at a much later unenhanced CT examination or errors in the accuracy of predicting final infarct with diffusion-weighted imaging at early time points. Despite these issues, by using a design in which infarcted white matter was defined at an unenhanced CT examination performed 5–7 days later, we have shown that results of admission CT perfusion imaging were able to characterize ischemic white matter that progressed to infarction 5–7 days later if recanalization was not achieved.
Another limitation of this study was that the presence or absence of recanalization was observed at admission and 24 hours later; therefore, the exact time of recanalization was unknown. Inherent limitations of the CT perfusion imaging technique used in our study are limited anatomic coverage, use of iodinated contrast agent, and exposure to x-rays. CT hardware and software improvements are enabling increased anatomic coverage and/or a reduction in radiation dose (21,22). Results of studies have shown that patients have a very good tolerance of iodinated contrast material (23), and the x-ray dose from a CT perfusion study is about double that of a whole-head unenhanced CT examination (unpublished data, 2005).
Evaluation of the ischemic penumbra in white matter and gray matter is critical for selection of appropriate patients for both thrombolytic and neuroprotection treatments; both of these treatments require a substantial volume of penumbra to achieve optimal results. CT imaging is readily accessible for stroke patients and is available around-the-clock in most hospitals; in most instances, it is the imaging modality of choice in the emergency room. Hemorrhage and early ischemic changes can be detected with unenhanced CT, while CT angiography provides information on the location of occlusion. Our study provides preliminary evidence that CT perfusion imaging can be used to segment white matter and that a combination of CBF and CBV can be used to characterize white matter tissue that progresses to infarction or recovers within the 1st week after stroke, with infarction defined at an unenhanced CT examination performed 5–7 days after the event. This technique needs to be tested for efficacy in a larger, randomized, prospective trial before we can speculate as to whether it could be used to guide treatment decisions or prospectively predict outcomes of specific tissue on an individual basis.
|
ADVANCES IN KNOWLEDGE |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
|
IMPLICATION FOR PATIENT CARE |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Abbreviations: CBF = cerebral blood flow • CBV = cerebral blood volume • NIHSS = National Institutes of Health Stroke Scale • ROI = region of interest
Author contributions: Guarantors of integrity of entire study, B.D.M., T.Y.L.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, B.D.M., M.J.H., S.B.C., T.Y.L.; clinical studies, A.J.F., D.H.L., D.J.S., S.E.B., M.J.H., S.B.C., A.M.D., M.G., R.I.A., S.S., I.B.G., D.P., T.Y.L., ; statistical analysis, B.D.M., T.Y.L.; and manuscript editing, B.D.M., A.J.F., D.H.L., S.E.B., M.J.H., S.B.C., A.M.D., R.I.A., S.S., V.B., D.P., R.K.C., T.Y.L.
See Materials and Methods for pertinent disclosures.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  L.-J. Zhang, Y.-E Zhao, S.-Y. Wu, B. M. Yeh, C.-S. Zhou, X.-B. Hu, Q.-J. Hu, and G.-M. Lu Pulmonary Embolism Detection with Dual-Energy CT: Experimental Study of Dual-Source CT in Rabbits Radiology, July 1, 2009; 252(1): 61 - 70. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A.A. Konstas, G.V. Goldmakher, T.-Y. Lee, and M.H. Lev Theoretic Basis and Technical Implementations of CT Perfusion in Acute Ischemic Stroke, Part 2: Technical Implementations AJNR Am. J. Neuroradiol., May 1, 2009; 30(5): 885 - 892. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A.S. Ahmed, M. Zellerhoff, C.M. Strother, K.A. Pulfer, T. Redel, Y. Deuerling-Zheng, K. Royalty, D. Consigny, and D.B. Niemann C-Arm CT Measurement of Cerebral Blood Volume: An Experimental Study in Canines AJNR Am. J. Neuroradiol., May 1, 2009; 30(5): 917 - 922. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I.Y.L. Tan, A.M. Demchuk, J. Hopyan, L. Zhang, D. Gladstone, K. Wong, M. Martin, S.P. Symons, A.J. Fox, and R.I. Aviv CT Angiography Clot Burden Score and Collateral Score: Correlation with Clinical and Radiologic Outcomes in Acute Middle Cerebral Artery Infarct AJNR Am. J. Neuroradiol., March 1, 2009; 30(3): 525 - 531. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. I. Aviv, C. D. d'Esterre, B. D. Murphy, J. J. Hopyan, B. Buck, G. Mallia, V. Li, L. Zhang, S. P. Symons, and T.-Y. Lee Hemorrhagic Transformation of Ischemic Stroke: Prediction with CT Perfusion Radiology, March 1, 2009; 250(3): 867 - 877. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T.J. Huynh, B. Murphy, J.A. Pettersen, H. Tu, D.J. Sahlas, L. Zhang, S.P. Symons, S. Black, T.-Y. Lee, and R.I. Aviv CT Perfusion Quantification of Small-Vessel Ischemic Severity AJNR Am. J. Neuroradiol., November 1, 2008; 29(10): 1831 - 1836. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
14 to 
100 mL/min/100 g is shown in light blue). Penumbra is identified as ischemic white matter tissue that did not show infarction on the 5–7-day unenhanced CT images (black outline in E–G) and is superimposed on CBF and CBV images. Infarct is characterized by a matched decrease in CBF (5.40 mL/min/100 g) and CBV (0.41 mL/100 g), while the penumbra region has reduced CBF (10.9 mL/min/100 g) but maintained CBV (1.04 mL/100 g).
