DIAGNOSTIC METHODSCORONARY ARTERY DISEASE

Videodensitometric analysis of human coronarystenoses: validation in vivo by intraoperative high-frequency epicardial echocardiographyMARYL R. JOHNSON, M.D., DAVID D. MCPHERSON, M.D., STEVEN R. FLEAGLE, B.S.E.E.,MICHELLE M. HUNT, B.S., LOREN F. HIRATZKA, M.D., RICHARD E. KERBER, M.D.,MELVIN L. MARCUS, M.D., STEVE M. COLLINS, PH.D., AND DAVID J. SKORTON, M.D.

ABSTRACT Videodensitometry is a nongeometric method of coronary angiographic analysis that canbe used to provide an index of coronary luminal area. However, there are few direct studies in vivoof the relationship of videodensitometric data to independent measures of luminal area in humans.Although videodensitometry is theoretically independent of angiographic projection and luminal shape,validation of these assumptions in vivo is also limited. We therefore used intraoperative high-frequencyepicardial echocardiography, a technique that can directly determine human coronary luminal area andshape in vivo, to further validate videodensitometry. A total of 36 arterial segments in the left anteriordescending and right coronary arteries were studied by videodensitometry and high-frequency echo-cardiography. Videodensitometry was performed on angiograms in which the arterial segment ofinterest was not markedly foreshortened and was uniformly filled with contrast. In 22 discrete lesions(13 with circular lumens and nine with oval or complex lumens), videodensitometric and echocar-diographic measures of luminal area correlated well (r = .86). In 33 coronary arterial segments, theeffect of angiographic projection on videodensitometry was determined by comparison of the resultsof videodensitometry performed on left anterior oblique vs right anterior oblique angiograms of thesegments. Here too, the correlation was good (r = .94, y = 1.04x + 0.002). The good correlationof left anterior oblique with right anterior oblique videodensitometric results held true for lesions withcircular and oval or complex lumens. This study further validates the ability of videodensitometry toprovide an index of coronary luminal area and confirms in vivo previous assumptions that the resultsof videodensitometric analysis are independent of angiographic projection and luminal shape.Circulation 77, No. 2, 328-336, 1988.

RECENTLY, substantial investigative effort has beendirected toward developing approaches to better assessthe anatomic and physiologic significance of coronarylesions based on their angiographic appearance. Thiseffort has stemmed from increasing recognition ofproblems associated with the standard method of angio-graphic analysis, that of visually estimating percentdiameter stenosis. Measurements of percent stenosisFrom the Cardiovascular Image Processing and Ultrasonic Imaging

Laboratories, Cardiovascular Center, and Departments of Internal Med-icine, Electrical and Computer Engineering, Surgery, Radiology, andBiomedical Engineering, University of Iowa and Iowa City VeteransAdministration Medical Center, Iowa City.

Supported by grants HL00916 (Clinical Investigator Award, Dr.Johnson), HL35267, HL32295 (Specialized Center of Research inIschemic Heart Disease), and HL01290 (Research Career DevelopmentAward, Dr. Skorton) from the National Heart, Lung, and Blood Insti-tute. Support also provided by the Canadian Heart Foundation and theAlberta Heritage Foundation (Dr. McPherson).

Address for correspondence: Maryl R. Johnson, M.D., Cardiovas-cular Division, Department of Internal Medicine, University of IowaHospitals and Clinics, Iowa City, IA 52242.

Received June 2, 1987; revision accepted Nov. 5, 1987.

328

may be associated with substantial intraobserver andinterobserver variability. 1-4 In addition, measurementsof percent stenosis in patients with multivessel coro-nary artery disease may not correlate with coronarylesion geometry, as defined by the method of quanti-tative coronary arteriography described by Brown etal.', or lesion physiologic significance, as defined byintraoperative Doppler reactive hyperemia studies.6 7

Percent stenosis may not accurately reflect lesionphysiologic significance because the angiographically"normal" segment of the coronary artery may not befree of atherosclerotic disease. This has been shown inpathologic studies,8' 9 and we have recently confirmedthat diffuse coronary disease is present in some humancoronary arterial segments that appear normal or nearlynormal on coronary angiograms. 10 Thus, an absoluterather than relative measure of luminal area may be abetter predictor of coronary lesion physiologic signif-icance. Indeed, previous studies have shown that the

CIRCULATION

by guest on April 13, 2017

http://circ.ahajournals.org/D

ownloaded from

DIAGNOSTIC METHODS-CORONARY ARTERY DISEASE

minimal luminal cross-sectional area is one of themost important determinants of lesion physiologicsignificance.11

Geometric methods can define absolute size of thecoronary lumen. We have shown that absolute luminalsizes determined with geometric methods (such asquantitative coronary arteriography or computer-basededge definition followed by determination of arterialminimal diameter) correlate much better with coronarylesion physiologic significance than do measures of

7, 12,1percent stenosis. 13 However, geometric methodsdepend on exact definition of vessel edges and requireassumptions as to luminal shape. Therefore, inaccu-racies might occur in the evaluation of lesions withindistinct vessel edges or irregular luminal shapes.Although investigators have been working to makegeometric methods of angiographic analysis easier andfaster,14 many geometric methods remain time con-suming and difficult, and cannot be readily applied atthe time of cardiac catheterization.

Videodensitometry is a nongeometric approach tothe analysis of coronary cineangiograms. Videodensi-tometry consists of measurement of the x-ray attenu-ation through a contrast-filled coronary artery and fromthis information estimating the amount of contrast inthe vessel, which is an index of luminal size. 15 Video-densitometry is potentially advantageous for the eval-uation of human coronary cineangiograms for severalreasons. First, it can be used to estimate an absolute aswell as relative (i.e., percent stenosis) index of coro-nary luminal size. 16-19 In addition, videodensitometryis not dependent on exact definition of vessel edges andis theoretically free of assumptions as to shape of thelumen. Such characteristics may be important in ana-lyzing lesions immediately after thrombolysis or angio-plasty, when vessel borders are indistinct and the lumi-nal shape is irregular.20 Finally, videodensitometricanalysis is simple and can be rapidly applied, partic-ularly to digital coronary angiographic information,which is becoming more widely available.We have reported that the results of videodensito-

metric analysis of lesions in all three coronary arteriescorrelate well with coronary luminal area, as defined byquantitative coronary arteriography by the method ofBrown et al. ,5 and with lesion physiologic significance,as defined by Doppler studies of vasodilatorreserve. 1619 However, before videodensitometry canbe recommended for widespread clinical application,more specific, assumption-free validation in humans invivo is necessary.

High-frequency epicardial echocardiography is anintraoperative, ultrasonic technique that provides

Vol. 77, No. 2, February 1988

tomographic images of the coronary arteries. Thesetomographic images allow direct visualization of cor-onary arterial lumens and walls, and from the imagesaccurate and reproducible measurements of coronarylesion luminal areas, diameters, wall thicknesses, andshapes can be made.2' Luminal areas defined by high-frequency echocardiography are not dependent onassumptions of a particular luminal shape as are manyof the geometric methods of angiographic analysis.SFor this reason, we chose high-frequency epicardialechocardiography as a nonangiographic method of fur-ther validating in vivo the ability of videodensitometryto provide an index of coronary luminal area. In addi-tion, although videodensitometry is theoretically inde-pendent of angiographic projection and luminal shape,proof of this assumption in vivo is limited. High-frequency echocardiography allows analysis of coro-nary luminal shape. Therefore, we also used this tech-nique to evaluate the effect of angiographic projectionand luminal shape on the results of videodensitometricanalysis.

MethodsPatient population. One-hundred eight coronary arterial

segments in 47 patients had been evaluated by use of high-frequency echocardiography and were available for consid-eration for inclusion in this study. Fifty-five segments wereexcluded because no discrete lesion was seen on the corre-sponding coronary angiogram. In eight cases, the segment stud-ied with high-frequency echocardiography was distal to anangiographic total occlusion. In one case the segment studiedwith echocardiography was distal to a subtotal occlusion andshowed competitive flow (antegrade and collateral) on the angio-gram. One segment was excluded due to the presence of calciumin the lesion, four due to incomplete opacification on the angio-gram, and three due to vessel overlap on the angiogram. Theremaining 36 coronary arterial segments in 23 patients wereincluded in the study.Of the 23 patients in the study, all but two had multivessel

coronary artery disease. (Considering lesions greater than 50%diameter stenosis as significant, 16 patients had three-vesseldisease, five patients had two-vessel disease, one had one-vesseldisease, and one had no lesions greater than 50%.) To beincluded in the study, patients had to have angiographicallydefined disease in the left anterior descending or right coronaryartery. Although many of the patients had significant lesions inthe circumflex coronary artery as well, these could not be studiedby ultrasound due to the posterior location of the artery and thesize of the high-frequency echocardiographic probe currentlyused (20 cm long, 2 cm wide). Coronary angiograms of thearterial region of interest had to be adequate for evaluation bycomputerized videodensitometry (that is, angiograms had toshow the arterial segment with the plane of the vessel as parallelto the image intensifier as possible and the artery uniformly filledwith contrast). The patients also had to have good quality,high-frequency epicardial echocardiographic images of the arte-rial segment of interest. For the portion of the study in whichthe effect of angiographic projection on videodensitometric datawas evaluated, there also had to be adequate angiographic stud-ies of the arterial region of interest in nearly orthogonal (left

329

by guest on April 13, 2017

http://circ.ahajournals.org/D

ownloaded from

JOHNSON et al.

anterior oblique and right anterior oblique) views. That is, leftanterior oblique and right anterior oblique angiograms showingthe arterial segment of interest nearly parallel to the imageintensifier and uniformly filled with contrast had to be available.(Cranial or caudal angulated views were necessary to meet thesecriteria in 13 of the vascular segments included in the study.) Allpatients gave informed consent before the high-frequency epi-cardial echocardiographic studies. The protocol for intra-operative echocardiographic study of human coronary arterieswas approved by the Human Subjects Review Committee of theUniversity of Iowa.

Videodensitometric analysisCoronary angiography. Premedications in the 23 patients

included 0.6 mg sc atropine (in 19 patients), a mild sedative(diazepam in 20 patients, secobarbitol in one patient), and anantihistamine (promethazine in 15 patients, diphenhydramine insix patients). Nitroglycerin (0.4 mg sublingual) was given to 16of the patients immediately before coronary angiography. Fourof the patients received sublingual nifedipine in the catheter-ization laboratory. Coronary angiography was performed by thepercutaneous femoral technique described by Judkins.22 Theelectrocardiogram and central aortic pressure (measured througha contrast-filled coronary catheter) were recorded just before andafter each coronary cineangiogram. Angiograms were obtainedin multiple, single-plane projections with use of a Siemens x-raysystem with a 7 inch image intensifier (20 patients) or a Philipssystem with a 6 inch image intensifier (three patients). Angio-grams were recorded on 35 mm cine film at 60 frames/sec. Filmprocessing was carefully controlled. End-diastolic angiographicframes showing the arterial region of interest well opacified withcontrast and as nearly parallel to the image intensifier as possiblewere selected for videodensitometric analysis.

Computerized coronary videodensitometry. Videodensito-metric analysis was performed as previously described by ourgroup.16 Cine frames were digitized by use of a modified Van-guard cine film transport. The projected angiographic image wascoupled through selected optical lenses to a high-quality vidiconcamera (Cohu Model 8000). Cine frames were digitized by aDeAnza 8500 image-array processor interfaced with a PDP1 1/44 computer.

Before digitization of individual cine frames, the sensitivityof the vidicon camera (target control) was adjusted to preventsaturation in the brighter portions of a typical image and toensure a wide dynamic range in the digitized image. At thiscamera setting, calibration data relating mean gray level (dig-itized value) and optical density of the film were measured witha Kodak gray level step wedge with 21 optical density levelsdistributed uniformly between 0.05 and 3.10. This range encom-passed the film densities we observed. These data were enteredinto the computer, and a continuous calibration curve was cal-culated by linear interpolation. Individual cineangiographicframes were then digitized at a magnification of 3.7 X into a 512x 512 pixel matrix with 8-bit (256 levels) gray level resolution.(This represents a digitization resolution of about 0.06mm/pixel, depending on the geometric magnification at whichthe angiogram was obtained.) The digitized images were sub-sequently displayed on a video monitor for videodensitometricanalysis with use of an operator-interactive program. The imageswere analyzed by an observer who was unaware of the resultsof the high-frequency echocardiographic analysis.The operator identified the approximate centerline of the

arterial segment of interest and the length of the segment to beanalyzed. The computer then calculated and displayed an aver-age gray level profile from left to right across the artery per-pendicular to the centerline. Using this profile as an aid, theoperator identified the arterial borders as well as backgroundregions on both sides of the artery. The gray level profile for each

lesion was converted to an optical density profile with the useof the previously determined calibration curve. The opticaldensity for each point on the profile within the artery wascorrected for the optical density of the corresponding back-ground point defined by linearly interpolating between the aver-age optical density values for the background regions on bothsides of the artery. An integrated optical density value for thearterial segment of interest was obtained by summing the result-ant background-corrected optical densities over all points withinthe artery. This value was reported in optical density units.Ignoring the effects of x-ray scatter and assuming that the imageintensifier operates linearly, that image data are recorded in thelinear portion of the film characteristic curve, and that there isuniform mixing of contrast media and blood in the arteriallumen, this measure of integrated optical density is proportionalto luminal cross-sectional area. We have previously found thatvalues of integrated optical density determined in this way havegood intraobserver (r = .98) and interobserver (r = .87)reproducibility. 17

For vessels included in the portion of the study assessing theefects of angiographic projection on videodensitometric results,the above analysis was performed by the same observer onorthogonal angiographic projections of the coronary arterialsegment of interest.

High-frequency epicardial echocardiographyIntraoperative protocol. Patients taken to the operating room

for elective open heart surgery were managed according tostandard operative procedures. They were anesthetized, intu-bated, and ventilated with a mechanical respirator. A variety ofanesthetic agents was used, including nitrous oxide, halothane,morphine, and fentanyl. Radial artery pressure and the elec-trocardiogram were recorded and left atrial pressure was con-tinuously monitored. All operations were performed through amidline sternotomy. Just before the echocardiographic studies,intravenous heparin was administered to raise the activatedclotting time to 480 sec or more. Preparations were made toinstitute cardiopulmonary bypass should hemodynamic insta-bility occur; however, all patients remained hemodynamicallystable during the echocardiographic studies.A Biosound Surgiscan unit (Biosound Inc., Indianapolis)

with a 12 MHz probe was used to obtain the coronary arterialimages. The ultrasonic probe was sterilized with a standard1300 C ethylene oxide sterilizing procedure for an 18 hr cyclebefore intraoperative use.The surgeon scanned the diseased segments of the left anterior

descending and right coronary arteries in cross and longitudinalsection with the hand-held, high-frequency echocardiographicprobe. Probe location was referenced to external landmarks,primarily vessel branch points, and imaging of diseased arterialsegments in cross section was emphasized. (As noted previously,the circumflex coronary artery could not be scanned due to thesize of the echocardiographic probe.) Ultrasound system gainswere adjusted to make the interface between the arterial lumenand the intima as sharp as possible without blooming of theintima into the lumen. As the ultrasonic probe was moved alongthe diseased coronary segment, echocardiographic images(including a 3 mm calibration bar) were recorded on videotapefor subsequent playback and analysis. Probe location was alsodescribed and recorded.Image analysis. Analysis was limited to high-quality, cross-

sectional arterial images. Freeze-frame arterial images wereanalyzed with an IREX (IREX Med. Systems, Ramsey, NJ)digitizing system. The echocardiographic image was displayedin cross section at the point in the cardiac cycle when the arterialdiameter appeared largest.

For measurement purposes, the inner edge of the bright inter-face between the coronary lumen and the intima was identified

CIRCULATION330

by guest on April 13, 2017

http://circ.ahajournals.org/D

ownloaded from

DIAGNOSTIC METHODS-CORONARY ARTERY DISEASE

and the lumen was outlined on the luminal side of this interface.From this outline, luminal area was directly calculated. Fourluminal diameters at equiangular intervals (i.e., every 45degrees) around the circumference of the lumen were measured.With these diameter measurements, luminal shape was classifiedby calculating the maximum to minimum luminal diameterratio. Lumens having a maximum to minimum diameter ratioof less than 1.5:1 were considered by convention to be circular,those with a ratio of greater than 1 .5:1 were considered oval, andmarkedly irregular lumens were classified as complex. Examplesof echocardiographic images of lesions with circular, oval, andcomplex lumens are shown in figure 1. Although multiplefreeze-frame images along the diseased coronary segment wereanalyzed, only those frames showing the lumen with the smallestcross-sectional area were used for the comparisons made in thisstudy.

Study protocol and data analysisVideodensitometry vs echocardiographic luminal area. The

integrated optical densities of arterial segments with single,discrete lesions on the angiogram were compared with mea-surements of minimal coronary luminal area defined by high-frequency echocardiography of the same arterial segments. Onlyarterial segments with single, discrete lesions were included inthis portion of the study so that the segment ofthe artery analyzedby videodensitometry and high-frequency echocardiographycould be matched as precisely as possible. Analyses were per-formed for the entire group of lesions studied, including lesionswith luminal shapes defined as both circular and oval/complex.The comparison of integrated optical density vs echocardio-graphic luminal area was performed by linear regression analysis

Effects of angiographic projection and luminal shape. Inarterial segments for which left anterior oblique and right ante-rior oblique orthogonal angiographic projections of the arterialsegment of interest suitable for videodensitometric analysis wereavailable, the results of videodensitometry performed on the leftvs the right anterior oblique projection were compared. Thisanalysis was performed to determine if the specific angiographicprojection, i.e., left anterior oblique versus right anterioroblique, affected the results of videodensitometric analysis. lnaddition, to determine whether the effect of projection on video-densitometric data was affected by luminal shape, similar anal-yses were performed in the subgroups of arterial segmentsdefined as having circular and oval/complex lumens on thehigh-frequency echocardiogram. All analyses were performedwith linear regression techniques. The slopes and intercepts ofthe regression lines calculated for the comparison of videoden-sitometric results in different projections were compared withone and zero respectively, by use of t statistics.

ATHS

ResultsPatient characteristics. A total of 36 arterial segments

(12 in the left anterior descending coronary artery and24 in the right coronary artery) in 23 patients were

evaluated. The study population included 15 men andeight women with ages ranging from 35 to 76 years anda mean age of 58 years. The interval between cardiaccatheterization and surgery varied from 2 days to 16months, with a mean interval of 36 + 98 (mean ± SD)days. In all but three patients the interval betweencatheterization and surgery was less than 1 month, andin all but one patient the interval was less than 3months. At the time of coronary angiography, mean

systemic blood pressure was 77 2 mm Hg (mean +SEM) and mean heart rate was 75 + 2 beats/min. Atthe time the high-frequency echocardiographic studieswere performed in the operating room, mean bloodpressure was 76 2 mm Hg and mean heart rate was71 ± 2 beats/min. The hemodynamic values duringangiography were not significantly different from thoserecorded intraoperatively (p = NS, paired t test). Sev-enteen patients were taking fl-blockers, 14 patientscalcium-channel blockers, and 14 patients nitrate prep-arations at the time of cardiac catheterization. Justbefore surgery, 20 patients were on fl-blockers, 18patients were on calcium-channel blockers, and 16patients were using nitrate preparations. Twenty-twodiscrete lesions in 18 patients were evaluated in theportion of the study evaluating the ability of video-densitometry to define an index of luminal area. Echo-cardiographic minimum luminal area varied from 0.2to 6.1 mm2 in these lesions of which 13 were classifiedas circular, eight as oval, and one as complex. Thirty-three arterial segments in 21 patients were studied toevaluate the effect of angiographic projection on video-densitometry. Of these 33 arterial segments, 19 were

classified as having circular lumens, 1 1 were classified

...~~~~~~~~~ ~ ~ ...

Oval Lesion Conmplex LesiFIGURE 1. Examples of echocardlographic images of lesions with circular (left), oval (center), and complex (right) lumens.Ath atheroma; L lumen.

Vol. 77, No. 2, February 1988 331

by guest on April 13, 2017

http://circ.ahajournals.org/D

ownloaded from

JOHNSON et al.

as having oval lumens, and three were classified ashaving complex lumens. (Nineteen arterial segments in16 patients were included in both portions of the study.)

Videodensitometry vs echocardiographic luminal area.The relationship between videodensitometric and echo-cardiographic measurements of coronary lumen area isshown in figure 2. For all 22 lesions, a correlationcoefficient of r = .86 was noted for the relationshipbetween integrated optical density defined by video-densitometry and luminal area defined by high-fre-quency echocardiography. The correlation coefficientsfor the relationship between integrated optical densityand echocardiographic luminal area in the subgroups ofthe 13 circular lesions (circles) and the nine oval orcomplex lesions (triangles) were r = .81 and r = .93,respectively.

Efects of angiographic projection and luminal shape.The correlation noted between integrated optical den-sity values defined in the left anterior oblique and rightanterior oblique angiographic projections is shown infigure 3. In the 33 arterial segments, the correlationcoefficient for this relationship was .94, with a regres-sion line of y = 1.04x + 0.002. This slope andintercept did not significantly differ from one and zero,respectively. Thus, regardless of whether the angio-grams analyzed were taken in a left anterior oblique orright anterior oblique projection, the results obtainedfor videodensitometric analysis of the lesions weresimilar.

Figure 3 also shows that the good correlationbetween integrated optical densities defined by video-densitometric analysis in two orthogonal angiographic

24

16

IntegratedOpticalDensity

8

0

A

1 2 3 4 5 6

Echo Minimum Lumen Area (mm2)

FIGURE 2. The relationship between lesion integrated optical density,defined by videodensitometric analysis and reported in optical densityunits, and minimum luminal area defined by high-frequency epicardialechocardiography. The circles represent data for 13 circular lesions andthe triangles represent data for nine lesions classified as oval or complex.

332

16

12

IntegratedOpticalDensity:LAO

8

4

0 [

* Circular LesionsA Oval/Complex Lesions

r = 0.94y = 1.04x + 0.002

- n = 33SEE= 1.208

A

A

A

A

*1 .

g I

4 8 12 16

Integrated Optical Density: RAOFIGURE 3. The relationship between integrated optical densitiesdefined by videodensitometric analysis of the same coronary segment on

orthogonal left anterior oblique (LAO) and right anterior oblique (RAO)angiographic projections. Data for 19 arterial segments with lumensdefined as circular on high-frequency echocardiographic images areplotted as circles. Similar data for 14 segments defined as having an ovalor complex lumen are plotted as triangles.

projections holds true for both circular lesions (circles)and oval/complex lesions (triangles). For the 19 circularlesions, the correlation coefficient was .95, with aregression line of y = 0.98x + 0.080. Similarly, forthe 14 oval/complex lesions, the correlation coefficientwas .94, with a regression line of y =1.09x - 0.022.The slopes and intercepts of these relationships werenot significantly different from one and zero, respec-tively. Thus, even in lesions with oval or complexshapes, for which geometric measurements of luminalsize in different projections might be variable, video-densitometry gives the same results in orthogonalangiographic projections and therefore is independentof luminal shape.

DiscussionThe major findings of the present study are that: (1)

videodensitometric analysis provides a reasonableindex of coronary lesion luminal area as defined witha nonangiographic, in vivo measurement technique(high-frequency epicardial echocardiography), and (2)the results ofvideodensitometric analysis are independ-ent of angiographic projection, as long as the vascularsegment of interest is nearly parallel to the image inten-sifier, and independent of luminal shape as defined byhigh-frequency epicardial echocardiography.

Prior videodensitometric studies. Other investigatorshave reported on the feasibility and accuracy of the useof videodensitometry to estimate coronary luminalsize. Kruger23 has reported that direct density measurescan be used to accurately define absolute and relative

CIRCULATION

l l l

by guest on April 13, 2017

http://circ.ahajournals.org/D

ownloaded from

DIAGNOSTIC METHODS-CORONARY ARTERY DISEASE

vessel diameters in Plexiglas and aluminum vesselphantoms.23 Similar data have been obtained byNichols et al.24 for relative densities (i.e., percentstenosis) of the stenotic and "normal" arterial segmentsin phantoms and postmortem human hearts. Wiesel etal.25 have used a videodensity method to define abso-lute luminal areas in cylindrical phantoms and in Delrincylinders placed in dog coronary arteries. Tobis et al.26have reported the use of videodensitometry to defineabsolute luminal diameters in lucite phantoms filledwith contrast. Mancini et al.14 have correlated video-densitometric determinations of luminal area in a dogpreparation ofcoronary stenosis with independent mea-sures of coronary flow reserve. Our previous work hasshown that the results of videodensitometric analysis ofhuman coronary stenoses correlate well with coronaryluminal area, as defined by quantitative coronary arte-riography (method of Brown et al.5), and lesion phys-iologic significance, as defined by Doppler studies ofvasodilator reserve.'l19 However, the present studyprovides the most direct validation of the videodensi-tometric technique in humans in vivo that is currentlyavailable.

Theoretically, videodensitometry should be inde-pendent of angiographic projection and luminal shape.Nichols et al.24 and Tobis et al.26 have reported thatdeterminations of percent stenosis derived from den-sitometric data are similar whether defined on the leftanterior oblique or right anterior oblique angiogram.(In the Tobis study,26 the right anterior oblique/leftanterior oblique correlation for videodensitometric per-cent stenosis was similar to that for geometric percentstenosis.) However, study of the effects of angiographicprojection on the derivation of indexes of absoluteluminal area by videodensitometry has been limited.Likewise, determination of the effects of luminal shapeon videodensitometric results in vivo has not previouslybeen reported.

Potential limitations of this studyLimitations of videodensitometry. Videodensitometric

analysis is based on the Lambert-Beer law, whichassumes that there is a direct relationship between themass of radioopaque material in a certain portion of thevascular system and fractional x-ray transmission.19Unfortunately, the Lambert-Beer law is an idealizedmodel of the interaction between the x-ray beam andthe contrast-filled vessel and does not take into accountthe effect of certain technical factors on videodensi-tometric analysis. These factors include beam harden-ing and x-ray scatter (which result directly from theprocess of x-ray generation and x-ray interaction withthe subject) as well as "veiling glare" (which results

Vol. 77, No. 2, February 1988

from light scatter and multiple reflections within theimage intensifier and the optics of the imagingsystem).27 Variation in focal spot size may also modifythe observed videodensitometric profile. Other inves-tigators have attempted to develop correction factorsfor some of these variables,28' 29 and phantom studiessuggest that the degree of inaccuracy resulting fromthese factors is small.30' 31 However, little is knownabout the effect of these factors on the evaluation ofhuman coronary stenoses, where the effect of suchtechnical variables may be greater, and these factorsmay have accounted for some of the variability in ourresults.

Another potential problem with videodensitometryis that variable contrast concentrations may occurwithin the vessel lumen, resulting from variability ininjection techniques or in the rate of coronary bloodflow. Since videodensitometry assumes a direct rela-tionship between the mass of radioopaque materialwithin the vascular lumen and fractional x-iay trans-mission, the concentration of contrast medium willaffect videodensitometric results. Phantom studieshave shown a linear relationship between the concen-tration of contrast medium and density measure-ments.32 In our clinical studies, we have simply ana-lyzed arteries that appear to be uniformly well opacifiedwith contrast and we have obtained good correlationsbetween the results of videodensitometry and quanti-tative coronary arteriography, Doppler studies of vaso-dilator reserve, and high-frequency echocardiography.However, our correlations may have been better, andthe scatter in our data less, had we been able to controlthe uniformity of contrast in the arterial lumen.

Another concern is that angulation of the vesselrelative to the image intensifier may significantly affectvideodensitometric results. Clearly, analysis of a vessel"end-on" would not result in an accurate determinationof lumen area because the x-ray attenuation would beproduced by contrast present in a "length" rather thana "cross section" of the vessel. Thus, in our studies wehave chosen angiographic projections in which thearterial segment of interest appears parallel to the imageintensifier. However, the degree of angulation of thevessel relative to the image intensifier that results insignificant inaccuracies has not been defined. In a pre-liminary study we found that the results of videoden-sitometric analysis of proximal left anterior descendinglesions in the right anterior oblique projection corre-lated quite well with lesion geometry and functionalsignificance, whereas results of analysis in -the leftanterior oblique projection did not correlate as well. 16It was presumed that the problem was due, at least in

333

by guest on April 13, 2017

http://circ.ahajournals.org/D

ownloaded from

JOHNSON et al.

part, to foreshortening of the proximal left anteriordescending coronary artery on the left anterior obliqueangiograms, which at the time of the study wererecorded without caudocranial angulation. The data inour current study, in which obviously foreshortenedvessels were excluded, strengthen this presumption.Angulation of the vessel relative to the image inten-sifier, resulting in foreshortening, should produce arather predictable increase in videodensitometric val-ues. Therefore, the degree of vessel angulation presentin selected views of the coronary angiogram and theeffect of vessel angulation on the results of videoden-sitometry need to be further analyzed.Our current method is also limited by the fact that

results are reported in terms of optical density unitsrather than absolute luminal cross-sectional area.Potential approaches to the derivation of absolute lumi-nal areas by videodensitometry include: (1) the use ofa contrast-filled calibration catheter or (2) comparisonof the density value of the narrowed coronary segmentto that of a normal appearing segment whose luminalsize could more easily be determined by quantitativeangiographic techniques. 14 25, 26 However, even if amethod is devised that allows the determination ofabsolute coronary luminal cross-sectional area byvideodensitometry, this defined area must then be com-pared with the normal area for that vessel in that par-ticular patient to determine the true severity of thecoronary disease. Defining this normal area may beproblematic, because the caliber of the normal coro-naray tree varies at different portions in its branchingstructure, and is also related to patient gender.33 34

Thus, the usefulness of knowing absolute coronaryluminal area, defined by videodensitometry or anyother quantitative technique, depends on knowledge ofthe normal size of the artery under study in an indi-vidual patient. In addition, Kirkeeide et al.35 haveemphasized that minimum luminal area is only one ofthe geometric characteristics influencing coronary flowreserve and that other lesion characteristics may alsoneed to be considered to accurately predict coronarylesion functional significance.35The particular approach that we have used - the

videodensitometric analysis of cineangiograms of cor-onary stenoses - has limitations in addition to thosementioned above. First, the digitization of angio-graphic frames increases the time required for video-densitometric analysis This step would not be requiredif digitally acquired angiograms were analyzed. Inaddition, variabilities in film processing could con-tribute to inaccuracies in the data. Despite these lim-

two reasons. First, most clinical catheterization labo-ratories still use angiographic film so the methods weused, if further refined and simplified, could be appliedclinically. In addition, at the time this study was done,we did not have the capability of performing digitalcoronary angiograms in our laboratory.

Limitations of high-frequency echocardiography. We havepreviously shown (using histology and sonomicrom-etry) that high-frequency epicardial echocardiog-raphy can be used to accurately determine coronaryluminal diameters and areas in open-chest animals andpostmortem human hearts.21 The reproducibility ofhigh-frequency echocardiographic determinations ofcoronary luminal diameters is good, with an intraob-server correlation of r = .99 and an interobservercorrelation of r = .97 for 10 human coronary arteriesin vitro.2' In a separate study, the interobserver repro-ducibility of measurements of luminal area of dogcoronary arteries in vivo was also found to be good (r= .95, y = 0.99x + 0.14, n 58).* However, thereare problems inherent in the echocardiographic tech-nique that may have influenced the results of our study.Ultrasound images characteristically have poor lateralresolution. Therefore, only images with clearly visu-alized lateral walls were analyzed in our study. Theultrasonic probe was hand-held, and although care wastaken to rotate the transducer 90 degrees from thelong-axis image ofthe artery to obtain true arterial crosssections, this does not exclude the possibility that someof the images analyzed may have been slightly oblique.It is also possible that despite analysis of the echocar-diographic image of the apparently smallest lumen, theactual minimum luminal area may have been missed.Finally, the size of the current probe limits echocar-diographic analysis to the left anterior descending andright coronary arteries.

Limitations of the study protocol. First, although weincluded only angiographically discrete coronary arte-rial stenoses for analysis of luminal areas to allow closematching of the segments analyzed by videodensitom-etry and echocardiography, any mismatch of the seg-ments analyzed would be problematic. Second, anydifferences in medical therapy (i.e., coronary vasodi-lators such as nitrates or calcium-channel blockers)between the time of catheterization and the time ofsurgery could have introduced variability into ourresults. Sixteen of the patients did receive nitroglycer-in and/or nifedipine in the cardiac catheterization lab-oratory, and coronary luminal areas do increase inresponse to such medications.36 In addition, although

itations of film-based videodensitometry we used it for

334

*McPherson DD, Kerber RE, et al: Unpublished observations.

CIRCULATION

by guest on April 13, 2017

http://circ.ahajournals.org/D

ownloaded from

DIAGNOSTIC METHODS-CORONARY ARTERY DISEASE

medical therapy at the time of cardiac catheterizationand surgery was similar, there was a trend towardincreased antianginal therapy before coronary bypasssurgery. Third, differences in the hemodynamics (i.e.,distending pressures) at catheterization vs surgerycould have influenced coronary arterial luminal size.However, for the entire patient group, pulse rate andblood pressure did not differ significantly between thetime of coronary angiography and the high-frequencyechocardiographic studies. Finally, although the timebetween angiography and surgery was short (less than1 month in all but three patients), disease progressionand/or regression or propagation of thrombus couldhave resulted in a change in luminal size between thetime of catheterization and the echocardiographic stud-ies. Several of these factors, as well as limitations ofeach technique as delineated above, probably contrib-uted to the variability in our correlation of data derivedby videodensitometry with luminal areas defined byhigh-frequency echocardiography.

Clinical implications. A need exists for a rapid andclinically feasible method of quantitative angiographicanalysis to allow the prediction of the geometry andphysiologic significance of stenoses seen on humancoronary angiograms. Although the angiogram re-mains the standard by which clinical decisions aremade in patients with coronary artery disease, thecommon clinical method of angiographic analysis, thatof estimating percent diameter stenosis, does not rep-resent an optimal method of interpretation.1-4 6, 7Although rapid, quantitative analysis of all coronaryangiograms would be ideal, such analysis may be par-ticularly important for the evaluation of lesions seen onangiograms obtained during interventional cardiologicprocedures such as thrombolysis and angioplasty. Har-rison et al.37 have shown that the residual luminal areaafter thrombolysis could be used to predict the likeli-hood of reocclusion, and therefore the likelihood of thenecessity of proceeding on an emergency basis toangioplasty or bypass surgery to maintain vesselpatency. Serruys et al.,38 as well as our group,39 haveshown the usefulness of videodensitometric analysisduring and after percutaneous transluminal coronaryangioplasty. Therefore, if the videodensitometricmethod can be made even more efficient, and partic-ularly if further validation of its accuracy in the analysisof digitally acquired human coronary angiographicimages becomes available, videodensitometry couldallow critical therapeutic decisions to be made in thecardiac catheterization laboratory rather than hours ordays later when angiograms can be analyzed by othersophisticated quantitative analysis procedures.

Vol. 77, No. 2, February 1988

In summary, this study further validates the abilityof videodensitometry to provide an index of coronaryluminal area, one of the important determinants of thephysiologic significance of a lesion. It also confirms invivo the assumptions that the results of videodensito-metric analysis are independent of angiographic pro-jection and luminal shape. Further studies evaluatingthis method of angiographic analysis, and particularlyevaluating some of the limitations of the method as itis now used, are appropriate and will be necessarybefore broad clinical application can be encouraged.

We thank Jill Christy for her expert preparation of this manu-script.

References1. Detre KM, Wright E, Murphy ML, Takaro T: Observer agreement

in evaluating coronary angiograms. Circulation 52: 979, 19752. Zir LM, Miller SW, Dinsmore RE, Gilbert JP, Harthorne JW:

Interobserver variability in coronary angiography. Circulation 53:627, 1976

3. DeRouen TA, Murray JA, Owen W: Variability in the analysis ofcoronary arteriograms. Circulation 55: 324, 1977

4. Galbraith JE, Murphy ML, deSoyza N: Coronary angiogram inter-pretation. Interobserver variability. JAMA 240: 2053, 1978

5. Brown BG, Bolson E, Frimer M, Dodge HT: Quantitative coronaryarteriography. Estimation of dimensions, hemodynamic resistance,and atheroma mass of coronary artery lesions using the arteriogramand digital computation. Circulation 55: 329, 1977

6. White CW, Wright CB, Doty DB, Hiratzka LF, Eastham CL,Harrison DG, Marcus ML: Does visual interpretation of the cor-onary arteriogram predict the physiologic importance of a coronarystenosis? N Engl J Med 310: 819, 1984

7. Harrison DG, White CW, Hiratzka LF, Doty DB, Barnes DH,Eastham CL, Marcus ML: The value of lesion cross-sectional areadetermined by quantitative coronary angiography in assessing thephysiologic significance of proximal left anterior descending cor-onary arterial stenoses. Circulation 69: 1111, 1984

8. Arnett EN, Isner JM, Redwood DR, Kent KM, Baker WP, Ack-erstein H, Roberts WC: Coronary artery narrowing in coronaryheart disease: comparison of cineangiographic and necropsy find-ings. Ann Intern Med 91: 350, 1979

9. Vlodaver Z, Frech R, VanTassel RA, Edwards JE: Correlation ofthe antemortem coronary arteriogram and the postmortem speci-men. Circulation 47: 162, 1973

10. McPherson DD, Hiratzka LF, Lamberth WC, Brandt B, Hunt M,Kieso RA, Marcus ML, Kerber RE: Delineation of the extent ofcoronary atherosclerosis by high-frequency epicardial echocardiog-raphy. N Engl J Med 316: 304, 1987

11. Mates RE, Gupta RL, Bell AC, Klocke FJ: Fluid dynamics ofcoronary artery stenosis. Circ Res 42: 152, 1978

12. Johnson MR, Fleagle SR, Aylward PE, Collins SM, Skorton DJ,Hiratzka LF, Harrison DG, White CW, Marcus ML: Digital proc-essing and analysis of coronary cineangiograms: geometric andphysiological assessment of coronary stenosis. Circulation 70(supplII): 11-324, 1984 (abst)

13. Collins SM, Fleagle SR, Johnson MR, Wilson RF, White CW,Marcus ML, Skorton DJ: Prediction of the physiological signifi-cance of a coronary stenosis using automated analysis of arterialgeometry. Circulation 74(suppl II): 11-485, 1986 (abst)

14. Mancini GBJ, Simon SB, McGillem MJ, LeFree MT, FriedmanHZ, Vogel RA: Automated quantitative arteriography: morphologicand physiologic validation in vivo of a rapid digital angiographicmethod. Circulation 75: 452, 1987

15. Reiber JHC, Slager CJ, Schuurbiers JCH, denBoer A, GerbrandsJJ, Troost GJ, Scholts B, Kooijman C, Serruys PW: Transferfunctions of the x-ray cine video chain applied to digital processingof coronary cineangiograms. In Heintzen PH, Brennecke R, editors:Digital imaging in cardiovascular radiology. Stuttgart, 1983,George Thieme Verlag, p 89

335

by guest on April 13, 2017

http://circ.ahajournals.org/D

ownloaded from

JOHNSON et al.

16. Collins SM, Skorton DJ, Harrison DG, White CW, Eastham CL,Hiratzka LF, Doty DB, Marcus ML: Quantitative computer-basedvideodensitometry and the physiological significance of a coronarystenosis. In Computers in cardiology. Long Beach, CA, 1982, IEEEComputer Society, p 219

17. Collins SM, Johnson MR, Ericksen EE, Fleagle SR, Hiratzka LF,Harrison DG, White CW, Skorton DJ: Videodensitometric analysisof coronary stenoses: methodologic considerations and geometricand functional validation. Circulation 70(suppl II): II-30, 1984 (abst)

18. Johnson MR, Fleagle SR, Aylward PE, Ripley JE, Wilson RF,White CW, Hiratzka L, Marcus ML, Collins SM, Skorton DJ:Analysis of right coronary artery stenoses using videodensitometry.Circulation 72(suppl III): 111-262, 1985 (abst)

19. Johnson MR, Fleagle SR, Wilson RF, White CW, Hiratzka LF,Marcus ML, Collins SM, Skorton DJ: Anatomic and functionalassessment of circumflex coronary artery stenoses using videoden-sitometry. J Am Coll Cardiol 7: 152A, 1986 (abst)

20. Holmes DR Jr, Vlietstra RE, Mock MB, Reeder GS, Smith HC,Bove AA, Bresnahan JF, Piehler JM, Schaff HV, Orszulak TA:Angiographic changes produced by percutaneous transluminal cor-onary angioplasty. Am J Cardiol 51: 676, 1983

21. McPherson DD, Armstrong M, Rose E, Kieso RA, Megan M, HuntM, Hite P, Marcus ML, Kerber RE: High frequency epicardialechocardiography for coronary artery evaluation: in-vitro and in-vivo validation of arterial lumen and wall thickness measurements.J Am Coll Cardiol 8: 600, 1986

22. Judkins MP: Selective coronary arteriography. Part I: A percuta-neous transfemoral technique. Radiology 89: 815, 1967

23. Kruger RA: Estimation of the diameter of and iodine concentrationwithin blood vessels using digital radiography devices. Med Phys8: 652, 1981

24. Nichols AB, Gabrieli CFO, Fenoglio JJ, Esser PD: Quantificationof relative coronary arterial stenosis by cinevideodensitometricanalysis of coronary arteriograms. Circulation 69: 512, 1984

25. Wiesel J, Grunwald AM, Tobiasz C, Robin B, Bodenheimer MM:Quantitation of absolute area of a coronary arterial stenosis: exper-imental validation with a preparation in vivo. Circulation 74: 1099,1986

26. Tobis J, Nalcioglu 0, Johnston WD, Qu L, Reese T, Sato D, RoeckW, Montelli S, Henry WL: Videodensitometric determination ofminimum coronary artery luminal diameter before and after angio-plasty. Am J Cardiol 59: 38, 1987

27. Kruger RA: X-ray digital cineangiography. In Collins SM, SkortonDJ, editors: Cardiac imaging and image processing. New York,1986, McGraw-Hill, Inc., p 24

28. Shaw C-G, Ergun DL, Myerowitz PD, VanLysel MS, MistrettaCA, Zarnstorff WC, Crummy AB: A technique of scatter and glarecorrection for videodensitometric studies in digital subtractionvideoangiography. Radiology 142: 209, 1982

29. Jacques P, DiBianca F, Pizer S, Kohout F, Lifshitz L, Delany D:Quantitative digital fluorography. Computer vs. human estimationof vascular stenoses. Invest Radiol 20: 45, 1985

30. Boone JM, Nalcioglu 0, Roeck WW, Wang Y, Lando AV: Theinfluence of point spread functions (PSF) in the determination ofcoronary stenosis by video densitometry. In Computers in cardi-ology. Long Beach, CA, 1982, IEEE Computer Society, p 227

31. Simons MA, Kruger RA: Vessel diameter measurement using dig-ital subtraction radiography. Invest Radiol 20: 510, 1985

32. Spears JR, Sandor T, Als AV, Malagold M, Markis JE, GrossmanW, Serur JR, Paulin S: Computerized image analysis for quanti-tative measurement of vessel diameter from cineangiograms. Cir-culation 68: 453, 1983

33. MacAlpin RN, Abbasi AS, Grollman JH Jr, Eber L: Human cor-onary artery size during life. Radiology 108: 567, 1973

34. Vieweg WVR, Alpert JS, Hagan AD: Caliber and distribution ofnormal coronary arterial anatomy. Cathet Cardiovasc Diagn 2: 269,1976

35. Kirkeeide RL, Gould KL, Parsel L: Assessment of coronary ste-noses by myocardial perfusion imaging during pharmacologiccoronary vasodilation. VII. Validation of coronary flow reserveas a single integrated functional measure of stenosis severity re-flecting all its geometric dimensions. J Am Coll Cardiol 7: 103,1986

36. Brown BG, Bolson E, Petersen RB, Pierce CD, Dodge HT: Themechanisms of nitroglycerin action: stenosis vasodilation as a majorcomponent of the drug response. Circulation 64: 1089, 1981

37. Harrison DG, Ferguson DW, Collins SM, Skorton DJ, EricksenEE, Kioschos JM, Marcus ML, White CW: Rethrombosis afterreperfusion with streptokinase: importance of geometry of residuallesions. Circulation 69: 991, 1984

38. Serruys PW, Reiber JHC, Wijns W, von der Brand M, KooijmanCJ, tenKaten HJ, Hugenholtz PG: Assessment of percutaneoustransluminal coronary angioplasty by quantitative coronary angiog-raphy: diameter versus densitometric area measurements. Am JCardiol 54: 482, 1984

39. Johnson MR, Brayden GP, Ericksen EE, Collins SM, Skorton DJ,Harrison DG, Marcus ML, White CW: Changes in cross-sectionalarea of the coronary lumen in the six months after angioplasty: aquantitative analysis of the variable response to percutaneous trans-luminal angioplasty. Circulation 73: 467, 1986

CIRCULATION336

by guest on April 13, 2017

http://circ.ahajournals.org/D

ownloaded from

Marcus, S M Collins and D J SkortonM R Johnson, D D McPherson, S R Fleagle, M M Hunt, L F Hiratzka, R E Kerber, M L

intraoperative high-frequency epicardial echocardiography.Videodensitometric analysis of human coronary stenoses: validation in vivo by

Print ISSN: 0009-7322. Online ISSN: 1524-4539 Copyright © 1988 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.77.2.328

1988;77:328-336Circulation. 

http://circ.ahajournals.org/content/77/2/328the 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:

by guest on April 13, 2017

http://circ.ahajournals.org/D

ownloaded from