42321078

8/13/2019 42321078

1/9

8/13/2019 42321078

2/9

Perfusion CT Improves DiagnosticAccuracy for Hyperacute Ischemic Stroke

Cerebrovasc Dis 2009;28:7279 73

the clinical status of the patient and the resources of anurgent care setting, MRI may not always be suitable oravailable. This has prompted continued interest in devel-oping CT-based modalities for triage given their generaladvantages of universal availability, speed, and conve-nience. In the current 3-hour therapeutic window for use

of intravenous tissue plasminogen activator (tPA), NCCTis performed primarily to exclude intracerebral hemor-rhage and ischemic stroke mimics. There have been sev-eral reports of improved accuracy using dynamic CT per-fusion (CTP) for hemispheric infarcts in the first 12 h [5,6] and 6 h [7, 8] after symptom onset. However, the use ofintravenous tPA outside the 3-hour window remains acontraindication to date. Our goal was to determine thediagnostic accuracy of CTP for hyperacute infarction in aretrospective cohort of patients presenting with strokesymptoms in the first 3 h and to compare this accuracy toNCCT, using follow-up DWI as the reference standard.

Methods

PatientsImaging and clinical data obtained as part of standard clinical

stroke care at our institution were retrospectively reviewed. Con-sent was waived for review of all images and charts according tothe guidelines of the Institutional Board of Research Associates.From January 2004 to May 2008, 151 consecutive patients pre-sented to our institutions emergency department within 3 h ofonset of symptoms that were suggestive of an acute ischemicstroke and received a CT stroke series consisting of NCCT, CTP,

and CT angiogram (CTA) as part of routine clinical care. Subsetsof these patients had been included in previously published stud-ies [911]. Patients were initially included for this analysis if a fol-low-up MRI of the brain with DWI was performed !7 days fromictus. This primary selection criterion included 109 patients.

Image AcquisitionsAll CT examinations were performed on 16-slice scanners

(Siemens Sensation, Siemens AG, Erlangen, Germany). NCCTwas obtained with 5-mm contiguous axial sections from vertex toskull base using imaging parameters of 120 kVp, 285 mAs, 1.5-mm slice collimation, and 1.0-second rotation. CTP was obtainedusing two contiguous 12-mm-thick axial sections centered at thelevel of the basal ganglia and internal capsules for a total of 24 mm

of z-axis coverage. This anatomical region was chosen to maxi-mize the inclusion of the middle cerebral artery (MCA) territory,while also including portions of the brain supplied by the ante-rior cerebral artery and posterior cerebral artery (PCA). A 60-sec-ond cine series was performed beginning 4 s after the intravenousadministration of 50 ml of iodinated contrast at 5 ml/s by a powerinjector into an antecubital vein (Omnipaque, 300 mg iodine/ml;Amersham Health, Princeton, N.J., USA). CTP parameters were80 kVp, 200 mAs, 0.5-second rotation, and 60 images per section.All follow-up MRI exams were performed on 1.5-tesla systems

(Siemens Vision or Symphony, Siemens AG). DWI parameterswere TR 3,400 ms, TE 95 ms, b values of 0, 100, 1,000 s/mm 2,5-mm images at 5-mm slice intervals, 128!128 matrix, and 6gradient directions.

Data AcquisitionRaw CTP source images were retrieved from our institutions

picture archiving and communication system (PACS; SiemensMedical Solutions, Erlangen, Germany) onto an offline worksta-tion (Siemens Leonardo, Siemens AG) equipped with postpro-cessing software that generates color overlay maps of dynamicenhancement data (Siemens Syngo Neuro Perfusion CT, SiemensAG). This software calculates the perfusion metrics of cerebralblood f low (CBF), cerebral blood volume (CBV), and time to peak(TTP) using tissue and blood time attenuation curves as previ-ously described [5, 12, 13]. Patients were further excluded if thedynamic CTP data was degraded by poor injection (n = 3), severemotion artifacts (n = 4), or cutoff of the venous input curve beforereturning to baseline (n = 2). The remaining dynamic datasetsfrom 100 unique patients were used to generate CBF, CBV, andTTP parametric color maps in 256!256 matrix. The color scaleswere set at 0100 ml/100 ml/s for all CBF maps, 06 ml/100 ml forall CBV maps, and 020 s for all TTP maps.

Data AnalysisTwo board-certified radiologists (each with approximately 1-

year experience in CTP interpretation) reviewed in consensuseach set of CTP color maps and NCCT, in alternating, random,patient order. The criterion for hyperacute infarction on CTPmaps (examined on the workstation) was a lesion with low relativeCBV in a focal region of elevated TTP or decreased CBF comparedto the homologous region in the contralateral hemisphere. Thecriteria for hyperacute infarction on NCCT (examined on ourPACS) were hypoattenuation or loss of gray-white distinction inthe basal ganglia, loss of the insula cortical ribbon or obscurationof the sylvian fissure, any cortical sulcal effacement, any focal hy-

poattenuation compared to the homologous structure in the con-tralateral hemisphere not compatible with chronic infarction, andhyperattenuation of an artery greater than that of any other vessel[14]. The readers were instructed to examine the NCCT in stan-dard brain windows (center = 40 Hounsfield units, width = 80Hounsfield units) followed by a second evaluation in which win-dow widths and center level settings were freely adjusted to opti-mize gray-white distinct ion [15]. The readers were aware that allpatients arrived to the emergency room!3 h from symptom onsetand that an infarct, if present, may involve more than one vascu-lar territory. They were blinded to all other clinical information,formally dictated reports, and follow-up imaging data. For bothCTP and NCCT, the readers indicated on a spreadsheet (Excel2003, version 11.6560, Microsoft, Seattle, Wash., USA) whether an

infarct was present, and if so, its laterality and vascular distribu-tion. One board-certif ied radiologist examined each patients fol-low-up whole-brain DWI and corresponding apparent diffusioncoefficient maps on PACS for evidence of acute infarction. Simi-larly, if an infarct was present, its laterality and vascular territorywere recorded on a spreadsheet.

Statistical AnalysisA detection of hyperacute infarct on NCCT or CTP was con-

sidered a true positive if assigned laterality and vascular territory

8/13/2019 42321078

3/9

Lin /Do /Ong /Shapiro /Babb /Siller /Pramanik

Cerebrovasc Dis 2009;28:727974

matched those recorded on DWI. In patients with DWI lesions inmore than one vascular territory (including bilateral disease), de-tection of one lesion by NCCT or CTP was sufficient to be consid-ered as a true positive. The Blyth-Still-Casella procedure was used

to construct exact 95% confidence intervals for sensitivity, speci-ficity, positive predictive value, negative predictive value, andoverall accuracy on NCCT and CTP. An exact McNemar test forthe comparison of paired binomial proportions was used to com-pare the sensitivity, specificity, and overall accuracy of NCCTversus CTP. Generalized estimating equations from a logistic re-gression model for correlated data were used to compare the pre-dictive values of NCCT versus CTP. All statistical computationswere performed on SAS (version 9.0, SAS Institute, Cary, N.C.,USA) and StatXact (version 6.0, Cytel Software Corporation,Cambridge, Mass., USA), with p values !0.05 being statisticallysignificant.

Results

Patient characteristics are summarized in table 1. Fol-low-up DWI occurred at a median of 27.5 h (range, 2.0158) after NCCT and CTP were performed. Thirty-five(35.0%) patients did not have restricted diffusion on fol-low-up exam to indicate acute infarction. The final dis-charge diagnoses for these 35 patients were transientischemic attack in 19 (54.3%), migraine in 3 (8.6%), men-ingitis in 1 (2.9%), posterior reversible encephalopathy in1 (2.9%), and undetermined in 11 (31.4%). The 65 (65.0%)patients with hyperacute infarction confirmed on follow-

up DWI were distributed in the following vascular terri-tories: 49 in MCA, 8 in PCA, 1 in unilateral MCA andPCA, 1 in bilateral MCA and PCA, and 6 vertebrobasi-lar.

On NCCT, 17 (26.2%) of the 65 hyperacute infarctswere detected without any false-positive errors. On CTP,42 (64.6%) of the 65 hyperacute infarcts were detectedwith one false-positive error (patient with a region of

chronic infarction). See figures 1 and 2 for examples offalse-negative NCCT and true-positive CTP detections.Of the 23 infarcts that were undetected on CTP, 10 (43.5%)were outside the volume of coverage (territories: 3 inMCA, 1 in PCA, 6 vertebrobasilar). All of the remaining13 (56.5%) infarcts undetected on CTP were deemed to

be within CTPs coverage and small (including lacunartype) measuring 15 mm in maximal dimension. Seefigure 3 for an example of an acute small infarct missedon CTP.

Section-selective CTP provided 23 (47.9%) additionalcorrect diagnoses of hyperacute infarct for the 48 infarctsundetected on whole-brain NCCT. On the other hand,NCCT did not reveal any of the 23 infarcts that weremissed on CTP, but would have corrected for CTPs singlefalse-positive error of chronic infarction. Diagnostic ac-curacies of NCCT and CTP are shown in table 2. CTPhad significantly better sensitivity (p !0.0001), negative

predictive value (p = 0.032), and overall accuracy (p !0.0001) than NCCT. CTP was not significantly differentfrom NCCT with respect to specif icity or positive predic-tive value.

Discussion

We performed a retrospective study of 100 emergencyroom patients who presented with an acute neurologicaldeficit within 3 h of onset, were evaluated with NCCTand CTP during triage, and had follow-up DWI to con-

firm presence or absence of an acute cerebral infarct. Ouraim was to determine if diagnostic accuracy can be im-proved by the additional evaluation of CTP despite itslimitation in z-axis coverage of the brain. The resultsshowed that CTP was significantly more sensitive (64.6vs. 26.2%) and accurate (76.0 vs. 52.0%) than NCCT alonein this critical time window. Nevertheless, 10 infarctswere found outside the CTP coverage volume which rep-resented 15.4% of the 65 confirmed stroke patients in ourcohort. Although these 10 infarcts were also undetectedon NCCT, it highlights section selectivity as being CTPsmajor drawback. The readers were limited by dynamic

datasets acquired on a 16-slice CT scanner which pro-vided only 24 mm of z-axis coverage. Youn et al. [16] re-cently showed that CTPs detection accuracy can be im-proved by using 64-slice scanners in combination withthe toggling table technique, allowing coverage of up to80 mm. It is worth noting that the sensitivity for infarcts115 mm and inside the 24-mm coverage was 100% in ourstudy.

Table 1.Patient and stroke characteristics

Gender (male:female) 53:47Median age (range), years 67 (1896)Median NIHSS score (range) 12 (428)Acute infarct, % 65.0 (65/100)Large-vessel infarct, % 69.2 (45/65)

Small infarct, % 30.8 (20/65)Vessel occlusion, % 47.7 (31/65)Treated by intravenous tPA, % 38.5 (25/65)Recanalization, % 58.3 (14/24a)

aSeven patients did not have follow-up angiographic data.

8/13/2019 42321078

4/9

8/13/2019 42321078

5/9



Table 2.Diagnostic accuracy of hyperacute infarct

8/13/2019 42321078

6/9



a b c

d e f

creased number of false-negatives to our study, but if thishad occurred, it was limited to the small infarcts. Patientselection bias may also have affected our results. We ex-

amined only patients who had DWI follow-up which ex-cluded 42 (27.8%) of the original 151 patients in the pop-ulation that presented in the 3-hour window during theinclusion period. Follow-up DWI was a necessary crite-rion because small infarcts can be difficult to visualize onearly follow-up NCCT [24]. Doing so also allowed us toselect a cohort that carried a uniform reference standard.Finally, in regard to disease prevalence in our retrospec-

tive cohort, acute infarction was present in 65% of pa-tients which compares favorably to a recent large pro-spective study in which the prevalence of an acute isch-

emic infarct was 62% in those who arrived in the 3-hourwindow with sudden neurological deficit [1].

The results of this study may provide compelling rea-sons why CTP should be performed in the 3-hour win-dow. We believe, as do others [1], that greater accuracy foracute stroke should logically result in an increased num-ber of patients being treated since more of them will beconfidently diagnosed at triage. While NCCT is adequate

Fig. 3.A 33-year-old male imaged at 2 h from sudden onset of lefthemiparesis. CTA showed no focal arterial stenosis or occlusion.aNCCT shows no signs of acute ischemic change. bMap of CBFin which the readers did not identify a focal area of hypoperfu-sion. cMap of TTP in which the readers failed to isolate the punc-tuate focus of prolongation in the left lentiform nucleus. dMap of

CBV in which the readers did not appreciate any focal abnormal-ity. eDWI at 24 h after ictus reveals an acute lacunar infarct inthe left lentiform nucleus. A nonspecific lesion on all three para-metric maps can be seen retrospectively at this location (arrowsin be). fCorresponding ADC map of acute infarct (arrow).

C

olorversiona

vailable

online

8/13/2019 42321078

7/9



for those patients who have a high pretest probability, thecases in which there is diagnostic uncertainty will resultin delay or nontreatment. For example, Barber et al. [25]have shown in their large cohort study that nearly a thirdof patients who arrived within the 3-hour window butwere not given tPA because of improving symptoms

(presumed transient ischemic attack) were left disabledor died during the hospital admission. A more definitivediagnosis of acute stroke also allows the stroke team tobetter initiate and apply hospital resources (e.g. admis-sion to intensive care facility/unit). Finally, NCCT usu-ally provides little information regarding vessel statuswhich can be critical since an ICA terminus or M1 clotmay not be recanalized with intravenous tPA alone, ne-cessitating intra-arterial therapy. Therefore, even in the3-hour window, we believe CTA is a necessary adjunct to

NCCT for triage and if the patient is already receiving abolus of iodinated contrast for CTA, CTP can easily beperformed as well with only a modest increase in contrastload.

We did not delve into the issue of salvageable tissue atrisk (penumbra) even though it is the primary target of

perfusion stroke imaging [2629]. We believe, as do oth-ers [30], that in the 3-hour window, it is most importantto determine whether acute ischemia is present that re-quires intervention. The question of penumbra is lesscritical than it is for patient presentation beyond the first3 h when a justification is needed for off-label usage ofintravenous tPA. Such was the motivation behind ourstudy of whether CTP improves stroke detection duringthis early time frame when the key to favorable patientoutcome remains early diagnosis and treatment [30].

References

1 Chalela JA, Kidwell CS, Nentwich LM, LubyM, Butman JA, Demchuk AM, Hill MD, Pa-tronas N, Latour L, Warach S: Magnetic res-onance imaging and computed tomographyin emergency assessment of patients withsuspected acute stroke: a prospective com-parison. Lancet 2007; 369: 293298.

2 Saur D, Kucinski T, Gryzyska U, Eckert B,Eggers C, Niesen W, Schoder V, Zeumer H,Weiller C, Rther J: Sensitivity and interrateragreement of CT and diffusion-weightedMR imaging in hyperacute stroke. AJNR Am

J Neuroradiol 2003; 24: 878885.3 Gonzalez RG, Schaefer PW, Buonanno FS,Schwamm LH, Budzik RF, Rordorf G, WangB, Sorensen AG, Koroshetz WJ: Diffusion-weighted MR imaging: diagnostic accuracyin patients imaged within 6 h of stroke symp-tom onset. Radiology 1999; 210: 155162.

4 Fiebach JB, Schellinger PD, Jansen O, MeyerM, Wilde P, Bender J, Schramm P, Jttler E,Oehler J, Hartmann M, Hhnel S, Knauth M,Hacke W, Sartor K: CT and dif fusion-weight-ed MR imaging in randomized order: diffu-sion-weighted imaging results in higher ac-curacy and lower interrater variability in thediagnosis of hyperacute ischemic stroke.Stroke 2002; 33: 22062210.

5 Mayer TE, Hamann GF, Baranczyk J, Rosen-garten B, Klotz E, Wiesmann M, Missler U,Schulte-Altedorneburg G, Brueckmann HJ:Dynamic CT perfusion imaging of acutestroke. AJNR Am J Neuroradiol 2000; 21:14411449.

6 Wintermark M, Fischbein NJ, Smith WS, KoNU, Quist M, Dillon WP: Accuracy of dy-namic perfusion CT with deconvolution indetecting acute hemispheric stroke. AJNRAm J Neuroradiol 2005; 26: 104112.

7 Scharf J, Brockmann MA, Daffertshofer M,Diepers M, Neumaier-Probst E, Weiss C,Paschke T, Groden C: Improvement of sensi-tivity and interrater reliability to detect acutestroke by dynamic perfusion computed to-mography and computed tomography angi-ography. J Comput Assist Tomogr 2006; 30:

105110.8 Pepper EM, Parsons MW, Bateman GA, LeviCR: CT perfusion source images improveidentification of early ischaemic change inhyperacute stroke. J Clin Neurosci 2006; 13:199205.

9 Lin K, Kazmi KS, Law M, Babb J, PeccerelliN, Pramanik BK: Measuring elevated micro-

vascular permeabil ity and predicting hem-orrhagic transformation in acute ischemicstroke using first-pass dynamic perfusionCT imaging. AJNR Am J Neuroradiol 2007;28: 12921298.

10 Lin K, Rapalino O, Law M, Babb JS, SillerKA, Pramanik BK: Accuracy of the A lbertaStroke Program Early CT Score during thefirst 3 h of middle cerebral artery stroke:comparison of noncontrast CT, CT an-giography source images, and CT perfu-sion. AJNR Am J Neuroradiol 20 08; 29: 931936.

11 Lin K, Rapalino O, Lee B, Do KG, SussmannAR, Law M, Pramanik BK: Correlation of

volumetric mismatch and mismatch of Al-berta Stroke Program Early CT Scores on CTperfusion maps. Neuroradiology 2009; 51:1723.

12 Klotz E, Knig M: Perfusion measurementsof the brain: using dynamic CT for the quan-titative assessment of cerebral ischemia inacute st roke. Eur J Radiol 1999; 30: 170184.

13 Koenig M, Kraus M, Theek C, Klotz E, Geh-len W, Heuser L: Quantitative assessment ofthe ischemic brain by means of perfusion-related parameters derived from perfusionCT. Stroke 2001; 32: 431437.

14 Wardlaw JM, Mielke O: Early signs of braininfarction at CT: observer reliability andoutcome after thrombolytic treatment-sys-

tematic review. Radiology 2005; 235: 444453.15 Lev MH, Farkas J, Gemmete JJ, Hossain ST,

Hunter GJ, Koroshetz WJ, Gonzalez RG:Acute stroke: improved nonenhanced CTdetection Benefits of soft-copy interpreta-tion by using variable window width andcenter level settings. Radiology 1999; 213:150155.

16 Youn SW, Kim JH, Weon YC, Kim SH, HanMK, Bae HJ: Perfusion CT of the brain using40-mm-wide detector and toggling tabletechnique for initial imaging of acute stroke.AJR Am J Roentgenol 2008; 191:W120W126.

17 Boiten J, Lodder J: Lacunar infarcts. Patho-genesis and validity of the clinical syn-dromes. Stroke 1991; 22: 13741378.

18 The NINDS t-PA Stroke Study Group: Gen-eralized efficacy of t-PA for acute stroke.Subgroup analysis of the NINDS t-PA StrokeTrial . St roke 1997; 28: 21192125.

19 Hsia AW, Sachdev HS, Tomlinson J, Hamil-ton SA, Tong DC: Efficacy of IV tissue plas-minogen activator in acute stroke: doesstroke subtype really matter? Neurology2003; 61: 7175.

8/13/2019 42321078

8/9



20 Cocho D, Belvs R, Mart-Fbregas J, BravoY, Aleu A, Pagonabarraga J, Molina-Porcel L,Daz-Manera J, San Roman L, Martnez-Lage M, Martnez A, Moreno M, Mart-Vilalta JL: Does thrombolysis benefit pa-tients with lacunar syndrome? Eur Neurol2006; 55: 7073.

21 Ay H, Oliveira-Filho J, Buonnanno FS, Ez-

zeddine M, Schaefer PW, Rordorf G,Schwamm LH, Gonzalez RG, Koroshetz WJ:Diffusion weighted imaging identifies a sub-set of lacunar infarction associated with em-bolic source. St roke 1999; 30: 26442650.

22 Baumgart ner RW, Sidler C, Mosso M, Geor-giadis D: Ischemic lacunar stroke in patientswith and without potential mechanism otherthan small-arter y disease. Stroke 2003; 34:653659.

23 Seifert T, Enzinger C, Storch MK, Pichler G,Niederkorn K, Fazekas F: Acute small sub-cortical infarctions on diffusion weightedMRI: clinical presentation and aetiology. JNeurol Neurosurg Psychiatry 2005; 76: 15201524.

24 De Reuck J, Hemelsoet D, Nieuwenhuis L,Van Maele G: Computed tomographicchanges in lacunar syndromes. Clin NeurolNeurosurg 2005; 108: 1824.

25 Barber PA, Zhang J, Demchuk AM, Hill MD,Buchan AM: Why are stroke pat ients exclud-ed from TPA therapy? An analysis of patienteligibilit y. Neurology 2001; 56: 10151020.

26 Parsons MW, Pepper EM, Chan V, SiddiqueS, Rajaratnam S, Bateman GA, Levi CR: Per-fusion computed tomography: prediction offinal infarct extent and stroke outcome. AnnNeurol 2005; 58: 672679.

27 Schaefer PW, Roccatagliata L, Ledezma C,Hoh B, Schwamm LH, Koroshetz W, Gonza-lez RG, Lev MH: First-pass quantitative CTperfusion identifies thresholds for salvage-able penumbra in acute stroke patients treat-ed with intra-arterial therapy. AJNR Am JNeuroradiol 2006; 27: 2025.

28 Wintermark M, Flanders AE, Velthuis B,Meuli R, van Leeuwen M, Goldsher D, Pine-da C, Serena J, van der Schaaf I, Waaijer A,Anderson J, Nesbit G, Gabriely I, Medina V,Quiles A, Pohlman S, Quist M, Schnyder P,Bogousslavsky J, Dillon WP, Pedraza S: Per-fusion-CT assessment of infarct core andpenumbra: receiver operating characteristic

curve analysis in 130 patients suspected ofacute hemispheric stroke. Stroke 2006; 37:979985.

29 Murphy BD, Fox AJ, Lee DH, Sahlas DJ,Black SE, Hogan MJ, Coutts SB, DemchukAM, Goyal M, Aviv RI, Symons S, Gulka IB,Beletsky V, Pelz D, Hachinski V, Chan R, LeeTY: Identification of penumbra and infarctin acute ischemic stroke using computed to-mography perfusion-derived blood flow andblood volume measurements. Stroke 2006;37: 17711777.

30 ATLANTIS ECASS and NINDS rt-PA StudyGroup Investigators: Association of outcomewith early stroke treatment: pooled analysisof ATLANTIS, ECASS, and NINDS rt-pa

stroke trials. Lancet 2004; 363: 768774.

8/13/2019 42321078

9/9

42321078

Documents