Diagnostic performance of PWI/DWI MRI parameters in discriminating hyperacute versus acute ischaemic...

Author's personal copy

Diagnostic performance of PWI/DWI MRI parameters indiscriminating hyperacute versus acute ischaemic stroke:Finding the best thresholdsE. Roldan-Valadez*, O. Gonzalez-Gutierrez, M. Martinez-LopezMagnetic Resonance Unit, Medica Sur Hospital & Clinic Foundation, Mexico City, Mexico

article information

Article history:Received 10 February 2011Received in revised form21 August 2011Accepted 25 August 2011

AIM: To determine the accuracy of the perfusion/diffusion-weighted imaging (PWI/DWI)parameters [time to peak (TTP), mean time to peak (MTT), relative cerebral blood volume(rCBV), and relative cerebral blood flow (rCBF) maps]; in the evaluation of acute versushyperacute ischaemic stroke.MATERIALS AND METHODS: Fifty-five patients with symptomatic hyperacute (first 6 h) or

acute (7e24 h) ischaemic stroke underwent diffusion and perfusion evaluation. Statisticalanalysis included Student’s t-test, receiver operating characteristics (ROC) analysis of apparentdiffusion coefficient (ADC), TTP, MTT, CBV, and CBF; correlation, linear, and logistic regressionanalysis.RESULTS: Area under receiver operating characteristics (AUROC) analysis identified the ADC

cut-off value 385! 10"6 mm2/s, MTT at 109.5%, TTP at 3.05 s, CBV at 129%, and CBF at 98.5%(the record of the time of onset was considered the reference standard). The best performancecorresponded to TTP, which showed a sensitivity of 0.94 and specificity of 0.88 (p< 0.001).CONCLUSIONS: Based on the present findings, hyperacute penumbra is reliably defined with

a TTP >3 s with no visible changes in diffusion. ADC, rCBF, and rCBV are not useful fordiscriminating between acute and hyperacute ischaemic stroke.

! 2011 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.

Introduction

Since 19951,2 the concept of the therapeutic window inthe early treatment of hyperacute cerebral ischaemia hasbeen established from 0e3 h for intravenous therapy and4e6 h for intra-arterial thrombolytic therapy.3,4 Severalstudies have shown that intravenous thrombolysis can besafe and effective up to 6 h5 or 9 h.6,7

Although dynamic susceptibility perfusion-weightedimaging (PWI) and diffusion-weighted imaging (DWI)quantitative maps have become routine examinations foracute stroke patients in many stroke centres,8 there is noconsensus about the best thresholds of the PWI/DWImismatch9 that should be used to identify the ischaemicallythreatened but viable area.10,11

DWI delineates regions of bio-energetic compromise,where sodiumepotassium adenosine triphosphatase(ATPase) activity has failed. Protons that are not moving(restricted diffusion/cytotoxic oedema) are associated witha higher signal on DWI, and the apparent diffusion coefficient(ADC) is its quantitative measure. In areas of more profoundenergetic compromise and sequestration of water moleculesADC values are lower.3

* Guarantor and correspondent: E. Roldan-Valadez, Magnetic ResonanceUnit, Medica Sur Hospital & Clinic Foundation, Puente de Piedra 150, TorielloGuerra, Tlalpan, CP 14050, Mexico City, Mexico. Tel.: þ52 55 5666 1765;fax: þ52 55 5424 4429.

E-mail address: [email protected] (E. Roldan-Valadez).

Contents lists available at SciVerse ScienceDirect

Clinical Radiology

journal homepage: www.cl in icalradiologyonl ine.net

0009-9260/$ e see front matter ! 2011 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.doi:10.1016/j.crad.2011.08.020

Clinical Radiology 67 (2012) 250e257


PWI images are acquired using the dynamic-susceptibilitycontrast imaging technique. A series of susceptibility-weighted images (known as T2*) are obtained every 1e2 sduring an injection of intravenous gadolinium contrastmedium. As the contrast medium transits in the cerebralcirculation, the T2* signal intensity of the images successivelydecreases due to the paramagnetic nature of the contrastmedium, and then returns to normal. This change in signalintensity is plotted as a function of time (the signal intensi-tyetime curve). The central volume principle is used tocalculate the relative cerebral blood flow (rCBF) and relativecerebral blood volume (rCBV) on a voxel-wise basis; rCBF isproportional to the amplitude of the signal intensityetimecurve, while rCBV is estimated from the area under the signalintensityetime curve.12 PWI allows the visualization of areasof altered blood flow.13 Areas of hypoperfusion can also bevisualized as the bolus arrival time or tissue with delayedtime to peak (TTP), the first moment of the curve corre-sponding to the apparent mean transit time (MTT). Voxel-wise values for each parameter are assigned a colour codeor intensity value, and maps are then constructed for therCBF, rCBV, TTP, and MTT.12,14 Such parameters are usefulespecially when the onset of stroke cannot be determinedexactly from medical records.

The aim of the present study was to establish the bestthresholds using the TTP, MTT rCBV, and rCBF maps thatenable the classification of a patient undergoing a hyper-acute-phase or acute-phase ischaemic stroke to investi-gate whether one of these parameters can be used to assessthe risks and benefits of therapeutic interventions.

Materials and methods

Patients

Patients with symptomatic acute ischaemic strokeunderwent PWI/DWI magnetic resonance imaging (MRI)evaluation. Based on the clinical history, the time of onset ofstroke was recorded. Patients were classified into one of twogroups: hyperacute strokewhen assessment occurredwithin6 h of the onset of stroke and acute stroke when symptomswere evident within 7e24 h of the assessment. Patients witha longer duration of symptoms or intracranial haemorrhageand also patients who woke up with stroke symptoms(unclear time of onset) were excluded.

MRI evaluation of ischaemic stroke

Conventional evaluation for stroke was performed usinga 3 T HDxt and 1.5 T HDx GE Signa MRI system (GeneralElectric Healthcare, Milwaukee, WI, USA). Images wereacquired with an eight-channel high-resolution brain arraycoil. The standard clinical sequences consisted of a sagittal,T1-weighted, fluid-attenuated inversion recovery (FLAIR);coronal, T2-weighted, fast spin-echo (FSE); axial, T2-weighted FLAIR; axial, diffusion-weighted, echo planarperfusion; three-dimensional (3D) time-of-flight; contrast-enhanced; axial spoiled gradient (SPGR) T1-weighted. Post-gadolinium T1 images used Gadovist" (Bayer Schering

Pharma, dose 0.1 mmol/kg). The average time of acquisitionwas 15 min at 1.5 T and 12 min at 3 T.

Diffusion and perfusion analysis

Post-processing of rawdiffusion andperfusion imageswasperformed by a single investigator using the softwarepackage FuncTool Performance release 4.3 (GE MedicalSystems). This software was used to plot the change in MRItransverse relaxivity,which is not linearly related toGd-DTPAconcentration, on a per-voxel basis over time. During theperfusionpost-processing, amaskof thediffusion imageswassuperposed over the TTP colour maps, outlining the infarctregions on the diffusion and hypoperfused tissue on TTPcolour maps, so that the penumbra tissue could be discrimi-nated from the infarct and normal brain tissue. Then regionsof interest (ROIs) within the penumbra and infarct allowedthe semiquantitative evaluation of perfusion maps includingTTP, MTT, rCBV, and rCBF.

Statistical analysis

Analyses were performed using SPSS software version17.0 (SPSS, Chicago, IL, USA) and MedCalc" version 11.6.0.(MedCalc Software, Broekstraat 52, 9030 Mariakerke,Belgium). Statistical significance was indicated by p< 0.05(two-tailed). Presentation of data was done according to theguidelines of the American Psychological Association.15,16

Variables were tested for normal distribution and homo-geneity of variance. Means for diffusion and perfussionmarkers were compared using a t-test for independentsamples, values of Cohen’s d and the effect-size r werecalculated17; we defined effect sizes as small (d¼ 0.2),medium (d¼ 0.5), and large (d¼ 0.8).18

The degree of associationwas assessed using the Pearsonor Spearman correlation coefficients; rCBV and rCBF werecompared with TTP; r-values close to or above 0.75 signifiedgood to excellent agreement.

Before linear regression (LR) analysis the non-normallydistributed data of dependent variables (rCBV and rCBF)were given a log transformation to achieve normal distri-bution (LogCBV and LogCBF, respectively). LR analyses wereperformed to assess significance of models to quantify rCBVand rCBF using TTP as an independent variable (individualp-values were obtained for each group, acute and hyper-acute; and, the group as a whole was also examined). Ageneral lineal model (GLM) was performed to evaluatesignificant interactions between TTP and the phase of onset.

ROC curves were used to evaluate the diagnostic perfor-mance of ADC, MTT, TTP, rCBV, and rCBF as diagnostic tests.19

The area under the ROC curve (AUROC) was considered themeasure of the overall performance of each variable20,21;standard error, p-values, and 95% confidence intervals (CIs)were also obtained. The recorded time of onset of stroke wasconsidered the reference standard for the ROC analysis. Cut-off ranges were calculated and represent the optimal cut-offto maximize sensitivity plus specificity for assessing thepresence of the hyperacute stage. For MTT, rCBV, and rCBF,cut-off ranges are expressed in percent units. For ADC, cut-off

E. Roldan-Valadez et al. / Clinical Radiology 67 (2012) 250e257 251


ranges are expressed in square millimetres per second, andfor TTP, cut-off ranges are expressed in seconds. The accuracyof each marker was determined using a traditionalacademic point-system classification: 0.90e1¼ excellent (A),0.80e0.90¼ good (B), 0.70e0.80¼ fair (C), 0.60e0.70¼ poor(D), and 0.50e0.60¼ fail (F).22 Values of sensitivity, speci-ficity, positive predictive value, negative predictive value, andpositive and negative likelihood ratios were also determined.Statistical significance for the AUROC assessment was set atp< 0.05.

Based on the results of the diagnostic performance of thePWI and DWI parameters, variables with a significant effectupon the distinction between hyperacute and acute strokeswere examined using binary logistic regression to find outwhether they would give a better prediction than a singlevariable alone.23

Results

Demographic and clinical data

Fifty-five patients were included in this analysis, 25 males(mean age% SD 54.28% 6.14 years; range 18e87 years) and30 females (72.70%18.41 years; range 33e95 years). Allpatients had experienced anterior circulation corticalsyndromes. Regarding the onset of the event, for males therewere 42.9% hyperacute and 57.1% acute strokes; for females54.5% hyperacute and 45.5% acute ischaemic strokes.

PWI/DWI markers of ischaemia

There was a non-significant difference in the ADC valuesbetween phases. The TTP depicted a higher value for thehyperacute phase with a large effect size. The MTT alsoshowed a significant difference between acute and hyper-acute phases with a moderate effect size. rCBV hada decreased value in the hyperacute phase with a non-significant higher value in the acute phase, with a smalleffect size. The rCBF showed a similar behaviour to rCBV forthe acute and hyperacute phases, but it was significant andwith a medium effect size (Table 1).

Diagnostic performance of DWI and PWI

A dichotomized time of onset (hyperacute or acute) wasconsidered the reference standard. AUROC analysis identi-fied the cut-off value of ADC at 385!10"6 mm2/s, MTT at109.5%, TTP at 3.05 s, rCBV at 129%, and rCBF at 98.5%. Fig 1

depicts graphs of the five AUROCs. Table 2 shows the valuesof the AUROCs and confidence intervals.

The highest performance corresponded to TTP(p< 0.001), an excellent fit with an AUROC of 0.917, sensi-tivity of 0.94, specificity of 0.88, and accuracy of 0.91,p< 0.001. Table 3 lists the diagnostic performance ofthe PWI/DWI markers. Figs 2 and 3 show examples of theDWI/PWI marker values after post-processing (Fig 4).

Relationship between rCBV-TTP and rCBF-TTP estimatedby LR analysis

There was a significant negative Spearman correlationbetween the TTP and the rCBV; rs¼"0.289, p¼ 0.033 andbetween TTP and rCBF, r¼"0.522, p< 0.001. The normalitytests of the log transformation to base 10 of the dependentvariables before LR analysis showed normal distribution(LogCBV: Z¼ 0.626, p¼ 0.829; LogCBF: Z¼ 0.707, p¼ 0.700;KolmogoroveSmirnov test), requisite for the simple LRanalysis. The LogCBV-TTP regression model was non-significant in the hyperacute phase (p¼ 0.827) and nearlysignificant in the acute onset group (p¼ 0.095); the rela-tionship was also examined for the group as a whole witha non-significant relationship (p¼ 0.068). In the LogCBF-TTP regression model, a non-significant value was alsoobserved in the hyperacute phase (p¼ 0.140), which wassignificant in the acute phase (p¼ 0.028); the analysis forthe group as a whole was significant (p¼ 0.001).

No significant interaction between TTP and the phase-of-onset (p¼ 0.500) was found using the general lineal model(GLM). The significant equations for estimating rCBF andrCBV using TTP within the acute stage were:rCBF¼ 10("0.058 "0.527 TTP) and rCBV¼ 10("0.056 "0.434 TTP);fitted lines met the assumptions of regression (normallydistributed residuals p¼ 0.875 and p¼ 0.461 from rCBV andrCBF, respectively). Fig 2 illustrates the curve fit for theLogCBV and LogCBF (for each group, acute and hyperacute;and for the group as a whole).

PWI model to predict time of onset in ischaemic stroke

Preliminary analyses before logistic regression ensured noviolation of the assumptions of normality, linearity, multi-collinearity, and homoscedasticity among the variables. Thefull model containing all predictors (TTP, MTT, LogCBV,LogCBF) was statistically significant, c2 (4, n¼ 55) 27.356,p< 0.001; indicating that the model was able to distinguishbetween hyperacute- and acute-phase in ischaemic stroke.

Table 1Results of apparent diffusion coefficient (ADC), time to peak (TTP), mean time to peak (MTT), relative cerebral blood volume (rCBV), and relative cerebral bloodflow (rCBF).

PWI/DWIanalysis

Hyperacute ischaemicstroke (0e6 h)

Acute ischaemic stroke (7e24 h) t-Statistic p-Value Cohen’s d Effect-size r

Mean SE Mean SE

ADC (mm2/s) 589.06! 10"6 32.143! 10"6 518.11! 10"6 36.041! 10"6 e1.474 .146 0.394 0.193MTT (%) 118.483 2.183 108.916 2.596 e2.837 0.006 0.779 0.363TTP (s) 8.116 0.770 2.175 0.355 e7.005 <0.001 1.924 0.693rCBV (%) 97.129 14.382 141.291 28.614 1.473 0.147 0.405 0.198rCBF (%) 79.516 5.468 133.375 21.026 2.479 0.020 0.681 0.322

E. Roldan-Valadez et al. / Clinical Radiology 67 (2012) 250e257252


The model as a whole explained between 39.2% (Cox andSnell R square) and 52.3% (Nagelkerke R squared) of thevariance inhyperacute phase, and correctly classified78.2% ofcases. As shown in Table 4, only one of the independentvariables made a unique statistically significant contributionto the model. TTP was the strongest predictor in reportinga hyperacute phase, with an odds ratio of 1.88. This valueindicated that, for every additional second increase in the TTPvalue, patients were over 1.88 times more likely to reporta hyperacute stage than those who did not have an increase,controlling for all other factors in the model.

Discussion

In this study, the diagnostic tests ADC, MTT, TTP, rCBVand rCBF were performed in patients with ischaemic stroketo indentify which biological marker best classifies patientsduring the therapeutic window (hyperacute versus acute),such analysis is scarce in imaging reports.

The recommended cut-off value for ADC was385!10"6 mm2/s, which was lower than the mean for

hyperacute and acute stroke (589.06!10"6 mm2/s versus518.11!10"6 mm2/s). It had the worst performance andwas not able to discriminate between hyperacute and acuteoligemic brain tissue.

Interesting data were found regarding the TTP analysis,which showed that a lower threshold might have a highersensitivity than previous published thresholds. With a cut-off value of 3 s the sensitivity of the test was 94% and thepositive predictive was 91%; meaning that the test coulddetect 91 out of 100 patients who could benefit fromrecanalization treatment. The effect size for the TTP was1.92 (Table 1) indicating that the mean of the hyperacutegroupwas at the 97.1 percentile of the acute group. This alsoindicated a no overlap of at least 79.4% in the two distri-butions of patients, which can be helpful in selecting whichpatients are still within the 6 h time-window. AlthoughMTT and CBV were significant in their evaluations as diag-nostic markers, their sensitivity and PPV were inferior tothose of TTP (Table 3). Some studies have reported betterperformance for theMTTmarker (88% good rating of MTTascompared to 68% of TTP).24 Butcher et al.25 found a delay inMTT in the penumbra area of 6.5% 0.4 s (range 2.2e11.7 s)in hyperacute-phase ischaemic stroke, whose inferior rangewas lower than the present results (Table 1). Other studieshave also shown that a TTP prolongation of >4 s correlateswith significant hypoperfusion (<20 ml/100 g/min).14,26

The results of the present study are in agreement withdiagnostic centres, which are already using TTP perfusion/ADC diffusion mismatch to evaluate stroke.27,28

The data from the present study depicted that a significantdecline in rCBV and rCBF appears only after the 6 h time-window (Fig 2b and d), explaining why a definition of abso-lute perfusion thresholds might be problematic.25 Thisdelayed decline is due to the compensatory vasodilatationthat occurs until the collateral circulation becomes

Figure 1 (a and b) Schematic illustration of a comparison of the sensitivities of the five ROC curves (ADC, MTT, TTP, rCBV, and rCBF) withdifferent areas under the curve. TTP had an AUROC of 0.917, which was the best marker. The chance diagonal (the line segment from 0.0 to 1.0)had an AUROC of 0.5. Complete AUROCs analyses are shown in Table 2.

Table 2Analysis of the area under the receiver operating characteristic curve(AUROC).

Diagnostic test AUROC SE 95% CI p-Value

Time to peak 0.917 0.041 0.837e0.997 <0.001Cerebral blood flow 0.745 0.072 0.604e0.886 0.002Mean transit time 0.683 0.075 0.537e0.830 0.021Cerebral blood volume 0.628 0.077 0.477e0.780 0.105Apparent diffusion coefficient 0.608 0.078 0.456e0.760 0.172

Taking the time of onset of stroke as the reference standard, the largest areacorresponded to time to peak, followed by mean transit time and cerebralblood flow; apparent diffusion coefficient, and cerebral blood volume did notshow significance.



insufficient.29 MRI evidence of ischaemia depends on whichvariable is chosen and its threshold to diagnose treatableischaemia.

Changes in rCBV and rCBF should not be used to recruitpatients in the hyperacute phase but to follow-up theprogression of the infarct in the first 7e24 h. Previousstudies30,31 havedescribed that ischaemia can appearwith aninitial negative DWI (5e25% of cases), which supports the

need to always perform PWI in suspected ischaemic stroke.The lack of interaction between TTP and phase-of-onset allows the behaviour of rCBF and rCBV in each phase tobe considered a confident reflection of the behaviour of thedata.

This study has a number of limitations. There was nocorrelation of the imaging evaluationwith the clinical follow-up, neither a volumetric estimation of the ischaemic or the

Table 3Evaluation of diagnostic tests.

Diagnostic Test TTP MTT rCBV rCBF ADC

Value CI Value CI Value CI Value CI Value CI

Sensitivity 93.5 78.5e99.2 82.7 64.2e94.1 86.2 68.3e96.1 75.8 56.4e89.7 24.1 10.3e43.5Specificity 87.5 67.6e97.3 61.5 40.5e79.7 38.4 20.2e59.4 46.1 26.5e66.6 72.4 52.7e87.2Positive predictive value 90.6 74.9e98.0 70.5 52.5e84.9 60.9 44.5e75.8 61.1 43.4e76.8 46.6 21.2e73.4Negative predictive value 91.3 71.9e98.9 76.1 52.8e91.7 71.4 41.9e91.6 63.1 38.3e83.7 48.8 33.3e64.5Positive likelihood ratio 7.4 2.5e21.6 2.1 1.2e3.6 1.40 1.00e1.96 1.41 0.93e2.12 0.87 0.37e2.1Negative likelihood ratio 0.07 0.02e0.28 0.28 0.12e0.66 0.36 0.13e1.01 0.52 0.24e1.13 1.05 0.7e1.4

With a cut-off value of 3.05 s, the best accuracy was achieved by time to peak (TTP) with 91%, followed by mean time to peak (MTT) and relative cerebral bloodvolume (rCBV); relative cerebral blood flow (rCBF), and apparent diffusion coefficient (ADC) were not significant.

Figure 2 Scatterplots with curve estimations (linear regression) between TTP-LogCBV and TTP-LogCBF. (a) and (c) Fitted lines for each group asa whole (LogCBV and LogCBF). (b) and (d) Fitted lines obtained for each phase, acute and hyperacute.



penumbra area because those were not aims of the presentstudy. Sequential follow-ups, recording the “hours” of theevent, and the different thrombolytic agents are necessary,considering previous observations of persistence of a PWI/DWI mismatch even 24 h after the onset of stroke.32e36

Despite previous reports about perfusion, there are stillsome unanswered questions; for example, in the predictionof haemorrhage during ischaemic stroke.37 The present studyprovides thresholds for the five markers used in DWI andPWI, which might help radiologists and clinicians performnot only an objective evaluation but also a quantitativeanalysis of the ischaemic penumbra so treatment can be

tailored for each patient. It is possible to build a significantpredictive model to diagnose patients within the therapeuticwindow by using the four parameters of PWI. It would bedesirable that stroke centres around the world could havetheir own predictive models, which are tailored for specificpopulations. In clinical practice, perfusion post-processingshould be done prior to diffusion analysis to first find thehypoperfused area and then overlap it with DWI to seewhether it matches with the infarcted tissue.

In conclusion, the evaluationof ischaemic stroke shouldbedone with PWI/DWI MRI; the parameter with best perfor-mance is TTP, followedbyMTT. ATTP cut-off valueof 3 s in the

Figure 3 Images from a patient with hyperacute infarct. (a) DWI shows the area of infarcted tissue involving the right corona radiata and body ofthe ipsilateral caudate (arrow head); (b) TTP map showed 2 s delay in the infarcted tissue (arrow head) and 6 s in the penumbra area (whiteasterisk; blue area); (c) rCBV map showing diminished volume values in the penumbra area (58%, white asterisk); (d) perfusion analysis showingthe typical shapes in normal (red curve) versus infarcted and penumbra tissues (green curves).



penumbra area has a positive predictive value of 91% todiscriminate between hyperacute and acute-phase ischaemicstroke. In a logistic-regressionpredictivemodel, the TTP is theonly value to significantly increase the odds of correctlyreporting a hyperacute stage stroke. MTT, rCBF, and rCBVthresholds are not useful parameters in this discrimination.Inclusionof patientsduring the therapeuticwindowmight bebased on TTP thresholds only, which would be useful espe-cially when the time of onset of stroke is unknown.

References

1. Tissue plasminogen activator for acute ischemic stroke The nationalinstitute of neurological disorders and stroke rt-PA stroke study group.N Engl J Med 1995;333:1581e7.

2. Hacke W, Kaste M, Fieschi C, et al. Intravenous thrombolysis withrecombinant tissue plasminogen activator for acute hemispheric stroke.The European cooperative acute stroke study (ECASS). JAMA 1995;274:1017e25.

3. Saini M, Butcher K. Advanced imaging in acute stroke manage-mentdpart II: magnetic resonance imaging. Neurol India 2009;57:550e8.

4. Sorensen AG, Copen WA, Ostergaard L, et al. Hyperacute stroke:simultaneous measurement of relative cerebral blood volume, relativecerebral blood flow, and mean tissue transit time. Radiology 1999;210:519e27.

5. Ribo M, Molina CA, Rovira A, et al. Safety and efficacy of intravenoustissue plasminogen activator stroke treatment in the 3- to 6-hourwindow using multimodal transcranial Doppler/MRI selection protocol.Stroke 2005;36:602e6.

6. Furlan AJ, Eyding D, Albers GW, et al. Dose escalation of desmoteplasefor acute ischemic stroke (DEDAS): evidence of safety and efficacy 3 to 9hours after stroke onset. Stroke 2006;37:1227e31.

7. Hacke W, Albers G, Al-Rawi Y, et al. The desmoteplase in acute ischemicstroke trial (DIAS): a phase II MRI-based 9-hour window acute strokethrombolysis trial with intravenous desmoteplase. Stroke 2005;36:66e73.

8. Lee DH, Kang DW, Ahn JS, et al. Imaging of the ischemic penumbra inacute stroke. Korean J Radiol 2005;6:64e74.

9. Butcher KS, Parsons M, MacGregor L, et al. Refining the perfu-sionediffusion mismatch hypothesis. Stroke 2005;36:1153e9.

Figure 4 Same patient from Fig 3. (a) Initial DWI shows a focal area of hyperacute infarct (white arrowhead) together with a superposed outlinerepresenting the penumbra area. (b) Post-processed ADC map showed restricted diffusion (value of 54%, white box) within the hyperacuteinfarct. (c) and (d) DWI and ADC maps from the follow-up study (24 h after onset) show the progression of the infarct area with its corre-sponding ADC values. Small green and pink ROIs show the decreased ADC values within the previous areas of penumbra and infarct.

Table 4Logistic regression model predicting likelihood of hyperacute phase in ischaemic stroke.

Variables B SE Wald df p-Value Odds ratio 95% CI for odds ratio

Lower Upper

TTP 0.634 0.206 9.493 1 0.002 1.885 1.259 2.821MTT e0.058 0.035 2.710 1 0.100 0.944 0.880 1.011LogCBV e2.851 2.355 1.466 1 0.226 0.058 0.001 5.840LogCBF 2.279 3.492 0.426 1 0.514 9.771 0.010 9163.185Constant 4.898 6.035 0.659 1 0.417 134.046

TTP, time to peak; MTT, mean time to peak; CBV, cerebral blood volume; CBF, cerebral blood flow.



10. Karonen JO, Liu Y, Vanninen RL, et al. Combined perfusion- and diffusion-weighted MR imaging in acute ischemic stroke during the 1st week:a longitudinal study. Radiology 2000;217:886e94.

11. Neumann-Haefelin T, Wittsack HJ, Wenserski F, et al. Diffusion- andperfusion-weighted MRI. The DWI/PWI mismatch region in acute stroke.Stroke 1999;30:1591e7.

12. Ostergaard L. Principles of cerebral perfusion imaging by bolus tracking.J Magn Reson Imaging 2005;22:710e7.

13. Albers GW, Thijs VN, Wechsler L, et al. Magnetic resonance imagingprofiles predict clinical response to early reperfusion: the diffusion andperfusion imaging evaluation for understanding stroke evolution(DEFUSE) study. Ann Neurol 2006;60:508e17.

14. Sobesky J, Zaro Weber O, Lehnhardt FG, et al. Which time-to-peakthreshold best identifies penumbral flow? A comparison of perfusion-weighted magnetic resonance imaging and positron emission tomog-raphy in acute ischemic stroke. Stroke 2004;35:2843e7.

15. Wilkinson L. Statistical methods in psychology journals. task force onstatistical inference. APA board of scientific affairs. American Psychologist1999;54:594e604.

16. APA. Publication manual of the American psychological association. 6th ed.American Psychological Association; Washington, DC, 2009.

17. Becker LA. Effect size, http://www.uccs.edu/wfaculty/lbecker/es.htm;2000 [accessed 30.01.11] Nunnally.

18. Cohen J. Statistical power analysis for the behavioral sciences. 2nd ed.Hillsdale, NJ: Lawrence Erlbaum Associates; 1988.

19. Park SH, Goo JM, Jo CH. Receiver operating characteristic (ROC) curve:practical review for radiologists. Korean J Radiol 2004;5:11e8.

20. Hanley JA, McNeil BJ. The meaning and use of the area under a receiveroperating characteristic (ROC) curve. Radiology 1982;143:29e36.

21. Obuchowski NA. Receiver operating characteristic curves and their usein radiology. Radiology 2003;229:3e8.

22. Metz CE. Basic principles of ROC analysis. Semin Nucl Med 1978;8:283e98.

23. Chan YH. Biostatistics 202: logistic regression analysis. Singapore Med J2004;45:149e53.

24. Kim J, Leira EC, Callison RC, et al. Toward fully automated processing ofdynamic susceptibility contrast perfusion MRI for acute ischemic cere-bral stroke. Comput Methods Programs Biomed 2010;98:204e13.

25. Butcher K, Parsons M, Baird T, et al. Perfusion thresholds in acute strokethrombolysis. Stroke 2003;34:2159e64.

26. Heiss WD, Sobesky J, Hesselmann V. Identifying thresholds forpenumbra and irreversible tissue damage. Stroke 2004;35:2671e4.

27. Campbell BC, Christensen S, Foster SJ, et al. Visual assessment of per-fusionediffusion mismatch is inadequate to select patients for throm-bolysis. Cerebrovasc Dis 2010;29:592e6.

28. Lee SY, Cha JK, Kang MJ. Regional cerebral blood volume ratio onperfusion MRI on the growth of infarct size in acute ischemic stroke. EurNeurol 2009;62:281e6.

29. Maeda M, Itoh S, Ide H, et al. Acute stroke in cats: comparison ofdynamic susceptibility-contrast MR imaging with T2- and diffusion-weighted MR imaging. Radiology 1993;189:227e32.

30. Sylaja PN, Coutts SB, Krol A, et al. When to expect negativediffusion-weighted images in stroke and transient ischemic attack. Stroke2008;39:1898e900.

31. Engelter ST, Wetzel SG, Bonati LH, et al. The clinical significance ofdiffusion-weighted MR imaging in stroke and TIA patients. Swiss MedWkly 2008;138:729e40.

32. Kidwell CS, Saver JL, Mattiello J, et al. Thrombolytic reversal of acutehuman cerebral ischemic injury shown by diffusion/perfusion magneticresonance imaging. Ann Neurol 2000;47:462e9.

33. Schellinger PD, Jansen O, Fiebach JB, et al. Monitoring intravenousrecombinant tissue plasminogen activator thrombolysis for acuteischemic stroke with diffusion and perfusion MRI. Stroke 2000;31:1318e28.

34. Schellinger PD, Fiebach JB, Jansen O, et al. Stroke magnetic resonanceimaging within 6 hours after onset of hyperacute cerebral ischemia.Ann Neurol 2001;49:460e9.

35. Jansen O, Schellinger P, Fiebach J, et al. Early recanalisation in acuteischaemic stroke saves tissue at risk defined by MRI. Lancet 1999;353:2036e7.

36. Marks MP, Tong DC, Beaulieu C, et al. Evaluation of early reperfusion andi.v. tPA therapy using diffusion- and perfusion-weighted MRI. Neurology1999;52:1792e8.

37. Kassner A, Roberts T, Taylor K, et al. Prediction of hemorrhage in acuteischemic stroke using permeability MR imaging. AJNR Am J Neuroradiol2005;26:2213e7.


Diagnostic performance of PWI/DWI MRI parameters in discriminating hyperacute versus acute ischaemic...

Documents

Transcript of Diagnostic performance of PWI/DWI MRI parameters in discriminating hyperacute versus acute ischaemic...