| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Faculty of Medicine, School of Medicine, National Yang Ming University,
Taipei, Taiwan, Republic of China.
2 Section of Thoracic and Circulation Imaging, Department of Radiology,
Kaohsiung Veterans General Hospital, 386 Ta-Chung 1st Rd., Kaohsiung, Taiwan
813, Republic of China.
3 Department of Psychiatry, Kaohsiung Medical University, Kaohsiung, Taiwan,
Republic of China.
4 Department of Mechanical and Electromechanical Engineering, National Sun-Yat
Sen University, Kaohsiung, Tainan, Taiwan, Republic of China.
5 Department of Radiology, National Cheng Kung University, Tainan, Taiwan,
Republic of China.
Received August 4, 2007;
accepted after revision November 12, 2007.
Supported by Kaohsiung Veterans General Hospital's Research Program of
Taiwan (VGHKS-96-60) and the National Science Council of Taiwan
(NSC95-2314-B-075B-010-MY2).
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. The subjects were 513 patients consecutively registered for health screening and undergoing both low-dose ungated (120 kVp, 20 mAs) and regular-dose ECG-gated MDCT (120 kVp, 150 mAs, retrospective ECG gating). The first 30 cases were used for protocol optimization and a training session. Agatston score on regular-dose ECG-gated and low-dose ungated MDCT in the other 483 cases (320 men; mean age, 62.2 ± 13.2 [SD] years) was calculated by two observers in a blinded manner. Interobserver and intertechnique scoring variability and concordance were calculated.
RESULTS. The mean of interobserver scoring variability for
regular-dose ECG-gated MDCT was 3.6% and for low-dose ungated MDCT was 9.6%.
Regular-dose ECG-gated MDCT depicted CAC in 221 (46%) of the subjects. With
low-dose ungated MDCT, observers 1 and 2, respectively, had five and seven
false-positive and five and four false-negative predictions. All the
miscategorized scores were 12 or less. The negative predictive values of CAC
on low-dose ungated MDCT were 98% and 99% for observers 1 and 2, respectively.
For patients with CAC, the mean intertechnique scoring variability was
40–43%. For all 483 subjects, the intertechnique concordance of the four
major score ranks (0, 1–100, 101–400, > 400) was high (
= 0.89 for the two observers).
CONCLUSION. Low-dose ungated MDCT with an optimized protocol is reliable for prediction of the presence of CAC and categorization of the four major Agatston score ranks. This technique may be useful for coronary artery disease risk stratification of persons undergoing low-dose ungated MDCT for lung cancer screening.
Keywords: coronary artery calcification • CT • lung cancer
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
For years, electron beam CT has been considered the standard for CAC scoring. However, because of its superior image quality ECG-gated MDCT has had better measurement results with regard to the accuracy and reproducibility of calcium scores [5–7]. Although ECG-gating technique appears to be advantageous for CAC measurement [3], results of experiments with a working heart phantom [8] have shown ungated helical CT reliable in the detection of CAC. The technique also has been used since 1995 to measure CAC in several clinical scenarios [9–11]. Therefore, ungated MDCT performed for lung cancer screening may also be useful in assessment of CAC.
The clinical importance of CAC detected with CT for lung cancer screening has been reported. With a routine protocol of single-detector CT [12], 4-MDCT, or 8-MDCT [13], results of visual assessment of CAC were found to be predictive of cardiovascular death [12] and to contribute to risk stratification of coronary artery disease (CAD) [13]. Nonetheless, the key issues regarding the sensitivity and reliability of low-dose ungated MDCT for CAC scoring have not been systematically investigated, to our knowledge.
We hypothesized that with an optimized protocol, low-dose ungated MDCT would be sensitive in the detection of CAC and that CAC scores would have good concordance with those obtained with dedicated cardiac CT. In this study, we aimed to test our hypothesis by performing a head-to-head comparison of CAC scoring on low-dose ungated MDCT with the scoring on regular-dose ECG-gated MDCT of subjects undergoing these two types of CT for lung cancer screening and CAC scoring in a health screening examination.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
MDCT Protocols
All subjects underwent two CT examinations with a 16-MDCT unit (Somatom
Sensation 16, Siemens Medical Solutions) in one session without change of body
position. For low-dose un gated MDCT, the scanning parameters were as follows:
no ECG gating; peak voltage, 120 kVp; tube current–exposure product, 30
mAs with online anthropomorphic tube current modulation (CareDose 4D, Siemens
Medical Solutions); rotation time, 0.5 second; detector collimation, 0.75 mm
x 16; pitch, 1.5; scan range, whole lung coverage. The reconstruction
protocol for CAC scoring is described later (Optimization of Reconstruction
Protocol).
For regular-dose ECG-gated MDCT, the scanning protocol was as follows: retrospective ECG gating with pulse modulation; peak voltage, 120 kVp; tube current–exposure product, 150 mAs with online anthropomorphic tube current modulation (CareDose 4D); rotation time, 0.42 second; collimation, 1.5 mm; pitch, 0.28; scan range, carina to cardiac apex; raw data reconstructed at early to middiastolic phase [3]; slice thickness, 3.0 mm; slice increment, 1.5 mm; field of view, 250 mm2; medium soft-tissue algorithm.
The radiation dose was estimated by multiplying the dose–length product by 0.017, the normalized value of the effective dose per dose–length product of the thorax [14]. The effective dose of low-dose ungated MDCT was estimated with length coverage from lung apex to diaphragm. For regular-dose ECG-gated MDCT, the length of coverage was carina to cardiac apex.
Optimization of Reconstruction Protocol and Training in CAC Score Reading
We used the first 30 CT data sets for protocol optimization and a training
session on CAC scoring. These 30 data sets were excluded from the statistical
analysis.
Optimization of reconstruction protocol—Because a lower radiation dose was expected to cause a high noise level that might have affected the results of CAC scoring [15, 16], we attempted to modify low-dose ungated MDCT with an optimal noise level. For this purpose, we compared low-dose ungated MDCT in two reconstruction algorithms: a medium soft-tissue algorithm (B35) used in standard CAC scoring and a smooth soft-tissue algorithm (B20). The mean ± SD CT attenuation values in regions of interest set in the aortic root at the level of the left coronary artery were measured. The SD of the region of interest represented the noise level of the images [7].
We found the SD of attenuation of the region of interest was significantly higher with the medium soft-tissue algorithm (26.9 H) than with the smooth soft-tissue algorithm (21.3 H) (p < 0.001), but no difference in mean attenuation (40.3 vs 40.7 H, p = 0.66) was found. CAC scoring on low-dose ungated MDCT with both algorithms had high agreement with the scoring on regular-dose ECG-gated MDCT (medium soft-tissue algorithm, intraclass R = 0.90; smooth soft-tissue algorithm, intraclass R = 0.91). Both observers preferred the smooth soft-tissue algorithm for scoring on low-dose ungated MDCT. Therefore, we chose the smooth soft-tissue algorithm for the optimized protocol.
Because it has been well documented [6, 17] that the use of slice thickness and overlapping reconstruction has great influence on CAC scoring, we used the same reconstruction protocol as for regular-dose ECG-gated MDCT except for the smooth soft-tissue algorithm: carina to cardiac apex; field of view, 250 mm; slice thickness, 3 mm; slice increment, 1.5 mm.
Training session for CAC scoring—To help the observers become familiar with the appearance of CAC on low-dose ungated MDCT, we arranged a training session for the observers to perform CAC scoring of regular-dose ECG-gated and low-dose ungated MDCT scans in a side-by-side manner with the assessment method that follows. CAC scoring on regular-dose ECG-gated MDCT was used for adjudication interpretation to improve low-dose ungated MDCT scoring.
CAC Assessment
CAC scoring was performed with a commercially available program (Heartview,
Leonardo, Siemens Medical Solutions). Two observers performed the
measurements. Observer 1 was a thoracic radiologist with 3 years of experience
in CAC scoring, and observer 2 was a chief cardiothoracic technologist with 6
years of experience in CAC scoring. The DICOM patient identification headers
and labels were masked. The two sets of CT scans were given to the observers
in a random blinded manner.
The threshold for CAC scoring was set at a CT attenuation value of 130 H indicating potential calcification. A region of interest was encircled manually in each coronary artery, and a computer-driven measurement of the lesion area was automatically highlighted as the individual volume of a lesion. An Agatston score was obtained by multiplying pixel area by density score (1, 130–199 H; 2, 200–299 H; 3, 300–399 H; 4, > 399 H) and summing the lesion scores [18]. Separate scores were calculated for the left main coronary artery, right coronary artery, left circumflex coronary artery, and left anterior descending coronary artery [16]. The CAC scores were rounded and classified as binary, either absence (Agatston score, 0) or presence (Agatston score, > 0). CAC scores were further categorized into four score ranks (0, 1–100, 101–400, and > 400), as for CAD risk stratification [4, 19].
Statistical Analysis
The variables were expressed as mean ± SD or as median (minimum to
maximum). Intraclass correlation coefficient, kappa value, and variability
were used to express interobserver and intertechnique (regular-dose ECG-gated
MDCT vs low-dose ungated MDCT) agreement of CAC scores. Variabilities between
two CAC scores were calculated according to the following equation: (2 x
absolute [score 1 – score 2] / [score 1 + score 2]) x 100%. The
binary value of CAC (presence or absence) on low-dose ungated MDCT was used to
calculate positive predictive and negative predictive values with the binary
value of regular-dose ECG-gated MDCT scoring as reference. Concordance of CAC
Agatston score ranking (0, 1–100, 101–400, > 400) on
regular-dose ECG-gated and low-dose ungated MDCT was expressed with kappa
value. The parameters affecting image quality (including heart rate, body
weight, and effective tube current–exposure time product) and noise
level between false-positive versus true-negative groups and between
false-negative versus true-positive groups were compared by use of the
Mann-Whitney U test. Values of p < 0.05, two-sided, were
considered statistically significant. All analyses were performed with the
SPSS program (version 11.0 for Windows).
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The results of CAC scoring on regular-dose ECG-gated and low-dose ungated MDCT are listed in Tables 1 and 2. The interobserver agreement of CAC scores on regular-dose ECG-gated MDCT was intraclass R = 0.99 and mean of variability, 3.6%. Agreement on low-dose ungated MDCT was intraclass R = 0.99 and variability, 9.6%. For all 483 subjects, agreement between CAC scores on low-dose ungated and regular-dose ECG-gated MDCT by observer 1 was intraclass R = 0.96; mean of variability, 23%; median, 0%. For observer 2 the values were intraclass R = 0.95; mean of variability, 26%; median, 0%.
|
|
Table 3 shows the binary
results of CAC scoring on low-dose ungated and regular-dose ECG-gated MDCT.
With the results of regular-dose ECG-gated MDCT as the reference, low-dose
ungated MDCT had 216 true-positive (Figs.
1A,
1B,
1C, and
1D), five false-negative (Figs.
2A and
2B), five false-positive (Figs.
3A and
3B), and 257 true-negative
predictions by observer 1 (= 0.95) and four false-negative and seven
false-positive predic tions by observer 2 ( = 0.95). Therefore,
low-dose ungated MDCT had positive predictive values of 98% and 97% and
negative predictive values of 98% and 99% for observers 1 and 2, respectively.
For subjects with images that showed CAC on either regular-dose ECG-gated or
low-dose ungated MDCT, the agree ment between the scores on low-dose ungated
versus regular-dose ECG-gated MDCT for observer 1 was intraclass R =
0.95; mean of variability, 40%; median, 27%. The values for observer 2 were
intraclass R = 0.95; mean of variability, 43%; median, 28%.
|
|
|
|
|
|
|
|
|
For the false-negative predictions on low-dose ungated MDCT, the median CAC score on regular-dose ECG-gated MDCT was 6.2 (range, 1–12). For the false-positive predictions on low-dose ungated MDCT, the median CAC score on low-dose ungated MDCT was 4.5 (range, 3–8). All miscategorized false CAC scores were 12 or less.
In the search for the potential parameters causing miscategorization, we found a significantly higher body weight (75.2 ± 3.1 kg) in the false-positive group than in the true-negative group (63.5 ± 12.7 kg) (p = 0.015) but no difference in effective tube current–exposure time product (14.7 ± 1.1 vs 15.8 ± 1.2 mAs, p = 0.07) or heart rate (65.3 ± 4.1 vs 61.7 ± 13.1 beats per minute, p = 0.11). Comparison of the false-negative and true-positive groups showed no significant differences in body weight (60.1 ± 14.5 vs 68.4 kg ± 12.9, p = 0.15), effective tube current–exposure time product (17.2 ± 15.6 vs 16.1 ± 2.84 mAs, p = 0.47), or heart rate (56.8 ± 8.1 vs 60.8 ± 10.1 beats per minute, p = 0.26). The noise level, represented by the SD of attenuation of the region of interest at the aortic root, was significantly higher in the false-positive group (27.2 ± 1.8) than in true-negative group (21.9 ± 5.0) (p = 0.031); there was no difference between the false-negative group and the true-positive group (p = 0.42).
Table 4 shows the
concordance of CAC scores on low-dose ungated and regular-dose ECG-gated MDCT
for observer 1 for the four score ranks used in risk stratification of CAD.
There was good concordance between the two score ranks ( = 0.89). The
results were similar for observer 2 ( = 0.89, data not shown).
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
To test the reliability of ungated helical CT for measuring CAC, Hopper et al. [8] used a working heart phantom with artifical coronary arteries as a reference standard for CAC quantification. They found ungated helical CT depicted CAC quantification better than did gated electron beam CT. Those authors attributed the greater sensitivity of helical CT, even ungated, to volumetric data consisting of thin-collimation images and incremental overlapping reconstruction. The results of this funda mental experiment provided support for our hypothesis, that is, ungated MDCT can be sensitive for detecting CAC.
Reliably determining the presence of CAC is important in the assessment of CAD. A meta-analysis [4] showed a summary relative risk ratio of 4.3 (95% CI, 3.5–5.2) for any measurable CAC compared with nondetectable CAC (p < 0.0001). Table 3 shows that low-dose ungated MDCT was reliable for predicting the presence of CAC in subjects without symptoms undergoing lung cancer screening. For false-negative prediction with low-dose ungated MDCT, the missed CAC scores were all 12 or less. Although the heart rates in this group were not different from those in the true-positive group, we assumed that cardiac motion was the cause of the false-negative predictions, which can happen randomly in subjects with low CAC scores.
For the false-positive predictions with low-dose ungated MDCT, the ghost CACs all were scored 8 or less. We postulated that these scores resulted from the higher noise level of low-dose ungated MDCT because, first, as shown in Figures 3A and 3B, the noise level of low-dose ungated MDCT was higher than that of regular-dose ECG-gated MDCT and, second, the noise level of low-dose ungated MDCT in the false-positive group was significantly higher than that in the true-positive group. We postulate that in the future, with a body weight–adapted dose reduction scanning protocol [7, 21], the image noise of low-dose ungated MDCT can be improved and the number of false-positive findings reduced, although at the expense of higher radiation dose.
For risk stratification of CAD, CAC scoring is most useful in terms of
incremental prognostic value for populations with an intermediate Framingham
risk score. The summary relative risk ratios were 2.1 (CAC score, 1–100)
and as high as 10 (CAC score, > 400) compared with the values among
patients without CAC (p < 0.0001)
[4]. Although the variability
among scores on low-dose ungated and regular-dose ECG-gated MDCT was as high
as 40–43% among subjects with CAC, we found that CAC in the four score
ranks was highly concordant between low-dose ungated and regular-dose
ECG-gated MDCT ( = 0.89 for both observers)
(Table 4). This finding
indicated that CAC scoring on low-dose ungated MDCT may not be highly accurate
but is reliable for CAD risk stratification.
In a cohort study starting in 1996 with a population of 6,120 Japanese subjects without symptoms, Itani et al. [12] found that visual assessment of CAC on ungated single-detector CT scans was predictive of future cardiovascular death. In a more recent CT screening for lung cancer with low-dose 4- or 8-MDCT among 4,250 subjects (with 90% white) without a history of CAD [13], it was concluded that visually assessed CAC grades can be derived and contribute to risk stratification of CAD. Our results give quantitative support to the results of the qualitative studies [12, 13] of CAC on CT for lung cancer screening.
The clinical importance of our method will be further enhanced when the
technique is integrated into a large public health project entailing the use
of low-dose ungated MDCT screening for lung cancer. With a high negative
predictive value for CAC, regular-dose ECG-gated MDCT is not needed in cases
in which no CAC is detected on low-dose ungated MDCT. The result would be a
54% reduction in performance of regular-dose ECG-gated MDCT. This reduction is
important in terms of examination cost, preparation, examination time, and
radiation dose to each patient (
3 mSv, Tables
1 and
2). It should be noted that the
impact of the method would be influenced by the prevalence of CAC in the
population. In a Japanese population (6,120 subjects; mean age, 61.4 ±
11.3 years) [12], the
prevalence of CAC was as low as 20%. Therefore 80% of subjects may not need
regular-dose ECG-gated MDCT if our method is used. The importance may be less
in a white population, in which a higher prevalence of CAC is expected
[4,
13].
In cases of a high CAC score ranking on low-dose ungated MDCT, further investigation, such as stress ECG testing or myocardial perfusion imaging, should be initiated immediately to search for subclinical CAD. In our sample, only 88 (18%) of the subjects had a CAC score greater than 400. Because of the high variability in CAC scores between low-dose ungated and regular-dose ECG-gated MDCT, subjects with middle CAC score ranks on low-dose ungated MDCT may be advised to undergo regular-dose ECG-gated MDCT to guide primary prevention strategies and to track CAC score [5–7].
This study had two limitations. First, CAC scoring on low-dose ungated MDCT scans may not be familiar to most CAC readers. We found, however, that with a training session, interobserver variability was not particularly high compared with that reported for regular cardiac CT [22, 23]. Second, it is unknown whether our findings can be reproduced on scanners made by other manufacturers. Future cross-system study is needed to validate whether our method is widely applicable.
CAC scoring on low-dose ungated MDCT with an optimized protocol is reliable for predicting the presence of CAC. The results are highly concordant with the score ranking on regular-dose ECG-gated MDCT. Cardiac CT is not needed by persons in whom CAC is not detected on low-dose ungated MDCT, but additional investigation of subclinical CAD may be triggered for persons with a high CAC score rank. The results of this study suggest that low-dose ungated MDCT is useful for CAD risk stratification as well as for lung cancer screening.
Acknowledgments
We thank the Statistical Analysis Laboratory, Department of Clinical
Research, Kaohsiung Medical University Chung-Ho Memorial Hospital.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
