| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Neuroradiology |
1 From the Departments of Neuroradiology (C.Z.S., L.R., P.V.P., C.G., L.Ø.) and Neurology (G.A.), Århus University Hospital, Nørrebrogade 44, DK-8000 Århus C, Denmark. Received July 23, 2001; revision requested August 31; final revision received March 29, 2002; accepted April 9. L.Ø. supported by the Danish Medical Research Council and the Danish National Research Foundation. Address correspondence to C.Z.S. (e-mail: claus@pet.auh.dk).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Twenty-three patients with symptoms of stroke underwent magnetic resonance (MR) imaging during the acute stage, and the tissue at risk was estimated from MTT maps and maps generated by means of detecting abnormal FH. Final infarct volumes were calculated from T2-weighted follow-up MR image measurement. The Wilcoxon signed rank test was performed to compare the two predictive maps (MTT and FH) with T2-weighted follow-up maps.
RESULTS: Eleven (48%) patients experienced infarct growth. Both the MTT and the FH maps enabled prediction of 10 of these cases. There were five false-positive cases with MTT measurement but three with FH measurement. In terms of predicting final infarct volumes, the final infarct size on the MTT maps was overestimated by 75%. The final infarct size on the FH maps also was overestimated, but by only 15%. MTT map measurements were significantly different from follow-up MR image measurements (P = .005), but FH map measurements were not (P = .059).
CONCLUSION: FH maps may enable more precise prediction of final infarct volume in stroke patients.
© RSNA, 2002
Index terms: Brain, diffusion, 13.12144 • Brain, infarction, 13.78 • Brain, perfusion, 13.12144 • Magnetic resonance (MR), diffusion study, 13.12144, 17.12144 • Magnetic resonance (MR), vascular studies, 13.12144, 17.12144
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Magnetic resonance (MR) imaging sequences are used to evaluate patients with symptoms of acute stroke (hereafter referred to as acute stroke patients) in research settings. Diffusion-weighted (DW) imaging can depict the diseased tissue in the acute phase after cerebral infarction because of the decreased self diffusion of water in the infarcted area (3). In an attempt to predict final infarct size and thereby guide and evaluate treatment, Welch et al (4) proposed a model based on T2-weighted and DW MR data. Perfusion-weighted (PW) imaging also has been investigated (5–7).
Combined PW and DW imaging has shown particular promise as a technique used to predict final infarct size, because a perfusion defect often is seen extending beyond the limits of the defect depicted at DW imaging; this phenomenon is referred to as an area of diffusion-perfusion mismatch. This mismatch has been demonstrated to be the site of subsequent infarct growth and is therefore referred to as the ischemic penumbra or tissue at risk (8). This tissue is a potential target for treatment during the acute-symptom stage, and it may be feasible to more effectively triage patients for therapy by performing MR imaging during the acute stage (9).
The parameters assessed at PW imaging are cerebral blood flow, cerebral blood volume, and mean transit time (MTT), which is defined as the cerebral blood volume–to–cerebral blood flow ratio (10). It appears that no single one of these parameters can indicate which tissue will subsequently become infarcted. MTT is often used because it is related to the cerebral perfusion pressure (11). However, MTT maps often yield overestimations of the final infarct size (5).
In an attempt to more precisely define tissue survival with PW imaging, Østergaard et al (12) suggested a model based on the fact that in states of decreased cerebral perfusion pressure, an important determinant of oxygen delivery is the heterogeneity of microscopic blood flow relative to the mean flow. In the normal state, flow is widely distributed around the mean flow. Experimental data suggest that with low cerebral perfusion pressure, flow becomes more homogeneous (13) and thus theoretically contributes to a more efficient oxygen extraction (12,14), as has been observed with strokes in humans (15). In accordance with this theory, homogeneous flow has been detected in acute stroke patients, and the areas displaying this behavior showed an extremely high risk of infarction (16).
Thus, the purpose of our study was to compare acute measurements of flow heterogeneity (FH) and MTT with follow-up MR data to determine which method enables better prediction of final infarct volume.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Patients who presented with symptoms of stroke of less than 12 hours duration and were referred to the Department of Neurology of Århus University Hospital were asked to participate in the study. First, transverse CT was performed in all patients to exclude hemorrhage before DW imaging and PW imaging were performed. A total of 33 consecutive patients underwent MR imaging. Seven of the 33 patients had negative DW imaging findings, which were consistent with the later diagnosis of transient ischemic attack in six patients and functional symptoms in one. A tumor caused the symptoms of one patient, one patient could not be examined owing to movement artifact during PW imaging, and one was excluded because of an operational error during PW imaging. The remaining 23 patients (13 women, 10 men; age range, 19–85 years), all of whom had a defect at DW imaging, were included in our analyses. Eighteen patients underwent MR imaging less than 6 hours after symptom onset. The time from stroke symptom onset to MR imaging ranged from 3.5 to 12.0 hours (average, 5.96 hours).
Three patients participated in an ongoing neuroprotective drug trial, which is a phase 2 study of recombinant neutrophil inhibitory factor. One of these patients was among the 10 who were excluded from our analysis (owing to operational error), whereas the other two, patients 8 and 9, participated in our study. It is unknown whether these patients received the actual drug or a placebo. After 1 month, T2-weighted MR imaging was performed to determine the final volume of the infarct.
MR Imaging Protocol
MR imaging was performed by using a 1.0-T unit (GE Signa 1.0-T; GE Medical Systems, Milwaukee, Wis). After a sagittal scout image was obtained, multisection DW imaging was performed by using a spin-echo single-shot echo-planar MR sequence and acquiring first a nonweighted image (ie, with a DW factor b of approximately 0 sec/mm2) and then three DW images in orthogonal directions (ie, x, y, and z directions). DW imaging was performed with a b of 1,000 sec/mm2.
Fourteen to 16 transverse sections were acquired to include the entire brain. The acquisition parameters used for DW imaging were as follows: 5,000/100 (repetition time msec/echo time msec), 96 x 96 matrix, 22.0 x 16.5-cm field of view, 5-mm section thickness, and 2-mm intersection gap. The total acquisition time was 20 seconds.
PW imaging was performed by using a dynamic gradient-echo echo-planar MR sequence and an MR-compatible power injector (Medrad, Pittsburgh, Pa) to inject a bolus of 0.1 mmol of gadodiamide (Omniscan; Nycomed Imaging, Oslo, Norway) per kilogram of body weight at a rate of 5 mL/sec. The bolus injection was immediately followed by an injection of an equal volume of physiologic saline, also at a rate of 5 mL/sec. Five or 10 sections of the lesion area depicted on the DW images were obtained. The sections were chosen at the same areas as those depicted on the DW images. Pulse sequences that were optimized after MR imaging in the first 12 patients facilitated 10-section acquisition. In cases in which only five PW images were obtained and the lesion extended beyond this volume, only five DW image sections (from the same locations) were used in the calculations.
Fifty single-shot dynamic gradient-echo echo-planar images were obtained in each section during the contrast agent bolus passage, and accordingly, 250 or 500 images were obtained during the 1.16-minute acquisition time. The acquisition parameters were 1,500/45, 45° flip angle, 96 x 96 matrix, 22.0 x 16.5-cm field of view, 5-mm section thickness, and 2-mm intersection gap. The total time to perform this MR imaging examination, including positioning and preparation (ie, venous catheter insertion into the cubital vein), was 30 minutes.
One month after having the stroke symptoms, the patients underwent T2-weighted follow-up MR imaging with the following parameters: 4,000/102, 256 x 256 matrix, 22-cm field of view, and 5-mm section thickness with a 2-mm intersection gap. No realignment was performed, so the follow-up images were not positioned at the exact same section positions as those used at initial acute-stage imaging. In four cases (patients 3, 6, 8, and 9), the infarct extended beyond the five sections covered at PW imaging, so only those sections acquired at T2-weighted follow-up MR imaging that corresponded to those acquired at initial acute-stage PW imaging were used.
Infarct Measurement Theory
FH is defined as the distribution of microscopic blood flow relative to the mean flow. Because microscopic flow velocities affect the delivery of oxygen to tissue, this quantity is important for assessing the metabolic derangement in ischemia. Because the passage of MR contrast agent through vessels is closely related to the passage of oxygen-carrying red blood cells, this quantity can be determined from PW MR imaging residue data. First, the tissue retention needs to be characterized in terms of the residue function, R(t), which is the fraction of tracer that is still present after the injection of an infinitely sharp contrast agent bolus into the tissue. Thus, the tissue concentration–time curve is described by the following formula (17): Ct(t) = F · Ca(t) 
R(t), where the tissue concentration of tracer, Ct(t), is directly determined by observing the bolus passage in a single pixel. (The symbol denotes the convolution operation.) The concentration of intravascular tracer can be quantified by assuming a linear relation between concentration and change in transverse relaxation rate (18). This assumption is dependent on the finding that small T1 changes due to vascular contrast agent can be ignored in the calculations (19). The arterial input function, Ca(t), is determined by choosing pixels at a location of the middle cerebral artery, where the bolus shape is high and steep and the full width at half maximum is small. R(t) times flow (F) can, with the tissue concentration–time curve equation just described, be calculated by using a deconvolution routine. Deconvolution is performed by means of singular value decomposition (20). Thus, cerebral blood flow is the initial height of the residue function. To determine the FH, the distribution of transit times, h(t), is then found as the slope of the residue function (Fig 1): h(t) = -dR/dt. (dR/dt is the derivative of R[t] with respect to the time, t.)
|
By using a nonparametric statistical test to compare the distribution of relative flow rates in the infarcted core and penumbra with the distribution in normal tissue, it is possible to quantify how much the heterogeneity of flow at infarction changes from the normal situation (Fig 1). As mentioned in the introduction, the theory behind using FH as a marker of ischemia is that homogenation of flow is considered one of the last ways to optimize oxygen extraction after MTT is prolonged.
Data Analysis
Maps of cerebral blood flow were constructed by using a noninvasively determined arterial input function and by using singular value decomposition to solve the tissue concentration–time curve formula, as described herein earlier. This approach has been shown to be robust (21), and the arterial input function was chosen in the diseased hemisphere so that the function would best reflect the actual input in the tissue of interest. Cerebral blood volume was determined by numerically integrating the area under the tissue concentration–time curve during the contrast agent bolus and thus normalizing it to the injected dose. MTT was calculated, as mentioned herein earlier, as the ratio of cerebral blood volume to cerebral blood flow. FH maps were constructed, as just described, by using the same arterial input function. The end point in the arterial input function was, in some cases, chosen differently from the end point in the MTT calculations to yield less noisy FH maps (see Discussion). Postprocessing was performed at a workstation (SUN SPARC 60; Sun Microsystems, Palo Alto, Calif).
The DW, MTT map, FH map, and T2-weighted follow-up MR images were then transferred to a personal computer and analyzed by using a commercially available software package (ALICE; Hayden Image Processing Solutions, Boulder, Colo). The initial infarct volume was determined on the DW image by means of visual inspection. The diseased area was outlined, and a volume was calculated. T2-weighted images acquired in the acute phase were evaluated to identify patients with chronic edema (ie, T2 shine through), and one patient was excluded owing to a tumor, as mentioned herein earlier. Areas with abnormal perfusion were then identified on both the MTT and the FH maps, also by means of visual inspection. These parameters were chosen to simulate how the technique probably would be used in clinical practice. Areas with abnormal perfusion were outlined, and volumes were calculated. The predicted final infarct size was measured as the union of the volume seen on the DW image and the volume seen on MTT and FH maps, respectively. The final infarct volume was determined on the T2-weighted follow-up image obtained 1 month later. Volumetric analyses of the initial and follow-up images were carried out independently at two separate times by one author (C.Z.S.).
Infarct growth was said to have occurred if the final infarcted area seen at follow-up T2-weighted MR imaging was more than 10% larger than the infarcted volume seen at initial DW imaging. Furthermore, the growth itself had to be larger than 1 cm3 to exceed the assumed uncertainty in the detection of infarcts on the nonaligned images.
The final infarct volume predicted by using the MTT and FH methods was compared with the actual final infarct volume determined at follow-up MR imaging by using the Wilcoxon signed rank test. The predicted volumes were also correlated with the follow-up volumes by using the Spearman rank correlation coefficient.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
|
|
Wilcoxon signed rank analysis of the paired data was performed to compare how well the two models predicted final infarct size. When the MTT volumes were compared with the follow-up T2-weighted MR volumes, a statistically significant difference was observed (P = .005); however, the difference between FH and T2-weighted MR volumes was not statistically significant (P = .059).
In terms of predicting growth, an MTT mismatch was observed in 10 of the 11 patients who had infarct growth: The MTT lesion was 10% larger than the DW imaging defect, and growth exceeded 1 cm3. The same 10 patients were identified with FH mapping. Five false-positive cases occurred with the MTT method—that is, the patients did not experience growth as predicted. Three false-positive cases occurred with the FH method.
In Figure 3, the estimated volumes predicted by using MTT and FH are plotted along the y axis, and the final infarct sizes measured at follow-up T2-weighted MR imaging are plotted along the x axis. Figure 4 shows the four types of images obtained in patient 22. In this patient, MTT abnormalities extended throughout the middle cerebral artery territory, whereas FH abnormalities were more localized and corresponded well with findings on the follow-up T2-weighted MR images.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
In this study, 48% of the patients had infarct growth, the infarcted volume at follow-up MR imaging exceeded that at acute-stage DW imaging by more than 10%, and in absolute terms, growth was more than 1 cm3. These results are in good agreement with those of Baird et al (22), who observed that 43% (12 of 28) of patients had an infarct growth of more than 20%. In other studies, 65% (15 of 23) of patients had a more than 15% increase (5) and 44% (seven of 16) had a more than 10% increase (6) in infarct growth.
In the patients who experienced growth in the present study, there was, on average, a 2.4-fold growth. This is a substantial growth, and such infarcted brain tissue is potentially salvageable if the group of patients is identified in time. It is noteworthy that the larger infarcts in the middle cerebral artery territory had the greatest growth. We speculate that if the growth observed after 6 hours is combined with the PW and DW imaging findings to predict growth, performing acute-stage PW imaging may expand the current time window for thrombolysis (23).
The patients excluded with this technique will avoid the potentially harmful side effects of thrombolytic treatment. This theory can be illustrated by our identification of patients with growth in the present study: At analysis of both the MTT maps and the FH maps, the same 10 of 11 patients were identified. In the one patient (patient 11) who was not correctly assigned, the infarct grew from 13 to 18 cm3, a growth of 43%; however, a growth of 3% was predicted with MTT maps and a growth of 9% was predicted with FH maps. In this patient, spatial resolution became an issue: Had the FH map shown just a few pixels more, the patient would have been in the growth group, although the measurements still would have been far from the actual final infarct size.
In terms of identifying patients in whom infarction had progressed to the end stage such that thrombolysis was contraindicated, the MTT maps yielded five false-positive cases, whereas the FH maps yielded only three. These results suggest that when the methods are equally sensitive in the identification of patients with a growing infarct, the FH method is slightly more specific for ruling out the possibility of growth. The false-positive cases typically were those of patients with very small infarcts, or they may have been those of patients with a perfusion deficit that had resolved, for example, after spontaneous reperfusion.
Linear regression analysis of the final versus predicted infarct volumes revealed a steeper slope on the MTT curve due to the larger overestimation. This is in agreement with the findings of Østergaard et al (16), who observed similar regression results in a smaller patient population. One patient (patient 3) in our study had unusual infarct behavior in that growth was much greater than anticipated. Because the occurrence of stroke is a major risk factor for the occurrence of another, this patient may have had another stroke during the 1 month follow-up period. In a study focusing on the early recurrence of stroke that involved a 30-day follow-up, the recurrence rate was 3.3% for all infarctions and 7.9% for atherothrombotic infarctions (24). Therefore, recurrent stroke might explain the large growth in patient 3 and the growing infarct that was not detected at PW imaging in patient 11.
FH and MTT had similar power in terms of assigning patients to growth or no growth categories, which is important in the selection of patients for therapy—for example, thrombolysis. Predicting the exact location and size of the final infarct, however, may be additionally important to the treating physician. If very large or important cortical areas are in danger of infarction, this could affect the diagnostic and therapeutic choices. The more precise delineation of final infarct size with FH—like that in patients 13, 18, 20, and 22 in our study—has the potential to improve the early selection of patients for invasive diagnostic or therapeutic procedures, such as hemicraniotomy for malignant middle cerebral artery infarcts (25) and intraarterial thrombolysis (9).
More precise delineation of the final infarct is also useful for the evaluation of new drugs to treat ischemic stroke (26). Parameters like survival, final infarct, and neurologic outcome score are currently being used as end points in the evaluation of these drugs, which typically requires the inclusion of a large patient cohort. The better this pseudo end point, the smaller the number of patients who will be needed in the trials; phase 2 studies especially would benefit from this improvement. Saver et al (26) observed a correlation between infarct size at CT and clinical outcome and suggested that MR findings may be more sensitive. Studies (27,28) have shown good correlation between the clinical outcome evaluated with different stroke scales and the MR findings and thus prompted hope for the idea of using MR findings as a pseudo end point in drug trials.
PW imaging seems to have additional power in terms of providing more efficient pseudo end points of disease. By applying probabilistic models to initial PW and DW imaging and combining these studies with follow-up data on a pixel-by-pixel basis, models to assess risk of infarction can be formed (29). These models have immense statistical power in that they are sensitive to even small changes in infarct probabilities in subsequently treated patient cohorts. FH mapping, by improving predictions of infarct size, may improve these models. Thus, drug effectiveness may be established and quantified in smaller patient populations.
In our study, we tried to use FH maps derived from PW imaging data to predict final infarct size. In other studies, the apparent diffusion coefficient has been examined to find a threshold between tissue that proceeds to survival and that which proceeds to infarction (30). Furthermore, areas with a low apparent diffusion coefficient have been observed to survive after early thrombolysis (31). Future models will probably include many parameters, such as perfusion and diffusion data, infarct age, initial infarct size, patient age, and body temperature, and incorporate these into a statistical model to predict final infarct size and clinical outcome.
In our experience, FH maps are reproducible and can be easily constructed. The arterial input function is needed, but it is also needed to calculate other parameters in PW imaging (ie, cerebral blood flood and volume and MTT). The time required for the calculations is less than a minute. In our study, these maps were shown to be somewhat sensitive to the choice of end point of the arterial input function, that is, the time during which the deconvolution routine is performed. The end point could be reproducibly identified on the basis of the appearance of a continuous lesion with clear boundaries as opposed to that of noise over the entire section. Here, it must be mentioned that when calculating FH, information is extracted from data that are inherently noisy. It is very important to have an optimal signal from the imaging unit. Our study was performed with a 1.0-T system, and the maps would have had better signal-to-noise ratios with the more commonly used 1.5-T system or even a 3.0-T system.
In this study, the overestimation on MTT maps was largely due to a few patients in whom the penumbra zone size was overestimated substantially. This finding may indicate that FH maps provide more information about the fate of the tissue: Prolonged MTT is a reflection of decreased perfusion pressure (11) and thus may not be—either theoretically or practically—the best measure of final infarct volume.
Although MTT is the first moment of the distribution of transit times, and hence a measure of how long red blood cells or contrast agent spends in the circulation, FH is a measurement of the distribution of probabilities that flow is in different ranges. This quantity is a measure of the probability that neighboring vessels have the same transit time and thereby of how efficient oxygen gradients are in the tissue. We suggest that determining FH may be an important step in determining oxygen metabolism with MR imaging. This is currently an area of intense research, validation, and analysis. However, the seemingly high predictive power of FH maps justifies the evaluation of this PW imaging methodology in further research.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, C.Z.S., L.Ø.; study concepts, C.Z.S., L.Ø., G.A.; study design, C.Z.S., L.Ø., G.A., C.G., P.V.P.; literature research, L.R., C.Z.S.; clinical studies, G.A., C.G., L.R., C.Z.S.; data acquisition, L.R., C.Z.S., P.V.P.; data analysis/interpretation, C.Z.S.; statistical analysis, C.Z.S., L.Ø., P.V.P.; manuscript preparation and definition of intellectual content, C.Z.S., L.Ø.; manuscript editing, revision/review, and final version approval, all authors.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  F CHEN, Q LIU, H WANG, Y SUZUKI, N NAGAI, J YU, G MARCHAL, and Y NI Comparing two methods for assessment of perfusion-diffusion mismatch in a rodent model of ischaemic stroke: a pilot study Br. J. Radiol., March 1, 2008; 81(963): 192 - 198. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C.S. Rivers, J.M. Wardlaw, P.A. Armitage, M.E. Bastin, T.K. Carpenter, V. Cvoro, P.J. Hand, and M.S. Dennis Do Acute Diffusion- and Perfusion-Weighted MRI Lesions Identify Final Infarct Volume in Ischemic Stroke? Stroke, January 1, 2006; 37(1): 98 - 104. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Perkio, L. Soinne, L. Ostergaard, J. Helenius, A. Kangasmaki, S. Martinkauppi, O. Salonen, S. Savolainen, M. Kaste, T. Tatlisumak, et al. Abnormal Intravoxel Cerebral Blood Flow Heterogeneity in Human Ischemic Stroke Determined by Dynamic Susceptibility Contrast Magnetic Resonance Imaging Stroke, January 1, 2005; 36(1): 44 - 49. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Warach Stroke Neuroimaging Stroke, February 1, 2003; 34(2): 345 - 347. [Full Text] [PDF]
|
|
|
|
|
|
M. H. Lev, W. J. Koroshetz, L. H. Schwamm, R. G. Gonzalez, and T. Tatlisumak CT or MRI for Imaging Patients with Acute Stroke: Visualization of "Tissue at Risk"? Stroke, December 1, 2002; 33(12): 2736 - 2737. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
