| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, St. Vincent's University Hospital, Elm Park, Dublin
4, Ireland.
2 Department of Radiology, Cardiac MRI-PET-CT Program, Massachusetts General
Hospital and Harvard Medical School, Boston, MA.
3 Division of Cardiology, Massachusetts General Hospital and Harvard Medical
School, Boston, MA.
4 Department of Medicine 2, University Hospital Erlangen, Erlangen,
Germany.
Received October 17, 2007;
accepted after revision February 20, 2008.
Address correspondence to J. D. Dodd
(j.dodd{at}st-vincents.ie).
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. Twenty-nine consecutive patients (23 men and
six women; mean age, 62 ± 10 years) presenting with acute coronary
syndrome (ACS) had nonculprit coronary lesions of 
30% stenosis quantified
on quantitative coronary angiography (QCA). Five 64-MDCT postprocessing
techniques (maximum intensity projection [MIP], multiplanar reformat [MPR],
cross-sectional area [CSA], and diameter and area derived from
semiquantitative coronary software) were used to grade lesions. Two separate
groups of two independent readers analyzed QCA and cardiac CT images using a
17-segment model. Coronary angiography was the reference standard.
RESULTS. Nonculprit lesions were identified in 46 analyzable
coronary segments. Subgrouping lesions on the basis of reference vessel
diameter resulted in strong correlations for quantifying nonculprit lesions in
vessels > 3 mm (R = 0.78–0.91, p < 0.01) but
poor correlations for nonculprit lesions in vessels 
3 mm (R =
0.1–0.07). Subgrouping lesions on the basis of plaque type resulted in
poor correlations for calcified plaques (R = 0.01–0.30) but
moderate to strong correlations for mixed (R = 0.58–0.75,
p < 0.01) and noncalcified (R = 0.44–0.61,
p < 0.01) plaques. The best overall correlation among all CT
techniques with QCA was CSA (R = 0.56, p < 0.01).
Interobserver agreement (kappa values) for MPR, MIP, coronary software
diameter and area were 0.6, 0.7, 0.62, and 0.57, respectively.
CONCLUSION. In patients presenting with ACS, 64-MDCT provided an
accurate grade of stenosis for nonculprit coronary lesions in proximal
coronary segments. Calcified plaques and lesions in coronary segments 3
mm diameter remained difficult to accurately quantify.
Keywords: coronary angiography • coronary stenosis • CT • myocardial infarction • nonculprit coronary lesion
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Current studies have shown that the sensitivity and specificity of cardiac 64-MDCT for the detection of significant stenosis in culprit vessels approach 93–95% and 85–90%, respectively [1, 13–16]. Quantification studies have also shown that cardiac CT can quantify the degree of luminal stenosis in culprit lesions with reasonable accuracy [3, 6–8]. However, these "quantification of stenosis" analyses are proof of concept studies, in which selected coronary segments with known culprit lesions are evaluated. Whether cardiac CT can provide a correct grade of stenosis in nonculprit lesions is currently unknown. The aim of this study was to evaluate cardiac 64-MDCT for assessing the grade of stenosis in nonculprit lesions using five of the most commonly used CT postprocessing imaging techniques, with invasive coronary angiography as the reference standard.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Selective coronary angiography was performed using standard techniques and a transfemoral approach. Images were evaluated for the presence of nonculprit lesions, and then quantitatively assessed with quantitative coronary angiography (QCA) software (QCAPlus, Sanders Data Sys tems) by two experienced interventional cardiologists independently. Readers were aware of the presence of a treated culprit lesion. Culprit lesions or stents were not the primary focus of our study and were not included in the analysis. The culprit lesion was defined on the basis of the association of the angiographic lesion appearance with ECG changes or myocardial ischemia determined by stress testing. Readers were unaware of the CT results.
Cardiac CT was acquired on a 64-MDCT scanner (Somatom Sensation 64, Siemens Medical Solutions). Images were acquired with a collimation of 0.6 mm, slice overlap of 0.4 mm, gantry rotation of 330 ms, 120 kVp and 850–950 mAs depending on patient body habitus, and a half-scan algorithm resulting in a temporal resolution of 165 milliseconds. Images were reconstructed using retrospective ECG gating. Twenty-two patients required between 5 and 25 mg of metoprolol administered IV to obtain a heart rate of < 65 beats per minute (bpm) before scanning. If the heart rate remained above 65 bpm, multisegment reconstruction was used. Immediately after the administration of sublingual glyceryl nitrate for coronary vasodilation, 70 mL of iodinated contrast medium ([iopamidol] Isovue 370, Bracco Diagnostics) was injected at a rate of 5 mL/s followed by a 20 mL saline bolus chaser injected at 5 mL/s. Images were reconstructed with a 512 x 512 matrix and a smooth kernel, with a 0.75-mm slice thickness and 0.5-mm slice overlap. Reconstructions were individually optimized to minimize coronary motion artifact and then transferred to a work station (Leonardo, Siemens Medical Solutions) for further analysis.
Images were independently read by two experienced cardiac CT readers.
Discrepant results were determined by a third independent reader. Images were
anonymized and read in random order. Readers were blinded to the grade of
nonculprit lesion on QCA but were aware of nonculprit lesion location.
Coronary segments were analyzed using the nomenclature of the American Heart
Association standard classification system
[17]. Coronary segments with
coronary stenosis 30% identified on QCA were included in the analysis.
All segments with calcified plaque were included except those in which no
lumen could be identified on CT despite adjustments in window width and
center.
Five postprocessing techniques were used to evaluate CT images: first, cross-sectional 1-mm slice thickness multiplanar reformat (MPR) manually measured diameters created perpendicular (in two planes) to the median centerline of the coronary segment; second, 5-mm slice thick ness maximum intensity projection (MIP) manually measured diameters parallel to the long axis of the coronary segment; third, coronary segment luminal cross-sectional areas (CSA) manually traced by a single reader from MPRs rendered per pendicular to the median centerline of the coronary segment [6, 18, 19]; semiquantitative coronary artery software (Circulation, Siemens Medical Solutions) was used to evaluate lesions by rendering a curved multiplanar reformat of each coronary artery using a region-growing algorithm through the lesion; the software requires the reader to choose proximal, distal, and maximal coronary stenosis points from which it derives; fourth, a cross-sectional diameter (coronary software-diameter) and; fifth, the area (coronary software-area) of each point, and calculates a grade of stenosis [20]. Failure of the software to completely detect the entire coronary tree was supplemented by manual augmentation. For all postprocessing methods, window width and center were adjusted to optimize coronary lumen visualization.
All CT techniques were compared with invasive QCA as the reference
standard. Reference vessel points were taken in normal coronary segments as
close to the coronary lesion as possible, as described in previous studies
[6–8].
Lesions were subgrouped on the basis of the grade of stenosis (divided into
30–50%, 51–70%, and > 70% stenosis), size of reference vessel
(divided into 3 mm and > 3 mm), and plaque type (divided into
calcified [but visible coronary lumen], mixed, or noncalcified). Plaque type
was categorized on the basis of visual CT assessment. Grade of stenosis was
calculated by subtracting the mean of the reference segment diameters
(proximal and distal) from the maximal stenosis diameter and then dividing the
difference by the mean reference segment diameter
[6]. If stenosis occurred at
the ostium of a coronary artery, the distal reference vessel point was used.
Similarly, if stenosis occurred at a vessel bifurcation point, the proximal
reference vessel point was used.
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
30% stenosis by QCA. Of these, 16 were
excluded on CT because of extensive calcified plaque that completely
obliterated the lumen despite adjustments in window width and center, and
three were excluded because of severe motion artifacts, leaving 46 nonculprit
lesions in the final analysis. Eight nonculprit lesions were detected in
segment 3; seven in segments 1, 2, and 7; five in segment 11; four in segment
13; two in segment 6; two in segment 12; and one in segments 8, 9, 15, and 16.
On QCA, 26 lesions were causing 30–50% stenosis, 14 were causing
51–70% stenosis, and six were causing > 70%. Thirty-four lesions were
in coronary segments < 3 mm, whereas 12 were in segments 3 mm.
Twenty-six lesions were noncalcified, 10 were mixed, and 10 were
calcified.
|
The correlation between QCA and all five CT postprocessing methods for mean reference luminal diameter and mean luminal stenosis diameter was R = 0.64–0.70 and 0.57–0.69, respectively, p < 0.0 (Table 2). For quantifying the grade of stenosis for all nonculprit lesions, the correlation between QCA and CSA was R = 0.5, p < 0.01 (Figs. 1, 2A, 2B, 2C, 2D, 2E, 2F, 2G, 2H, 2I, 2J, and 2K) but poor for the remaining CT techniques.
|
|
|
|
|
|
|
|
|
|
|
|
When nonculprit lesions were subgrouped on the basis of grade of stenosis, correlation between QCA and CT was poor for lesions causing 30–50% and 51–70% except for CSA, which showed a moderate correlation in grading lesions causing 51–70% luminal narrowing (R = 0.59, p < 0.05). Correlation was very strong (R = 1.0, p < 0.05) for nonculprit lesions causing > 70% luminal stenosis (Table 3).
|
When nonculprit lesions were subgrouped on the basis of size of the mean
reference vessel diameter, divided into 3 mm (32 lesions) and > 3 mm
(12 lesions), correlation between QCA and CT was poor for nonculprit lesions
in vessels 3 mm (R = 0.34–0.1) but strong for nonculprit
lesions in vessels > 3 mm (R = 0.78–0.91, p <
0.01) (Table 4).
|
When nonculprit lesions were subgrouped on the basis of plaque type, the correlation between QCA and CT was poor for calcified plaques (R = 0.01–0.20) (Table 5). For noncalcified plaques, the correlation was moderate (R = 0.43–0.60, p < 0.05); for mixed plaques the correlation was moderate to strong (R = 0.58–0.75, p < 0.05).
|
Interobserver agreement for QCA was excellent (R = 0.9, p
< 0.0001). Interobserver agreement for MPR, MIP, coronary software
diameter, and coronary software area were 0.6, 0.7, 0.62, and 0.57,
respectively, p < 0.01. When nonculprit lesions were subgrouped
according to grade of stenosis, the kappa value for interobserver agreement
for MPR, MIP, coronary software diameter, and coronary software area varied
considerably (0.49, 0.35, 0.28, and 0.22, respectively). When nonculprit
lesions were subgrouped according to plaque type, interobserver agreement was
good (
= 0.65, p < 0.001).
Bland-Altman analysis showed MIP and MPR to have minimal bias (0.2%
± 8.7% [2SD] and 0.8% ± 26%, respectively); CSA, a systematic
overestimation bias (7.6% ± 34%); coronary software diameter, a
systematic underestimation bias (–7.6% ± 31%); and coronary
software area, a systematic overestimation bias (12.2% ± 35%). When
nonculprit lesions were subgrouped on the basis of mean reference vessel
diameter, for vessels 3 mm, MIP showed a systematic bias toward the
overestimation of grade of stenosis (3.2% ± 8%) but no systematic bias
for grade of stenosis in vessels > 3 mm.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
3 mm diameter) remained difficult to accurately quantify. The original concept in the pathogenesis of angina or AMI was the progressive growth of an atherosclerotic plaque until it caused sufficient lumen narrowing to significantly decrease blood flow [21, 22]. It is now thought that many culprit lesions do not cause hemodynamically significant luminal narrowing until vulnerable plaque ruptures and a clot forms, causing vessel occlusion [23]. It has also become clear that many patients will have more than one plaque and that the atherosclerotic process is a diffuse one, affecting nonculprit as well as culprit vessels [12]. This has clinical relevance because those patients with nonculprit lesions have a higher incidence of subsequent recurrent cardiac events and a poorer prognosis [9, 10].
When all nonculprit lesions were analyzed together, with no subgroup analysis, our results indicated poor to moderate cardiac CT performance for accurately quantifying the grade of stenosis. Such results are weaker than in previous studies evaluating CT quantification of purely culprit lesions. However, purely nonculprit lesions characteristically cause milder grades of stenosis [9]. We also included all calcified and mixed plaques unless the coronary lumen was completely obliterated by calcification despite adjustments to window center and width, and we therefore had several lesions that were inherently difficult to grade. Finally, our patient population had a high prevalence of multiple risk factors for coronary artery disease (mean body mass index = 26.6 [kg/m2], 69% of patients were diabetic) that inevitably resulted in a prominent number of calcified plaques and increased image noise [24]. Thus the analysis of nonculprit lesions, demographics of our patient group, and inclusion where possible of calcified plaques resulted in a more challenging imaging cohort [6–8].
Although CT provided strong correlations for nonculprit lesions in proximal
coronary segments, lesions measuring 3 mm in more distal segments yielded
poorer correlations with QCA. However, Schoenhagan et al.
[10], in 105 patients with
culprit lesions, found a high prevalence of additional atherosclerotic plaques
with mild to moderate stenosis proximal to the culprit lesion. In patients
with AMI, these nonculprit plaques had a fivefold higher frequency of
ulceration, suggesting multifocal plaque vulnerability. In a study by
Goldstein et al. [9], patients
with multivessel disease and multiple complex plaques had a higher incidence
of recurrent ischemia requiring repeat catheterization, recurrent ACS,
subsequent revascularization, and coronary artery bypass surgery compared with
those with single lesions. Thus it is still useful to quantify proximal
nonculprit lesions because these are the most commonly associated lesions with
adverse outcomes after fibrinolytic therapy.
Our study corroborates several findings from previous work. That CSA gave
the strongest correlations with QCA in nonculprit lesions is similar to the
findings of Moselewski et al.
[18], who assessed CSA and
intravascular sonography in 100 coronary artery segments. Significant
correlations were found between CSA and mean luminal area (R = 0.92,
p < 0.001) and plaque area (R = 0.55, p <
0.001). We found MIP yielded the smallest interobserver variability, which was
in keeping with the previous findings of Cury et al.
[6], who found excellent
interobserver agreement for grading 69 culprit lesions with 16-MDCT ( =
0.7, p < 0.001). Poor correlations were found between cardiac CT
and QCA for calcified plaques and lesions in distal coronary segments, similar
to Leber et al. [19], who
found in 73 patients with culprit lesions that incorrectly classified lesions
were all in distal segments or were extensively calcified.
The ability of coronary artery software to rapidly produce curved MPR images and quantify coronary stenosis was very compelling, and the time advantages of such computer-aided stenosis quantification have been emphasized by others. Busch et al. [20] evaluated 57 coronary lesions by visual assessment (MPR and MIP) and automated coronary software. Their results suggested that the automated software produced the most robust correlation with QCA (R = 0.82, p < 0.00001). However, their patient population is likely to have been different from ours, no patient having undergone coronary stent insertion before CT acquisition, and there were probably differing proportions of lesion locations and plaque types.
We found several technical limitations using the semiautomated software. Most important, it showed an inability to correctly grade several partially calcified lesions. CT image gray-scale threshold adjustments to improve semiautomated detection of the coronary lumen did not avoid this problem. A further limitation of the software was its inability to grade ostial lesions. The software mandated proximal and distal reference points to calculate a grade of stenosis; because the only possible reference diameter for an ostial lesion is distal, the software was unable to provide a grade of stenosis. A related limitation was its inability to precisely grade lesions at bifurcation points, although to be fair this was also a limitation for QCA and CT. Coronary software vessel diameters tended to systematically underestimate grades of stenosis, whereas coronary software areas systematically overestimated them. Finally, in several cases considerable time was required to manually augment the coronary tree algorithm, particularly for distal vessel depiction.
Although the coronary software provided good interobserver agreement for overall coronary stenosis, when lesions were subgrouped on the basis of grade of stenosis divided into 30–50%, 51–70%, and > 70%, very poor interobserver agreement resulted for both coronary software diameter and coronary software area. We identified two causes for this. First, there was considerable user subjectivity inherent in manually augmenting vessel lumen outlines using the software tools. Second, for many lesions, the grade of stenosis calculated by the software appeared to be rounded to the nearest multiple of five. Thus, if one observer graded a lesion as 52% and the second observer graded the lesion as 54%, the software rounded the former result to 50% and the latter result to 55%, resulting in the lesion being categorized differently by the two observers. We verified this finding by manually calculating grades of stenosis using the diameters and areas obtained by the software and comparing them with the grade of stenosis calculated by the software. Thus, of all the CT techniques, the semiautomated software showed the overall weakest accuracy and performance and cannot be recommended based on the data of this study at this time. Manual reconstruction of luminal cross-sectional areas provided the strongest correlation with invasive QCA. Future coronary software upgrades may overcome these issues.
Some limitations of our study should be noted. The overall number of nonculprit lesions in our study was small. A considerable number of heavily calcified plaques were unevaluable in our study, and such lesions remain a significant limitation to nonculprit lesion quantification with current CT technology. It would have been interesting to correlate our CT findings with intravascular sonography to assess the ability of cardiac CT to characterize nonculprit plaques in this patient population. Recent work suggests cardiac CT can identify differences in lesion morphology and plaque composition between culprit lesions in ACS and stable lesions in ACS or angina when compared with invasive angiography [25]. We stress that the primary aim of this study was not nonculprit lesion detection but quantification, the former already having been evaluated by others [26].
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  R. J. Gibbons, P. A. Araoz, and E. E. Williamson The Year in Cardiac Imaging J. Am. Coll. Cardiol., February 2, 2010; 55(5): 483 - 495. [Full Text] [PDF]
|
|
|
|
 |
|
 A. J. Duerinckx The Nonculprit Coronary Lesion as Seen by Coronary CT Angiography: What Should We Be Looking For? Am. J. Roentgenol., August 1, 2008; 191(2): 439 - 440. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
