Coronary artery stenoses

Journal of Clinical Imaging 25 (2001) 95 – 100

Coronary artery stenoses: A phantom study using contrast enhanced three-dimensional electron beam tomography Bin Lua, RuPing Daia, Hua Baia, Sha Hea, BaoLian Jinga, ShiLiang Jianga, Nan Zhuanga, XiaoGang Sunb,c, Matthew J. Budoff d,* a Department of Radiology, FuWai Cardiovascular Institute and Hospital, Peking Union Medical College, Beijing 100037, China Department of Heart and Vascular Surgery, FuWai Cardiovascular Institute and Hospital, Peking Union Medical College, Beijing 100037, China c Chinese Academy of Medical Sciences, Beijing 100037, China d Saint John’s Cardiovascular Research Center, Harbor-UCLA Research and Education Institute, 1124 West Carson Street, RB-2, Torrance, CA 90502, USA b

Received 20 July 2000

Abstract This paper evaluated the accuracy of electron beam tomographic angiography (EBA) with conventional coronary arteriography (CCA) using four graded artificial stenoses in a postmortem swine coronary phantom model. The sensitivity, specificity, and accuracy of EBA for diagnosing significant stenosis (  50% stenosis) were 94.3%, 96.7%, and 95.8%, respectively. The diagnostic accuracy of EBA had no significant difference with CCA (c2 = 0.0162; P > .05). EBA three-dimensional (3D) procedures had high interobserver reproducibility (k =.92 – .95, P >.05). Maximum intensity projection (MIP) was the most sensitive and curved planar reformation (CPR) was the most accurate 3D procedure for quantitatively identifying coronary stenosis. EBA yields promising results concerning the visualization of coronary artery stenosis with high accuracy for stenoses >50%. D 2001 Elsevier Science Inc. All rights reserved. Keywords: Coronary artery disease; Angiography; Phantom study; Three-dimensional; Computed tomography; Electron-beam

1. Introduction An invasive procedure, conventional coronary arteriography (CCA) is still the diagnostic ‘‘gold standard’’ for establishing the presence, location, and severity of coronary artery disease (CAD) [1]. Noninvasive and less expensive tests need to be found to evaluate coronary artery anatomy or screen its luminal stenosis. Electron beam tomographic angiography (EBA), with its ability of three-dimensional (3D) reformatting, is an emerging technology with the potential for obtaining essentially noninvasive coronary arteriograms [2]. EBA images, with its unique combination of high spatial and temporal resolution and ability to trigger image acquisition to the electrocardiogram, can be reformatted or rendered by using different 3D

protocols, such as shaded surface display (SSD) [2– 4], maximum intensity projection (MIP) [2,4], multiple or curved planar reformation (MPR or CPR) [5 – 7], and volume rendering (VR) [8]. Using these protocols for reformatting EBA – 3D images, the stenosis of the coronary artery lumina could be demonstrated with the sensitivity of 74 – 89%, the specificity of 79 – 94%, and accuracy of about 87% [2– 8]. The objectives of our phantom study were to test the accuracy of EBA for identifying significant coronary artery stenosis and evaluate the value of different 3D protocols, by comparison to CCA.

2. Materials and methods 2.1. Phantom design

* Corresponding author. Tel.: +1-310-2224107; fax: +1-310-7870448. E-mail address: [email protected] (M.J. Budoff).

The author (BL) built a new model of coronary artery stenoses using 28 fresh normal postmortem swine hearts

0899-7071/01/$ – see front matter D 2001 Elsevier Science Inc. All rights reserved. PII: S 0 8 9 9 - 7 0 7 1 ( 0 1 ) 0 0 2 4 8 - 0

96

B. Lu et al. / Journal of Clinical Imaging 25 (2001) 95–100

Table 1 Comparison between 3D electron beam tomography and coronary artery angiography in demonstrating 60 lesions CCA 1

MIP

2

3

CPR

SSD

4 1 2

3 4 1 2

2 3 2 7

2 2 0 4

2 9 1 3 1 1 4 13

4 4 0 8

2 1 0 3

5 6 1 3 2 3 8 12

3 1 1 3 5 6 9 17

0 0 2 2

0 3 0 2 0 9 0 14

0 3 4 7

0 0 1 1

1 0 0 1

4 1 2

LAD-P 5 7 6 LAD-M 3 3 4 LAD-D 2 0 1 Total 10 10 11

2 1 0 3

4 8 3 2 2 4 9 14

RCA-P RCA-M RCA-D Total

0 0 1 1

0 1 2 3

2 2 2 6

2 2 4 3 1 7 7 12

3

3

4

2 2 1 5

2 2 0 4

2 0 1 4 3 9 5 13

0 0 2 2

The coronary artery phantoms were positioned head first, scanned in the supine position with the heart apex direct to the right side at an angle of about 30 with the table longaxis (just like the heart position of human beings in the thorax cavity). A power injector was used to inject nonionic contrast media (Ultravist 300 [Schering, Berlin, Germany] was diluted into 2% using normal saline) into the coronary arteries with a rate of 0.5 ml/s and total volume of 10– 14 ml. The balloon at the top of the catheter

CCA = conventional coronary arteriography; MIP = maximum intensity projection; CPR = curved planar reformation; SSD = shaded surface display; LAD = left anterior descending; RCA = right coronary artery; P = proximal; M = middle; D = distal; and ‘‘1’’= < 50% stenosis; ‘‘2’’= 50 – 74% stenosis; ‘‘3’’= 75 – 99% stenosis, and ‘‘4’’= 100% stenosis.

(weight range 350– 540 g, mean 461.9 ± 50.6 g). Both the left anterior descending (LAD) and right coronary artery (RCA) were used to make physiologic shaped coronary arteries (the LAD and RCA of normal swine heart were similar with human being’s but no significant left circumflex branch exists). The swine heart coronary arteries were washed by saline solution before use, and established that these coronary arteries had not been broken or damaged. There were a total of 26 branches of LAD and 22 branches of RCA made into 60 stenosis, by using 5– 7 prolene suture to make irregular stenosis of the coronary wall (1 – 99% stenosis) and ligate the coronary artery (100% occlusion). All these 60 artificial stenoses were made successfully, including 18 stenoses (2 were 100% stenoses) in the proximal LAD, 11 stenoses (1 was 100% stenosis) in the mid-LAD, and 3 stenosis in the distal LAD; 6 stenoses in proximal RCA, 9 in mid-RCA, and 11 in distal RCA (1 was 100% stenosis). 2.2. Electron beam tomography scans This study was performed with an electron beam tomography scanner (Imatron C-150XL, San Francisco, CA).

Table 2 Evaluation of EBA 3D protocols for identifying significant coronary artery stenosis (  50% stenosis) (n = 48) MIP CPR SSD Overall

Sen (%)

Spe (%)

Accu (%)

PPV (%)

NPV (%)

96.3 94.2 92.5 94.3

93.3 98.9 97.8 96.7

94.4 97.3 95.8 95.8

89.7 98.0 96.1 94.6

97.7 96.8 95.7 96.7

Sen = sensitivity; Spe = specificity; Accu = accuracy; PPV = positive predictive value; NPV = negative predictive value; MIP = maximum intensity projection; CPR = curved planar reformation; and SSD = shaded surface display.

Fig. 1. Postmortem swine heart phantom of LAD was scanned by using EBA (slice-thickness was 1.5 mm) and the two-dimensional transverse images were rendered with MIP. (A) Conventional coronary arteriographic image showed the moderate (Category 2) and severe (Category 3) stenosis in proximal and middle portion of the LAD (arrows), respectively. (B) The same significant stenosis (  50% stenosis) were identified by EBA-MIP images (large arrows), however, a distal false positive lesion was diagnosed (small arrow) because the image was degraded by partial volume effects.

B. Lu et al. / Journal of Clinical Imaging 25 (2001) 95–100

97

2.3. 3D reconstruction EBA images were transferred to ISG workstation (Allegro, Toronto, Ontario, Canada). 3D images of the coronary arteries were reconstructed by a combination of three spatial

Fig. 2. (A) CCA showed a mild to moderate stenosis both in the proximal (white arrow) and distal portion of the LAD (black arrow). (B) The same stenosis phantom was scanned and reformatted using EBA-MIP method. Both of the two artificial stenoses (large arrows) were demonstrated but overestimated because of a higher windowing level (113 Hu) was selected. The diagonal branches and the distal portion of the LAD near the apex of the heart could not be evaluated (small arrows).

was inflated and put into the orifice of the coronary arteries to prevent the top of the catheter withdrawing and the contrast materials leaking. EBA scanning was performed using single slice mode with 1.5 mm slice-thickness, 100 ms acquisition time, without electrocardiographically triggering. About 45 –60 slices were scanned from the bottom to the apex of the heart phantoms. The electron beam images were reconstructed by using normal kernel and algorithm, 12.7 cm as field-of-view (FOV), 512  512 as matrix, and the pixel size was 0.25  0.25 mm2.

Fig. 3. One of the swine heart phantoms was scanned using the same EBA protocols. (A) CCA showed a localized moderate to severe stenosis in the proximal portion of the LAD (arrow). (B) The EBA axial images were reformatted by using SSD, and the proximal stenosis on LAD was demonstrated but overestimated as a 100% occlusion (arrow). This phenomenon caused by a relatively higher gray-scale threshold was used to reconstruct the stenosis.

98

B. Lu et al. / Journal of Clinical Imaging 25 (2001) 95–100

reconstruction algorithms: MIP, CPR, and SSD. The techniques of MIP, CPR, and SSD were used and described by other writers [2– 6]. In our study, the window levels and width for observing MIP and CPR images were 50 and 500 Hu, respectively; the reconstructive threshold of SSD was 50 –70 Hu. 2.4. Coronary artery angiography After acquiring EBA data on these coronary phantoms, conventional coronary angiography was performed immediately. Coronary arteriograms were obtained in the standard fashion with two orthogonal views (left anterior oblique 60 and right anterior oblique 30) of each coronary artery. The same contrast agents (Ultravist 300, Shering, Berlin, Germany) were injected into coronary arteries by using the same rate of 0.5 ml/s. Each coronary vessel was assessed, and the visual estimation of the percent luminal reduction for each lesion in four categories was reported: < 50% (luminal irregularities), 50 – 74% (moderate stenosis), 75 – 99% (severe stenosis), and 100% (complete occlusion). 2.5. Comparison of EBA and coronary artery angiography EBA 3D images and coronary angiograms were analyzed by two experienced readers unaware of each other’s findings. Significant coronary artery stenosis was defined as  50% luminal diameter narrowing in that vessel for both techniques. Diagnostic sensitivity and specificity of different 3D EBA procedures were calculated for hemodynamically significant stenoses, and interobserver variability was evaluated by using Cohen’s kappa statistic. Statistical significance between EBA and coronary angiography was calculated using the chi-square test. The lesion severities by EBA were compared with CCA.

3. Results In this group, 60 artificial coronary artery stenoses were made successfully by using surgical ligation procedures, and all of them had been identified on EBA 3D images. According to the severity of the coronary luminal narrowing and distribution of these stenoses, Table 1 demonstrates the consensus reading between 3D EBA and coronary angiography in the 60 lesions. The results of EBA evaluation of 3D protocols for demonstrating significant coronary stenosis (  50% stenosis) were summarized in Table 2. The overall sensitivity, specificity, and accuracy were 94.3%, 96.7%, and 95.8%, respectively. MIP had the highest sensitivity of 96.3%; CPR had the highest specificity and accuracy of 98.9% and 97.2%, respectively. The false positive result of MIP (n = 6) was higher than CPR (n = 1) and SSD (n = 2) (Figs. 1 and 2). The false negative result was higher in SSD (n = 4) than CPR (n = 3) and MIP (n = 2) (Fig. 3). Electron beam angiography using MIP, CPR, and SSD 3D protocols for identifying the 60 coronary artery lesions on phantoms had no significant difference with coronary angiography (c2 = 0.016; P >.05); the Cohen’s kappa statistic for interobserver variability on EBA was .92 – .95 ( P >.05) for segmental classification (grouping normals with luminal stenoses). For analyzing the luminal stenosis in four categories, MIP, CPR, and SSD images accurately identified 30 (50.0%), 35 (58.3%), and 33 (55.0%) of 60 lesions, respectively (overall 54.4%); however, 16 (26.7%), 14 (23.3%), and 14 (22.3%) of 60 lesions were overestimated (overall 24.4%), and 14 (23.3%), 11 (18.3%), and 13 (21.7%) of 60 lesions were underestimated (overall 21.1%), respectively (Fig. 4). The CPR protocol was better than MIP and SSD protocols for quantitatively diagnosing coronary luminal stenosis.

Fig. 4. The results of 3D protocols of EBA were compared with CCA for quantitatively identifying 60 coronary luminal stenoses in four categories. MIP = maximum intensity projection; CPR = curved planar reformation; SSD = shaded surface display.

B. Lu et al. / Journal of Clinical Imaging 25 (2001) 95–100

4. Discussion Our phantoms utilizing real postmortem swine hearts were ideal because the physical characteristics were very like human heart tissues, the geometric properties (size, length, diameter, volume, and bifurcations or small branches) were known and realistic. Renandin et al. and Davros et al. [9,10] reported their phantom studies on coronary artery and renal artery, respectively. Their anthropomorphic phantoms were made of liquid resin by using computer-aided design and computer-aided manufacturing, and could be imaged on helical CT and MRI three-dimensionally both statically and dynamically by injecting different contrast media. However, the attenuation characteristics of these plastic phantoms were slightly different with human heart tissue, and the accuracy of the building process and the viscosity of the liquid resin do not yet allow the manufacture of phantoms of small arteries less than 1 mm in diameter without errors [9]. Gould et al. and Farmer et al. [11,12] studied regional myocardial blood flow and assessment of left ventricular wall and chamber dynamics during transient myocardial ischemia using cine computed tomography on dogs. Their phantoms were made by ligating the LAD with a slipknot. Marcus reported postmortem coronary angiography on monkey coronary phantoms, which were induced by feeding high aliphatic food [13]. EBA can routinely be used to visualize the major vessels of the proximal and middle human coronary artery tree in normal subjects [14], and identify significant stenosis with a considerable sensitivity, specificity, and accuracy [2 – 8]. Early EBA scanners could maximally perform 40 slices acquisition, the proximal and middle coronary lesions were identified and yielded a sensitivity of 74 –89%, and specificity of 79 – 94%; however, the lesions at the distal portion of each coronary branches were missed [4 – 7]. Current generation of electron beam scanners have no limitations of spatial coverage acquisition, the small or distal lesions could be visualized and diagnosed with the sensitivity of 77 –82% and specificity of 88 – 94% [2,3,8]. Our results showed a higher sensitivity of 94.3%, specificity of 96.7%, and accuracy of 95.8% because the phantoms used in this study had no motion artifacts caused by cardiac pulsation or respiration, and the contrast material was directly injected into coronary arteries (not intravenously). We sought to identify differences in accuracy using different 3D reconstructions. No reports on EBA focused on this research field, and most EBA researchers used one of these 3D procedures independently or two of them in combination without comparison [2 –8]. Nakanishi et al. [5] reported the results of their group by using MPR, the diagnostic sensitivity and specificity were 74% and 91%. Achenbach et al. [7] used the method of CPR, the sensitivity was 89% and the specificity was 92%. Reddy et al. [6] used MIP as their 3D method, finding a sensitivity of 88% and specificity of 79%. Budoff et al.

99

[2], using a combination of two techniques (60% MIP and 40% SSD), demonstrated an overall sensitivity of 78%, specificity of 91%, and accuracy of 87%. Our results confirmed that MIP was the most sensitive 3D procedure for identifying coronary lesions (sensitivity was 96.3%), but with the lowest positive predictive value (89.7%). CPR was an ideal 3D procedure for having the highest specificity (98.9%) and accuracy values (97.2%). SSD had the lowest negative predictive value (95.7%) in our phantom study. According to our phantom study, EBA had similar accuracy with CCA (chi-square test, P >.05), and the kappa statistics of interobserver variability was high (.92 – .95, P >.05). The clinical studies also demonstrate that EBA was a reliable method with a good interstudy and interreader reproducibility for predicting  50% coronary stenosis (kappa values .81 to .88) [2 –4]. From the results of our phantom study, 60 lesions were accurately identified by EBA in overall 54.4% (33 lesions), and 24.4% (15 lesions) were overestimated, 21.1% (12 lesions) were underestimated. When four categories of stenosis were used in our study and compared with coronary angiography results, CPR was the most accurate procedure for identifying coronary stenosis, while MIP most easily overestimated or underestimated coronary stenosis. But in identifying which have or do not have  50% stenosis (two categories), EBA had no significant difference with angiography in this group ( P >.05). Another study found that the image quality of MIP was quite similar to that of 3D images (SSD), and MIP overestimated the 75% stenotic segments on their 4 mm-diameter acryl vessel phantoms; however, MIP slightly underestimated the 25% stenotic phantoms [15]. In studies of spiral CT phantom studies for identifying peripheral vessel stenosis, such as renal artery [16,17] and carotid artery [18], MIP was more sensitive and accurate than SSD [16]. EBA was very reliable and accurate for differentiating normal or minimally narrowed coronary artery (0 – 49% stenosis) from coronary artery occlusion (100% stenosis), but very difficult in making an accurate diagnosis of 50– 99% stenosis. The reason was that 3D procedures would over- or underestimate this stenosis as a misregistration error, or the editing procedure and gray-scale threshold or window level used will change the image quality and affect the stenosis being visualized. The reason for overestimation of stenosis is likely when low subject contrast is present between the vessel and background tissue, either due to reduction of vessel attenuation or elevation of background attenuation. This effect is proportional to the number of voxels through which the projection rays are passed on MIP images and what thresholding are selected on SSD images, and this is the reason that the CPR image is more accurate. Conversely, volume averaging tends to cause underestimation of stenoses if the contrast-to-noise ratio is sufficiently large to permit depiction of the blurred vessel lumen against background attenuation [17].

100

B. Lu et al. / Journal of Clinical Imaging 25 (2001) 95–100

4.1. Study limitations: Our study was based on a nonpulsatile postmortem phantom of swine heart, and the artificial coronary stenoses were not perfectly realistic without pathologic characteristics of atherosclerosis. Our results, including the diagnostic sensitivity, specificity, accuracy, and positive or negative predictive value, were better than some reports on clinics [2– 8], partly because our images had no motion artifacts caused by cardiac and respiratory motion and poor electrocardiographic gating, which detract from image quality [2]. Also, the scanner used did not have the latest software and hardware upgrades (high-resolution detector system with increased spatial resolution), thus the images are suboptimal compared to newer electron beam scanners. Our conclusion was that coronary angiography performed on electron beam tomography was a relatively accurate method for the noninvasive detection and definition of coronary artery stenosis. Our study confirmed that EBA has a potential for identifying significant disease and ruling out obstructive versus nonobstructive coronary artery disease. MIP 3D procedure was the most sensitive, and CPR was the most accurate procedure for quantitatively evaluating coronary artery stenosis. EBA 3D images provided highly reproducible results, but sometimes over- or underestimate the coronary artery stenosis.

References [1] Marcus ML, Skorton DJ, Johnson MR, Collins SM, Harrison DG, Kerber RE. Visual estimates of percent diameter coronary stenosis: ‘‘a battered gold standard’’. J Am Coll Cardiol 1988;11:882 – 5. [2] Budoff MJ, Oudiz RJ, Zalace CP, Bakhsheshi H, Goldberg SL, French WJ, Rami TG, Brundage BH. Intravenous three-dimensional coronary angiography using contrast enhanced electron beam computed tomography. Am J Cardiol 1999;83:840 – 5. [3] Schmermund A, Rensing B, Sheedy P, Bell MR, Rumberger JA. Intravenous electron-beam computed tomography coronary angiography for segmental analysis of coronary artery stenosis. J Am Coll Cardiol 1998;31:1547 – 54. [4] Moshage WE, Achenbach S, Seese B, Bachmann K, Kirchgeorg M. Coronary artery stenosis: three-dimensional imaging with electrocar-

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

diographically triggered, contrast agent enhanced, electron-beam CT. Radiology 1995;196:707 – 14. Nakanishi T, Ito K, Imazu M, Yamakito M. Evaluation of coronary artery stenoses using electron beam CT and multiplanar reformation. J of Comput-Assisted Tomogr 1997;21:121 – 7. Reddy GP, Chernoff DM, Adams JR, Higgins CB. Coronary artery stenoses: assessment with contrast-enhanced electron-beam CT and axial reconstructions. Radiology 1998;208:167 – 72. Achenbach S, Moshage W, Ropers D, Bachmann K. Curved multiplanar reconstructions for the evaluation of contrast-enhanced electron beam CT of the coronary arteries. AJR, Am J Roentgenol 1998;170: 895 – 9. Rensing BJ, Bongaerts A, Van Geuns RJ, Van Ooijen P, Oudkerk M, DeFeyter PJ. Intravenous coronary angiography by electron beam computed tomography: a clinical evaluation. Circulation 1998;98:2509 – 12. Renandin CP, Barbier B, Roriz R, Revel D, Amiel M. Coronary arteries: new design for three-dimensional arterial phantoms. Radiology 1994;190:579 – 82. Davros WJ, Obuchowski NA, Berman PM, Zeman RK. A phantom study: evaluation of renal artery stenosis using helical CT and 3D reconstructions. J of Comput-Assisted Tomogr 1997;21:156 – 61. Gould RG, Lipton MJ, Mcnamara MT, Sievers RE, Koshold S, Higgins CB. Measurement of regional myocardial blood flow in dogs by ultrafast CT. Invest Radiol 1988;23:348 – 53. Farmer D, Lipton MJ, Higgins CB, Ringertz H, Dean PB, Sievers R, Boyd DP. In vivo assessment of left ventricular wall and chamber dynamics during transient myocardial ischemia using cine computed tomography. Am J Cardiol 1985;55:560 – 5. Marcus ML, Armstrong ML, Heistad DD, Eastham CL, Mark AL. Comparison of three methods of evaluating coronary obstructive lesions: postmortem arteriography, pathologic examination and measurement of regional myocardial perfusion during maximal vasodilation. Am J of xCardiol 1982;49:1699 – 705. Chernoff DM, Ritchie CJ, Higgins CB. Evaluation of electron beam CT coronary angiography in healthy subjects. AJR, Am J Roentgenol 1997;169:93 – 9. Uchino A, Kato A, Kudo S. CT angiography using electron-beam computed tomography (EBCT): a phantom study. Radiat Med 1997;15:273 – 6. Rubin GD, Dake MD, Napel S, Brooke Jeffrey R, McDonnell CH, Graham Sommer F, Wexler L, Williams DM. Spiral CT of renal artery stenosis: comparison of three-dimensional rendering techniques. Radiology 1994;190:181 – 9. Brink JA, Lim JT, Wang G, Heiken JP, Deyoe LA, Vannier MW. Technical optimization of spiral CT for depiction of renal artery stenosis: in vitro analysis. Radiology 1995;194:157 – 63. Dillon EH, Van Leeuwen MS, Fernandez MA, Eikelboom BC, Mali WPTM. CT angiography: application to the evaluation of carotid artery stenosis. Radiology 1993;189:211 – 9.