Carr, J. Jeffrey M.D., M.S
CORONARY CALCIUM AS A QUANTITATIVE MEASURE OF ATHEROSCLEROSIS WITH CT
Coronary calcium is an expression of atherosclerosis in the coronary arteries, and cardiac gated CT can be used to measure the amount of calcium present (1). In asymptomatic individuals, coronary calcium is a measure of subclinical atherosclerosis. Cardiac gated CT, whether EBCT or helical CT (HCT), creates a cardiac phase-specific, volumetric image dataset of the heart. Typically, a window in late diastole, although this point remains an area of active debate, is used to generate the dataset for subsequent quantitative analysis. I refer to this dataset as the diastolic image set. The diastolic image set, be it with EBCT or HCT, is then used to measure the amount of calcium plaque related to the coronary arterial vessels. This number is commonly referred to as the calcium score. Several common methods and variations are used in calculating the calcium score from the diastolic image set. These include variations in the traditional Agatston scoring method (minimum lesion size, slice thickness, display field-of-view) as well as novel scoring methods based on calcium volume (2). The impact of variations in how the calcium score is calculated is discussed only briefly as it relates to scan acquisition with single and multislice HCT.
OBJECTIVE: OBTAINING A CONTIGUOUS DIASTOLIC VOLUME OF THE HEART AND CORONARY ARTERIES
The motion of the heart is complex, and similarly complex is the motion of the coronary arteries. To measure coronary calcium in living humans with precision and a high degree of reproducibility requires that the coronary circulation be imaged at a fixed time in the cardiac cycle. Methods have centered on imaging the coronary arteries during diastole when the velocity of the coronary arteries is reduced relative to the systolic ejection phase. In essence, the goal is to create a diastolic volume of the heart, including the coronary arteries. Ideally, the diastolic images would represent the entire coronary circulation, without overlap or gaps related to slice positioning, table motion, or beat-to-beat variation in the cardiac cycle. The heart moves in a complex twisting motion relative to the thorax. The electrocardiogram (ECG) provides an electrical timing diagram of cardiac activity. This map of electrical cardiac activity is used to select a specific time within the cardiac cycle for imaging, under the assumption that the heart (and coronary arteries) will return to this same location. If there was no variation in the cardiac cycle and each level of the heart was obtained at the same time delay from the R-wave, the result would be a contiguous volume of the heart at the given point in the cardiac cycle.
THE ULTIMATE CT SYSTEM FOR CARDIAC IMAGING
I propose the following for the ultimate CT system for measuring coronary calcium (and potentially this system could be used for other applications). The following are a short list of characteristics that illustrate key aspects of cardiac CT. Rapid temporal resolution to stop coronary motion is required. High-performance ECG waveform analysis is seamlessly integrated into scanner operation to allow prospective cardiac triggering or retrospective gating in a near real-time mode during scan acquisition. Sufficient signal-to-noise characteristics are needed in the images to reliably distinguish between small calcium plaques and noise, independent of patient size. Early detection of smaller plaques may be especially important in women and some ethnic groups. Improved spatial resolution, especially in the z-axis or slice thickness, reduces the partial volume effect. Most of the calcium volume scoring algorithms take advantage of mathematical interpolation to create isotropic voxels (volume elements in which each dimension (height, width, and depth) are the same and have shown improved reproducibility of the measure of calcium. In the ultimate cardiac CT system, isotropic or near isotropic voxels would be obtained directly without the need for mathematical estimation. Increased efficiency in image acquisition would result in reduced breath hold times. Short breath hold time reduces respiratory related motion and slice misregistration. Methods for addressing beat-to-beat physiologic variation in the cardiac cycle are needed as well.
In evaluating the reproducibility of single-slice HCT using retrospective cardiac gating, we and others observed that the plot of absolute percent change in calcium score against the mean score of the replicated scans was nearly identical to that of previously published articles with EBCT despite the differences in scanner technology and temporal resolution (3-5). Closer scrutiny of the data shows that for most individuals with sequentially repeated cardiac CT scans, the test has a reasonable reproducibility. However, a subset of individuals, across the spectrum of calcium scores, vary significantly in the measured calcium score from the first to the second scan. Review of the images, in most cases, shows that the plaque was not imaged or was imaged in a different orientation. In essence, there are differences in the two diastolic image sets secondary to differences in the orientation of the coronary arteries relative to the imaging plane, as well as gaps or overlaps along the z-axis. Our hypothesis is that beat-to-beat variation in coronary artery location is a major factor in variations in calcium scores with replicated scans and that reduction of this component of variation will improve the reproducibility of coronary calcium measures.
The ultimate cardiac CT system would image the entire coronary circulation in one heartbeat (or cardiac cycle). This would eliminate beat-to-beat cardiac variation and result in a diastolic image set derived from one cardiac cycle as opposed to the current situation with single-slice EBCT and HCT systems in which 35 different cardiac cycles are used to generate the 35 diastolic images to be used to measure coronary calcium. In addition, the ultimate cardiac CT system would have near isotropic spatial resolution, excellent signal-to-noise characteristics, rapid image acquisition to stop cardiac motion, and an integrated real-time prospective or retrospective cardiac gating system. Most individuals can have their entire heart imaged with a set of thirty-five 3-mm slices or a total of 105 mm z-axis coverage. If we desired 1-mm slice thickness to result in a near isotropic voxel of 0.5 mm × 0.5 mm × 1.0 mm, this would necessitate acquiring 105 contiguous 1-mm slices in one heart beat! A temporal resolution of 25 msec for each image would be adequate so long as this was coupled with sufficient photon flux to maintain adequate signal-to-noise within the image.
The objective with CT measurement of coronary calcium is straightforward: measure the amount of calcium (above a certain threshold) in the coronary arterial vessels with a high degree of accuracy and precision. A high overall image quality, which includes temporal resolution, spatial resolution, contrast resolution, and signal-to-noise, must be reproducibly obtained across a spectrum of patients. To achieve this, a contiguous, volumetric image of the coronary circulation, in a fixed point of the cardiac cycle, must be acquired and then analyzed to determine the presence and amount of coronary calcium. The challenge of measuring coronary calcium is compounded by respiratory, cardiac, and gross (voluntary and involuntary) patient motion. In addition, there are significant variations across the spectrum in patient size, cardiac contractility, and breath-holding ability that impact image quality and the ultimate measure of coronary calcium. CT scanner capabilities as far as temporal resolution, cardiac gating, and signal-to-noise all impact the calcium measurement. Overall, study quality is dependent on the composite of all of these factors.
Temporal (or time) resolution is critical to “freezing” the motion of the coronary arteries. The velocity and motion of the coronary arteries varies between the left and right circulations. The right coronary artery moves twice as far and has greater velocity than the left, and these characteristics vary significantly between and within individuals (6). Motion artifacts are thus more common in the right coronary artery. Detailed mapping of coronary artery motion by use of biplane angiography have shown considerable variability in coronary motion and variation in the coronary artery “rest period” within diastole from 66 to 333 msec (6). For measuring coronary flow with magnetic resonance imaging, acquisition times as short as 25 milliseconds have been recommended (7). Before the “rest period” of coronary motion in diastole is a time of gradual coronary motion. By combining the scan data from the rest and gradual motion periods, current-generation HCT systems are able to obtain images of the coronary arteries. The variability of the location of the rest period between individuals likewise may explain how retrospectively gated HCT images are able to select the optimum window to reduce coronary motion (Fig. 1).
Ideally, with cardiac CT or magnetic resonance imaging, imaging windows would be as short as possible while maintaining the other domains of image quality. Electron-beam CT, unlike current versions of helical CT scanners, can acquire images of the heart in a remarkably short window in time. Single images of the heart can be acquired in 50 msec; however, to improve signal-to-noise, an acquisition time of 100 milliseconds is routinely employed for coronary calcium determination. However, as suggested by the magnetic resonance imaging coronary velocity data, the 100 milliseconds acquisition time images are not devoid of motion artifacts (Fig. 2). The development of subsecond helical scanners, initially with 800 milliseconds and now with 500 milliseconds gantry rotation times, allowed the acquisition of images with sufficient temporal resolution to image the coronary arteries relatively free of motion. The faster gantry rotation times are coupled with a half scan (aka “partial” or “segmented”) scan reconstruction algorithm to result in images with temporal resolutions of 500 and 300 milliseconds, respectively. The addition of multislice (aka “multidetector” or “multirow”) CT systems allows for even further improvements in temporal resolution by taking further advantage of the detector array. The “sector-scan” multi-slice reconstruction algorithm (8), currently under development, can obtain cardiac gated images with a temporal resolution of 100 milliseconds through pooling views of raw scan data from each of the four detector rows and assembling them based on the concurrent ECG tracing.
Image noise impacts the ability to measure coronary calcium. Image noise is responsible for setting the threshold for calcium to be considered “scoreable.” The 130 HU was set based on the value exceeding 2 SDs above the average attenuation value of the blood in the aorta on the first-generation EBCT systems. The CT measurement of coronary calcium requires that lesions meet predefined criteria (i.e., 130 H.U. and larger than a minimum size) to be considered measurable coronary calcium. Small lesions may be very important in identifying early atherosclerosis in middle-aged individuals, as well as in women and minorities. These lesions may be composed of relatively few pixels (or voxels). High image signal-to-noise ratio improves the ability to reliably discriminate small but real calcified coronary plaques from noise. Interactive windowing of a CT scan with coronary calcium will rapidly show that the size and thus volume of a coronary plaque changes as the window and level values are changed (Fig. 3).
In addition, the measurement of coronary calcium should be independent of other physical characteristics of the individual, namely weight or body mass index. Larger individuals attenuate more x-ray photons, resulting in fewer photons reaching the CT detectors, be it with EBCT or HCT. Fewer photons result in more image noise (i.e., a reduction in the signal-to-noise ratio). Single-and multi-slice HCT systems, unlike EBCT, can adjust the tube current (mA) such that image quality at the center of the thorax can be maintained independent of the individual's weight or girth. This is standard practice for diagnostic imaging with helical CT in other applications. Maintaining a consistent signal-to-noise ratio in the region of the heart is important to the accuracy and precision of the calculated coronary calcium score. High amounts of noise degrade the ability to identify small plaques and increase the likelihood of falsely identifying noise as coronary calcium. Broderick et al. (9) were the first to take advantage of the higher signal-to-noise possible with HCT by using an Elscint dual-slice scanner and showed how scoring coronary calcium at a lower threshold of 90 H.U. was feasible (9). Helical CT system manufacturers have competed and continue to compete on image quality. The abilities of helical CT detector arrays to have superior imaging characteristics, both signal-to-noise and contrast-to-noise, have proved pivotal for modern diagnostic imaging and are taken advantage of in cardiac imaging applications with these systems.
The coronary arteries are relatively small, fast-moving structures, and sufficient spatial resolution must be achieved to measure the calcified plaque. Single-slice CT systems, both EBCT and HCT, have similar spatial resolution characteristics. Both typically scan with a 3-mm slice thickness and a display field of view between 26 and 35 cm. In both instances, a nonisotropic voxel is created. The slice thickness is nearly 6 times greater than the in-plane resolution of 0.50 to 0.7 mm. Multi-slice HCT systems acquire data, regardless of whether in the helical or axial (aka “sequential”) mode, as a relatively thick slab, which is then collimated to the slice thickness by the detector. Current multislice systems are capable of four slices (or data streams) at once. In the typical coronary calcium scan, four 2.5-mm slices are obtained simultaneously (Fig. 4). In essence, one 10-mm slab of data is interrogated and subsequently divided into four 2.5-mm slices based on the configuration of the detector array (post-patient detector collimation). The benefits of this acquisition mode are that gaps and overlaps within the four slices of the slab are eliminated. Because the entire slab is obtained over the same instant in time, the possibility for variation in coronary artery location over time or between cardiac cycles is also eliminated. Multislice HCT systems were introduced initially with a two-and now a four-slice mode. The rapid development of eight-slice and beyond scanners is pending and will likely have a further dramatic impact on cardiac CT. A multislice HCT system capable of eight slices would cover 20 mm of the z-axis per cardiac cycle (8 slices × 2.5 mm/slice). Thus, the entire heart, or 100 mm of coverage, could be obtained in only five heartbeats as opposed to 35 heartbeats with single-slice scanners, and bring us closer to the ideal of a volumetric image of the heart within a single cardiac cycle.
The ECG provides a timing diagram for synchronizing the CT image data with the relative position of the heart during the cardiac cycle. Our ultimate goal is to obtain a contiguous diastolic volume of the coronary circulation. Thus, the diastolic volume must not only be contiguous along the z-axis from superior to inferior but also within the cardiac cycle. In the absence of breathing and motion, the thorax can be held nearly motionless; however, the heart is fortunately continuously pumping blood, resulting in the coronary arteries moving in a complex motion.
PROSPECTIVE ECG TRIGGERING
ECG triggering works by activating the x-ray source and thus image acquisition after a predetermined delay time from the R wave. To determine the delay time, the average RR interval is calculated based on the preceding cardiac cycles. This method of cardiac synchronization is currently used with EBCT and multislice HCT systems and was initially evaluated with single-slice HCT (10). Single-slice HCT systems using prospective ECG triggering are operated in the axial (or sequential, i.e., nonhelical) mode. Like the initial implementation with EBCT, several heartbeats are required to move the table between slice locations. As a result, the total scan time exceeds that possible with a single breath hold, and thus a minimum of two separate acquisitions are required to cover the entire heart. Multislice HCT systems improve the efficiency of scan acquisition. Thus, a four-slice scanner, operating in the axial or cine mode, can acquire images at 4 times the rate of a single-slice HCT scanner, all other parameters being constant. The increased efficiency makes it possible to image the entire heart in a single breath hold. Immediately after the scan, the 35 to 40 diastolic images of the heart are immediately ready for measuring coronary calcium without further postprocessing. The primary disadvantage with ECG triggering is beat-to-beat cardiac variability. Because the offset for each scan is based on the preceding RR intervals, an aberrant beat can result in obtaining a slice in a different portion of the cardiac cycle and thus an unintended location. The unintended slice location results in a corresponding gap and overlap with the adjacent slices above and below the level in question. This can lead to both missing and double counting of plaques.
RETROSPECTIVE ECG GATING
With retrospective ECG gating, the CT image data and ECG data are simultaneously recorded during scan acquisition. Postprocessing, either on the CT scan console or on a separate computer workstation, aligns the two data sources for diastolic image selection (11,12). Because image selection is performed after the scan, the appropriate diastolic phase image can be selected at each level regardless of potential beat-to-beat variations in the location or the coronary rest period in the cardiac cycle (10). In addition, by scanning in the helical mode, a volumetric image, including the domains of time and space, is created, allowing the review of the heart throughout the cardiac cycle (approximately 10 frames per cycle) at each level (Figs. 1, 4). The major disadvantages are that with 10 images per cardiac cycle, the CT image series can range in size from 350 to 500 images, and an additional post-processing step of reviewing the automated diastolic selection and accepting or modifying is necessary to create the diastolic image set.
SCANNING MODES FOR HCT
Helical scanner consist of an x-ray tube and opposing detector array that rotate around the patient within the scanner gantry. Two basic modes of operation are possible: In axial mode, the x-ray tube and detectors rotate around a stationary patient. The tube is activated for a prescribed time interval and then turned off. The CT couch increments to the next slice location, and the tube is then activated. In helical mode, the tube is activated and the CT couch moves at a constant velocity. With single slice HCT systems, axial slice acquisition can be coupled with prospective ECG triggering to obtain a diastolic image set of the heart. Time constraints, as discussed earlier, result in a minimum of two breath holds to image the entire coronary circulation (10). A single-slice HCT system can image in the helical mode by using retrospective cardiac gating and thus image the entire heart in a single breath-hold and create a diastolic image set for measuring coronary calcium (11). Multi-slice scanners can scan in both axial and helical modes. By acquiring four slices at one scan, time is reduced by a factor of four, and thus prospective ECG trigger within the axial mode is possible with the entire heart imaged in a single breath hold. Retrospective cardiac gating can be coupled with a four-slice helical acquisition, similar to its use in single-slice HCT systems. However, reconstruction algorithms for cardiac gated four-slice helical mode are computationally demanding.
The detector array on a Multislice CT HCT system can be used to improve temporal resolution at the expense of z-axis coverage. The sector scan reconstruction algorithm combines sectors of Radon space from the four detector rows to create images with improved temporal resolution (100–150 msec) (8).
THE CASE FOR HELICAL CT
Cardiac gated HCT can accurately and reproducibly measure coronary calcium. Single-slice CT scanners, both electron beam and helical, have been shown to accurately and reproducibly measure the amount of coronary calcium, a measure of subclinical coronary atherosclerosis. Coronary calcium scores measured on HCT systems are highly correlated with EBCT derived scores (R > 0.97) (5,10,11). High agreement is seen in categorizing individuals into strata based on calcium scores measured with the two instruments, EBCT and HCT. Depending on your tendency toward optimism or pessimism, single-slice CT systems have either reasonable or poor reproducibility when using the traditional scoring method. This has resulted in Wang et al. concluding, “EBCT is not sufficiently reproducible to allow serial quantitation of coronary calcium in individual patients over relatively short periods (<2 years),” (13) and similar conclusions by Callister et al. (2) : “The results of previous analyzes of the reproducibility of the TCS [Total Calcium Score] have confirmed the large interscan variability of this measure.” High degrees of variability are seen in traditional calcium scores below 100. Even some individuals with extensive coronary calcium and scores as high as 1,000 have up to 20% change with replicated scans with both single-slice EBCT and HCT (3,4). The observed variability in repeated scans with EBCT or HCT is problematic for the measurement of progression or regression of calcium over relatively short intervals (12 to 18 months). Several means of improving scan reproducibility have been proposed, including increasing slice thickness from 3 to 6 mm (14), increasing the minimum size for a score-able plaque (3), and developing more robust measurement algorithms. Callister et al. (2), through implementation of a volume scoring algorithm, was able to reduce error by 40% compared with the traditional scoring method (2).
Cardiac gated multislice CT will improve reproducibility by reducing the amount of cardiac variability inherent in the diastolic image set and by being closer to a true volumetric acquisition. When scanning in the axial mode, each 10-mm slab is obtained within the same cardiac cycle. This slab is then divided into four 2.5-mm slices for measuring coronary calcium (4 slices × 2.5 mm/slice = 10 mm acquisition slab/heartbeat). With this technique, there are no gaps or overlaps between the four slices within the acquisition slab. The possibility remains for gaps and overlaps between slabs as with the single-slice acquisition. If, for example, a single-slice EBCT or HCT system required 32 slices to image the entire heart, 32 different cardiac cycles are used to acquire the diastolic image set. Using a four-slice HCT system, the same amount of coverage could be obtained with imaging during only eight heart beats (4 × 8 = 32). Each slab of four slices is a true volume acquisition with the benefits discussed earlier, and quantification of this improvement is an area of active research by our laboratory and others. By combining a true volumetric scan acquisition with volumetric scoring, further improvements in reproducibility should be expected (Figs. 5, 6).
HCT IS WIDELY AVAILABLE
HCT systems are the workhorses of medical imaging, with more than 10,000 systems installed in the United States. The additional (or marginal) cost of upgrading the systems for cardiac scanning is low. Furthermore, the capability of multislice scanners in other diagnostic imaging applications is resulting in rapid acquisition of this new technology. More than 300 multislice scanners have been installed by one major manufacturer alone (personal communication, General Electric Medical Systems). As with any new diagnostic test or intervention, the initial studies of coronary calcium generally have been performed in a limited population at single centers. The research by Arad et al. (15) showed a high predictive ability of coronary calcium for future cardiovascular events. However, the predictive ability of calcium scores in a high-risk population was similar to that of traditional cardiovascular risk factors (16). Prospective, population based, multi-site studies are necessary to determine the true predictive ability of coronary calcium and which individuals could benefit from asymptomatic screening for coronary calcium. If coronary calcium screening is justifiable, then the ability to measure coronary calcium accurately and reproducibly with HCT will make population-based screening possible in the United States (Figs. 7, 8).
THE CONTINUED EVOLUTION OF HCT
There has been and will continue to be rapid technologic advances in HCT in general and with respect to cardiac imaging in particular. Within the past 36 months, cardiac gating techniques were developed and implemented, gantry rotation times were reduced to 0.5 seconds, temporal image resolution was reduced to 300 milliseconds, and multislice HCT systems were introduced. Advanced reconstruction of multislice HCT data, such as the previously discussed sector reconstruction algorithm, can further improve temporal resolution to the 100-milliseconds range. Within the next 12 months, it is likely that 8-and possibly 12-or 16-slice systems may be available. Prototype CT systems that use high spatial resolution panel detectors (on the order of microns), similar to what is implemented for digital radiography, are being designed and tested for feasibility. It is highly likely that there will be further improvements in cardiac imaging with multislice CT systems.
I believe that it is critically important to remember and remain fixed on the ultimate objective, reducing the impact of cardiovascular disease on our society and worldwide. Current-generation multislice HCT systems have several characteristics that lend them to high-quality coronary imaging and improved measurement of coronary calcium. Although there are small but real differences between techniques of measuring coronary calcium, the spectrum of the disease process is large. It is inevitable that scientific debate and evaluation of the techniques will continue for the foreseeable future, there is no doubt that both EBCT and HCT systems will continue to improve. The CT measurement of coronary calcium provides another means of assessing subclinical atherosclerosis, and initial studies indicate that it might be an important risk factor for cardiovascular events. A remaining unsolved piece of the puzzle is how well coronary calcium predicts cardiovascular events independent of established cardiovascular risk factors or emerging inflammatory marker such as C-reactive protein (17). A methodical and evidence-based approach to introducing technology to cardiovascular disease prevention will allow the development of both medically effective and cost-effective preventive strategies for cardiovascular disease.
1. Wexler L, Brundage B, Crouse J, et al. Coronary artery calcification: pathophysiology epidemiology, imaging methods, and clinical implications. A statement for health professionals from the American Heart Association. Writing Group. Circulation 1996; 94:1175–92.
2. Callister TQ, Cooil B, Raya SP, Lippolis NJ, Russo V, Raggi P. Coronary artery disease: improved reproducibility of calcium scoring with an electron-beam CT volumetric method. Radiology 1998; 208 (3):807–14.
3. Bielak LF, Kaufmann RB, Moll PP, McCollough CH, Schwartz RS, Sheedy II. PF Small lesions in the heart identified at electron beam CT: calcification or noise? Radiology 1994; 192 (3):631–6.
4. Carr JJ, Danitschek J, Mitsa T, Terry JG, Crouse JRI, Buchanan R. Retrospective cardiac gating improves measurement of coronary artery calcifications with fast helical computed tomography. Radiology 1999; 213:352.
5. Goldin JG, Heinze SB, Emerick AM, Yoon C, Kolodaro G, Suh RD. Comparison of coronary artery calcification by electron beam computed tomography and subsecond helical computed tomography. Radiology 1999; 213:351.
6. Wang Y, Vidan E, Bergman GW. Cardiac motion of coronary arteries: variability in the rest period and implications for coronary MR angiography. Radiology 1999; 213 (3):751–8.
7. Hofman MB, Wickline SA, Lorenz CH. Quantification of in-plane motion of the coronary arteries during the cardiac cycle: implications for acquisition window duration for MR flow quantification. J Magn Reson Imaging 1998; 8 (3):568–76.
8. Pan T, Foley WD, Woodford M, Shen Y, Acharya KC, Hu H. A hybrid helical reconstruction for high-pitch multislice cardiac CT. Radiology 1999; 213:451.
9. Broderick LS, Shemesh J, Wilensky RL, et al. Measurement of coronary artery calcium with dual-slice helical CT compared with coronary angiography: evaluation of CT scoring methods, interobserver variations, and reproducibility. AJR 1996; 167:439–44.
10. Becker CR, Jakobs TF, Aydemir S, et al. Helical and single-slice conventional CT versus electron beam CT for quantification of coronary artery calcification. AJR 2000; 174:543–7.
11. Carr JJ, Crouse JR, Goff DCJ, D'Agostino Jr, RB Peterson MP, Burke GL. Evaluation of subsecond gated helical CT for quantification of coronary artery calcium and comparison with electron beam CT. AJR 2000; 174:915–21.
12. Woodhouse CE, Janowitz WR, Viamonte Jr. M Coronary arteries: retrospective cardiac gating technique to reduce cardiac motion artifact at spiral CT. Radiology 1997; 204 (2):566–9.
13. Wang S, Detrano RC, Secci A, et al. Detection of coronary calcification with electron-beam computed tomography: evaluation of interexamination reproducibility and comparison of three image-acquisition protocols. Am Heart J 1996; 132 (3):550–8.
14. Secci A, Wong N, Tang W, Wang S, Doherty T, Detrano R. Electron beam computed tomographic coronary calcium as a predictor of coronary events: comparison of two protocols. Circulation 1997; 96 (4):1122–9.
15. Arad Y, Spadaro LA, Goodman K, Liedo-Perez A, Sherman S, Lerner G, Guerci AD. Predictive value of electron beam computed tomography of the coronary arteries. Circulation 1996; 93 (11):1951–3.
16. Detrano RC, Wong ND, Doherty TM, et al. Coronary calcium does not accurately predict near-term future coronary events in high-risk adults. Circulation 1999; 99 (20):2633–8.
17. Greenland P, Abrams J, Aurigemma GP, et al. Prevention Conference V: Beyond Secondary Prevention: Identifying the high-risk patient for primary prevention: noninvasive tests of atherosclerotic burden. Circulation 2000; 101:e16–e22.
Editor: Jeffrey S. Klein
Associate Editors: Ann Leung, MD
David Lynch, MD, Jung-Gi Im, MD
Michio Kono, MD, Charles White, MD
Guest Editor: William Stanford, M.D.
© 2001 Lippincott Williams & Wilkins, Inc.