Intravenous Coronary Angiography UsingElectron Beam Computed Tomography

Benno J. Rensing, Alfons H.H. Bongaerts, Robert-Jan M. van Geuns,Peter M.A. van Ooijen, Matthijs Oudkerk, and Pim J. de Feyter

Intravenous coronary angiography with electronbeam computed tomography (EBCT) allows for thenoninvasive visualisation of coronary arteries. Withdedicated computer hardware and software, three-dimensional renderings of the coronary arteries,veins, and other cardiac structures can be con-structed from the individual transaxial tomograms.Interest in this technique is growing, and recently anumber of clinical studies have been publishedcomparing EBCT coronary angiography with conven-tional cine-coronary angiography. In this article,image acquisition, postprocessing techniques, andthe results of recently published clinical studies arediscussed. EBCT coronary angiography is a promis-ing imaging technique of coronary arteries. Cur-rently, it is a reasonably robust technique for thevisualization and assessment of the left main andleft anterior descending coronary artery. However,at the moment a relatively high proportion of theright and circumflex coronary angiograms are nonin-terpretable. Improvements in image acquisition andpostprocessing techniques are expected to improvevisualization and diagnostic accuracy of the tech-nique.Copyright � 1999 by W.B. Saunders Company

The small size and fast movement of humancoronary arteries pose a formidable challenge

for an imaging technique. For accurate visualiza-tion, it has to combine a high spatial resolutionwith a high temporal resolution. Modern coro-nary cine-angiography with direct intracoronarycontrast injection has a spatial resolution of 5linepairs per mm and a temporal resolution of upto 50 frames/s. This allows the accurate visualiza-tion of the epicardial luminal coronary trajectoryduring the complete cardiac cycle and has be-come the undisputed reference technique forcoronary imaging. These specifications come at a

price, however: the procedure is costly, invasive,and carries a small risk of serious adverse events.

Recently, electron beam computed tomography(EBCT) after intravenous injection of contrastmedium has emerged as an imaging technique ofcoronary arteries.1-7 Using modern image process-ing technology, three-dimensional (3D) recon-struction of the heart and coronary arteries can bemade from 40 to 60 consecutive tomograms. Theindividual tomograms have a resolution of 4 to 6linepairs/cm, and the scan time for each tomo-gram is 100 ms. Although this falls short of thespatial and temporal resolution of conventionalcoronary angiography, proximal and middle partsof the coronary arteries can be visualized with thistechnique.

In this article, image acquisition, image process-ing techniques, and clinical studies that com-pared EBCT coronary angiography to conven-tional coronary angiography will be reviewed.

Methodology

EBCT Coronary Angiography

The EBCT scanner, also called ultrafast-CT orcine-CT (Siemens Evolution, Munich, Germany),is a CT scanner that allows the acquisition ofhigh-resolution electrocardiogram (ECG)-trig-

From the Department of Cardiology, Thoraxcenter, andthe Department of Radiology, Dr Daniel den Hoed Kliniek,University Hospital Rotterdam, Rotterdam, The Nether-lands.

Address reprint requests to Benno J. Rensing, MD,Department of Cardiology, Thoraxcenter, BD 416, DrMolewaterplein 40, 3015 GD Rotterdam, The Netherlands;e-mail: rensing@card.azr.nl.

Copyright � 1999 by W.B. Saunders Company0033-0620/99/4202-0002$10.00/0

Progress in Cardiovascular Diseases, Vol. 42, No. 2 (September/October), 1999: pp 139-148 139

gered tomograms in 100 ms. This is fast enoughto reduce cardiac motion artefacts and to reliablyvisualize the fast-moving coronary arteries, espe-cially when the acquisition window is set duringdiastasis. The scanner was originally designed forcardiac function studies and was later applied toquantify coronary calcifications. A detailed de-scription of scanner specifications can be foundelsewhere.8 Scanning is performed with the pa-tient in the supine position. To ensure optimalcoronary opacification during image acquisition,the circulation time has to be determined. There-fore, 10 mL of contrast medium (iopromide 350mg/mL) is injected through an antecubital vein at4 mL/s. The passage of the contrast bolus throughthe ascending aorta is visualized by tomograms atthe aortic root level. The time from contrastinjection to peak aorta density is considered thecirculation time.

Next several localizing scans over the aorticroot are made at inspiration to determine theexact level of the left main coronary artery.

Acquisition of the 3D dataset starts with theinjection 120 to 180 mL of contrast medium at 3to 4 mL/s through an antecubital vein. During thecirculation time, patients are asked to mildlyhyperventilate by breathing in and out at theword of the CT technician. Just before the circula-tion time, the patient is asked to hold his/herbreath at a comfortable inspiratory level. At thecirculation time, scanning commences just proxi-mal to the left main coronary artery after an ECGtrigger at 80% of the RR (wave) interval (diasta-

sis). Tomogram thickness is set at 1.5 or 3 mm.Table increment after each tomogram is set at 1.5mm or 2 mm, resulting in contiguous nonoverlap-ping slices or slices with 1 mm overlap. A total of40 to 60 transaxial tomograms are made during asingle breathhold (Fig 1). Breath holding is neces-sary during data acquisition to avoid respiratorymotion artefacts. Field of view size is generally setat 18 cm with a matrix size of 512 � 512 pixels,yielding a pixel size of 0.35 � 0.35 mm. To speedup image acquisition and to shorten breath-holdtime, atropine 0.5 to 1 mg intravenously can beadministered when heart rate is less than 60min�1. Breath-holding time in the cited studieswas between 20 and 50 seconds and is dependenton the heart rate and the number of tomogramsmade. Effective radiation dose is estimated to beless than 10 mSv.9,10 This is approximately onethird to one fifth of the radiation dose at diagnos-tic coronary angiography.9,11

Image Processing

For accurate assessment of the anatomy andintegrity of the epicardial coronary arteries, merelyevaluating the axial tomograms is inadequate.The 3D trajectory of the coronary arteries spansalmost the complete set of 40 to 60 tomograms,and therefore some form of 3D reconstruction ofthe tomographic data is necessary to keep track ofthe exact course of the arteries and to accuratelyread the EBCT angiogram. Furthermore, respira-tion and rhythm artefacts are readily apparent

Fig 1. Transaxial tomograms at different levels. (A) At the level of the ostium of the left main coronary artery. (B) Atthe level of the ostium of the right coronary artery (RCA, arrowhead). LAD, proximal left anterior descendingcoronary artery. (C) Proximal circumflex coronary artery (arrow). Ao, ascending aorta; RVOT, right ventricularoutflow tract; SCV, superior caval vein.

RENSING ET AL140

with most 3D reconstruction methods. Indeed, allpublished studies used 3D reconstruction meth-ods to assess the EBCT coronary angiograms. 3Dreconstruction starts with transferring the two-dimensional (2D) transaxial tomograms to a com-puter workstation where they are stacked andinterpolated to form a 3D volume using specialcomputer software. This process creates a data-base representation of a 3D object. This databaseconsists of millions of volume elements (voxels),each with its own Hounsfield unit or densityvalue. With special rendering techniques, thisdatabase representation of the 3D object can beprojected onto a viewing surface. This creates theillusion of looking at a 3D object.

Rendering Techniques

Volume rendering. Volume rendering is a ren-dering technique that retains all the voxels fromthe original 3D dataset during each rendering.This process is computationally intensive but hasthe advantage that no information is discardedduring the process and that several image-processing techniques are still possible.12 Forinstance, each voxel can be assigned an opacityand color value. Based on these values, theintensity of each voxel is calculated and used forthe display of the voxel. Volume rendering allowscertain parts of the 3D object to be transparent sothat structures behind it are still visible. It can, byassigning the right color and opacity values todifferent tissues, simulate real anatomy in muchthe same way as pictures in an anatomic atlas(Figs 2 and 3).13

The development of dedicated computer hard-ware and software now allows volume renderingto be performed on a desktop computer worksta-tion, whereas only a few years ago the techniquewas reserved for large mainframe computers.Computation time for a single rendering is now amatter of seconds, which is fast enough to allowrapid and comfortable user interaction. It isexpected that volume rendering on a personalcomputer platform will be possible in the nearfuture.

Image processing techniques such as segmenta-tion and filtering have to be applied to bettervisualize the coronary arteries among the othercontrast-enhanced and sometimes overlappingcardiac structures.14 For instance, the thorax wall,lung vessels, and cardiac structures such as theleft and right auricle have to be removed from thedataset (Fig 4) to be able to see the coronaryarteries. This segmentation has to be done on theindividual tomographic level. Although someautomation is possible, a large part still has to bedone manually. On average, an experienced tech-nician can perform this segmentation in 15 to 20minutes. Dedicated segmentation software is un-der development and is expected to substantiallyshorten total image-processing time.

Surface rendering. This is a rendering tech-nique that displays only a fraction of the 3D data(Fig 5). In a first step, threshold values are set soonly voxels within a prespecified density intervalare retained. After this selection process a simpli-fied approximation of the surface of the selectedobject is calculated by assigning polygons (eg,

Fig 2. (A) Pressure fixed ana-tomical specimen showingthe left coronary artery. Re-printed with permission fromMcAlpine WA: Heart and Coro-nary Arteries. Heidelberg, Ger-many, Springer-Verlag, 1975,p 159, fig 1. Copyright � bySpringer-Verlag. (B) 3D render-ing of the left coronary arteryof a patient with a severe ste-nosis (arrow) of the bifurca-tion of the proximal left ante-rior descending artery anddiagonal artery (LAD andDiag).Ao,Aorta; LAA, left atrialappendage; LA, left atrium; RV,right ventricle; LV, left ven-tricle.

EBT CORONARY ANGIOGRAPHY 141

triangles) to this surface. As in volume rendering,segmentation steps to remove objects that ob-scure the outlook on the coronary arteries are alsonecessary with surface rendering. An advantage ofthe technique is that after the initial computa-tional intensive rendering steps, further interac-

tive viewing of the mathematically defined sur-face is very fast. Disadvantages are the loss ofsurface details by the polygon approximation, theinability to display internal structures, and thetime-consuming optimization of the thresholdsettings.

Fig 3. Volume rendering showing an occluded circumflex artery (arrows). (A) View from the top. (B) View from amore anterior angle. (C) Corresponding coronary angiogram. LAD, left anterior descending artery; Int, intermediatebranch.

Fig 4. Visualization of con-tour segmentation of theheart within the thorax. (A)Volume rendering of theheart and part of the chestwall and pulmonary vessels.(B) The circles depict thecontour lines drawn on theindividual axial tomograms.(C) The computer can be or-dered to delete all image datathat is outside the contours,here depicted by the grayvolumes. (D) Final result af-ter segmentation. The coro-nary arteries can now beseen without interference ofthe chestwall, pulmonary ar-teries and veins, and cardiacvenous structures.

RENSING ET AL142

Multiplanar reformatting (MPR). For this ren-dering technique, coronal, saggital, transversal, ordouble oblique sections are made through the 3Dvolume to target the course of the coronary artery.A set of tomograms parallel to the double obliqueslice can then be reconstructed to completelyvisualize and appreciate the often tortuous trajec-tory of the coronary arteries. A related techniqueis curved multiplanar reformatting. Here theoperator indicates the artery on each of theoriginal axial tomograms or on the reconstructeddouble oblique slices (Fig 6). The computerconnects the artery segments on every tomogramor slice and projects the reconstructed courseonto a 2D surface. Both techniques are simple andfast and do not need manual segmentation of

structures that obscure the outlook on the coro-nary arteries. On the other hand, MPR is operator-dependent and requires a thorough understand-ing of coronary anatomy.

Maximum intensity projection (MIP). MIP is arendering technique in which imaginary rays arecast through the 3D dataset from any viewpointthe image-analyst chooses (Fig 5). The highestintensity voxel encountered by each ray is used toconstruct a 2D image of the 3D dataset. Usuallythese are voxels within the contrast-filled vessels.Images resemble classic angiograms. This render-ing technique is fast and generally gives a gooddifferentiation between vascular and nonvascularstructures. For coronary angiography, however,this technique is not very well suited because

Fig 5. (A) Maximum intensity projection of the left coronary artery. Before the coronary arteries can be visualizedwith this technique, denser structures in the path of the imaginary rays (see text for explanation) have to be removedfirst. After extensive segmentation only the aortic root and the coronary arteries are retained in the volume. The leftmain coronary artery cannot be discerned in this projection because of overlap with the much denser aortic root.The dotted appearance of the LAD is due to partial volume effects. (B) Surface rendering showing the LAD. Thepulmonary artery and right ventricular outflow tract have been segmented out on the individual tomograms. On thesurface rendering at this area, a hole in the surface can be seen (arrow). (C) Same dataset now visualized withvolume rendering.

Fig 6. Curved multiplanarreformation. (A) Single proxi-mal transaxial tomogram,showing the left main coro-nary artery and the proximalLAD. The dotted line is theprojection on this tomogramof the line connecting theright coronary artery, theaorta, and the left anteriordescending coronary arterythrough different tomo-graphic levels. (B) Actualcurved multiplanar reforma-tion. The 3D course of thevessels is projected onto a2D plane.

EBT CORONARY ANGIOGRAPHY 143

denser noncoronary structures, such as contrast-filled cavities (left ventricle, right ventricle, atria)and bony structures, have to be removed firstfrom the dataset to ensure visualization of thecoronary arteries. Furthermore, the technique issensitive to partial volume effects and uses only10% of the available imaging data.

EBCT Angiography of NativeCoronary Arteries

Several studies have recently been publishedcomparing intravenous EBCT coronary angiogra-phy to conventional angiography2-7 (Table 1).The image acquisition protocols for all thesestudies were basically the same as described in the

methodology section. Differences were the num-ber of tomograms made (40 or 60) and the tomo-gram thickness/table increment used (either 1.5mm thickness, 1.5 mm table increment, or 3 mmthickness with 2 mm table increment). Three-dimensional image rendering techniques also dif-fered among the studies (Table 1). All used boththe original axial tomographic data and the recon-structed 3D data to assess the EBCT coronaryangiograms.

Visualization

Overall, the technique allowed interpretable visu-alization of the proximal and mid coronary arter-ies in approximately 80% to 90% of cases (Table

Fig 7. Example of 3D rendering of venous bypass grafts. Shown are volume renderings from different angles. Thispatient has a single venous bypass graft to the LAD (arrow), a single venous graft to the right coronary artery(arrowhead), and a single graft to the circumflex coronary artery (small white arrows). In the graft to the RCA thecontours of a stent can be appreciated (asterisk). The graft to the circumflex area is totally occluded at theanastomosis with the aorta. Three stents had been implanted earlier and can still be appreciated on the 3Drenderings. No contrast material is present between the stents, proving that the graft is totally occluded. Ao,ascending aorta; LAD, native left anterior descending artery; Pa, pulmonary artery; RCA, native right coronaryartery; SCV � superior caval vein.

TABLE 1. Diagnostic Accuracy for Detection of Greater Than 50% Diameter Stenosis by EBCTAngiography in Several Published Studies

PatientsRenderingTechnique

ArteriesInterpretable

Sens/Spec (%)

LMInt/Sens/Spec(%)

LADInt/Sens/Spec (%)

LCXInt/Sens/Spec (%)

RCAInt/Sens/Spec (%)

Schmermund et al2 28 SR/MPR/MIP 88% 83/91 100/�/100 90/82/98 71/75/81 91/90/83Rensing et al3 37 VR 81% 77/94 97/100/100 95/82/92 76/83/89 66/60/97Achenbach et al4 125 SR 75% 92/94 84/�/99 80/98/88 66/78/88 70/93/96Nakanishi et al5 37 MPR NR 74/91 NR/100/100 NR/83/84 NR/67/96 NR/63/79Reddy et al6 23 MIP 90% 88/79 100/�/96 100/93/63 95/100/67 74/67/77Budoff et al7 52 MIP/SR 90% 78/91 100/�/100 92/82/90 85/67/86 79/83/79

Abbreviations: Int, interpretable, good enough visualisation to allow classification; LM, left main coronary artery; Sens,sensitivity; Spec, specificity; NR, not reported.

RENSING ET AL144

1). Clearly, the right coronary artery (RCA) andleft circumflex coronary artery (LCX) were lessoften assessable than the left anterior descendingcoronary artery (LAD). Most often this was causedby cardiac motion artifacts. The RCA and LCX arevulnerable to motion artefacts because a large partof their course is in the atrioventricular groove inclose vicinity to the atria. Atrial contractionduring the 100-ms image acquisition window cancause these motion artefacts. Other reportedcauses of nonassessability are small size of thearteries, partial overlap with other contrast-filledstructures such as the coronary sinus and theatrial appendages, and poor opacification of thesmaller, distal coronary arteries at the end of thecontrast injection. Also, heavy, circular calcifica-tion of the coronary artery wall can obscure thecoronary lumen and render the coronary segmentnoninterpretable. If not recognized, circular calci-fication can be a cause of false negative2-4 or falsepositive6 results. Achenbach4 reported respiratorymotion artefacts to be the main cause of nonassess-ability in his patient population. This was, however,not found to be a major factor in the other studies.

Diagnostic Accuracy

Diagnostic accuracy for the detection of signifi-cant coronary artery disease was best for the leftmain coronary artery. Not all studies includedpatients with left main stenoses, but false positivetest results were very rare (Table 1). Reportedspecificities ranged from 96% to 100%. The 3 leftmain stenoses that were reported3,5 were allcorrectly diagnosed by EBCT angiography (sensi-tivity 100%).

Diagnostic accuracy for the detection of signifi-cant stenoses in the proximal and middle LADwas uniformly found to be good. Sensitivityvalues ranged from 82% to 98% and specificityvalues ranged from 63%6 to 98%. The highamount of false-positive test results in the LADreported in the study by Reddy et al, was due tomisinterpretation of areas with heavy calcifica-tion. They used maximum intensity projectionsfor interpretation, and it is known that partialvolume effects on MIP renderings can give theimpression of a stenosis in heavily calcified partsof the coronary arteries.6 Furthermore, it is knownfrom EBCT calcification studies that heavily calci-

fied sites do not necessarily coincide with asignificant stenosis.15,16

In general, diagnostic accuracy was poorer forthe RCA and LCX as compared to the LAD. Thelower specificity can be attributed to the smallersize of these vessels, especially the LCX.2-4 Fur-ther, poor contrast opacification and partial vol-ume effects can give the impression of a stenosisin these smaller arteries and lead to a false-positive result. The lower sensitivity can be ex-plained by the fact that a large part of the courseof the RCA and LCX is perpendicular to theimaging plane.4 Resolution in the scanning direc-tion is lower than in the tomographic plane(tomograms are 1.5-mm thick), and these vesselsare thus for a large part visualized with a lowerspatial resolution. Short stenoses can thus bemissed, resulting in a higher false-negative rate.

The reported numbers of false-positive testswere low, which means a high specificity and highnegative predictive value. This potentially makesEBCT coronary angiography a very useful test torule out significant coronary artery disease andthus to avoid conventional coronary angiographyin a large group of patients. It is important torealize, however, that diagnostic accuracy calcula-tions can only be based on the coronary segmentsthat were assessable in the first place. The factthat on average only 70% to 75% of EBCTangiograms were of sufficient quality to allowassessment of the right and circumflex coronaryartery makes this technique at the moment un-suited as an alternative for conventional coronaryangiography.

EBCT Angiography to Assess BypassGraft Patency

In the 1980s, several EBCT studies had beenconducted to assess coronary venous and arterialbypass graft patency17-19 (Table 2). Three-dimen-sional rendering techniques were not available atthat time and therefore only patency of the graftscould be determined by assessment of the indi-vidual transaxial tomograms. Only a few arterialbypass grafts were included in these studies, butdiagnostic accuracy appears to be equal for bothtype of conduits (Table 2). With the advent of 3Dreconstruction techniques (Fig 7), a renewedinterest in EBCT bypass graft visualization arose.

EBT CORONARY ANGIOGRAPHY 145

Achenbach et al20 and Jong-Won et al21 usedsurface rendering to assess bypass graft patency(Table 2). Diagnostic accuracy was similar to thatreported in the older studies using tomographicassessment. However, 3D rendering offers thepossibility not only to assess patency but also tovisualize hemodynamically significant stenoses inbypass grafts. This was studied by Achenbach etal. They were able to evaluate 84% of patentgrafts. Sensitivity and specificity to detect a signifi-cant stenosis was 100% and 97%, respectively.The main reasons for noninterpretability werebreathing artefacts and misplacement of the imag-ing volume, which caused part of the graft not tobe visualized.

Jong-Won et al found a lower diagnostic accu-racy with 3D renderings for the assessment ofarterial grafts as compared with venous grafts.They attribute that to the smaller size of thesegrafts and the fact that metal surgical clips aremore abundantly used. These clips are known tocause beam hardening and stellate artefacts.

Technical Improvements

Technical improvements should primarily be di-rected toward reducing the amount of noninter-pretable EBCT angiograms. Currently work isbeing conducted to reduce the tomogram acquisi-tion time from 100 ms to 50 ms. In combinationwith better ECG triggering protocols, end-diastolic motion artefacts can probably be re-duced. Recently a new detector array was intro-duced that is expected to improve the in-planeresolution by 30% to 10 linepairs/cm. It is ex-pected that this will increase assessability anddiagnostic accuracy in smaller vessels and im-prove discrimination between the coronary arter-ies and overlapping structures.

Breath-holding limitations and single tomo-

gram acquisition per heartcycle limit the volumeof the heart that can be visualized during 1contrast injection. Therefore, the distal coronaryarteries can not be visualized during a singlecontrast injection in all patients. This can besolved by a second contrast injection at theexpense of an increased contrast volume. Further-more, tomograms that are obtained during thelater phase of the contrast injection suffer from areduced contrast resolution between the contrast-enhanced myocardium and (small) coronary arter-ies.7

Ideally, the complete heart should be scannedwithin a few heartbeats. This would shortenbreath-holding time, decrease the probability ofartefacts caused by arrhythmias, and reduce thetotal amount of contrast medium necessary foropacification. Therefore, to decrease image acqui-sition time, a tomographic imaging techniquemust be able to perform multiple, simultaneous,parallel tomograms and thus become more of avolume scanner. Thomas et al used the multislicemode of the EBCT scanner to obtain 8 tomograms(slice thickness 8 mm) of a thorax phantomwithin 224 ms.22 The complete heart and coro-nary arteries could be visualized in 6 heartbeats.However, very extensive and time-consumingcomputations were necessary to correct for thenoncoplanar orientation of the targetrings relativeto the detector ring and to deconvolve the overlap-ping 8-mm slices into thinner slices. Although theimage acquisition window is probably too longand the deconvolved slice thickness still too highto reliably depict coronary arteries in vivo, thisreport shows that multislice EBCT coronary angi-ography is possible and represents a promisingalternative to single slice angiography. A solutionto the problem of circular calcification of thevessel wall (a major source of false negatives)

Table 2. Patency Assessment of Bypass Grafts by EBCT

No. ofVenous Grafts

No. ofArterial Grafts Analysis Technique

Sens/SpecVenous Grafts

Sens/SpecArterial Grafts

Stanford17 116 11 Tomograms 94/88 88/100Bateman18 39 — Tomograms 95/86 —Bateman19 68 12 Tomograms 95/97 100/100Achenbach20 55 1 3D 100/100 NVJong-Won21 57 22 3D 92/91 80/82

Abbreviations: NV, not visualized; Sens, sensitivity; Spec, specificity.

RENSING ET AL146

might be visualizing the coronary artery from theinside. A post processing technique called ‘‘flythrough virtual angioscopy’’ tracks the contrast-enhanced lumen of the artery on the individualtomograms. After stacking and interpolation thecomputer constructs a movie that gives the illu-sion of travelling through the artery. Using thistechnique van Ooijen et al were able to distin-guish stenosed and calcified areas.23

Conclusion

Intravenous EBCT coronary angiography in con-junction with 3D reconstruction and renderingtechniques is a promising imaging modality thatallows for noninvasive visualization of the proxi-mal and middle parts of the coronary arteries.Currently it is a reasonably robust technique forthe visualization and assessment of the left mainand left anterior descending coronary artery. Animportant problem at the moment is the relativelyhigh proportion of noninterpretable angiogramsof the right and circumflex coronary artery. Activework is being conducted to improve image acqui-sition and postprocessing techniques. This isexpected to improve visualization and diagnosticaccuracy of the technique.

References

1. Moshage WE, Achenbach S, Seese B, et al: Coronaryartery stenoses: Three dimensional imaging with elec-trocardiographically triggered, contrast enhanced, elec-tron beam CT. Radiology 196:707-714, 1995

2. Schmermund A, Rensing BJ, Sheedy PF, et al: Intrave-nous electron beam CT coronary angiography forsegmental analysis of significant coronary artery ste-noses: Feasibility and limitations. J Am Coll Cardiol31:1547-1554, 1998

3. Rensing BJ, Bongaerts A, van Geuns RJ, et al:Intravenous coronary angiography by electron beamcomputed tomography. A clinical evaluation. Circula-tion 98:2509-2512, 1998

4. Achenbach S, Moshage W, Ropers D, et al: Valueof electron-beam computed tomography for the nonin-vasive detection of high-grade coronary-artery steno-ses and occlusions. N Engl J Med 339:1964-1971,1998

5. Nakanishi T, Ito K, Imazu M, Yamakido M: Evaluationof coronary artery stenoses using electron-beam CTand multiplanar reformation. J Comput Assist Tomogr21:121-127, 1997

6. Reddy PR, Chernoff DM, Adams JR, Higgins CB:Coronary artery stenoses: Assessment with contrast

enhanced electron beam CT and axial reconstruc-tions. Radiology 208:167-172, 1998

7. Budoff MJ, Oudiz RJ, Zalace CP, et al: Intravenousthree-dimensional coronary angiography using con-trast enhanced electron beam computed tomography.Am J Cardiol 83:840-845, 1999

8. Gould RG: Principles of ultrafast computed tomogra-phy: Historical aspects, mechanisms and scannercharacteristics, in Stanford W, Rumberger JA (eds):Ultrafast Computed Tomography in Cardiac Imaging:Principles and Practice. Mt Kisco, NY, Futura, 1993,pp 1-16

9. Assessment of the effective dose for routine protocolsin conventional CT, electron beam CT and coronaryangiography. Rofo Fortschr Geb Rontgenstr NeuenBildgeb Verfahr 170:99-104, 1999

10. Janowitz WR, Agatston AS, Viamonte M: Comparisonof serial quantitative evaluation of calcified coronaryartery plaque by ultrafast computed tomography inpersons with and without obstructive coronary arterydisease. Am J Cardiol 68:1-6, 1991

11. Stehling MK, von SmekalA, Reise M: MRI der Koronar-arterieen Moglichkeiten und Grenzen. Radiologe 34:462-468, 1994

12. Ney D, Fishman E, Magid D: Volumetric rendering ofcomputed tomography data: Principles and tech-niques. Comput Graphics Applications 10:24-32, 1990

13. McAlpine WA: Heart and Coronary Arteries. Heidel-berg, Germany, Springer Verlag, 1975

14. Van Ooyen PMA, de Feyter PJ, Oudkerk M: Anintroduction to three dimensional cardiac image render-ing and processing. Cardiologie 4:312-319, 1997

15. Simons DB, Schwartz RS, Edwards WD, et al: Nonin-vasive definition of anatomic coronary artery diseaseby Ultrafast computed tomographic scanning: A quan-titative pathologic comparison study. J Am Coll Cardiol20:1118-1126, 1992

16. Mautner SL, Mautner GC, Froehlich J, et al: Coronaryartery calcification: Assessment with electron beamCT and histomorphometric correlation. Radiology 192:619-623, 1994

17. Stanford W, Brundage BH, Macmillan R, et al: Sensitiv-ity and specificity of assessing coronary bypass graftpatency with ultrafast computed tomography. J AmColl Cardiol 12:1-7, 1988

18. Bateman TM, Gray RJ, Whiting JS, et al: Cine com-puted tomographic evaluation of aortocoronary by-pass graft patency. J Am Coll Cardiol 8:693-698,1986

19. Bateman TM, Gray RJ, Whiting JS, et al: Prospectiveevaluation of ultrafast cardiac computed tomographyfor determination of coronary artery bypass graftpatency. Circulation 75:1018-1024, 1987

20. Achenbach S, Moshage W, Ropers D, et al: Non-invasive, three-dimensional visualization of coronarybypass grafts by electron beam tomography. Am JCardiol 79:856-851, 1997

21. Jong-Won H, Seung-Yun C, Won-Heum S, et al:

EBT CORONARY ANGIOGRAPHY 147

Noninvasive evaluation of coronary artery bypassgraft patency using three dimensional angiographyobtained with contrast enhanced electron beam CT.Am J Radiol 172:1055-1059, 1999

22. Thomas PJ, McCollough CH, Ritman EL: An electron-

beam CT approach for transvenous coronary angiog-raphy. J Comput Assist Tomogr 19:383-389, 1995

23. Van Ooijen PMA, de Feijter PJ, Oudkerk M: Virtualcoronary angioscopy using electron beam computedtomography. Eur Radiol 9:S38, 1999 (abstr, suppl 1)

RENSING ET AL148