Determination Coronary FlowReserve Parametric …circ.ahajournals.org/content/circulationaha/82/4/1438.full.pdfmagnetic flowmeters and myocardial flow reserve ... Microspheres were

1438

Determination of Coronary Flow Reserveby Parametric Imaging

Otto M. Hess, MD, Mark J. McGillem, BS, Scott F. DeBoe, BS, Ibraim M.F. Pinto, MD,

Kim P. Gallagher, PhD, and G.B. John Mancini, MD

Nine mongrel dogs were instrumented with electromagnetic flow probes (EMF) to measurecoronary blood flow through the left anterior descending (LAD) and left circumflex (LCx)coronary arteries at rest and after maximal coronary vasodilation (1 mg/kg/min adenosine).Relative coronary blood flow was determined by parametric imaging in the left posterior obliqueprojection using digital subtraction angiography (DSA). Transmural myocardial perfusion ofthe LAD and LCx beds was determined with tracer-labeled microspheres. Coronary flow reserve(maximal coronary blood flow divided by resting blood flow) was calculated under controlconditions and after constriction of the proximal LAD or LCx by a screw occluder. Heart ratedecreased significantly from 140 beats/min at rest to 122 beats/min after adenosine (p<0.001)and from 134 (rest) to 120 beats/min (adenosine; p<0.05) after coronary constriction. Peaksystolic pressure was kept constant with an aortic constrictor. Left ventricular end-diastolicpressure increased significantly from 18 mm Hg at rest to 23 mm Hg (p<0.05) after coronary

constriction. At baseline, coronary flow reserve was 4.2 with DSA, 3.8 with EMF, and 3.7 withmicrospheres; after coronary constriction,it was 2.6 (DSA), 1.9 (EMF), and 1.5 (microspheres)(all p<0.001 versus baseline). Coronary blood flow showed a good correlation between EMFand microspheres (r=0.87,p<0.001), with a standard error of estimate (SEE) of 0.78 ml/g/min.Coronary flow reserve also showed a good correlation between EMF and microspheres (r=0.82,p<0.00l), with an SEE of 0.93. There was a moderate correlation between EMF and DSA(r=0.68, p<0.001), with an SEE of 1.35 (40% of mean coronary flow reserve). The correlationcoefficient between microspheres and DSA was 0.54 (p<0.01), with an SEE of 1.46 (39%o ofmean coronary flow reserve). The mean difference (accuracy) and standard deviation ofdifference (precision) were 0.2±1.0 between EMF and microspheres, -0.1±1.4 between EMFand DSA, and -0.6±1.7 between microspheres and DSA. We conclude that determination ofcoronary flow reserve by parametric imaging is associated with large variations that are greaterthan variations also inherent in the two reference techniques. Parametric imaging allowsrelatively accurate assessment of coronary flow reserve (small mean difference), but precisionis low (large standard deviation of mean differences). This low precision probably is due to thesuperposition of different cardiac structures in the two-dimensional display of a three-dimensional perfusion zone, potentially inhomogeneous contrast distribution, poor temporalresolution of the once-per-cycle imaging, inadequate displacement of blood by contrastmaterial, and perturbations of flow caused by contrast material. Thus, in this animal model,only changes in coronary flow reserve that are clearly greater than 40%1 can be expected to bemeasured correctly with the currently available technique. (Circulation 1990;82:1438-1448)

oronary arteriography, used in the past as the nary stenosis solely in terms of anatomy. MoreC gold standard for assessing coronary artery recently, functional measurements reflecting the

disease, determines the severity of a coro- physiological significance of stenosed coronary arter-

ies have become available. For example, digital sub-From the Department of Internal Medicine, Division of Cardi-

ology, Veterans Administration Medical Center, University of See 1533

Michigan, Ann Arbor, Mich.Supported in part by funds from the Veterans Administration

and the American Heart Association of Michigan. traction angiography has been used successfully to

Address for correspondence: G.B. John Mancini, MD, Cardiol- determine coronary flow (CFR) in patients

ogy Section (lllA), Veterans Administration Medical Center, with coronary artery disease,' 5but the requirement2215 Fuller Road, Ann Arbor, MI 48105.

Received December 20, 1989; revision accepted May 29, 1990. for expensive and complex techniques has limited its

by guest on May 28, 2018

http://circ.ahajournals.org/D

ownloaded from

http://circ.ahajournals.org/

Hess et al Parametric Imaging and Coronary Flow Reserve 1439

clinical application. Intracoronary Doppler flowprobes also have been used to measure CFR,6,7butcatheter positioning, changes in coronary luminaldiameter, limited access to distal stenoses, and inabil-ity to measure flow velocities in different myocardialregions simultaneously have restricted its use forassessing CFR in daily practice. Moreover, this tech-nique measures only epicardial arterial flow reserve,which may not reflect myocardial flow reserve in thepresence of collateral recruitment and infarction.Positron emission tomography is another means forevaluating CFR in patients with coronary arterydisease,7-9 and trials are underway to determine itssuitability in the clinical setting.The purpose of the present study was to determine

the accuracy and reproducibility of a specific digitalsubtraction angiography method',2 for measuringrelative coronary blood flow and CFR in the experi-mental animal. We compared results with regionalcoronary arterial flow reserve measured with electro-magnetic flowmeters and myocardial flow reservemeasured with tracer-labeled microspheres.

MethodsAnimal Preparation

Mongrel dogs were anesthetized with 35 mg/kgpentobarbital i.v. and were intubated and ventilatedwith a Harvard ventilator pump (Harvard Apparatus,South Natick, Mass.). A left thoracotomy was per-formed in the fifth intercostal space, and the heartwas exposed and supported in a pericardial cradle.The proximal left anterior descending (LAD) andthe proximal left circumflex (LCx) coronary arterieswere dissected free, and an appropriately sized andcalibrated electromagnetic flow probe (CarolinaMedical Electronics, King, N.C.) was placed on eachvessel. A screw occluder was placed distal to the flowprobe either on the LAD (n=2) or LCx (n=4)arteries or both (n =3). Vascular sheaths were intro-duced into the left and right carotid arteries and theleft jugular vein for vascular access. A calibrated 5Fmicromanometer (Millar Instruments, Houston) waspassed into the left ventricle via an apical stab woundand secured with a purse-string suture. A second SFmicromanometer was passed through the rightcarotid artery sheath into the ascending aorta.A 7F Amplatz catheter was introduced through the

left carotid artery into the ascending aorta andpositioned into the ostium of the left coronary artery;care was taken to avoid subselective placement. Sixmilliliters of a nonionic contrast medium (iohexol,Winthrop-Breon, New York) was injected at a flowrate of 4 ml/sec using a power injector with electro-cardiographic gating (Mark IV, Medrad, Pittsburgh).Immediately after intracoronary injection of contrastmaterial, cancellation of the flow signal was observedbecause of the nonionic nature of the contrast fol-lowed by a short hyperemic response that lastedusually up to 30 seconds in the control state (Figure1). A polyethylene tube was placed in the left atrial

appendage for injection of tracer-labeled micro-spheres; a second catheter was placed in the femoralartery for withdrawal of reference blood samplesusing a Harvard withdrawal pump. A Blalock clampwas placed around the descending aorta for constric-tion of the aorta to maintain peak systolic pressure ata constant level during the experiment.

Coronary Blood Flow MeasurementsElectromagnetic flow probe. Coronary blood flow

through the LAD and LCx arteries (Figure 1) wasmeasured with electromagnetic flowmeters. Theywere calibrated by timed collection of blood, and thezero reference was checked several times during theexperiments by brief occlusion of the coronary artery.Any alterations in waveform suggesting poor place-ment or instability of the probes prompted reposi-tioning or selection of a probe of different size.Instability due to potential collapse of the vesselwithin the probe during the hypotension inducedduring maximal hyperemia was minimized by main-taining aortic pressure with a constrictive Blalockclamp on the aorta. Normalization of coronary bloodflow was carried out at the conclusion of the experi-ment by dividing the electromagnetic flow data by thecorresponding muscle mass of the LAD or LCx.10The perfusion beds of the two arteries were deter-mined by selective intracoronary infusion with 2,3,5-triphenyltetrazolium chloride (Sigma Chemical Co.,St. Louis) and Evan's blue (Sigma), respectively. Thecoronary arteries were cannulated separately andperfused simultaneously with warm isotonic saline.Dye perfusion was performed at 100 mm Hg for 5minutes to guarantee optimal staining of the twoperfusion beds. The perfusion beds were dissectedfree and weighed. These weights were used to calcu-late normalized coronary blood flow in milliliters pergram per minute.

Parametric digital angiography. Two images percardiac cycle (at 50% and 75% of the RR interval)were acquired in the left posterior oblique orienta-tion on a Philips Optimus M200 angiographic system(Philips Nederland, Eindhoven, The Netherlands)that was interfaced directly to a digital radiographiccomputer (DPS-4100C, ADAC, Milpitas, Calif.).Radiographic parameters were kept constant (peakkilovoltage, milliamperes) for each image run. Nocorrection of scatter or veiling glare was undertaken.Logarithmic analog-to-digital conversion was used.The images were stored on a digital disk with amatrix of 512 x512x 8 bits. All images were processedby mask-mode subtraction, whereby the last imagebefore contrast administration was chosen as themask. The respirator pump was turned off during theimage acquisition to minimize motion artifacts duringimage subtraction. Five to eight consecutive imageswere selected as the image subset (only one imageper cardiac cycle). A threshold generally less than orequal to 25% of the available grey levels was selectedto minimize background noise. Pixel densitiesexceeding the threshold were used to generate a



ownloaded from


1440 Circulation Vol 82, No 4, October 1990

IFIGURE 1. Original recordings of ventricular pressure (LVP), its first derivative (dP/dt), aortic pressure (AOP), electromagneticblood flow ofthe left anterior descending (LAD FLOW) and left circumflex coronary (CXFLOW) arteries. Data are shown at rest(BASAL), after adenosine infusion (HYPEREMIC), and after coronary stenosis with (HYPEREMIC) and without (BASAL)adenosine infusion. Left ventricular pressure was kept constant by aortic constriction. Coronary blood flow increased three to fivetimes after adenosine infusion and was decreased in the perfusion zone ofthe LAD after coronary constriction (LAD STENOSIS).Microspheres were injected before coronary angiography under steady-state conditions (asterisks). Immediately after microsphereinjection, coronary angiography was performed with a triggered power injector (arrows). ECG, electrocardiogram.

functional image. Red was assigned to pixels that hadan intensity that surpassed the threshold in the firstpostinjection cycle, yellow for the second cycle, whitefor the third, green for the fourth, and so on.10 Aregion of interest was defined by two observers in theperfusion area of the LAD and LCx arteries. Theseregions contained between 500 and 1,500 pixels andwere placed carefully to exclude the large andmiddle-sized coronary arteries, the aortic root, thecoronary sinus, and areas that were perfused by boththe LAD and LCx arteries. The same pair of regionsand the same threshold were used for baseline,hyperemia, and stenosis images.Mean arrival time (AT) and peak contrast density

(CD) were calculated from the regions of interest.The mean AT obtained in cardiac cycle units wasconverted into seconds by multiplying the units by thetrue cycle time. Pixels appearing in the first cardiaccycle after contrast administration were assigned anAT of 0.5 cycle, those in the second cycle an AT of1.5 cycles, and so on. Peak CD was defined as themaximal density achieved during the entire series.From these measurements, a relative coronary flowindex (F1) was obtained 1-4,10 as FI=CD/AT (I/sec).CFR was defined as hyperemia-to-baseline Fl and

was calculated 1,2,7,8 as:

Flhyper CDhyper ATbasal

Flbasal CDbasal AThyperMicrosphere technique. Myocardial perfusion was

determined with the reference withdrawal techniqueusing commercially available microspheres (15-,mdiameter; New England Nuclear, Boston) labeledwith "3Sn, 46Sc, 14"Ce, or 85Sr. For each flow determi-nation, approximately 3 x 106 microspheres sus-pended in 10 ml of 370 C saline was used. The samplewas carefully ultrasonicated and vortexed and theninjected over 60 seconds through the left atrial line.At the same time, a reference arterial sample waswithdrawn from the femoral artery (7.6 ml/min) usinga Harvard withdrawal pump. Nine tissue sampleswere obtained from each perfusion bed (LAD andLCx perfusion areas), and transmural as well assubendocardial, midmyocardial, and subepicardialblood flows were determined. All tissue samples wereweighed on a Sartorius digital balance (Goettingen,FRG). Blood samples were hemolyzed with KOHand desiccated at 700 C for 3-5 days. Radioactivecounts were determined with a Tracor Model 1185gamma scintillation counter (Tracor Instruments,Austin, Tex.), and blood flow was calculated" as



ownloaded from



Qm= (cmQr)/cr, where Qm is myocardial blood flow(ml/min), cm is counts in the tissue sample (counts/min), Qr is withdrawal rate of the reference bloodsample (ml/min), and Cr iS counts in the referenceblood sample (counts/min). Blood flow per gram oftissue was calculated by dividing Qm by the weight ofthe tissue sample.10

Study ProtocolBaseline measurements were carried out after

instrumentation was completed. Left ventricularpressure, its first derivative (dP/dt), aortic pressure,electromagnetic coronary blood flow of the LAD andLCx arteries, and a standard lead of the electrocar-diogram were recorded (Figure 1) on a Gouldrecorder (model 2800S, Gould Electronics, Cleve-land), which was interfaced to an IBM-AT modifiedfor on-line signal digitization at 200 Hz per channel.In each animal, 10 beats were averaged and stored ondisk for further analysis. Microspheres then wereinjected over 60 seconds, and hemodynamic data of10 cardiac cycles were digitized and averaged duringthe injection period. After microsphere injection wascompleted, digital angiography was performed. Therespirator pump was turned off during angiography,and only those sequences with good quality coronaryangiograms and without premature beats wereselected for further analysis.Hyperemia was induced by infusion through the

jugular sheath of 1 mg/kg/min adenosine (Sigma).Adenosine was dissolved in warmed saline and washeated during infusion to prevent precipitation. Theinfusion rate was considered to cause maximal coro-nary vasodilation because contrast material injectionhad no further hyperemic effect during adenosineinfusion (Figure 1). Because adenosine causedperipheral vasodilation with a significant drop inaortic pressure, the descending aorta was constrictedwith a Blalock clamp to maintain peak aortic pres-sure at a constant level during the experiment.

Coronary stenosis was induced by a screw occluderduring adenosine infusion. This occluder was tight-ened until electromagnetically measured coronaryblood flow had fallen to values approximating controlflow values before adenosine infusion. After a steady-state pressure was reached, flow data were recordedand stored on disk. Microspheres then were injectedand digital angiography was carried out as describedpreviously.12 A second control run in the presence ofcoronary stenosis was obtained after turning off theadenosine infusion (Figure 1). Aortic constrictionusually was released at this stage to keep peak aorticpressure constant. Then pressure and flow data of 10cardiac cycles were digitized, averaged, and stored ondisk. Microsphere injection and digital angiographywere performed thereafter. The experiments wereconcluded by the administration of an overdose ofpentobarbital and potassium chloride.

Thirty-four mongrel dogs were studied, but theprotocol was completed in only 22 (44 arterial beds).Of these, data from 24 beds were excluded because

of changes in electromagnetic blood flow measure-ments greater than 10% during microsphere injec-tions. In three dogs (six arterial beds), technicaldifficulties occurred with the withdrawal of a refer-ence sample during microsphere injection; one dog(two arterial beds) was excluded because of a tech-nically inadequate parametric image. Thus, thisreport is based on 12 arterial beds from nine dogs.

StatisticsComparisons of hemodynamic variables and flow

data at rest and during hyperemia under controlconditions and after coronary constriction were per-formed by a two-way analysis of variance for repeatedmeasures. CFR at rest and after coronary constric-tion were compared using a Wilcoxon rank sum test.A least-squares linear regression was used to com-pare the CFR from the three techniques (digitalangiography, electromagnetic flow probe, and micro-spheres). The intercept and slope, the correlationcoefficient, and the standard error of estimate (SEE)as well as the mean difference (accuracy) and thestandard deviation of the mean difference (precision)were determined for each comparison. Interobservervariability was calculated from the least-squares lin-ear regression analysis of results obtained by twoindependent observers. Results were considered tobe significant at p<0.05. Results in tables arereported as mean±+1 SD.

ResultsRepresentative digital angiograms are shown in

Figure 2. Functional images are shown at baseline,during adenosine infusion, and after coronary con-striction of the LCx artery at rest and during adeno-sine administration.

HemodynamicsTable 1 shows standard hemodynamic data taken

at rest (baseline), during adenosine infusion, andafter coronary constriction at rest and during aden-osine infusion. Heart rate decreased significantlyduring adenosine infusion under baseline conditionsand after coronary constriction. Parallel to thedecrease in heart rate, left ventricular end-diastolicpressure increased (p<0.001) during adenosine infu-sion; after coronary constriction, left ventricular end-diastolic pressure increased significantly (p<0.05)when compared with baseline and tended to increasefurther, but not significantly, after adenosine infu-sion. Left ventricular peak systolic pressure remainedunchanged during the experiment through use of theaortic constrictor. Mean aortic pressure, however,decreased significantly during adenosine infusionbecause of peripheral vasodilation.

Coronary Flow MeasurementsTable 2 shows coronary blood flow data under

control conditions and after coronary constriction atrest (baseline) and during adenosine infusion. Coro-nary blood flow (ml/g/min) increased significantly



ownloaded from



FIGURE 2. Representative digitalangiograms in the left posterioroblique projection at baseline(upper left panel), during adeno-sine infusion (upper right panel),and after coronary constriction ofthe circumflex coronary artery at rest(lower left panel) and duringadenosine infusion (lower rightpanel).

during adenosine infusion. Flow data were in thesame range (NS) for the electromagnetic flowmeterand microsphere techniques. Absolute coronaryblood flow cannot be measured with parametricimaging and, therefore, a coronary Fl (as defined in"Materials and Methods") is reported for parametricimaging in Table 2. CFR on the average was similar(NS) with all three techniques, ranging from 3.71(microspheres) to 4.17 (parametric imaging). Coronaryconstriction was associated with a significant reductionin hyperemic blood flow, but resting coronary flow wasnot diflerent from baseline. CFR was reduced signifi-cantly by all measurement methods after coronaryconstriction when compared with baseline.

Correlations Between Electromagnetic Flow Probe,Microspheres, and Parametric Imaging

Coronary blood flow. Normalized coronary bloodflow (electromagnetic flowmeter) and transmuralmyocardial blood flow (microspheres) showed a goodcorrelation (Figure 3) (r=0.87), with an SEE of 0.78ml/g/min (36% of mean blood flow). Correlationbetween coronary Fl (parametric imaging) and trans-mural blood flow (microspheres) was good (Figure 3),but parametric imaging underestimated high bloodflows. When a semilogarithmic plot was used, the

correlation coefficient between coronary Fl and thenatural logarithm of transmural blood flow was 0.81,with an SEE of 0.32 (40% of the mean blood flow).

Coronaryflow reserve. There was a good correlation(Figure 4) between CFR measured with the electro-magnetic flowmeter and microspheres (r=0.82), withan SEE of 0.93 (28% of the mean CFR). Thecorrelation of CFR determined with parametricimaging and microspheres was only moderate(r=0.54), with an SEE of 1.46 (39% of the meanCFR). The correlation of electromagnetically derivedCFR and parametric imaging CFR was somewhatbetter (r=0.68) and was associated with an SEE of1.35 (40% of the mean CFR) (Figure 5).The angiographic flow reserve parameters (Figure

6) such as contrast density ratio, arrival time ratio,and Fl ratio (CFR) also showed a weak correlationwith the CFR estimates based on microsphere-derived data; the correlation coefficient was 0.48 forthe contrast density ratio, 0.47 for the arrival timeratio, and 0.54 for the Fl ratio.

Accuracy and PrecisionTable 3 shows accuracy and precision among the

three techniques. Coronary blood flow was signifi-cantly less when it was determined by the flowmeter



ownloaded from


Hess et al Parametric Imaging and Coronary Flow Reserve

TABLE 1. Hemodynamic Data

Heart rate LVEDP LVSP MAP(beats/min) (mm Hg) (mm Hg) (mm Hg)

BaselineDSA 140±10 18±7 155±18 137±17Microspheres 140±11 18±9 154±17 135±16

HyperemiaDSA 122±11* 23±9t 145±22 114±21tMicrospheres 121±12* 21±8 147±19 116±20t

Stenosis baselineDSA 134±12 21±11 154±19 133±20Microspheres 136±13 22±11 156±18 134±20

Stenosis hyperemiaDSA 120±13* 25±13 146±29 118±28Microspheres 121±13* 24±13 147±29 120±25

Data were acquired at rest (baseline), during adenosine infusion (hyperemia), and after coronary constriction at rest(stenosis baseline) and during adenosine infusion (stenosis hyperemia). Data are reported just before acquisition ofdigital angiograms (DSA) and during microsphere injection. LVEDP, left ventricular end-diastolic pressure; LVSP, leftventricular peak systolic pressure; MAP, mean aortic pressure.

*p<0.001 vs. rest.tp<0.01 vs. rest.tp<0.05 vs. rest.

than by the microsphere technique. The mean dif-ference (accuracy) of coronary blood flow betweenthe electromagnetic flow probe and microsphereswas -0.68 ml/g/min (p<0.001 versus zero) with an

SD of difference (precision) of 0.94 ml/g/min. Themean difference in CFR between the electromag-netic flowmeter and parametric imaging measure-ments was smaller (-0.10) than between the flow-meter and microspheres (0.22) or microspheres andparametric imaging (-0.64), but none of these was

significantly different from zero; that is, both theelectromagnetic flowmeter and the parametricimages showed good accuracy. The SD of differenceor precision for CFR measurements was betterbetween the electromagnetic flowmeter and micro-sphere measurements (1.0) than between micro-spheres and parametric imaging (1.7).

Observer VartiabilityThere were excellent correlations between the

two independent observers for intensity fraction(r=0.97), arrival time (r= 0.93), coronary Fl (r=0.95),and CFR (r=0.90) as assessed by parametric imaging(Figure 7). The SEE for CFR was 20% betweenobserver 1 and 2, the mean difference was 0.34, andthe SD of difference was 0.79.

DiscussionEstimates of stenosis severity using conventional

coronary arteriography are based on anatomic cri-teria that may not reflect the physiological signifi-cance of a coronary stenosis. The effect of diffuseatherosclerosis or eccentric lesions on CFR is dif-ficult to estimate solely from the anatomic informa-

TABLE 2. Coronary Blood Flow Data

EMF Microspheres DSA(ml/g/min) (ml/g/min) (sec-')

Control conditions (n=17)Baseline 1.08+0.35 1.54±0.58 0.41±0.23Hyperemia 3.90±1.24 4.99±1.48 1.40±0.32CFR 3.84±1.59 3.71±1.76 4.17±1.75

Coronary stenosis (n =7)Baseline 0.98±0.25 1.50±0.36 0.33±0.18

Hyperemia 1.83±0.63* 2.25±1.02* 0.78±0.19*

CFR 1.93±0.68t 1.53±0.75t 2.63±0.92tData were acquired from nine dogs under control conditions and after coronary constriction at rest (baseline) and

during adenosine infusion (hyperemia). Flow data for electromagnetic flow probe (EMF) are reported duringmicrosphere injection. DSA, digital subtraction angiography (parametric imaging); CFR, coronary flow reserve

(hyperemia/rest).*p<0.001 vs. control conditions.tp<0.Ol vs. control conditions.

1443



ownloaded from



8.0- Y=.72X+0.13 R=.87 X 2.5- 0o SEE=0.78 P(.0001 W Y=.66LnX+0.24 R-All0~~~~~~~~~~SEE=0.32 P(.0001

w 6.0- * Z2E ~ ** 1.5--

0.0 0.00~ 11.L2.0 1.0---E~~2.0--

Li- ~~~~~~~~<0.51

0.0 0.00.0 2.0 4.0 6.0 8.0 0.0 2.0 4.0 6.0 8.0

MYOCARDIAL PERFUSION(mi/gm/min)

tion derived from coronary arteriography. Severalattempts have been made to directly quantify CFRin patients with coronary artery disease by gasclearance methods, thermodilution, digital angiog-raphy, Doppler flow techniques, positron emissiontomography, and newer approaches such as con-trast echocardiography, ultrafast computed tomog-raphy, and magnetic resonance imaging.7,8 Eachtechnique is characterized by certain limitations,and none has been used thus far to assess thefunctional significance of a coronary stenosis indaily clinical practice.Use of parametric imaging to assess CFR is concep-

tually appealing, because the same technique can beused to determine anatomy and the functional signif-icance of a stenosis. Coronary arteriography remainsthe gold standard for determining the location andseverity of coronary artery disease and provides thebasic information necessary for performing coronarybypass surgery. Thus, the combination of quantitativecoronary arteriography and parametric imaging wouldseem to have considerable potential for widespreadclinical application. There has been, however, somedebate on the value and limitations of parametricimaging for assessing CFR.7,8,12-14 For example, Nishi-mura et al14 observed that videodensitometric mea-surements of CFR do not accurately reflect CFRmeasured by microspheres.14 Accordingly, the pur-pose of the present study was to carefully reevaluate

10.0-EMF PROBE CMAPY=.72X+ 1.07 R=.82 Y=.50X+2.19 R=.54

8.0- SEE=0.93 PC0001 SEE=1.46 P=.0068

(94.0-6 *

4.0 6.0 8.0 0.0 2.0 4.0 6.0 8.0 14

MICROSPHERE CFR

FIGURE 3. Correlation between transmuralblood flow (microspheres) and nornalizedcoronary blood flow (electromagnetic flowprobe [EMF]) (left panel) and coronaryflowindex (parametric imaging [CMAP]) (rightpanel). Correlation was fairfor both compar-isons, with a correlation coefficient of 0.87and 0.81, respectively. Because large transmu-ral blood flows were underestimated by para-metric imaging, a semilogarithmic plotbetween parametric imaging and microspheredata was chosen. SEE was 0.78 ml/glmin and0.32 ml/g/min, respectively.

the accuracy and precision of a specific parametricimaging method for assessing CFR in the experimen-tal animal and to compare this method to two stan-dard techniques, electromagnetic flowmeters andmicrosphere measurements.

Coronary Blood Flow and Coronary Flow ReserveIn the present study, coronary blood flow was

examined over a wide range in normal and stenoticconditions both at rest and during hyperemiainduced by adenosine. Comparisons were madebetween parametric imaging, electromagnetic flow-meters, and microspheres under steady-state condi-tions with minimal changes in the hemodynamicdeterminants of coronary blood flow (Table 1).Radiographic measurements of coronary Fl andCFR were made during the first 1.5-5 seconds ofcontrast material injection in an effort to minimizethe effect of contrast medium on coronary flow.'There was some scatter between the electromag-

netic flowmeter and microsphere measurements ofabsolute flow (Figure 3), with an SEE of 0.8 ml/g/minand an SD of difference of 0.9 ml/g/min. The scattermay be explained, at least in part, by the fact thatmuscle mass of the perfusion bed of the LAD andLCx arteries had to be determined for normalizationof electromagnetic flow data and by the fact thatcollateral blood flow is not taken into account by theelectromagnetic flowmeter. Some scatter also was

FIGURE 4. Correlation of coronary flowreserve (CFR) between electromagnetic flowprobe (EMF probe) and microspheres (leftpanel) and between parametrtic imaging(CAMP) and microspheres (right panel).Correlation between EMF and microspheredata is clearly better (r=0.82) than betweenCMAP and microspheres (r=0.54). SEE is0.93 and 1.46, respectively.



ownloaded from



10.0 Y=76X+O.89 successful acquisition of adequate microsphere dataY=1.76X089 R=.68 and digital angiograms.SEE=1.35 P=.0002

8.0 In general, some inaccuracy and imprecisionLY / results whenever radiographic techniques are used tolv * *-<0analyze two-dimensional images of a three-dimen-(9 6.0--- sional perfusion bed. In the left posterior obliquen * 4 projection used in most studies, the area of interest< 4.0 for measuring coronary blood flow represents an

0* / integral of different perfusion areas in the postero-2.0 t = wZ lateral wall, including subendocardial and subepicar-

J0 dial perfusion zones. Consequently, defining the pre-

a cise three-dimensional location of the myocardialo0. - I , perfusion zone under investigation is difficult. Addi-

0.0 2.0 4.0 6.0 8.0 10.0 tionally, recirculation of contrast through the coro-

EMF PROBE CFR nary sinus may cause difficulties with density mea-

FIGURE 5. Correlation of coronary flow reserve (CFR)between parametric imaging (CMAP) and electromagneticflow probe (EMF). The correlation is moderate, with a zcorrelation coefficient of 0.68 and an SEE of 1.35. 0 3.0 Y-0X+1.46 Rn.48

SEE-0.46 P-.0171

reported between coronary blood flow measurements < 2.5using the microsphere technique and flow measure- 0 * *ments by collection of coronary venous return in dogs*and sheep.15 The comparison of coronary Fl (para- _r 2.0metric imaging) and microsphere flow data also zshowed some variations, with an SEE of 0.32. Per- 1.5haps most importantly, high coronary blood flows Z 0were underestimated by the Fl (Figure 3). 1.0CFR estimates based on the electromagnetic flow-

meter data paralleled the microsphere-derived data 3.0 Y-i.13X+1.41 R-.47well but tended to underestimate CFR measured SEE-0.46 P-.0194with microspheres at higher flow rates. The correla-tion coefficient was 0.82, with an SEE of 0.93 (Figure ° i4). The correlation between parametric imaging and t*microspheres was characterized by greater scatter 55er 2.0 0 i(Figure 4) with a larger SEE (1.46). The correlationbetween parametric imaging and electromagnetic 1.5flowmeter measurements was slightly better (Figure l5). The correlations clearly are not as strong as those 1.0that have been reported previously.2 The mean dif- 8.0 *Y^ X+2.149R-S4ference (accuracy) was, however, good (-0.1),although the SD of difference (precision) was poor er 6 0 *(1.4) (Table 3). Interobserver variability, however,was excellent and cannot explain the scatter in mea- 0suring CFR by parametric imaging. CL 4.0---

Other angiographic flow reserve parameters such 2 *as intensity fraction ratio and arrival time ratio werecharacterized by weaker correlations with the micro- 2.0 emsphere data (Figure 6) than the ratio of the two.Similarly poor correlations and wide scatter between o.o . 4. , . , .0videodensitometric measurements and microsphere 0.0 2.0 4.0 6.0 8.0data were reported by Nishimura and coworkers.14 MICROSPHERE CFR

FIGURE 6. Correlations of coronary flow reserve (CER)Limitations between parametric imaging (CMAP) (ordinate) and micro-The limitations of this study relate to three broad spheres (abscissa). CFR was determined from the intensity

areas: the use of radiographic techniques, the unique fraction ratio (top panel), the arnival time ratio (middlelimitations of the specific parametric imaging modal- panel), and the ratio of the intensity fraction divided by theity chosen for study, and the limitations imposed by arrival time (bottom panel). All three comparisons showedthe model and protocol that were necessary for the weak correlations.



ownloaded from



TABLE 3. Accuracy and Precision

Mean absoluteMean difference SD difference

Coronary blood flow (n=48)EMF vs. microspheres -0.68* 0.94 0.90+0.73*

Coronary flow reserve (n=24)EMF vs. microspheres 0.22 1.04 0.71+±0.78*EMF vs. DSA -0.10 1.38 1.05±0.88*DSA vs. microspheres -0.64 1.70 1.39±1.14*

Mean differences (accuracy) and standard deviation (SD) of difference (precision) for comparison of coronary bloodflow and coronary flow reserve between electromagnetic flow probe (EMF), microspheres, and digital subtractionangiography (DSA). Mean absolute difference values are mean±SD.

*p<0.001 versus zero (paired t test).

surements. Although it has been assumed that theinaccuracies of videodensitometry should cancel eachother out when density ratios are calculated (i.e.,densities at baseline and hyperemia), this may notalways be the case, because the amount of contrastmaterial in the myocardium during hyperemic imagesis likely to be greater. Furthermore, the spillover canvary from injection to injection depending on abso-lute coronary blood flow and coronary vascular resis-tance. These factors will cause differences in theeffects of scatter and beam hardening on the densityvalues.The specific parametric method we used is limited

by the temporal imprecision inherent in once-per-cardiac-cycle image analysis of appearance time anddensity. This factor probably is most important inexplaining the systematic underestimation of highflows noted in Figure 3. These errors would tend tocause overestimation of appearance time and under-estimation of contrast density. Cusma et a14 sug-gested that improvement in the temporal resolutionof the method could be achieved by interpolationmethods and that this could simultaneously improve

determination of actual appearance time and peakdensity. Such methods were not used in this studybecause we set out to examine the methodologyoriginally proposed in an earlier report2 in which thisfactor was taken into account only by assigning valuesof 0.5, 1.5, and 2.5 cycles to the first, second, andthird images, etc. The limitation imposed by thisfactor may be more notable in the current study thanin prior studies because maximal flow was sustainedby continuous infusion of adenosine and mainte-nance of perfusion pressure, whereas prior validationattempts used submaximal stimuli such as atrialpacing1 and graded contrast injections.2An underlying prerequisite for this method of

parametric CFR imaging is that contrast materialmust virtually displace blood. The need for this wasrecognized in prior studies2 and was explicitly eluci-dated by Cusma et al.4 This prerequisite can beapproximated, but it is virtually impossible to guar-antee it, especially during hyperemic flow, despitethe use of maximally tolerable doses of contrastinjected rapidly. Although studies with overt stream-ing and/or subselective contrast injection were omit-

Z .0 SEE-0.07 P(.0001 *

LL 0.6 - z ;

04

zLii- 0.2z

0.00.0 0.2 0.4 0.6 0.8 1.0

2.5---Y-1.07X-0.01 R-.95SEE-0.17 P(.0001

21.0---vLiL01.5

o 1.0. .

0.00.0 0.5 1.0 1.5 2.0 2.5

2.0.0a'In 1.5.

il.o1

z 0.5

0.00.

8.0-

6.0-

Lo- 4.00

2.0

0.04

Y-.91XT0.05 R-.93SEE-0.15 P(0001

0

.0 0.5 1.0 1.5 2.0Y-.84X+0. R-.90 *SEE-0.75 P(.0001

* *0 *

0

FIGURE 7. Interobserver variability for intensityfraction (upper left panel), arrival time (upperright panel), flow index (lower left panel), andcoronary flow reserve (CFR) (lower right panel) as

measured with parametric imaging.

0.0 2.0 4.0 6.0 8.0

OBSERVER 1

CN

ulJ

LUJ

m0



ownloaded from



ted from analysis, the possibility of more subtlestreaming or preferential flow of contrast into one orthe other coronary arteries when contrast wasinjected into the main stem of the arterial bed couldnot be completely eliminated. It is possible that thislimitation might be overcome by relating the peakdensity in the myocardial bed to the peak density inthe proximal coronary arteries so that the input ofcontrast and the actual degree of blood displacementachieved by the bolus injection can be taken intoaccount. This approach would require more preciseregistration of the proximal arteries and avoidance ofsegments superimposed by reflux into the aortic rootto avoid invalidating the intraluminal density values.The experimental model and protocol we used

also presented some unavoidable limitations. Themicrosphere technique requires a prolonged steadystate for injection and collection of samples. Thisprecluded acquisition of the parametric imagessimultaneously with the microsphere injections.Moreover, the prolonged, systemic infusion ofadenosine and the maximal hyperemia of the entirevascular bed may have promoted substantial collat-eral flow, a factor known to diminish the CFRvalue.16 This possibility is supported by the generalunderestimation of microsphere-determined perfu-sion by flowmeter measurements. In theory, mea-surement of contrast density in the myocardiumrather than in the arteries should have reducederror due to collateral perfusion. The effects of thecontrast agent on red cell morphology and theresultant effects on the rheological characteristicsof collateral flow, however, may have led to anunderrepresentation of collateral flow by the para-metric imaging technique, which always was per-formed after the microsphere determination ofblood flow. In addition, although nonionic contrastwas used, such agents can cause a diminution ofblood flow that may have caused underestimation ofmaximal CFR.17 Capillary solubility, uptake by myo-cardial cells, extravasation into the extracellularspace, and the effect of the high viscosity andosmotic pressure of the dye itself on the smallvessels all may influence the densitometric assess-ment of perfusion.'4 Finally, the variability ofmicrosphere measurements11 and the regional het-erogeneity of CFR measures by microspheres18cannot be ignored, although this technique is con-sidered the gold standard for measures of myocar-dial perfusion.

Comparison to Prior Validation StudiesWith these limitations in mind, a more meaningful

comparison of this validation study can be made toprior studies. The initial clinical validation of theappearance time concept for measures of flowchanges was undertaken using atrial pacing as thehyperemic stimulus and coronary sinus blood flow asthe reference measure.' Because atrial pacing doesnot induce maximal hyperemia, CFR estimates prob-ably were not as adversely afected by the factors of

high flow and collateralization, which may haveplayed a major role in the present study. On the otherhand, the coronary sinus flow measurements cannotbe considered a true gold standard for measuringmyocardial perfusion. Finally, because the study didnot include an analysis of density, the inaccuracies ofvideodensitometry could not have affected theresults.

Subsequently, another validation study was under-taken in dogs, using contrast material as the vasodi-lating substance.2 A range of CFR values was createdby graded doses of contrast material, whereas in ourcurrent study, flow was altered by imposing a stenosison a fully vasodilated bed. This means that onlysubmaximally dilated states were evaluated in theearlier study and that problems potentially related tohigh flow and collateralization probably were notcritical. Furthermore, in the absence of maximalvasodilation, blood may have been more adequatelydisplaced by contrast material, thereby minimizingthe errors of videodensitometry and the errorscaused by potentially insufficient contrast injection.

In the study of Cusma et al,4 electromagneticflowmeter measurements were used as the only ref-erence standard. A curve interpolation method, dis-cussed above, was used to improve density andtemporal resolution.4 Intracoronary boluses of aden-osine were used, and the transient hyperemia thatresulted may have minimized difficulties attributableto collateral flow.4 In addition, contrast doses wereinjected more rapidly (5-10 ml/sec) than in thecurrent study (4 ml/sec). Thus, methodological dif-ferences may explain part of the reason we failed toobserve correlations similar to those previouslyreported.Nishimura et al'4 undertook a study comparing

microsphere-derived blood flow estimates withparametric imaging. Differences from our methodsinclude use of gamma variate curve fitting to con-struct time-density curves, use of the one-half max-imum density value to determine the time parame-ter, use of interpolation to determine peak density,use of highly disparate stimuli to induce changes inblood flow (intracoronary adenosine and vasopres-sin, rapid atrial pacing), postprocessing from video-taped images, and inclusion of a paucity of high flowreserve values (only three values greater than 2.0).Nonetheless, the conclusion of Nishimura et al'4 islargely similar to our own; that is, parametric imageanalysis is reproducible, but the CFR values corre-late only moderately well with microsphere-determined CFR ratios.

Clinical ImplicationsParametric imaging enables estimation of CFR

that correlates with microsphere-derived measure offlow reserve only moderately well over a wide rangeof values in this animal model. On the average,however, CFR results obtained by parametric imag-ing were not different from values measured by theelectromagnetic flowmeter and microsphere tech-



ownloaded from



niques. The specific parametric technique has inher-ent limitations that lead to imprecision in the mea-surements. At the present time, only changes of 40%or greater in CFR can be expected to be demon-strated reproducibly with this specific parametricmethod. Efforts directed at improving the accuracy ofboth time and density measurements, minimizing theeffects induced by contrast material and collateralflow, and avoiding the need for full displacement ofblood by contrast material will be required to makethe methodology more precise.

References1. Vogel RA, LeFree MT, Bates ER, O'Neill WW, Roster R,

Kirlin P, Smith D, Pitt B: Application of digital techniques toselective coronary arteriography: Use of myocardial contrastappearance time to measure coronary flow reserve. Am HeartJ 1984;107:153-164

2. Hodgson JMcB, LeGrand V, Bates ER, Mancini GBJ, AueronFM, O'Neill WW, Simon SB, Beauman GJ, LeFree MT, VogelRA: Validation in dogs of a rapid digital angiographic tech-nique to measure relative coronary blood flow during routinecardiac catheterization. Am J Cardiol 1985;55:188-193

3. LeGrand V, Aueron FM, Bates ER, O'Neill WW, HodgsonJMcB, Mancini GBJ, Vogel RA: Value of exercise radionu-clide ventriculography and thallium-201 scintigraphy in evalu-ating successful coronary angioplasty: Comparison with coro-nary flow reserve, translesional gradient and percent diameterstenosis. Eur Heart J 1987;8:329-339

4. Cusma JT, Toggart EJ, Folts JD, Peppler WW, HangiandreouNJ, Lee C, Mistretta CA: Digital subtraction angiographicimaging of coronary flow reserve. Circulation 1987;75:461-472

5. Zijlstra F, van Ommeren J, Reiber JHC, Serruys PW: Does thequantitative assessment of coronary artery dimensions predictthe physiologic significance of a coronary stenosis? Circulation1987;75:1154-1161

6. Wilson RF, Marcus ML, White CW: Prediction of the physi-ologic significance of coronary arterial lesions by quantitativelesion geometry in patients with limited coronary artery dis-ease. Circulation 1987;75:723-732

7. Marcus ML, Wilson RF, White CW: Methods of measurementof myocardial blood flow in patients: A critical review. Circu-lation 1987;76:245-253

8. Klocke FJ: Measurements of coronary flow reserve: Definingpathophysiology versus making decisions about patient care.Circulation 1987;76:1183-1189

9. Gould KL, Goldstein RA, Mullani NA, Kirkeeide RL, WongWH, Tewson TJ, Berridge MS, Bolomey LA, Hartz RK,Smalling RW, Fuentes F, Nishikawa A: Noninvasive assess-ment of coronary stenoses by myocardial perfusion imagingduring pharmacologic coronary vasodilation. VIII: Clinicalfeasibility of positron cardiac imaging without a cyclotronusing generator-produced rubidium-82. J Am Coll Cardiol1986;7:775-778

10. Mancini GBJ, McGillem MJ, DeBoe SF, Gallagher KP: Thediastolic hyperemic flow vs. pressure relation in anesthetizeddogs: A new index of coronary stenosis severity and flowreserve. Circulation 1989;80:941-950

11. Heyman MA, Payne BD, Hoffman JIE, Rudolph AM: Bloodflow measurements with radionuclide-labeled particles. ProgCardiovasc Dis 1977;20:55-79

12. Mancini GBJ, Higgins CB: Digital subtraction angiography: Areview of cardiac applications. Prog Cardiovasc Dis 1985;18:111-141

13. Nissen SE, Elion JL, Booth LE, Evans J, DeMaria AN: Valueand limitations of computer analysis of digital subtractionangiography in the assessment of coronary flow reserve. Cir-culation 1986;73:562-571

14. Nishimura RA, Rogers PJ, Holmes DR, Gehring DG, BoveAA: Assessment of myocardial perfusion by videodensitome-try in the canine model. JAm Coll Cardiol 1987;9:891-897

15. Utley J, Carlson EL, Hoffmann JIE, Martinez HM, BuckbergGD: Total and regional myocardial blood flow measurementswith 25 ,u, 15 ,u 9 g and filtered 1-10 g diameter microspheresand antipyrine in dogs and sheep. Circ Res 1974;34:391-405

16. LeGrand V, Mancini GBJ, Bates ER, Hodgson JMcB, GrossMD, Vogel RA: Comparative study of coronary flow reserve,coronary anatomy and results of radionuclide exercise tests inpatients with coronary artery disease. J Am Coil Cardiol1986;8:1022-1032

17. Friedman HZ, DeBoe SF, McGillem MJ, Mancini GBJ: Theimmediate effects of iohexol on coronary blood flow andmyocardial function. Circulation 1986;74:1416-1423

18. Austin RE, Aldea GS, Coggins DL, Flynn AE, Hoffman JIE:Profound variation of regional coronary flow reserve(abstract). Circulation 1988;78(suppl II):II-82

KEY WORDS * coronary flow reserve * digital subtractionangiography * microspheres * electromagnetic flow measurements



ownloaded from


O M Hess, M J McGillem, S F DeBoe, I M Pinto, K P Gallagher and G B ManciniDetermination of coronary flow reserve by parametric imaging.

Print ISSN: 0009-7322. Online ISSN: 1524-4539 Copyright © 1990 American Heart Association, Inc. All rights reserved.

is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231Circulation doi: 10.1161/01.CIR.82.4.1438

1990;82:1438-1448Circulation.

http://circ.ahajournals.org/content/82/4/1438the World Wide Web at:

The online version of this article, along with updated information and services, is located on

http://circ.ahajournals.org//subscriptions/

is online at: Circulation Information about subscribing to Subscriptions:

http://www.lww.com/reprints Information about reprints can be found online at: Reprints:

document. Permissions and Rights Question and Answer information about this process is available in the

located, click Request Permissions in the middle column of the Web page under Services. FurtherEditorial Office. Once the online version of the published article for which permission is being requested is

can be obtained via RightsLink, a service of the Copyright Clearance Center, not theCirculationpublished in Requests for permissions to reproduce figures, tables, or portions of articles originallyPermissions:



ownloaded from

http://circ.ahajournals.org/content/82/4/1438

http://www.ahajournals.org/site/rights/

http://www.lww.com/reprints

http://circ.ahajournals.org//subscriptions/


Documents

Determination Coronary FlowReserve Parametric …circ.ahajournals.org/content/circulationaha/82/4/1438.full.pdfmagnetic flowmeters and myocardial flow reserve ... Microspheres were