at SciVerse ScienceDirect

Clinical Radiology 68 (2013) e659ee668

Contents lists available

Clinical Radiology

journal homepage: www.cl inicalradiologyonl ine.net

Additional value of adenosine-stress dynamic CTmyocardial perfusion imaging in thereclassification of severity of coronary arterystenosis at coronary CT angiographyS.M. Kim a,b, J.-H. Choi b,c, S.-A. Chang b,c, Y.H. Choe a,b,*

aDepartment of Radiology, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul,Republic of KoreabCardiovascular Imaging Center, Samsung Medical Center, Sungkyunkwan University School of Medicine,Seoul, Republic of KoreacDivision of Cardiology, Department of Medicine, Samsung Medical Center, Sungkyunkwan University School ofMedicine, Seoul, Republic of Korea

article information

Article history:Received 19 April 2013Received in revised form8 July 2013Accepted 15 July 2013

* Guarantor and correspondent: Y.H. Choe, DepCardiovascular Imaging Center, Cardiac and VMedical Center, Sungkyunkwan University Schodong, Gangnam-gu, Seoul 135-710, Republic of2519; fax: þ82 2 3410 2559.

E-mail addresses: yhchoe@skku.edu, yh.choe@

0009-9260/$ e see front matter � 2013 The Royal Cohttp://dx.doi.org/10.1016/j.crad.2013.07.015

AIM: To assess the additional value of adenosine-stress dynamic computed tomography (CT)perfusion (ASDCTP) imaging compared with coronary CT angiography (CCTA) alone to detectsignificant coronary artery stenosis for each threshold of 50% and 70% diameter stenosis.MATERIALS AND METHODS: The study included 34 patients (65 � 11 years, 79% men) with

suspected coronary artery diseases who underwent ASDCTP imaging using a 128-section dual-source CT (DSCT) and invasive coronary angiography (ICA). Two investigators classified coro-nary artery stenosis on CCTA as severe or not. If appropriate image quality could not be ac-quired due to artefacts, the segment was classified as a lesion with significant stenosis. Afterthe interpretation of ASDCTP imaging, the degree of stenosis was reclassified. All parameters ofdiagnostic accuracy were calculated before and after ASDCTP analysis for detection of signif-icant coronary artery stenosis with ICA as the reference standard.RESULTS: The diagnostic accuracy parameters per vessel for the detection of �50% stenosis

before and after ASDCTP analysis changed as follows: sensitivity, from 80% to 83%; specificity,from 83% to 98%; positive predictive value (PPV), from 87% to 98%; and negative predictivevalue (NPV), from 75% to 80%. The addition of ASDCTP resulted in reclassification from oneclass of stenosis severity to another in a significant number of vessels with threshold of 50%stenosis [net reclassification improvement (NRI), 0.176; p < 0.01]. Conversely, the addition ofASDCTP did not result in significant reclassification of stenosis severity in vessels withthreshold of 70% stenosis (NRI, 0.034; p ¼ 0.51).CONCLUSIONS: ASDCTP imaging provides incremental value in the detection of significant

coronary artery stenosis using a threshold of 50%.� 2013 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.

artment of Radiology andascular Center, Samsungol of Medicine, 50 Ilwon-Korea. Tel.: þ82 2 3410

samsung.com (Y.H. Choe).

llege of Radiologists. Published by

Introduction

Multidetector computed tomography (MDCT) has highsensitivity for the detection of coronary artery disease,

Elsevier Ltd. All rights reserved.

S.M. Kim et al. / Clinical Radiology 68 (2013) e659ee668e660

with good negative predictive values.1e5 However, it islimited for evaluating the physiological significance ofparticular stenotic lesions. Currently, invasive techniques,such as fractional flow reserve (FFR), as well as non-invasive examinations, such as single photon-emissioncomputed tomography (SPECT) and positron-emissiontomography (PET) are used for this task. However,although PET is capable of quantitatively measuringmyocardial blood flow (MBF) and coronary flow reserve,neither SPECT nor PET are able to provide anatomicalassessment of coronary arteries. Invasive coronary angi-ography (ICA) with FFR provides information on the de-gree of arterial stenosis and its haemodynamicsignificance, not visualizing myocardial perfusion directly.Cardiac magnetic resonance imaging (MRI) is also used formyocardial perfusion imaging (MPI). However, cardiacMRI is expensive and time-consuming compared to car-diac CT. Furthermore, the number of sections in myocar-dial perfusion MRI is limited and dark-rim artefactsdecrease its diagnostic accuracy.6

In most myocardial CT perfusion (CTP) research to date, a“snapshot” protocol has been used with the acquisition of aCTP dataset in a single phase both at rest and duringstress.7e13 After the introduction of “shuttle mode” of 128-section dual-source CT (DSCT), adenosine-stress dynamic-stress myocardial perfusion CT (ASDCTP) and quantitativeanalysis of MBF14e20 have recently been adopted. The pur-pose of the present study was to evaluate the additionalvalue of ASDCTP using a second-generation DSCT for thedetection of significant coronary artery stenosis comparedto coronary CT angiography (CCTA).

Materials and methods

Subjects

This study included 34 consecutive patients (65 � 11years, 79% men) who underwent ASDCTP using a 128-section DSCT with subsequent ICA. They were referred toradiology department for the evaluation of clinically sus-pected CAD. All patients were examined with ICA within 3months of the CTP examination (22 � 18 days; range 0e75days). In our institution, adenosine-induced CTP or MRperfusion have routinely replaced SPECT. Patients werescreened for contraindications to adenosine administra-tion. Contraindications included history of asthma or se-vere obstructive lung disease, second- or third-degreeatrioventricular block without a functioning pacemaker,acute myocardial infarction or unstable coronary syn-drome (<24 h), systolic blood pressure <90 mmHg, hy-persensitivity to adenosine and intake of caffeine orxanthine-containing compounds within the last 12 h.Metformin was discontinued at the time of the CT exami-nations and withheld for at least 48 h in patients with anestimated glomerular filtration rate <60 ml/min/1.73 m2.The institutional review board approved this retrospectivestudy and waived the requirement for written informedconsent.

CT protocol

All patients underwent cardiac CT using a DSCT system(SOMATOM Definition Flash, Siemens Medical Solution, For-chheim, Germany) with a 2 � 64 � 0.6 mm detector colli-mation and the z-axis flying focal spot technique, resulting in2 � 128 sections. Bilateral antecubital intravenous catheterswere placed with adenosine infused on the left and contrastinjected on the right. Patient preparation included theplacement of electrocardiography (ECG) leads on the pa-tient’s chest and of a blood pressure cuff on the patient’slower extremity as well as a detailed explanation of the CTexamination, including instructions for breath-holds. Nobeta-blockers were administered prior to the examination.

First, single heartbeat CT calcium scoring was acquiredwith the following parameters: 280ms gantry rotation time,120 kV tube potential, and 80 reference mAs per rotationtube currentetime product with automatic tube currentmodulation technique (ATCM; CAREDose4D, SiemensMedical Systems, Forchheim, Germany).

Subsequently, stress CTP and rest CT examinations wereundertaken (Fig 1). Details of the CT protocol and contrastinjection techniques were previously described.14 For stressCTP, after adenosine (0.14mg/kg/min) infusion for 3min,MPIwas performed in dynamic acquisition mode for 30 s whilemaintaining adenosine infusion. Injection of contrast me-dium into the opposite arm began at 2min 54 s after the startof adenosine infusion. Image acquisition was initiated 3 minafter the start of adenosine infusion. Images were obtainedduring end inspiration with a standardized acquisition timeof 30 s. Image acquisition parameters were 100 kV tubevoltage and automatic tube current modulation technique(CAREDose4D) with 350 reference mAs per rotation.

Immediately after the stress CT, the adenosine infusionwas discontinued. After a brief rest period of 5e10 min untilthe return of the patient’s heart rate to baseline, a rest CTexamination was performed 1 min after sublingualadministration of 0.4 mg nitroglycerine. A retrospectiveECG-gated spiral scan technique was used and reduceddoses (4% of dose during acquisition window) were appliedto the rest of the ReR interval to minimize radiation dose.Image acquisition parameters were 330 mAs per rotationtube current time product and a 100 kV tube potential.

ICA

ICA was performed using a standard technique. Quanti-tative coronary angiography (QCA) analysis was performedusing QCA software (CAAS, version 2.0, Pie Medical Imaging,Maastricht, The Netherlands) by a single observer with 10years of experience in ICA analysis. The degree of coronaryartery stenosis was classified by two different thresholds(�50% and �70% diameter stenosis).

Image analysis of CTP for the detection of significantcoronary artery stenosis

Coronary CT angiographic and CTP images were evalu-ated by two experienced investigators (onewith 2 years and

Figure 1 Dynamic stress CT perfusion imaging protocol.

S.M. Kim et al. / Clinical Radiology 68 (2013) e659ee668 e661

one with 4 years of adenosine-stress dynamic perfusion CTexperience), both of whom were blinded to each patient’sclinical data and identifiers. Readers independently ana-lysed the coronary CT angiographic and CTP images.Discordant findings were reconciled during consensusreading.

Step 1: classification of coronary artery stenosis on CCTACoronary CT angiographic images were reconstructed

using 0.6 mm section thickness and 0.4 mm reconstructionincrement during the systolic or diastolic phases of the ReRinterval of electrocardiograms (ECGs), depending on theheart rate. CCTA datasets were reconstructed using a softkernel (B26f). If a patient had calcified plaques, a sharpkernel (B46f) was also used for image reconstruction. Thepresence of significant coronary artery stenosis with twothresholds (greater than 50% diameter stenosis and greaterthan 70% diameter stenosis) was identified on curved-planar reformatted images, multiplanar images, vesselcross-section images, and transverse images by using amodified 17 segment model of the coronary artery tree.21 Ifimage quality was not appropriate for the diagnosis ofcoronary artery disease due to blooming artefacts of calci-fied plaques or motion artefacts, the segment was classifiedas a lesion with significant coronary artery stenosis in eachthreshold. The segments with stenosis were visuallyassessed first. Then, except minimal stenosis (less than 25%stenosis), readers calculated diameter stenosis on vesselcross-section images of stenotic lesions and reference seg-ments using commercial software (Terarecon Intuition,version, 4.4.7, TeraRecon, Foster City, CA, USA). Readerconfidence in the perceived degree of stenosis of each vesselwas scored by a four-point scale. A score of 1 meant poorconfidence; 2, low confidence; 3, moderate confidence; and4, high confidence.

Step 2: reclassification of the severity of coronary arterystenosis after visual interpretation of CTP images

Stress CT images were reconstructed at 280 ms of theReR interval by using a B23f kernel. The stress images wereread in the cardiac short-axis view at a 4 mm sectionthickness with averaged reconstruction. For detection ofsubtle perfusion defects, observers used a user-definednarrow window width and window level setting, as

described previously.22 Dynamic stress CT images wereevaluated visually according to the 16myocardial segments,excluding the apical segment.23 Myocardial segments thatcovered 50% or less of the thickness of the myocardiumwere also excluded due to limited scan coverage (73 mm).Myocardial segments were considered to show perfusiondefects when hypoperfusion lasted for more than sixheartbeats on dynamic datasets.19

Quantitative analysis of CTP

The MBF and myocardial blood volume (MBV) werecalculated using volume perfusion software (Leonardo;Siemens Medical Solutions, Erlangen, Germany). For quan-titative analysis, a dedicated parametric deconvolutiontechnique was used to fit the timeeattenuation curves. Thedeconvolution technique was based on a two-compartmentmodel of intra- and extravascular space.24 Double samplingof the arterial input function (AIF) was performed to in-crease the precision of the fit. In the descending thoracicaorta, the input function was sampled at every table posi-tion and was integrated into one AIF. Therefore, AIF hadtwice the sampling rate compared to the tissue timeeattenuation curve. The algorithm established the maximumslope using the fit model curve for every voxel andmeasured MBF from the following relationship:MBF ¼ maximum slope/maximum AIF, where themaximum slope indicates the tissue timeeattenuationcurve.19,22 MBF and MBV were determined in each of the 16myocardial segments, excluding the apical segment.23

Reference standard and radiation dose

ICA was considered the reference standard for theassessment of coronary artery stenosis. The effective radi-ation dose was derived by multiplying the doseelengthproduct by the conservative constant k (k ¼ 0.014 mSv/mGy/cm).25

Statistical analysis

All continuous data were expressed as mean � standarddeviation, whereas categorical variables were expressed aspercentages. The diagnostic accuracy of CCTA for thedetection of significant coronary stenosis both before and

S.M. Kim et al. / Clinical Radiology 68 (2013) e659ee668e662

after the analysis of ASDCTP imaging, with ICA as thereference standard, was compared with determined sensi-tivity, specificity, positive predictive value (PPV), andnegative predictive value (NPV); 95% confidence intervalswere provided for each estimate. Calculations were per-formed on both a per-patient basis and a per-vessel (leftanterior descending, left circumflex, and right coronary ar-teries) basis. Significant stenosis was considered to exist attwo thresholds (�50% or �70% luminal diameter stenosis)in each analysis. Sensitivity and specificity before and afterCTP analysis were compared by using the McNemar test.The improvement of diagnostic performance after additionof CTP to the coronary assessment in CCTA was evaluatedusing the net reclassification improvement index.

For interobserver agreements, the Cohen kappa valuewas used. All reported diagnostic values are based onconsensus between the two observers. Reader confidence ofa categorical scale was compared before and after CTPanalysis by theWilcoxon sign-rank test. Comparisons of thecontinuous variables in two different groups were per-formed using independent t-tests. A p-value of <0.05 wasconsidered to indicate statistical significance. Statisticalanalysis was performed using MedCalc� software (MedCalcSoftware, Ostend, Belgium).

Results

Patient characteristics

Patient characteristics are listed in Table 1. Of the 34patients (mean 65�11 years old), 79% were men, 94% had abody mass index (BMI) below 30, 65% had experiencedangina pectoris previously, 41% were former or currentsmokers, and 50% had diabetes mellitus. The pretest prob-ability of coronary artery disease of those patients wasclassified as follows: low,12 patients (35%); intermediate,16patients (47%); high, six patients (18%). In patients with lowpretest probability of coronary artery disease, nine patientshad non-anginal chest pain. Seven of the nine patients were

Table 1Baseline characteristics of 34 patients.

Age, yearsa 65 � 11Male 27 (79%)Body mass index, kg/m2 a 24.7 � 3.1b

Body weight, kga 65.8 � 10.7Coronary risk profileArterial hypertension 18 (53%)Cigarette smoking 14 (41%)Positive family history 4 (12%)Hyperlipidaemia 4 (12%)Diabetes 17 (50%)

Medical historyPrevious angina pectoris 22 (65%)Prior myocardial infarction 3 (9%)Prior coronary revascularization 5 (15%)

Unless otherwise noted, data are numbers of patients. Coronary risk profile,diabetes, and medical history were classified according to documentation inthe cardiologist’s notes.

a Mean value � standard deviation.b Only two patients had a BMI above 30.

unable to exercise and two patients had exercise treadmilltest with positive results. The remaining three patients withlow pretest probability had peripheral arterial occlusivediseases and underwent cardiac CT for preoperative riskevaluation.

CT imaging parameters

CT imaging parameters are summarized in Table 2. Themean heart rate was 73 � 14 beats per minute at stress and69 � 11 at rest (p ¼ 0.01). Mean effective radiation doses forthe stress and rest acquisitions were 5.1 � 2.2 mSv and3.7 � 0.7 mSv, respectively. The mean total effective radia-tion exposure was 9.7 � 2.4 mSv.

Diagnostic accuracy for the detection of significantcoronary artery stenosis before and after perfusionanalysis

Values of diagnostic accuracy for the detection of sig-nificant coronary artery stenosis are listed in Table 3. RestCCTA revealed that 55 vessels (54%) had 50% or greaterstenosis and 32 vessels (31%) had 70% or greater stenosisamong a total of 102 coronary arteries. We found that 18(18%) of the 102 vessels had blooming artefacts and threevessels (3%) had motion artefacts, and they were countedpositive for significant coronary artery stenosis in eachanalysis with two thresholds. With ICA, 60 (59%) of 102coronary arteries had 50% or greater stenosis and 50 vessels(49%) had 70% or greater stenosis.

With a cut-off value of>50% diameter stenosis at ICA andCCTA, the stenosis degree of 10 (10%) of the 102 vessels wasreclassified after perfusion analysis (Fig 2). All the lesions ofthe 10 vessels were focal discrete lesions. Five vessels withblooming artefacts on CCTA without perfusion defects onASDCTP were reclassified as insignificant stenosis (Fig 3);three vessels with <50% stenosis with perfusion defects,reclassified as significant stenosis (Fig 4); two vessels witharound50% stenosiswithout perfusion defects, reclassified asinsignificant stenosis. Among 10 reclassified vessels, the ninevessels (90%) were in agreement with ICA findings and onevessel was not. The vessel was one of five vessels that had

Table 2Adenosine-stress dynamic computed tomography (CT) perfusion imagingparameters.

Parameter Stress CTP Rest CCTA

Heart rate (beats/min)Minimum 61 � 16 66 � 11Maximum 92 � 24 74 � 18Mean 73 � 14 69 � 11

Scan mode Axial shuttle HelicalTube voltage (kV) 100 100Tube current (mAs) 350a 330Effective radiation exposure (mSv) 5.1 � 2.2 3.7 � 0.7Contrast mediumVolume (ml) 44.8 � 4.6 77.5 � 4.3Flow rate (ml/s) 4.5 � 0.5 4.3 � 0.4

Unless otherwise noted, data are mean values � standard deviation.CCTA, coronary CT angiography; CTP, CT perfusion.

a 350 reference mAs per rotation with tube current modulation.

Table 3Diagnostic accuracy of coronary computed tomography (CT) angiography (CCTA) for detection of significant stenosis before and after analysis of perfusionimages with use of invasive coronary angiography (ICA) as the reference standard.

Per-vessel analysis Per-patient analysis

Diagnostic estimates CCTA alone Combined CCTA and CTP CCTA alone Combined CCTA and CTP

Stenosis �50%Sensitivity 80 (48/60) [70e90]a 83 (50/60) [74e93]a 97 (32/33) 97 (32/33)Specificity 83 (35/42) [72e95]b 98 (41/42) [93e100]b 100 (1/1) 100 (1/1)PPV 87 (48/55) [78e96] 98 (50/51) [94e100] 100 (32/32) 100 (32/32)NPV 75 (35/47) [63e86] 80 (41/51) [70e91] 50 (1/2) 50 (1/2)Stenosis �70%Sensitivity 68 (34/50) [55e81]c 70 (35/50) [57e83]c 88 (28/32) 84 (27/32)Specificity 83 (43/52) [72e93]d 92 (48/52) [85e100]d 100 (2/2) 100 (2/2)PPV 79 (34/43) [67e91] 90 (35/39) [80e99] 100 (28/28) 100 (27/27)NPV 73 (43/59) [62e84] 76 (48/63) [66e87] 33 (2/6) 29 (2/7)

All data are percentages. The absolute numbers used to calculate the percentages are in parentheses. Ninety-five percent confidence intervals are provided insquare brackets. Per-vessel analysis was based on data from a total of 102 vessels. Per-patient analysis was based on data from a total of 34 patients. Significantstenosis was calculated based on luminal diameter.CTP, CT perfusion; NPV, negative predictive value; PPV, positive predictive value.

a p ¼ 0.63.b p ¼ 0.03.c p ¼ 1.d p ¼ 0.06.

S.M. Kim et al. / Clinical Radiology 68 (2013) e659ee668 e663

blooming artefacts without perfusion defects and ICArevealed a 60% focal stenosis in the proximal left circumflexartery.

The stenosis severity of 82 (89%) of the 92 vessels thatwere not reclassified was ultimately in agreement with ICA,and the stenosis severity of 10 (11%) vessels was not. Of 10vessels, nine vessels measured <50% stenosis on CCTA andhad no perfusion defect on CTP. These lesions were signifi-cantly stenotic (�50%) at ICA; five vessels had stenotic lesion

Figure 2 Classification of significant coronary artery stenosis before and afArtefacts in 21 vessels consisted of 18 calcific blooming artefacts and threeartery stenosis. CT perfusion positive indicates a hypoperfused territory wICA; I, insignificant stenosis; RI, reclassified as insignificant stenosis; RS, re‘�’, absence.

at distal coronary arteries or branch vessels with <2 mmdiameter; three vessels had focal discrete lesions at proximalleft anterior descending (LAD), obtuse marginal (OM), anddistal left circumflex artery (LCx); one vessel had an ostiallesion at LCx. Except for three lesions thatwere distal arteriesor branch vessels, six lesions in six vessels were treated withstent implantation (n ¼ 5) or coronary artery bypass graft(n ¼ 1). All nine vessels were from the patients with multi-vessel diseases. The remaining one vessel had a segmental

ter CT perfusion analysis using coronary CT angiography (102 vessels).motion artefacts. They were counted positive for significant coronaryith an associated significant coronary stenosis. *C, concordance withclassified as significant stenosis; S, significant stenosis; ‘þ’, presence;

Figure 3 Reclassification of a false-positive lesion with calcific blooming artefacts in a 49-year-old male with chest pain. (a) CTA image showssignificant three-vessel coronary artery stenosis. (b) The left circumflex artery (LCX) has a calcified plaque with a severe blooming artefact. Thislesion was considered to be a significant stenosis. (c) However, CT perfusion (part of the dynamic acquisition) shows perfusion defects in the leftanterior descending (LAD) and right coronary artery (RCA) territories, and not the left circumflex territory (red ROI). Therefore, the LCX lesionwas reclassified as having no significant stenosis. (d) Invasive coronary angiography demonstrates that the LCX does not have a significantstenosis. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

S.M. Kim et al. / Clinical Radiology 68 (2013) e659ee668e664

Figure 4 Reclassification of a false-negative lesion due to overlooking of a significant stenosis at initial CTA analysis in a 55-year-old womanwithchest pain. (a) On initial CT angiography analysis, a fibrocalcific lesion (arrow) was regarded as insignificant (45% stenosis). (b) However, colour-coded CT perfusion shows a perfusion defect (arrows) in the RCA territory. The difference of colour means the difference in MBF. The red isindicative of normal MBF and the blue is indicative of decreased MBF. (c) On second-look analysis of CT angiography, significant stenosis (arrow)was found at the mid-RCA. (d) Findings at ICA (arrow) correlated with that (arrow) of the second-look analysis of CTA. (For interpretation of thereferences to color in this figure legend, the reader is referred to the web version of this article.)

S.M. Kim et al. / Clinical Radiology 68 (2013) e659ee668 e665

lesion with maximum diameter stenosis of 55% at proximaland mid-LAD on CCTA. CTP showed subtle hypoperfusionarea at apical septal wall. However, this lesion measured 40%stenosis using QCA. Percutaneous coronary intervention orcoronary artery bypass graft was not performed.Mean valuesof the diagnostic accuracy of CCTA before ASDCTP analysiswere listed inTable 3. Onper-vessel analysis, ASDCTPanalysisincreased specificity significantly (p ¼ 0.03). However,sensitivity did not change significantly (p ¼ 0.63). The addi-tion of ASDCTP resulted in reclassification from one class ofstenosis severity to another in a significant number of vesselswith threshold of 50% stenosis [net reclassificationimprovement (NRI) 0.176, net proportion of patients reclas-sified 7.84%; p<0.01].

With a threshold of 70% diameter stenosis at ICA andCCTA for significant stenosis, the stenosis degrees of eight(8%) of the 102 vessels were reclassified after ASDCTPanalysis. Six vessels with blooming artefacts on CCTA andwithout perfusion defects on ASDCTP were reclassified asinsignificant stenosis; two vessels, <70% stenosis withperfusion defects, reclassified as significant stenosis. Amongreclassified vessels, seven vessels were in agreement withICA findings and one vessel was not. The vessel hadblooming artefacts without perfusion defects and wasreclassified as insignificant stenosis. However, ICA showedsevere stenosis (>70%) in the lesion. The stenosis severity of76 (81%) of the 94 vessels that were not reclassified wasultimately in agreement with ICA, and the stenosis severity

S.M. Kim et al. / Clinical Radiology 68 (2013) e659ee668e666

of 18 (19%) vessels was not. The stenosis degrees of 18vessels were under- or overestimated on CCTA as comparedwith the results of ICA. Mean values of the diagnostic ac-curacy of CCTAwere listed in Table 3. On per-vessel analysis,ASDCTP analysis did not improve diagnostic estimates ofCCTA, significantly [sensitivity (p ¼ 1.0), specificity(p ¼ 0.06)]. The addition of ASDCTP did not result in sig-nificant stenosis severity reclassification in vessels withthreshold of 70% stenosis (NRI 0.034; p ¼ 0.51).

In the ICA analysis using a threshold of 50% stenosis, 13patients had significant stenosis in one vessel. Twenty pa-tients hadmulti-vessel involvement and one patient had nostenosis in the coronary arteries. With a threshold of 70%stenosis, 19 patients had significant stenosis in one vessel.Thirteen patients had multi-vessel disease and two patientshad no stenosis in the coronary arteries.

Quantitative analysis of CTP

Measurements of MBF and MBV are listed in Table 4.Seventeen (3%) of 544 segments were excluded from anal-ysis because of the limited anatomical coverage (73 mm) ofaxial shuttle scan mode. Excluded segments consisted of 10basal anterior and seven basal inferior segments. Of 527included segments, 193 segments (37%) had perfusion ab-normalities on ASDCTP with significant stenosis confirmedby ICA, and 12 (2%) of the 527 segments without significantcoronary stenosis in the associated territories wereconsidered as artefacts. The mean MBF of all myocardialsegments at ASDCTP was 104.6 � 26.4 ml/100 ml/min.There was significant difference in MBF values betweennormal (116.8 � 30.2 ml/100 ml/min) and hypoperfused(61.4 � 17.8 ml/100 ml/min) myocardial segments(p < 0.01). The mean MBV of all myocardial segments atASDCTP was 18.2 � 3.7 ml/100 ml. There was a significantdifference in MBV values between normal (21.4 � 4.3 ml/100 ml) and hypoperfused (12.7 � 3.8 ml/100 ml) myocar-dial segments (p < 0.01).

Reader confidence and interobserver variability

In the classification of stenosis before and after perfusionanalysis, the median score of confidence improved signifi-cantly, from 3 to 4 (p < 0.01). After ASDCTP analysis, agree-ment between the blinded independent readers regardingCCTA findings improved from 80% to 92% and l valuesimproved from 0.72 (95% CI: 0.61e0.83) to 0.85 (95% CI:0.76e0.94).

Table 4Quantitative analysis of computed tomography perfusion.

Normalmyocardium

Abnormalmyocardium

p-Valuea Wholemyocardium

MBF (ml/100ml/min)

116.8 � 30.2 61.4 � 17.8 <0.001 104.6 � 26.4

MBV (ml/100 ml) 21.4 � 4.3 12.7 � 3.8 <0.001 18.2 � 3.7

Unless otherwise noted, data are mean values � standard deviation.NMBF, myocardial blood flow, MBV, myocardial blood volume.

a Independent t-test.

Discussion

In the present study, the diagnostic performances ofCCTA for the detection of significant stenosis werecompared at thresholds of 50% and 70% diameter stenosisbefore and after perfusion analysis, using ICA as the refer-ence standard. After ASDCTP interpretation with a cut-offvalue of 50% luminal stenosis, the addition of ASDCTPshowed significant improvement in diagnostic estimates ona vessel-based analysis after reclassification (p < 0.01).However, at a cut-off value of 70% stenosis, sensitivity andspecificity did not increase significantly. Recent studiesshowed incremental diagnostic values of CTP over CCTAalone with single threshold18,20,26 or two different thresh-olds 9,27 of stenosis. Although Rocha-Filho et al.9 reportedthat the area-under-the-curve and all parameters of diag-nostic accuracy increased significantly with both thresholdsof 50% and 70% luminal stenosis (p < 0.05), the incrementalvalue of the diagnostic performance of CTP with a thresholdof 70% stenosis was lower than that of a threshold of 50%stenosis. They reported that 35 vessels (85%) of 39 non-interpretable vessels were reclassified correctly. The re-sults of the present study showed that specificity and netreclassification index increased significantly after perfusionanalysis in the evaluation of significant stenosis with athreshold of 50% diameter stenosis. Considering nine ves-sels (90%) that were reclassified correctly after ASDCTPanalysis, it is assumed that ASDCTP could provide additionalinformation on the vessels with artefacts, intermediatedegree stenosis, and <50% stenosis. Based on the results ofRocha-Filho et al. and those of the present study, the diag-nosis of severe stenosis (�70%) is less influenced by theresult of CTP than the diagnosis of>50% stenosis. Therefore,the incremental diagnostic value of CTP was particularlyuseful in evaluating coronary artery stenosis with <70%stenosis or non-interpretable vessels with artefacts. How-ever, Bettencourt et al.27 showed that incremental diag-nostic value of CTP was more beneficial in the evaluation ofhigher-grade stenosis (�70%). The differences in the resultsamong the studies may be ascribed to the fact that twoprevious studies used “snapshot” CT techniques, whereasthe present study used dynamic perfusion. Further studiesare needed to identify themost appropriate cut-off value forthe evaluation of significant coronary stenosis using com-bined CCTA and CTP.

The present data indicate that dynamic perfusion usingDSCT was limited in scan range and could not cover theentire left ventricle, especially when patients could not holdtheir breath during dynamic imaging or in cases where theheart was relatively large. However, all excluded segmentswere located in the basal anterior and/or basal inferiorsegments. Haemodynamically significant stenosis seldominvolves only one myocardial segment of basal anterior orbasal inferior wall. Therefore, scan coverage of DSCT dy-namic perfusion did not affect the diagnostic accuracy ofASDCTP in the detection of perfusion defects.

In the present study, the MBFs of normal and abnormalmyocardium were 116.8 � 30.2 ml/100 ml/min and

S.M. Kim et al. / Clinical Radiology 68 (2013) e659ee668 e667

61.4 � 17.8 ml/100 ml/min, respectively. Previous stud-ies18e20,28 reported that the MBFs of normal myocardiumranged from 104.8e142.9 ml/100 ml/min and the MBFs ofhypoperfused myocardium ranged from 52e96 ml/100 ml/min. Bamberg et al.18 also reported that the presence ofcoronary stenosis with a corresponding MBF <75 ml/100 ml/min indicated high risk of haemodynamic signifi-cance compared with FFR.18 Further studies are needed tovalidate the threshold MBF values for myocardial ischaemiaby quantitative analysis of CTP.

The radiation dose is the major concern of dynamicperfusionCT. Previous studies and the present study used thechest conversion factor (0.014 or 0.017) to calculate theeffective radiation doses. Because cardiac CT (with theexception of bypass graft evaluation and triple-rule-out CT)covers the lower chest and upper abdomen, including thebreasts, the imaging technique of cardiac CT (i.e., ECG-basedtube current modulation) is different from that of chest CT.Considering these differences, Gosling et al.29 calculated theeffective radiation dose with a conversion factor of 0.028.Mean effective radiation doses in the present studywould be10.2mSv for dynamic scanning and 7.4mSv for rest scanningby using a conversion factor of 0.028. Therefore, technicaldevelopments and amended protocols may be needed toreduce radiation dose of dynamic myocardial perfusion CT.

The present study had some limitations. The resultshowed that sensitivity and NPV before and after CTPanalysis were lower than those of previous studies. Leberet al.30 reported relatively lower sensitivity of 73e80% forthe detection of stenosis <50%, stenosis >50%, and stenosis>75% using 64-section CT without excluding distal coro-nary segments and side branches. They also suggested thattheir results were explained by partial volume effectsoccurring at the lumen/plaque border either caused bycalcium or the dense contrast agent and by the fact thatdensity values of opacified lumen and plaque may overlapin a certain range. The data were also affected by using QCAas a reference study. Lack of functional assessment, such asFFR or cardiac MRI, as a reference standard might not allowfor accurate physiological assessment in the present study.The second limitation is that no premedication with beta-blockers was administered. This might hamper the diag-nostic accuracy of CCTA, although 128-section DSCT has thecapability of providing diagnostic CCTA images in patientswith fast heart rates. The third limitation is classification ofuninterpretable lesions as having significant coronary ar-tery stenosis. It also might affect the diagnostic accuracyand net reclassification statistics. The final limitation is thatthe number of cases was relatively small and selection biasmight be present in the study’s design. In the present study,patients who had undergone ASDCTP and ICA in clinicalpracticewere included and a large proportion of the presentcases had multi-vessel diseases. Thus, the results areapplicable to a population with a similar prevalence to thepresent study.

In conclusion, the present results show that ASDCTPimaging provides incremental value in the detection ofsignificant coronary artery stenosis using a threshold of 50%stenosis compared with a threshold of 70% stenosis at CCTA.

Acknowledgement

The authors appreciate the contribution of Ms Sang EunLee for the illustration in Fig. 2.

This study was supported by a grant of Korea HealthTechnology R&D Project, Ministry for Health, Welfare &Family Affairs, Republic of Korea (A102065-26).

References

1. Budoff MJ, Dowe D, Jollis JG, et al. Diagnostic performance of 64-multidetector row coronary computed tomographic angiography forevaluation of coronary artery stenosis in individuals without knowncoronary artery disease: results from the prospective multicenter AC-CURACY (Assessment by Coronary Computed Tomographic Angiographyof Individuals Undergoing Invasive Coronary Angiography) trial. J AmColl Cardiol 2008;52:1724e32.

2. Maffei E, Palumbo A, Martini C, et al. Diagnostic accuracy of 64-slicecomputed tomography coronary angiography in a large population ofpatients without revascularisation: registry data and review of multi-centre trials. Radiol Med 2010;115:368e84.

3. Marano R, De Cobelli F, Floriani I, et al. Italian multicenter, prospectivestudy to evaluate the negative predictive value of 16- and 64-slice MDCTimaging in patients scheduled for coronary angiography (NIMISCAD-Non Invasive Multicenter Italian Study for Coronary Artery Disease). EurRadiol 2009;19:1114e23.

4. Meijboom WB, Meijs MF, Schuijf JD, et al. Diagnostic accuracy of 64-slicecomputed tomography coronary angiography: a prospective, multi-center, multivendor study. J Am Coll Cardiol 2008;52:2135e44.

5. Miller JM, Dewey M, Vavere AL, et al. Coronary CT angiography using 64detector rows: methods and design of the multi-centre trial CORE-64.Eur Radiol 2009;19:816e28.

6. Gebker R, Jahnke C, Paetsch I, et al. Diagnostic performance ofmyocardial perfusion MR at 3 T in patients with coronary artery disease.Radiology 2008;247:57e63.

7. Kurata A, Mochizuki T, Koyama Y, et al. Myocardial perfusion imagingusing adenosine triphosphate stress multi-slice spiral computed to-mography: alternative to stress myocardial perfusion scintigraphy. Circ J2005;69:550e7.

8. George RT, Silva C, Cordeiro MA, et al. Multidetector computed tomog-raphy myocardial perfusion imaging during adenosine stress. J Am CollCardiol 2006;48:153e60.

9. Rocha-Filho JA, Blankstein R, Shturman LD, et al. Incremental value ofadenosine-induced stress myocardial perfusion imaging with dual-source CT at cardiac CT angiography. Radiology 2010;254:410e9.

10. Blankstein R, Shturman LD, Rogers IS, et al. Adenosine-induced stressmyocardial perfusion imaging using dual-source cardiac computed to-mography. J Am Coll Cardiol 2009;54:1072e84.

11. George RT, Arbab-Zadeh A, Miller JM, et al. Computed tomographymyocardial perfusion imaging with 320-row detector computed to-mography accurately detects myocardial ischemia in patients withobstructive coronary artery disease. Circ Cardiovasc Imaging2012;5:333e40.

12. Ko BS, Cameron JD, Meredith IT, et al. Computed tomography stressmyocardial perfusion imaging in patients considered for revasculariza-tion: a comparison with fractional flow reserve. Eur Heart J2012;33:67e77.

13. Nasis A, Ko BS, Leung MC, et al. Diagnostic accuracy of combined cor-onary angiography and adenosine stress myocardial perfusion imagingusing 320-detector computed tomography: pilot study. Eur Radiol2013;23:1812e21.

14. Kim SM, Kim YN, Choe YH. Adenosine-stress dynamic myocardialperfusion imaging using 128-slice dual-source CT: optimization of theCT protocol to reduce the radiation dose. Int J Cardiovasc Imaging2013;29:875e84.

15. Weininger M, Schoepf UJ, Ramachandra A, et al. Adenosine-stress dy-namic real-time myocardial perfusion CT and adenosine-stress first-passdual-energy myocardial perfusion CT for the assessment of acute chestpain: initial results. Eur J Radiol 2012;81:3703e10.

S.M. Kim et al. / Clinical Radiology 68 (2013) e659ee668e668

16. Bamberg F, Hinkel R, Schwarz F, et al. Accuracy of dynamic computedtomography adenosine stress myocardial perfusion imaging in esti-mating myocardial blood flow at various degrees of coronary arterystenosis using a porcine animal model. Invest Radiol 2012;47:71e7.

17. Ho KT, Chua KC, Klotz E, et al. Stress and rest dynamic myocardialperfusion imaging by evaluation of complete timeeattenuation curveswith dual-source CT. JACC Cardiovasc Imaging 2010;3:811e20.

18. Bamberg F, Becker A, Schwarz F, et al. Detection of hemodynamically sig-nificant coronary artery stenosis: incremental diagnostic value of dynamicCT-based myocardial perfusion imaging. Radiology 2011;260:689e98.

19. Bastarrika G, Ramos-Duran L, Rosenblum MA, et al. Adenosine-stressdynamic myocardial CT perfusion imaging: initial clinical experience.Invest Radiol 2010;45:306e13.

20. Wang Y, Qin L, Shi X, et al. Adenosine-stress dynamic myocardialperfusion imaging with second-generation dual-source CT: comparisonwith conventional catheter coronary angiography and SPECT nuclearmyocardial perfusion imaging. AJR Am J Roentgenol 2012;198:521e9.

21. Austen WG, Edwards JE, Frye RL, et al. A reporting system on patientsevaluated for coronary artery disease. Report of the Ad Hoc Committeefor Grading of Coronary Artery Disease, Council on Cardiovascular Sur-gery, American Heart Association. Circulation 1975;51:5e40.

22. Mahnken AH, Klotz E, Pietsch H, et al. Quantitative whole heart stressperfusion CT imaging as noninvasive assessment of hemodynamics incoronary artery stenosis: preliminary animal experience. Invest Radiol2010;45:298e305.

23. Cerqueira MD, Weissman NJ, Dilsizian V, et al. Standardized myocardialsegmentation and nomenclature for tomographic imaging of the heart:a statement for healthcare professionals from the Cardiac Imaging

Committee of the Council on Clinical Cardiology of the American HeartAssociation. Circulation 2002;105:539e42.

24. Bruder H, Raupach R, Klotz E, et al. Spatio-temporal filtration of dynamicCT data using diffusion filters. In: Samei E, Hsieh J, editors. Proceedings ofSPIE: medical imaging 2009-physics of medical imaging vol. 7258.Bellingham,WA: SPIEe The International Society forOptical Engineering;2009. p. 725810e57.

25. Shrimpton PC, Hillier MC, Lewis MA, et al. National survey of doses fromCT in the UK: 2003. Br J Radiol 2006;79:968e80.

26. Feuchtner G, Goetti R, Plass A, et al. Adenosine stress high-pitch 128-slice dual-source myocardial computed tomography perfusion for im-aging of reversible myocardial ischemia: comparison with magneticresonance imaging. Circ Cardiovasc Imaging 2011;4:540e9.

27. Bettencourt N, Rocha J, Ferreira N, et al. Incremental value of an inte-grated adenosine stress-rest MDCT perfusion protocol for detection ofobstructive coronary artery disease. J Cardiovasc Comput Tomogr2011;5:392e405.

28. Nakauchi Y, Iwanaga Y, Ikuta S, et al. Quantitative myocardial perfusionanalysis using multi-row detector CT in acute myocardial infarction.Heart 2012;98:566e72.

29. Gosling O, Loader R, Venables P, et al. Cardiac CT: are we under-estimating the dose? A radiation dose study utilizing the 2007 ICRPtissue weighting factors and a cardiac specific scan volume. Clin Radiol2010;65:1013e7.

30. Leber AW, Knez A, von Ziegler F, et al. Quantification of obstructive andnonobstructive coronary lesions by 64-slice computed tomography: acomparative study with quantitative coronary angiography and intra-vascular ultrasound. J Am Coll Cardiol 2005;46:147e54.