Author + information
- Received December 28, 2006
- Revision received April 18, 2007
- Accepted April 24, 2007
- Published online August 21, 2007.
- Annick C. Weustink, MD⁎,†,
- Willem B. Meijboom, MD⁎,†,
- Nico R. Mollet, MD, PhD⁎,†,
- Masato Otsuka, MD⁎,
- Fransesca Pugliese, MD⁎,†,
- Carlos van Mieghem, MD⁎,†,
- Roberto Malago, MD†,
- Niels van Pelt, MD⁎,†,
- Marcel L. Dijkshoorn, BSc†,
- Filippo Cademartiri, MD, PhD⁎,†,
- Gabriel P. Krestin, MD, PhD† and
- Pim J. de Feyter, MD, PhD⁎,†,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Pim J. de Feyter Department of Cardiology and Radiology, Thoraxcenter, Room Ba 589, ’s Gravendijkwal 230, 3000 CA Rotterdam, the Netherlands.
Objectives Our objective was to prospectively evaluate the diagnostic performance of the high-speed dual-source computed tomography scanner (DSCT), with an increased temporal resolution (83 ms), for the detection of significant coronary lesions (≥50% lumen diameter reduction) in a clinically wide range of patients.
Background Cardiac motion artifacts may decrease coronary image quality with use of earlier computed tomography scanners that have a limited temporal resolution.
Methods We prospectively studied 100 symptomatic patients (79 men, 21 women, mean age 61 ± 11 years) with atypical (18%) or typical (55%) angina pectoris, or unstable coronary artery disease (27%) scheduled for conventional coronary angiography. Mean scan time was 8.58 ± 1.52 s. Mean heart rate was 68 ± 11 beats/min. Quantitative coronary angiography was used as the standard of reference. Irrespective of image quality or vessel size, all segments were included for analysis.
Results Invasive coronary angiography demonstrated no significant disease in 23%, single-vessel disease in 31%, and multivessel disease in 46% of patients; 1,489 coronary segments, containing 220 significant (14.8%) stenoses, were available for analysis. Sensitivity, specificity, and positive and negative predictive values of DSCT coronary angiography for the detection of significant lesions on a segment-by-segment analysis were 95% (95% confidence interval [CI] 90 to 97), 95% (95% CI 93 to 96), 75% (95% CI 69 to 80), 99% (95% CI 98 to 99), respectively, and on a patient-based analysis 99% (95% CI 92 to 100), 87% (95% CI 65 to 97), 96% (95% CI 89 to 99), and 95% (95% CI 74 to 100), respectively.
Conclusions Noninvasive DSCT coronary angiography is highly sensitive to detect and to reliably rule out the presence of a significant coronary stenosis in patients presenting with atypical or typical angina pectoris, or unstable coronary artery disease.
For almost 50 years, invasive coronary angiography has been the standard of reference for diagnosing coronary artery disease. However, noninvasive coronary imaging with computed tomography (CT) has rapidly emerged, and initial experience with 4-, 16-, and 64-slice CT coronary angiography has been reported (1–13). Despite technical advances in CT technology, a substantial number of coronary segments remain unevaluable due to presence of motion artifacts and a limited image resolution, which seriously hampered clinical implementation of CT coronary angiography (4,14).
A newly introduced dual-source computed tomography (DSCT) system, with an improved temporal resolution of 83 ms independent of patient’s heart rate, allows for scanning of the coronaries without the use of prescan beta-blockers. The pitch is adapted to the heart rate, and scan times are reduced at higher heart rates. Shorter scan times allow for reduction of radiation exposure to the patient. We now report the diagnostic performance of DSCT coronary angiography to detect or rule out significant coronary stenoses in the clinically relevant coronary tree in 100 patients with a wide spectrum of symptomatic coronary artery disease.
After an initial 3-week test period during which scan protocols were optimized, we subsequently included during a 10-week period 111 symptomatic patients with atypical angina, typical angina, and unstable coronary artery disease (unstable angina or non–ST-segment elevation myocardial infarction) scheduled for conventional coronary angiography (CCA). All CT examinations were performed before CCA. Only patients in sinus heart rhythm without previous history of percutaneous coronary intervention or bypass surgery were included. Excluded were 11 patients with known allergy to iodinated contrast material (n = 1), impaired renal function (serum creatinine >120 μmol/l) (n = 5), persistent arrhythmias (n = 3), or logistic inability to perform a CT scan before CCA (n = 2). Thus, the study population comprised 100 patients (79 men, 21 women, mean age 61 ± 10.9 years; range 28 to 87 years). The institutional review board approved the study, and all patients gave informed consent.
No oral or intravenous prescan beta-blockers were administered before the scan.
Scan protocol and image reconstruction
All patients were scanned using a DSCT (Somatom Definition, Siemens Medical Solutions, Forcheim, Germany). The system combines 2 arrays each consisting of an X-ray tube plus detector (64 slices) mounted on a single gantry with an angular offset of 90° and a gantry rotation time of 330 ms. Dual source computed tomography permits spiral CT scanning of the coronary arteries with an improved temporal resolution of 83 ms using single-segment reconstruction (15).
In DSCT, radiation exposure has been reduced by the application of an additional cardiac bowtie filter, a smaller field of vision of the second detector, and an increased pitch in higher heart rates. All patients underwent a nonenhanced CT scan for calcium scoring before DSCT coronary angiography. All patients received nitroglycerin (0.4 mg/dose) sublingually just before scanning.
Calcium scoring scan parameters were a tube current of 84 mAs/rot (maximum), and full X-ray tube current was given during 50% to 70% of the RR interval. A single dataset was reconstructed using electrocardiogram (ECG) gating with a slice thickness of 3 mm and increment of 1.5 mm using a medium convolution kernel (B35f) during 60% of the RR interval.
Dual source computed tomography scanner angiographic scan parameters were number of slices/rotation 32 × 2 with z-flying focal spot for each detector, individual detector width 0.6 mm, rotation time 330 ms, tube voltage 120 kV. Pitch values were adapted to heart rate after an estimation based on the last 10 heartbeats preceding the scan. Each tube provided 412 mAs/rot (maximum), and full X-ray tube current was given during 25% to 70% of the RR interval.
The volume of iodinated contrast material (Ultravist 370 mgl/ml, Schering AG, Berlin, Germany) was adapted to the scan time. A contrast bolus (60 to 90 mgl) was injected in an antecubital vein at a flow rate of 5.5 ml/s followed by a saline chaser of 40 ml at 5.5 ml/s. A bolus tracking technique was applied to synchronize the arrival of contrast in the coronary arteries and the start of the scan. All CT coronary angiography datasets were reconstructed with a slice thickness of 0.75 mm and increment of 0.4 mm using medium-to-smooth convolution kernel (B26f), resulting in a spatial resolution of 0.6 mm to 0.7 mm in-plane and 0.5 mm through-plane (15).
The reconstruction algorithm uses data from a single heart beat, obtained during a quarter X-ray tube rotation by 2 separate X-ray tubes, resulting in a temporal resolution of 83 ms. Images were reconstructed after a stepwise pattern depending on patient’s heart rate during scanning. Initially, a single dataset was reconstructed during the mid- to end-diastolic phase (350 ms before the next R-wave) in patients with low heart rates (<60 beats/min), during both the mid- to end-diastolic phase and end-systolic phase (275 ms after the next R-wave) in patients with intermediate heart rates (60 to 80 beats/min), and during the end-systolic phase in patients with high heart rates (>80 beats/min). Image quality was assessed on a per-segment level. In case of persistent coronary motion artifacts in patients with low and high heart rates, additional datasets were reconstructed in end-systolic and mid- to end-diastolic phase, respectively. If necessary, multiple datasets of a single patient were used separately in order to obtain optimal image quality of all available coronary segments.
The effective dose for DSCT coronary angiography was estimated based on Monte Carlo calculations (ImPACT, version 0.99x, St. George’s Hospital, Tooting, London, United Kingdom).
Quantitative coronary angiography (QCA)
One experienced cardiologist, unaware of the results of DSCT coronary angiography, identified all available coronary segments using a 17-segment modified American Heart Association classification (16). All segments, irrespective of size, were included for comparison with DSCT coronary angiography, except for segments distal to a total occlusion.
Segments were classified as normal (smooth parallel or tapering borders), as having nonsignificant disease (luminal irregularities or <50% diameter stenosis), or as having significant stenoses (≥50% diameter stenosis). Stenoses were evaluated in 2 orthogonal views, and classified as significant if the mean lumen diameter reduction was ≥50% using a validated QCA algorithm (CAAS, Pie Medical, Maastricht, the Netherlands).
DSCT image evaluation
One experienced observer, unaware of the results of CCA, calculated total calcium scores as Agatston scores, using validated software (Syngo MMWP VE20A, Siemens, Forcheim, Germany).
One observer evaluated image quality on a per-segment level and classified as good image quality (defined as absence or presence of any image-degrading artifacts related to motion, calcification, or noise, but evaluations possible with good-to-moderate confidence), or poor (presence of image-degrading artifacts and evaluation only possible with low confidence). Irrespective of image quality, all available coronary segments (including poor image quality) were included for comparison of DSCT with CCA.
Two experienced observers, unaware of the results of CCA, scored all DSCT coronary angiography datasets. Axial views and maximum intensity projections were used to identify coronary lesions. In addition, (curved) multiplanar reconstructions were used to classify coronary lesions into significantly diseased or not. Interobserver disagreements were resolved by consensus in a joint session.
The diagnostic performance of DSCT coronary angiography for the detection of significant lesions in coronary arteries with QCA as the standard of reference is presented as sensitivity, specificity, positive predictive value and negative predictive values, and positive and negative likelihood ratios with the corresponding 95% confidence intervals (CIs). Comparison between DSCT coronary angiography and QCA was performed on 3 levels: segment-by-segment, vessel-by-vessel (no or any significant stenosis per vessel), and patient-by-patient (no or any significant stenosis per patient). An additional sensitivity analysis to detect significant stenoses was performed after random selection of a single segment per patient to explore the effect of nesting. Inter- and intraobserver variability for the detection of significant coronary artery stenosis was calculated using κ statistics. To determine the intraobserver variability, 1 observer evaluated 30 (33%, 30 of 100) CT datasets twice with a time interval of 3 weeks.
Patient demographics are shown in Table 1.The mean interval between conventional and DSCT coronary angiography was 4.0 ± 4.8 days (range 0 to 17 days). All scans were performed without the use of oral or intravenous beta-blockers.
Mean scan range was 11.9 ± 1.1 cm (range 9.3 to 13.8 cm). Mean CT acquisition time was 8.6 ± 1.5 s (range 5.7 to 12.7 s). Pitch varied between 0.20 and 0.53. Mean heart rate was 68 ± 11 beats/min (range 44 to 107 beats/min). The overall radiation exposure for CT coronary angiography was estimated as 11.1 to 14.4 (men to women) mSv; 71 (71%, 71 of 100) patients had long-term beta-blocker medication. The estimated radiation exposure of DSCT coronary angiography was 13.5 to 16.9 mSv (men to women) in low heart rates (mean 56 beats/min), 10.7 to 13.8 mSv (men to women) in moderate heart rates (mean 68 beats/min), and 8.3 to 9.6 mSv (men to women) in high heart rates (mean 81 beats/min). In 5% (5 of 100) of patients with a ventricular extrasystole and in 3% (3 of 100) of patients with a premature atrial complex, ECG editing was successful.
A single dataset for the assessment of significant stenoses was used in 81%, 2 datasets in 16%, and 3 datasets in 3% of patients in order to obtain optimal image quality on a per-segment level.
Image quality was classified as good in 94% (1,400 of 1,489) and poor in 6% (89 of 1,489) on a per-segment level. Reasons for poor image quality were breathing motion artifacts (33%, 29 of 89), cardiac motion artifacts (14%, 12 of 89), severe calcifications (46%, 41 of 89), or low contrast-to-noise (8%, 7 of 89).
Diagnostic performance of DSCT coronary angiography
Sixteen patients with either an angiographically normal coronary angiogram (n = 16) or with nonsignificant disease (n = 4) were correctly identified with DSCT. Three patients were incorrectly classified as having single-vessel disease. One patient with significant disease was incorrectly classified as having nonsignificant disease with DSCT. Agreement between DSCT coronary angiography and QCA on a per-patient (no or any disease) level was good (κ value 0.89). Agreement between both techniques for classifying patients as having no, single-, or multivessel disease was very good (κ value 0.85).
One significantly diseased left anterior descending artery and 2 significantly diseased right coronary arteries were incorrectly classified as nonsignificantly diseased on the CT scan. Sensitivity for the detection of significantly diseased left anterior descending coronary arteries was 98%, for the right coronary arteries 96%, and for the left main and circumflex coronary arteries 100%. Agreement between CT coronary angiography and QCA on a per-vessel level was very good (κ value 0.85).
A total of 1,489 segments that were visualized with invasive coronary angiography were analyzed with DSCT coronary angiography. There were 12 (5.5%, 12 of 220) segments, which were incorrectly classified as having nonsignificant stenosis by DSCT, of which 3 segments demonstrated poor image quality due to cardiac motion artifacts in 2 segments (mean heart rates 65 and 78 beats/min) and due to severe calcifications in 1 segment. There were 69 (5.4%, 69 of 1,269) segments, which were incorrectly classified as having a significant stenosis by DSCT, of which 19 segments demonstrated poor image quality due to severe calcifications in 16 segments, a cardiac motion artifact in 1 segment (mean heart rate 68 beats/min), a breathing artifact in 1 segment, and low contrast-to-noise in 1 segment.
Agreement between CT coronary angiography and QCA on a per-segment level was very good (κ value 0.81).
The κ value of inter- and intraobserver variability for the detection of a significant stenosis per segment was 0.83 and 0.85, respectively.
Table 3shows the diagnostic accuracy of DSCT to detect significant coronary stenoses in patients with low (17 ± 27), intermediate (198 ± 96), and high (927 ± 727) Agatston calcium scores based on per-segment–based analysis.
Patients were divided into 3 groups based on the mean heart rate during DSCT. There was no significant difference in diagnostic accuracy on a segment-based analysis between these groups (Table 4).
In patients with low heart rates (mean 56.1 beats/min), optimal datasets reconstructed during the mid- to end-diastolic phase were selected in 94% (31 of 33) of patients, and additional datasets during the end-systolic phase were needed in 6% (2 of 33) of patients. In patients with intermediate heart rates (mean 67.9 beats/min), optimal datasets reconstructed during the mid- to end-diastolic phase were selected in 74% (25 of 34) of patients, and additional datasets in the end-systolic phase were needed in 26% (9 of 34) of patients. In patients with high heart rates (mean 80.7 beats/min), optimal datasets reconstructed during the end-systolic phase were selected in 91% (30 of 33) of patients, and additional datasets in the mid- to end-diastolic phase were needed in 9% (3 of 33) of patients.
Table 5demonstrates the diagnostic accuracy of DSCT to detect significant coronary stenoses in patients with atypical and typical angina, and unstable coronary artery disease based on a segment-based analysis.
A sensitivity analysis was performed after random selection of a single segment per patient. The sensitivity was calculated as 92% (12 of 13, 95% CI 87 to 96); specificity was 94% (82 of 87, 95% CI 90 to 99); positive predictive value was 71% (12 of 17, 95% CI 62 to 80); and negative predictive value was 99% (82 of 83, 95% CI 97 to 100).
Earlier studies using 4- and 16-slice CT scanners reported moderate-to-good diagnostic accuracy to detect significant lesions (1–8,14), but the technique was seriously limited by the presence of unevaluable segments that were, on average, 22% and 9% for the 4- and 16-slice CT, respectively (14). In a recent multicenter study using 16-slice CT scanners, the percentage of unevaluable coronary segments was 29% (4).
The development of 64-slice CT scanners involved a significant improvement in image quality and robustness of CT coronary angiography; however, on average, 5% and in one report even 12% of segments were reported to be unevaluable, and diagnostic accuracy was reduced at higher heart rates (9,10,17,18).
The introduction of DSCT is another step forward. This scanner is equipped with 2 X-ray tubes (dual source) thereby significantly reducing the temporal resolution to 83 ms independent of heart rate, using single-segment reconstruction. In non-DSCT systems, multisegment algorithms are used to improve temporal resolution. However, this approach is very dependent on a regular heart rate. Minor variation in the time interval between consecutive heart beats can result in interpolation artifacts and image blurring. Furthermore, multisegment reconstruction algorithms require a lower pitch thus longer scan times, more contrast material, and a higher radiation exposure. Multisegment approaches can also be applied in DSCT, resulting in a mean temporal resolution of up to 40 to 60 ms at 0.33 s gantry rotation time. This approach is not recommended for coronary angiography examinations, but may be useful for advanced functional evaluation (15).
With the DSCT scanner we were able to evaluate all coronary segments irrespective of heart rate and image quality. Despite the use of the high-speed DSCT scanner, poor image quality due to cardiac motion artifacts was observed in 14% of the coronary segments. However, the incidence of poor image quality occurred independent of heart rate, and good image quality could also be obtained in high heart rates.
We demonstrated that DSCT coronary angiography had a high diagnostic accuracy to detect significant coronary lesions on a per-segment-based level as compared with QCA. We selected a ≥50% diameter stenosis as the cutoff criterion for significant coronary artery disease to allow comparison with the majority of previous published reports (19). A segmental analysis is clinically useful in patients referred for coronary angiography to assess location (proximal, mid, distal, right coronary artery, left anterior descending artery, circumflex artery); severity (luminal narrowing ≥50%); and extent (1-, 2-, or 3-vessel disease) of coronary artery disease, which determines the value of CT scanning as an alternative to invasive coronary angiography. The patient-based diagnostic accuracy was high (96%), and a negative DSCT scan reliably ruled out the presence of a significant coronary stenosis in patients with atypical and typical angina, and unstable coronary artery disease (Table 5). These findings indicate that DSCT scanning is reliable as a gatekeeper of invasive coronary angiography.
In patients with a positive CT scan showing a severe (>70% diameter stenosis) lesion or a totally occluded vessel, no further evaluation is necessary. However, a positive CT scan with an estimated lesion severity of around 50% has limited value since it poorly discriminates functionally significant lesions from the ones that are not hemodynamically important (20). In this situation an additional functional imaging test such as myocardial perfusion scintigraphy or stress echocardiography would be a logical step before referring the patient for an invasive angiogram and possible revascularization. In patients deemed necessary to undergo revascularization, direct referral to the cathlab may be more logical with invasive assessment of the functional relevance of a lesion using fractional flow reserve and performance of percutaneous coronary intervention in the same session.
Lastly, new developments in CT coronary angiography are desirable for further improvement in clinical performance. Increased gantry rotation speed can further improve temporal resolution, but structural modifications will be required to account for a substantial increase in mechanical forces on the gantry. An alternative concept is the use of multiple (>2) X-ray sources and detectors within a single gantry, thereby obviating the need for an increased gantry rotation speed to improve temporal resolution. Further improved spatial resolution of less than 0.6 mm can be achieved by the use of smaller detector rows. However, an equal contrast-to-noise ratio requires an exponential increase in X-ray power, which will result in an excessive X-ray radiation exposure. Thus, new detector technology is needed to further improve spatial resolution.
Dual source computed tomography scanner coronary angiography should not be performed in patients with significant renal dysfunction or contrast intolerance. This further restricts the use of CT coronary angiography to selected patients, which should be taken into account when the technique is going to be applied in general clinical practice.
One advantage of DSCT is that patients with higher heart rates do not require premedication with beta-blockers because necessary treatment with beta-blockers before CT scanning hampers the CT throughput. The majority of patients (73%) in our study population already received long-term beta-blocker treatment and, therefore, did not benefit from an increased workflow. However, the use of DSCT in low or intermediate risk patient groups with expected lower use of chronic beta-blockers could be more efficient in terms of diagnostic throughput.
The rather high radiation exposure with CT coronary angiography is of concern. In our study, the overall effective dose for DSCT coronary angiography was estimated as 11.1 to 14.4 mSv (men/women), which is lower than the reported effective dose in 64-slice CT angiography (15.2 to 21.4 mSv men/women) (21). The significant reduction of effective radiation dose (8.3 to 9.6 mSv men/women) in high (>80 beats/min) heart rates as compared with low (<60 beats/min) heart rates (13.5 to 16.9 mSv men/women) can mainly be ascribed to an increased pitch and, therefore, shorter scan times in patients with high heart rates. However, compared with the effective dose in diagnostic coronary angiography (3 to 10 mSv) (22), the effective dose in DSCT coronary angiography still remains relatively high.
In this initial experience with the DSCT scanner, we selected a relatively wide pulsing window (25% to 70% of the RR interval), which allows for reconstruction of datasets during both the mid- to end-diastolic phase and end-systolic phase to obtain optimal image quality. However, there is a delicate balance between the width of the pulsing window and radiation exposure to the patient. Earlier technical feasibility studies demonstrated a significant reduction of the effective radiation dose by using a smaller width of the pulsing window (15). Further clinical studies should establish which pulsing window provides the optimal balance between radiation exposure and image quality, and the effect of a small pulsing window on diagnostic accuracy.
Persistent arrhythmias preclude accurate assessment with DSCT. For the purpose of this study, we excluded patients with persistent arrhythmias, which was also an exclusion criteria in studies using 64-slice scanners. However, our results demonstrate that DSCT technology enables us to scan patients with minor heart rate irregularities, such as a ventricular extrasystole or a premature atrial complex by automatically switching off ECG pulsing during irregular heartbeats. This enables the operator to perform ECG editing to correct for minor heart rate irregularities.
Severe calcifications remain problematic. Calcifications obscure the underlying lumen and preclude judgment of coronary lumen integrity resulting in overestimation of the severity of a coronary stenosis. This explains the observation that in 84% (16 of 19) of segments, which were incorrectly classified as having a significant stenosis by DSCT, severe calcifications resulted in poor image quality. In patients with high (mean 927 ± 727) Agatston scores, diagnostic accuracy was lower (91%) as compared with patient with low (mean 17 ± 27) Agatston scores (98%) (Table 4).
Our study was performed in a selected population consisting of symptomatic patients who were referred for conventional coronary angiography. This was evidenced by the fact that our study population had a high prevalence of coronary disease (77%, 77 of 100), and that a fairly large population had multivessel disease (46%, 46 of 100). In this population, DSCT coronary angiography performed well to excellent, but it remains to be demonstrated that such a high diagnostic accuracy will be achieved in a symptomatic patient population with a low-to-intermediate prevalence of disease or in a nonchest-pain population.
Our study was performed in a high-risk population with a wide range of symptoms who were referred for conventional coronary angiography. Dual source computed tomography scanner coronary angiography demonstrated a high diagnostic accuracy for the detection or exclusion of significant stenoses in patients with various heart rates without exclusion of unevaluable segments. These results indicate that the technique may now be tested in a cohort with a low-to-intermediate pretest probability of coronary artery disease or in patients with nonanginal chest pain to establish the role of DSCT coronary angiography in the management of patients with suspected coronary artery disease.
- Abbreviations and Acronyms
- conventional coronary angiography
- confidence interval
- computed tomography
- dual-source computed tomography scanner
- quantitative coronary angiography
- Received December 28, 2006.
- Revision received April 18, 2007.
- Accepted April 24, 2007.
- American College of Cardiology Foundation
- Achenbach S.,
- Ropers D.,
- Pohle F.K.,
- et al.
- Kuettner A.,
- Beck T.,
- Drosch T.,
- et al.
- Ropers D.,
- Baum U.,
- Pohle K.,
- et al.
- Mollet N.R.,
- Cademartiri F.,
- Nieman K.,
- et al.
- Leschka S.,
- Alkadhi H.,
- Plass A.,
- et al.
- Raff G.L.,
- Gallagher M.J.,
- O’Neill W.W.,
- Goldstein J.A.
- Leber A.W.,
- Knez A.,
- von Ziegler F.,
- et al.
- Hamon M.,
- Biondi-Zoccai G.G.,
- Malagutti P.,
- Agostoni P.,
- Morello R.,
- Valgimigli M.
- Schuijf J.D.,
- Wijns W.,
- Jukema J.W.,
- et al.
- Mollet N.R.,
- Cademartiri F.,
- van Mieghem C.A.,
- et al.
- Morin R.L.,
- Gerber T.C.,
- McCollough C.H.