OBJECTIVE. The purpose of this study was to compare the accuracy of 64-MDCT with that of intravascular ultrasound (IVUS) for the identification and quantitative analysis of coronary atherosclerotic plaques.
MATERIALS AND METHODS. Twenty-six patients (17 men, nine women; mean age, 56 years) with suspected coronary atherosclerotic disease were studied using contrast-enhanced 64-MDCT and IVUS. The coronary arteries were divided into 10-mm segments and vascular cross-sectional area (CSA), luminal CSA, and plaque burden were measured in each segment. Plaque analysis software was used to automatically detect both plaques and vessel walls on CT images. Two investigators who were blinded to IVUS results independently determined the presence, classification, and quantitative measurement of atherosclerotic plaques on the CT images, which were then compared with the IVUS images.
RESULTS. Of 40 coronary arteries, 247 of 263 segments were imaged and analyzed by both contrast-enhanced 64-MDCT and IVUS. Sixteen segments were ruled out because of poor CT image quality. Compared with IVUS, 64-MDCT enabled correct detection in 86 of 89 (96.6%) segments containing noncalcified plaques, 25 of 27 (92.6%) segments containing calcified plaques, and 118 of 131 (90.1%) segments without atherosclerotic plaques. Sensitivity, specificity, positive predictive value, and negative predictive value for the detection of plaques by 64-MDCT were 97.4%, 90.1%, 89.7%, and 97.5%, respectively. Plaque analysis software using predetermined Hounsfield unit ranges for different components of plaque was able to distinguish between fibrous, fibrous-soft, and calcified plaques to a significant degree, but was less able to distinguish between soft and fibrous, and between soft and fibrous-soft plaque. Cohen's kappa coefficient for the sole detection of atherosclerotic segments by observers was 0.91. The correlation coefficients to determine vascular CSA, luminal CSA, and plaque burden were r = 0.85, 0.82, and 0.77, respectively (p < 0.01).
CONCLUSION. Compared with IVUS, contrast-enhanced 64-MDCT has a good ability to identify and quantify coronary atherosclerotic plaques. However, the reliable differentiation of the composition of noncalcified plaques is still limited.
Detection, characterization, and quantification of coronary atherosclerotic plaque can improve individualized risk stratification. Invasive coronary angiography, the clinical gold standard for assessing coronary arteries, provides excellent data regarding luminal size but is limited to detection of plaques in remodeling vessels and is not useful for the evaluation of wall thickness or plaque composition. Instead, intravascular ultrasound (IVUS) is considered an accurate method for visualizing the wall of the coronary artery . However, IVUS is invasive and expensive and is unsuited for routine applications related to risk stratification. MDCT is a rapid, noninvasive imaging technique capable of providing high-resolution images for the assessment of the coronary arteries. Several studies using 16-and, more recently, 64-MDCT scanners have confirmed the ability of this technique to detect both calcified and noncalcified coronary artery plaques [2, 3]. In addition, once they are detected, it may be clinically important to determine the components of individual plaques and to measure the overall plaque burden. The aim of this study was to determine the accuracy and reliability of 64-MDCT, using new plaque analysis software, for the identification, classification, and quantification of coronary atherosclerotic plaques compared with IVUS.
Materials and Methods
From July 2006 to February 2007, 26 consecutive patients (17 men, nine women; mean age, 56 years) with coronary atherosclerotic disease underwent 64-MDCT and IVUS. Eleven patients had acute coronary syndrome and 15 patients had stable angina. All patients had abnormal ECG results. Coronary risk was assessed using the Framingham database . The Framingham risk score of these patients ranged from 12 to 17 (average, 14.7). CT was performed within 3 days before or after the catheterization procedure. Patients with arrhythmias, previous bypass surgery, unstable clinical presentation, and contraindications to iodinated contrast media were excluded from enrollment. The institutional ethics committee approved the research protocols, and all patients gave informed consent.
Patients with heart rates greater than 70 beats per minute received 25–50 mg of metoprolol orally 60 minutes before MDCT (Aquilion 64, Toshiba). MDCT data were acquired using an X-ray beam collimation width of 32 mm (0.5-mm slice thickness × 64 rows) with the spatial resolution in x-, y-, and z-axes of 0.35 mm, a gantry rotation of 400 milliseconds, a pitch of 11.2–13.2, an electric current of 350–450 mAs, and a tube voltage of 120–135 kV (depending on patient size).
Contrast medium (70–85 mL, 370 mg I/mL, depending on patient size) was injected IV at 4 mL/s followed by a 40-mL saline flush at the same injection rate. The CT value of the area of interest in the ascending aorta was monitored from the start of the injection. As soon as the CT value in the ascending aorta reached 130–150 H, scanning was manually initiated, and imaging of the entire volume of the heart was acquired during one breath-hold with simultaneous recording of the ECG tracing.
Axial images (slice thickness, 0.5 mm; increment, 0.3 mm) were reconstructed using a multisector reconstruction algorithm. Using retrospective ECG gating, we routinely performed reconstructions at 75% of the R-R interval. If motion artifacts existed at these reconstructions, another, more optimal ECG phase was chosen to provide better image quality.
64-MDCT Image Analysis
The baseline axial images were postprocessed and analyzed by plaque analysis software (Sure Plaque, Toshiba) at the workstation (HP XW8200, Hewlett-Packard). The software detected both plaques and vessel walls on CT images. Different CT density ranges hypothesized to represent different components of plaque were displayed with various colors. Specifically, the lipid composition (range, –100 to 29 H) was represented by red; fibrous composition (range, 30–189 H) was represented by blue; calcification (range, 350–1,000 H) was represented by yellow; lumen (range, 190–349 H) was represented by green; and vessel wall (range, 30–189 H; same range as for fibrous composition) was represented by luteous. The Hounsfield unit ranges for plaque were preset by software and could be manually altered by the operator. The ranges should be adjusted slightly to avoid the influence of lumen density and obtain more accurate quantitative measurements. The mean density in the contrast-enhanced lumen of the proximal coronary artery (American Heart Association [AHA] segments 1, 5, 6, and 11) and the Hounsfield unit range of the vessel wall, adjusted to equal the IVUS image in size, were recorded. The results of this analysis revealed that when the average intensity in the lumen of the proximal artery is > 380 H, the optimal setting to detect plaque and the outer vessel boundaries is obtained at the CT vessel wall range of 30–189 H; however, when the average intensity of the proximal lumen is < 380 H, a CT vessel wall range of 20–179 H is optimal.
For comparisons with IVUS, the target coronary arteries were divided into 10-mm segments from the ostium or onset [2, 3]. Segments with artifacts or blur images were excluded. Plaque qualitative analysis included determinations of the presence and classification of plaque in each coronary segment. Atherosclerotic plaques were considered present when the vessel wall thickness exceeded 0.5 mm. Each plaque was classified according to the color displayed by the software. Plaques displayed as red or green were classified as noncalcified plaque. Plaques with yellow present were defined as calcified plaque without considering the calcified size. Luminal cross-sectional area (CSA) (mm2), vascular CSA—that is, external elastic membrane CSA in IVUS—and plaque burden (%) (defined as external elastic membrane CSA minus the luminal CSA and then divided by the external elastic membrane CSA again) were measured in each 10-mm interval cross section. Two investigators blinded to the IVUS results who were not involved in the later comparative analysis, performed these analyses independently .
64-MDCT Density Measurements
Five well-distributed regions of interest (ROIs), each with a 1 mm2 area, were selected in each plaque. The CT density of each ROI area was measured and recorded. The mean density was calculated by averaging the density measurements. Density measurements for calcification were performed only in hyperdense parts of the calcified plaques. To obtain accurate and complete density of noncalcified plaques, entire plaques were measured instead of hypodense tissue areas only . Densities of different types (soft, fibrous, and fibrous-soft plaques) of noncalcified plaques corresponding to IVUS detection were retrospectively calculated.
IVUS was performed as part of the invasive diagnostic procedure. After coronary angiography, the patients were given intracoronary administration of 0.2 mg of nitroglycerin before introduction of the IVUS catheter. A Volcano Eagle Eye IVUS catheter (20-MHz, electronic transducer, Volcano) was used. The ultrasound catheter was positioned sufficiently distal (30 mm distal) to the targeted site, and motorized pullback at a speed of 0.5 mm/s was used.
The IVUS data were stored digitally and assessed offline by the software configuration of the IVUS system. As with MDCT, 10-mm segments were analyzed. The external elastic membrane CSA, equal to the vascular CSA in CT analysis, luminal CSA, and plaque burden were calculated by software after the borders of vessel intima and adventitia were depicted manually in each detectable cross section. These quantitative measurements were performed in accordance with the IVUS interpretation recommendations of the American College of Cardiology (ACC) . According to the ACC recommendations, atherosclerotic plaques were defined as lesions located between the media and the intima with a thickness of at least 0.5 mm. Calcified plaques were defined as plaque tissue containing any tissue with an echogenicity as bright as, or brighter than, the adventitia causing acoustic shadows. Noncalcified plaques were defined as plaque tissue without acoustic shadows, including soft (hypoechoic) plaques, defined as plaque tissue revealing an echogenicity lower than the adventitia, and fibrous (hyperechoic) plaques, having an echogenicity as bright as, or brighter than, the adventitia. Plaques containing more than one acoustical subtype were considered mixed plaques. Fibrous-soft plaques were defined as plaques in which hypoechoic and hyperechoic lesion occupied approximately 50% of the plaque area, respectively, Two investigators, blinded to the CT results, independently determined the presence, classification, and quantitative measurement of athero-sclerotic plaque in IVUS images.
Comparison of IVUS Versus 64-MDCT
In this study, both 64-MDCT and IVUS were performed in 40 coronary arteries with moderate (50–75%) to significant (> 75%) coronary artery stenoses on conventional angiography, including three right coronary arteries (RCA), 14 left main arteries (LMA), 20 left anterior descending arteries (LAD), and three left circumflex arteries (LCX). Coronary atherosclerotic plaques often involve a large segment or the entire coronary vessel; therefore, the target coronary trees were divided into 10-mm segments on the basis of their distance to the ostium [2, 6]. The complete distance of the target artery from the distal reference point to the coronary ostium was measured using the longitudinally reconstructed IVUS data set and multiplanar reconstructions of the CT data set. Furthermore, we selected fiduciary points, such as side branches, characteristic calcifications, or stents, to ensure that the same corresponding coronary sections were compared. Each segment was morphologically classified according to IVUS and CT criteria. Noncalcified plaques defined by CT images included the soft, fibrous, and fibrous-soft plaques in IVUS; and calcified plaques defined by CT images were equal to the ones in IVUS, including fibrous calcified plaques. The comparison between 64-MDCT and IVUS was on a segmental basis for qualitative analysis and on a site-by-site basis for quantitative analysis. A measurement discrepancy of 1 mm per vessel section was tolerated.
Sensitivity and specificity of MDCT for detection of any plaque were calculated on a segment-per-segment basis. Cohen's kappa statistic was calculated to determine interobserver agreement in detecting plaque. The interobserver variability for 64-MDCT quantitative measurements was calculated . To compare the mean density values of different plaque types, the nonparametric Kruskal-Wallis test was used. For comparison of the external elastic membrane CSA, luminal CSA, and plaque burden obtained by 64-MDCT and IVUS, the Spearman's rank correlation coefficient was determined. The paired Student's t test was used to test for potential differences between those parameters obtained by both methods. A p value < 0.05 was considered statistically significant. All calculations were performed using the SPSS13 software package (SPSS).
No patients were excluded because of reduced MDCT image quality in this study. Of 40 coronary arteries, 247 of 263 segments were imaged by both contrast-enhanced 64-MDCT and IVUS. Sixteen segments were ruled out due to poor CT image quality. Nine sections with large or circular calcifications were excluded from quantitative analysis because of partial volume effects that made it impossible to measure the lumen. Table 1 shows that, compared with IVUS, 64-MDCT enabled a correct detection in 86 of 89 (96.6%) segments containing noncalcified plaques and 25 of 27 (92.6%) segments containing calcified plaques. In 118 of 131 (90.1%) segments, atherosclerotic lesions were correctly excluded. Eighteen of 247 segments were incorrectly classified: 13 were classified as normal by IVUS but misclassified as containing noncalcified plaque by CT; three classified as noncalcified plaque by IVUS were misclassified by CT as normal; and two containing small and superficial calcifications were misclassified as containing noncalcified plaque by CT. The sensitivity, specificity, positive predictive value, and negative predictive value for the detection of any plaque by 64-MDCT were 97.4%, 90.1%, 89.7%, and 97.5%, respectively.
TABLE 1: Comparison of 64-MDCT and Intravascular Ultrasound for Detecting and Differentiating Coronary Plaque
Note—Data are numbers of 10-mm coronary artery segments.
In 64-MDCT, mean contrast enhancement in the coronary lumen was 398 ± 74 H in normal proximal segments (AHA 1, 5, 6, and 11). The mean CT density values for soft (hypoechoic), fibrous (hyperechoic), fibrous-soft, and calcified plaques were 79 ± 34 H (range, 7–149 H), 90 ± 27 H (range, 22–154 H), 72 ± 36 H (range, 3–152 H), and 772 ± 251 H (range, 295–1,325 H), respectively (Fig. 1). The differences of the mean CT density values between fibrous, fibrous-soft, and calcified plaques were significant, with a p value < 0.01; however, the differences between soft and fibrous, and between soft and fibrous-soft, were not significant. Use of 64-MDCT enabled the visualization of lipid pools (hypodense spots) in three of five (60%) segments; and in these lipid pools, the hypodense area was at least 50 H less than the highest value of surrounding noncalcified plaque (Figs. 2A, 2B, 2C, 2D, 2E, 2F and 3A, 3B, 3C, 3D, 3E). Fourteen segments that were homogeneous and hypo- or hyperechoic on IVUS were identified as hypodense areas on 64-MDCT, and all had a difference in CT density of less than 50 H.
The Spearman's rank correlation coefficients of vascular CSA (external elastic membrane CSA), luminal CSA, and plaque burden determined by 64-MDCT and IVUS were r = 0.85, 0.82, and 0.77, respectively (p < 0.01; Fig. 4A, 4B, 4C). The paired Student's t test showed that luminal CSA was underestimated by 64-MDCT (5.81 ± 3.32 mm2 vs 6.32 ± 3.04 mm2; p < 0.01). External elastic membrane CSA and plaque burden were overestimated by 64-MDCT (17.02 ± 5.79 mm2 vs 16.07 ±6.30 mm2 and 61.02% ± 16.02% vs 56.25% ± 16.07%, respectively; p < 0.05). The interobserver variability for plaque quantitative measurements by 64-MDCT was 12%. Cohen's kappa coefficient for the sole detection of atherosclerotic segments was 0.91, indicating good agreement.
Several studies using 16-MDCT scanners have confirmed the ability to detect noncalcified and calcified coronary artery plaques [2, 3]. The improved spatial and temporal resolution of 64-MDCT has facilitated evaluation of atherosclerotic plaques in coronary arteries. Achenbach et al.  reported a sensitivity of 78% and a specificity of 87% for the detection of noncalcified plaque in coronary artery segments by 16-MDCT. Leber et al.  found a sensitivity of 84% and a specificity of 91% in the detection of proximal coronary plaques by 64-MDCT. Our studyshowed excellent sensitivity (97.4%) and specificity (90.1%) of 64-MDCT for detection of coronary plaques and a sensitivity of 96.6% for noncalcified plaque in coronary arteries without considering degree of stenosis, representing a significantly increased accuracy compared with several previously reported studies.
Better results obtained in our study depended not only on high spatial and temporal resolution of 64-MDCT but also on the application of plaque analysis software. This semiautomatic tool can accurately detect plaque compared with IVUS and can also distinguish between calcified and noncalcified plaque. The morphology of coronary plaques could be visualized and quantified. Software assessment allowed detailed investigation of plaques by color-coding on preset Hounsfield unit ranges. Hence, contrast-enhanced lumen, vessel wall, and even plaque composition (soft, fibrous, or calcified) could be depicted in different colors according to their corresponding CT Hounsfield unit ranges. Furthermore, this software was able to describe plaque morphology more clearly and to perform quantitative analysis of plaque. Note that the lumen and vessel wall of coronary arteries had no obvious border on CT images because of limited spatial resolution and because of smaller vascular sections compared with the carotid and aortic arteries. Thus, software could provide more rapid and accurate detection of plaques by CT range settings than by a manual method of depicting the inner and outer borders of the vessel wall; the software was also less observer-dependent than manual adjustments of window and level [3, 8].
Noncalcified plaque may be more unstable and prone to rupture than calcified plaque. Recent studies have shown that CT can identify noncalcified plaques in the coronary arteries in vivo [9–11]; moreover, CT features of noncalcified and calcified plaques correlate well with histopathologic stages of atherosclerosis . Research by Ehara et al.  suggested that spotty distribution of calcium is of greater predictive value for acute coronary syndromes than the overall quantity. Twenty-five of 27 (92.6%) calcified plaques were found in our study, which missed two plaques containing small calcifications. Such high sensitivity for the detection of calcium with 64-MDCT is helpful for risk stratification in patients with coronary atherosclerotic disease.
CT density of plaque correlates with echogenicity of IVUS, and CT attenuation reflects major plaque composition. 64-MDCT can assess the thickness of the athero-sclerotic wall directly and can identify a large lipid core, superficial calcified nodules, and outward remodeling as morphology characteristics of vulnerable plaque [13–16]. Granted, improving CT detection of plaque with automated software is progress toward understanding plaque composition and its natural history. However, it is many steps away from identifying the high-risk “vulnerable” plaque.
Lipid-rich, fibrous, and calcified plaques displayed different CT densities and can be reliably differentiated. In our results, the mean CT density values for soft, fibrous, and calcified plaques were 72 ± 36 H, 90 ± 27 H, and 772 ± 251 H, respectively. Similar investigations showed these values to be 14 ± 26 H, 91 ± 21 H, and 419 ± 194 H, respectively , and 49 ± 22 H, 91 ± 22 H, and 391 ± 156 H . However, it remains difficult to determine the definite composition of individual noncalcified plaques, even if there is a significant difference in CT Hounsfield units among components. For example, hypodense areas in plaque could represent various tissue types, including lipid pool, less dense fibrous tissue, neovascularized areas, and thrombus. Moreover, in our study, CT found some noncalcified plaques with homogeneous echogenicity in IVUS that showed inhomogeneous density on CT, detected lipid pools (hypodense spots) in three of five (60%) segments, and could not discriminate the fibrous cap. Therefore, the accuracy for identifying noncalcified plaque components such as lipid-rich (soft) or fibrotic components remains low [17, 18].
The mean CT density values measured in our study were higher than those of other studies [2, 8]. The reason may be that the concentration of contrast material we used (370 mg I/mL) was higher than that used in other studies (300 or 350 mg I/mL). With 370 mg I/mL, we obtained a mean contrast enhancement in the coronary lumen of normal proximal segments of 398 ± 74 H, higher than the 343 ± 55 H measured by Achenbach et al.  using 350 mg I/mL contrast medium. It was assumed that plaque density is partly influenced by contrast enhancement in the coronary lumen; the CT densities of smaller and soft plaques were often elevated by greater enhancement in the lumen.
A significant trend was seen to underestimate luminal CSA and overestimate plaque burden. However, this study was designed to determine the accuracy of 64-MDCT without excluding significant stenosis. Significant stenosis causes decreased contrast enhancement of the lumen in corresponding segments. If we analyze proximal coronary arteries without significant stenosis, it is possible that better correlation between IVUS and CT would be obtained.
Sure Plaque software has the following technical limitations. First, the maximum number of images on which Sure Plaque software can be performed is only 512. Second, the z-direction size is less than the x, y-direction size. There are no standard recommendations for absolute density results of Sure Plaque tools. It is difficult to further differentiate or classify the noncalcified plaque because of the overlap of CT values of different composition in noncalcified plaque. Several limitations are also inherent in contrast-enhanced 64-MDCT itself, including restrictions of its spatial and temporal resolution [2, 19, 20]. Partial volume effects caused by large coronary calcifications and lumen contrast enhancement are frequently considered to be false-positive and false-negative results. Furthermore, contrast enhancement of the lumen may overlap, and thus conceal, superficial and smaller areas of calcium with 64-MDCT.
In summary, 64-MDCT is a useful noninvasive method and has good ability to detect, differentiate, and quantify coronary athero-sclerotic plaques. It may be used to improve risk stratification in patients with suspected or known coronary atherosclerotic disease and to monitor progression or regression of coronary atherosclerosis. However, the reliable differentiation of the composition of individual noncalcified plaques is still limited.
We are indebted to the medical and technical staff members of the CT and ultrasound departments for their invaluable contribution, and in particular to Yike Zhao, Hong Jiang, Haixia Yang, and Yanhong Wang for their expert technical assistance.
Supported in part by the Wards of Cardiology in An Zhen Hospital, Capital Medical University.
Mintz GS, Nissen SE, Anderson WD, et al. American College of Cardiology clinical expert consensus document on standards for acquisition, measurement and reporting of intravascular ultrasound studies (IVUS): a report of the American College of Cardiology Task Force on Clinical Expert Consensus Documents. J Am Coll Cardiol 2001; 37:1478 –1492
Leber AW, Knez A, Becker A, et al. Accuracy of multidetector spiral computed tomography in identifying and differentiating the composition of coronary atherosclerotic plaques: a comparative study with intracoronary ultrasound. J Am Coll Cardiol 2004; 43:1241 –1247
Leber AW, Becker A, Knez A, et al. Accuracy of 64-slice computed tomography to classify and quantify plaque volumes in the proximal coronary system: comparative study using intravascular ultrasound. J Am Coll Cardiol 2006; 47:672 –677
Leber AW, Knez A, Ziegler F, et al. Quantification of obstructive and nonobstructive coronary lesions by 64-slice computed tomography: a comparative study with quantitative coronary angiography and intravascular ultrasound. J Am Coll Cardiol 2005; 46:147–154
Schmermund A, Baumgart D, Adamzik M, et al. Comparison of electron-beam computed tomography and intracoronary ultrasound in detecting calcified and noncalcified plaques in patients with acute coronary syndromes and no or minimal to moderate angiographic coronary artery disease. Am J Cardiol 1998; 81:141–146
Budoff MJ, Cohen MC, Garcia MJ, et al. ACCF/AHA clinical competence statement on cardiac imaging with computed tomography and magnetic resonance: a report of the American College of Cardiology Foundation/American Heart Association/American College of Physicians Task Force on Clinical Competence and Training. J Am Coll Cardiol 2005; 46:383–402
Nikolaou K, Sagmeister S, Knez A, et al. Multidetector-row computed tomography of the coronary arteries: predictive value and quantitative assessment of non-calcified vessel-wall changes. Eur Radiol 2003; 13:2505 –2512
Schroeder S, Kuettner A, Kopp AF, et al. Noninvasive evaluation of the prevalence of noncalcified atherosclerotic plaques by multi-slice detector computed tomography: results of a pilot study. Int J Cardiol 2003; 92:151 –155
Ehara S, Kobayashi Y, Yoshiyama M, et al. Spotty calcification typifies the culprit plaque in patients with acute myocardial infarction: an intravascular ultrasound study. Circulation 2004; 110:3424 –3429
Nieman K, van der Lugt A, Pattynama PM, de Feyter PJ. Noninvasive visualization of athero-sclerotic plaque with electron beam and multislice spiral computed tomography. J Interv Cardiol 2003; 16:123 –128
Schroeder S, Kuettner A, Wojak T, et al. Noninvasive evaluation of atherosclerosis with contrast enhanced 16 slice spiral computed tomography: results of ex vivo investigations. Heart 2004; 90:1471 –1475
Nikolaou K, Becker CR, Muders M, et al. Multi-detector-row computed tomography and magnetic resonance imaging of atherosclerotic lesions in human ex vivo coronary arteries. Atherosclerosis 2004; 174:243 –252