Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

DOI: 10.1148/radiol.2432050808
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by McCollough, C. H.
Right arrow Articles by Kalender, W. A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by McCollough, C. H.
Right arrow Articles by Kalender, W. A.
(Radiology 2007;243:527-538.)
© RSNA, 2007

Special Reports

Coronary Artery Calcium: A Multi-institutional, Multimanufacturer International Standard for Quantification at Cardiac CT1

Cynthia H. McCollough, PhD, Stefan Ulzheimer, PhD2, Sandra S. Halliburton, PhD, Kaiss Shanneik, MS, Richard D. White, MD3, and Willi A. Kalender, PhD For the International Consortium on Standardization in Cardiac CT

1 From the Department of Radiology, Mayo Clinic College of Medicine, 200 First St SW, Rochester, MN 55905 (C.H.M.); Institute of Medical Physics, University of Erlangen, Erlangen, Germany (S.U., K.S., W.A.K.); and Division of Radiology, Cleveland Clinic, Cleveland, Ohio (S.S.H., R.D.W.). From the 2003 RSNA Annual Meeting. Supported by GE Healthcare, Philips Medical Systems, Siemens Medical Solutions, and Toshiba Medical Systems. Received May 11, 2005; revision requested July 12; revision received September 12, 2006; accepted October 17; final version accepted November 1. Address correspondence to C.H.M. (e-mail: mccollough.cynthia{at}mayo.edu).


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 References
 
Purpose: To develop a consensus standard for quantification of coronary artery calcium (CAC).

Materials and Methods: A standard for CAC quantification was developed by a multi-institutional, multimanufacturer international consortium of cardiac radiologists, medical physicists, and industry representatives. This report specifically describes the standardization of scan acquisition and reconstruction parameters, the use of patient size–specific tube current values to achieve a prescribed image noise, and the use of the calcium mass score to eliminate scanner- and patient size–based variations. An anthropomorphic phantom containing calibration inserts and additional phantom rings were used to simulate small, medium-size, and large patients. The three phantoms were scanned by using the recommended protocols for various computed tomography (CT) systems to determine the calibration factors that relate measured CT numbers to calcium hydroxyapatite density and to determine the tube current values that yield comparable noise values. Calculation of the calcium mass score was standardized, and the variance in Agatston, volume, and mass scores was compared among CT systems.

Results: Use of the recommended scanning parameters resulted in similar noise for small, medium-size, and large phantoms with all multi–detector row CT scanners. Volume scores had greater interscanner variance than did Agatston and calcium mass scores. Use of a fixed calcium hydroxyapatite density threshold (100 mg/cm3), as compared with use of a fixed CT number threshold (130 HU), reduced interscanner variability in Agatston and calcium mass scores. With use of a density segmentation threshold, the calcium mass score had the smallest variance as a function of patient size.

Conclusion: Standardized quantification of CAC yielded comparable image noise, spatial resolution, and mass scores among different patient sizes and different CT systems and facilitated reduced radiation dose for small and medium-size patients.

© RSNA, 2007


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 References
 
The advent of multi–detector row computed tomography (CT) has expanded the potential of and the range of applications for CT (1,2). In combination with subsecond rotation times and techniques for prospective triggering or retrospective gating according to the cardiac cycle, advances have been made in heart imaging, particularly for quantification of coronary artery calcium (CAC) (3–6). However, the quantitative nature of this task demands that care be taken to accurately calibrate and standardize the measurement of CAC (7). In addition, the clinical success of multi–detector row CT for CAC detection and quantification requires some method of assessing patient risk according to calcium burden and patient age and sex. Because the extrapolation of data from electron-beam CT calcium score databases has not been adequately validated, a multi–detector row CT database based on standardized acquisition and scoring protocols is necessary. Thus, the purpose of this study was to develop a consensus standard for quantification of CAC.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 References
 
In this article, we describe the composition of a cardiac calcification thoracic phantom and the methods used to scan and quantify (score) the calcifications in this phantom. The utility of this technique for standardizing calcium scores among multiple scan acquisitions and multiple scanners is demonstrated by using data collected with scanners from five manufacturers. The use of such a standardization process will enable the accumulation of multi–detector row CT CAC scores in a reference database and allow meaningful comparisons of scores acquired with different equipment and at different times. We hope to provide valuable clinical information for patient risk stratification that is currently not available and facilitate ongoing research of the epidemiology, progression, and regression of atherosclerotic coronary artery disease (5,8).

International Consortium on Standardization in Cardiac CT
This report reflects the work of the Physics Task Group of the International Consortium on Standardization in Cardiac CT, which was formed in late 2000 by Richard D. White, MD, and other interested parties from the academic and manufacturing communities to bring consensus to the field of CAC scanning and scoring by using multi–detector row CT systems. Additional information can be found on the consortium Web site (9). This report describes the precise consortium-recommended definitions and procedures used to acquire CAC image data and quantify CAC mass scores (10,11). A Web-based database was developed at the Cleveland Clinic by Sandra S. Halliburton, PhD, and colleagues (9,12) for use by members of the consortium. We expect to allow use of this database soon to others interested in participating. Such an expansion of membership will allow a large amount of patient data to accrue in a relatively short time and enable statistical assessment of risk on the basis of multi–detector row CT calcium mass scores and patient age and sex.

GE Healthcare (Milwaukee, Wis), Philips Medical Systems (Bothell, Wash), Siemens Medical Solutions (Malvern, Pa), and Toshiba Medical Systems (Tustin, Calif) provided support for this project in the form of liaison representation at meetings and financial contributions to defray the costs of meeting rooms and phantom materials. However, the physics committee members (W.A.K., C.H.M., S.U., S.S.H., K.S.) retained control of the data obtained at all times.

Design and Composition of the Phantom
The cardiac CT thoracic phantom used for calibration and verification purposes was developed by two of the authors (W.A.K., S.U.) in cooperation with QRM (Möhrendorf, Germany, http://www.qrm.de) in 1999 (4,7,13) and consists of two parts: the anthropomorphic thorax phantom and the calibration insert. The thorax phantom consists of artificial lungs, spine, and soft tissue–equivalent material. At the position of the heart is a cylindrical hole in which the calibration insert can be placed (Fig 1a).


Figure 1A
View larger version (51K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1a: (a) Diagram of frontal view of anthropomorphic phantom body with the calibration insert. (b) Diagrams of frontal (left) and side (right) views of calibration insert, with the nine different calcifications and the two large calibration inserts (0-HU water and 200 mg/cm3 calcium hydroxyapatite [HA]).

 

Figure 1B
View larger version (18K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1b: (a) Diagram of frontal view of anthropomorphic phantom body with the calibration insert. (b) Diagrams of frontal (left) and side (right) views of calibration insert, with the nine different calcifications and the two large calibration inserts (0-HU water and 200 mg/cm3 calcium hydroxyapatite [HA]).

 
The materials used in the anthropomorphic thoracic phantom mimic the tissues in the thorax in terms of density and attenuation. Because coronary calcifications consist of calcium HA (6), calcium HA is used in the thoracic phantom to mimic CAC and for the calibration insert. The calibration insert contains nine small cylindrical calcifications that vary in size and calcium HA density and two large calibration cylinders: one composed of water-equivalent material and the other composed of 200 mg/cm3 calcium HA (Fig 1b). The calcium HA densities and masses and the volumes and dimensions of the cylindrical calcifications are detailed in Table 1.


View this table:
[in this window]
[in a new window]

  		Table 1. Properties of the Cylindrical Calcium HA Calcifications inside the Calibration Insert

 
The measured values of the various properties of the cylindrical calcifications depend strongly on the scan acquisition parameters and scanner characteristics—in particular, the x-ray beam spectrum and the section thickness. The calibration process, in which the known values are compared with the measured values, is used to establish appropriate calibration parameters and thus obtain more accurate quantitative measurements of CAC that can be more robustly compared among measurements obtained at different times and with different equipment.

A specific goal was to establish size-specific tube current–time product values (in milliampere-seconds) for acquiring image data such that comparable image noise could be generated independently of patient size. This further ensures that small and medium-size patients do not receive unnecessary radiation doses. To achieve this goal, two additional oval rings were designed to fit over the primary thoracic phantom. The use of these rings further enabled the determination of calibration factors that corresponded to various levels of patient attenuation. The small tube current–time products required to meet the noise target and the small calibration factors were derived from scan acquisitions in the 30-cm-wide thoracic phantom. The medium tube current–time products and medium calibration factors were derived from acquisitions in the 35-cm-wide thoracic phantom (original phantom and medium attenuation ring). The large tube current–time products and large calibration factors were derived from acquisitions in the 40-cm-wide thoracic phantom (original phantom and large attenuation ring) (Fig 2).


Figure 2
View larger version (97K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2: Photograph of semianthropomorphic thoracic phantom, calibration insert, and two attenuation rings.

 
To determine the small phantom size, measurements of the upper abdomen–thorax region were obtained (W.A.K.) from a European young adult population in a previous study (7). On the basis of lateral dimensions at the level of the middle region of the liver assessed (C.H.M.) in more than 100 U.S. adults at CT, the medium-size and large extension rings were deemed necessary to appropriately represent a spectrum of patient sizes. The most common reconstruction field-of-view size in these patients was found to be approximately 38 cm. The lateral dimension of the typical-size (80-kg) adult was approximately 35 cm and was established as the medium phantom dimension. To represent large patients, a 40-cm phantom dimension was used.

Our experience has shown that lateral patient dimension is a better predictive factor of x-ray attenuation than surrogate measures such as weight and body mass index. Thus, we chose lateral width in the anatomic region of interest as the most relevant predictive measure for appropriate selection of the amperage and calibration factor.

Data Acquisition
To test the utility of the proposed standards for CAC data acquisition and scoring, the CAC thoracic phantoms (small, medium, and large sizes) were scanned by using scanner-specific acquisition protocols with 10 scanners (total of six models) (Table 2). The majority of the data were measured by using multi–detector row CT systems equipped with four data channels. Because these scanners are no longer considered to be state of the art, data for a 64-channel multi–detector row CT system were added. For all scan acquisitions, the center of the phantom insert was placed in the center of the field of measurement and the entire phantom was scanned by using the recommended CAC scanning protocols and a test electrocardiographic signal.


View this table:
[in this window]
[in a new window]

  		Table 2. Recommended Scan Acquisition Parameters for Measurement of CAC in Small, Medium-Size, and Large Patients

 
The primary differences between the protocols suggested by the different manufacturers were variances in tube current (in milliamperes) or tube current–time product (in milliampere-seconds) setting and variances in reconstruction kernels. All of these parameters strongly affect the image noise level. A similar noise level on the images used for CAC scoring across three general patient sizes (small, medium, and large) was a desired outcome of the consortium project. However, because the many acquisition and reconstruction parameters that affect image noise are not comparable across different scanners, the tube current or tube current–time product that resulted in approximately the same noise level on all scanners needed to be determined experimentally for the three patient sizes. In Figure 3, images of the three phantoms demonstrate the variation in noise that occurs when a constant tube current or tube current–time product setting is used, regardless of patient size. After appropriate adjustment of the tube current setting based on patient size, these images had the same noise level and hence a similar appearance.


Figure 3A
View larger version (76K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3a: Transverse CT images of semianthropomorphic thoracic phantom (a) without an attenuation ring (small), (b) with the 35-cm attenuation ring (medium size), and (c) with the 40-cm attenuation ring (large). All images were acquired with the same tube current–time product (in milliampere-seconds) setting. The increase in noise is due to the additional attenuation of the added rings and represents the variation in image quality that occurs when a fixed technique is used, regardless of patient size.

 

Figure 3B
View larger version (114K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3b: Transverse CT images of semianthropomorphic thoracic phantom (a) without an attenuation ring (small), (b) with the 35-cm attenuation ring (medium size), and (c) with the 40-cm attenuation ring (large). All images were acquired with the same tube current–time product (in milliampere-seconds) setting. The increase in noise is due to the additional attenuation of the added rings and represents the variation in image quality that occurs when a fixed technique is used, regardless of patient size.

 

Figure 3C
View larger version (126K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3c: Transverse CT images of semianthropomorphic thoracic phantom (a) without an attenuation ring (small), (b) with the 35-cm attenuation ring (medium size), and (c) with the 40-cm attenuation ring (large). All images were acquired with the same tube current–time product (in milliampere-seconds) setting. The increase in noise is due to the additional attenuation of the added rings and represents the variation in image quality that occurs when a fixed technique is used, regardless of patient size.

 
After review of the experimental data, a noise level target of 20 HU in the water region of the phantom insert was selected for small and medium-size patients. For large patients, a noise level target of 23 HU was chosen because a 20-HU noise target in large patients would have required operation of the scanner at the upper limits of x-ray tube power, which was not possible with all of the CT systems. These noise levels (20 and 23 HU) were below the noise level of electron-beam CT images, approximately 24 HU with use of the sharp reconstruction kernel, which was believed to be at the upper limit of acceptable noise levels. To keep radiation doses at a reasonable level, lower noise targets (<20 HU) were not selected. (Lower image noise would require higher radiation dose.)

The target tube current values were determined for small, medium-size, and large patients by using the three corresponding phantom sizes: First, by using the standardized acquisition protocol (eg, peak voltage, rotation time, image width, and reconstruction kernel) for each scanner model, images of each phantom were acquired as the tube current was varied from the maximum to the minimum value in 25–50-mA intervals. On each image, a circular region of interest with an area of 2.00 cm2 ± 0.10 (mean ± standard deviation) was placed over the water-equivalent cylinder in the cardiac portion of the phantom and the image noise (standard deviation of pixel values, in Hounsfield units) was recorded (S.U., S.S.H., C.H.M.). The noise data were plotted against the tube current–time product, and a power fit was performed to determine the equation of the curve that best fit the data points. By using these equations, the tube current–time product required to achieve the 20- or 23-HU noise target in the given phantom was determined (S.U., S.S.H., C.H.M.).

Clinically, the appropriate patient size category (small, medium, or large) must be determined before the CAC scan acquisition so that the appropriate tube current–time product will be used. This requires measurement of the lateral width of the patient by using the digital calipers on the user interface and an anteroposterior CT radiograph. The three size categories were defined as a lateral thickness at the top of the liver of less than 32.0 cm for small, of 32.0–38.0 cm for medium, and of greater than 38.0 cm for large. The lateral thickness measurement should be performed at the appropriate window and level values so that the skin-to-skin distance can be well visualized and measured.

Scoring Methods
Agatston score.—In 1990, Agatston et al (14) introduced a system of quantifying the CAC measured with electron-beam CT. Although this scoring system has several limitations, it has been used for many years. Thus, a large amount of data that were acquired and analyzed by using Agatston scores are available. Published risk tables derived from these data for subsequent coronary events are available (5,8).

Originally, the Agatston score was based on a data set of 20 contiguous electron-beam CT sections with 3-mm thickness (no gaps or overlap between sections). On each of the 20 images, calcifications are identified by setting a threshold of 130 HU and ignoring structures smaller than 1 mm2 to reduce the influence of image noise in the evaluation. A region of interest is placed around each lesion on each of the 20 images, and the area and maximum CT number of the calcifications are determined. The calcium score for each region of interest (CSi) is calculated by multiplying the area of the lesion (Ai) by a weighting factor (wi) that depends on the maximum CT number (CTmax) in the region of interest: CSi = wi · Ai, where

Formula 1(1)

Agatston scores for each artery, each calcification, or the entire heart—sometimes called total calcium score (TCS)—are calculated by summing the respective values for the regions of interest:

Formula 2(2)

The most obvious drawbacks of using the Agatston score are its relative complexity, dependence on the number of images per total scanning length, nonlinearity with respect to the amount of calcium (ie, the values of CTmax), and strong dependence on noise because the score is based on maximum CT numbers. For variations in CT scanning parameters (such as different section thickness and overlapping sections) or equipment (eg, electron-beam or multi–detector row systems), one has to adapt the score to obtain results that are comparable to the original published scores.

Volume score.—The volume score became more popular as some studies revealed that it was more robust than the Agatston score in terms of reproducibility (15). The volume score represents the volume of the calcification (V). It is calculated as the number of voxels (Nvox) in the volume data set that belong to the calcification, multiplied by the volume of one voxel (Vvox): V = Nvox · Vvox.

The easiest approach to determining which voxels belong to the calcification is to use all voxels with an attenuation value above a certain threshold, typically 130 HU. However, this scoring method is particularly sensitive to partial volume averaging. For example, imaging a small but high-density calcification at a section thickness greater than the calcification dimension would result in a volume score larger than the actual calcification size. This would occur because the highly attenuating calcification would be partial volume averaged to fill the entire voxel with a CT number above the calcification threshold. Thus, volume scores are overestimations of the calcium content and are subject to substantial variability between repeat acquisitions, depending on the position of the small calcification within the image plane.

Calcium mass score.—On the basis of literature review (15–18) and our independent work, we believe that the best measure of the amount of CAC in a CT data set is the calcium mass score. To determine the relative mass score, the mean CT number of the calcification depicted on each image (CTi) is multiplied by the volume of the calcification on that image (Vi). The product is directly proportional to the calcium mass depicted on that image (mi): mi {propto} (CTi · Vi). With this method, one automatically corrects for the effects of linear partial volume averaging, because objects smaller than the section thickness are represented by accordingly decreased mean CT numbers. To obtain the total relative mass of the calcification, the single mass scores for all images are summed. Although the product of CTi · Vi could serve as a relative measure of the calcium mass, across different CT scanners and scan acquisition protocols it is more robust to use an absolute value of calcium mass.

To obtain absolute values for calcium mass (m), a calibration measurement is necessary. The CT number of a calcification with known HA density ({rho}HA) is measured, and a calibration factor (c) is determined such that mi = c · CTi · Vi. By using the relationship m = {rho}HA · V, the calibration factor is calculated as follows:

Formula 3(3)
where CTC is the measured CT number of the known calcium calibration object. With this equation, one assumes that the CT number of water (CTW) is zero, as would be expected according to the definition of CT number. In practice, however, the CT number of water often differs from 0, making its value relevant to the calibration equation. If an exact CT number of water is measured from the same calibration acquisition used to determine the CT number of the known quantity of HA, then the CT number of water can be taken into account by subtracting the CT number of water from the CT number of the known calcification:

Formula 4(4)
The calibration factor is therefore derived by dividing the HA density of the known calcification by the difference of the mean CT number of the known calcification minus the mean CT number of water.

Segmentation of Calcium Voxels
For any computation of CAC score, the voxels containing calcium must be identified. This has generally been done by using a CT number threshold of 130 HU (14). However, the variation in calibration factors among different scanners and among patients of different sizes—and even within a given patient (19,20)—indicates that a given CT number does not necessarily indicate the same amount of calcium. Thus, we have adopted a threshold segmentation approach that is based instead on a fixed density (concentration) of calcium HA. The recommended density threshold is 100 mg/cm3 calcium HA, which corresponds closely to the typical CT number threshold of 130 HU. All voxels (pixel values) having a calcium HA density of 100 mg/cm3 or greater are considered calcium voxels. Thus, the correct unit for reporting calcium mass score is milligrams of calcium HA over the segmentation threshold of 100 mg/cm3.

Determination of Calcium Calibration Factor
The calcium calibration factors for the various scanner models and three patient sizes were measured by using the calibration insert in the described thoracic phantom. After scanning the phantom with the clinically relevant protocol, the mean CT numbers for the large 200 mg/cm3 calcium HA and 0-HU water calibration inserts were measured and the calcium calibration factor determined by using Equation (4). To determine mean CT numbers, circular regions of interest with a mean area of 2.00 cm2 ± 0.10 were used.

Assessment of Calcium Score Accuracy and Precision
The reproducibility (ie, precision) of quantitative CAC measurements is very important for follow-up examinations and ensuring the reliability of individual measurements. For follow-up examinations to investigate the evolution of a disease, the reproducibility of the measurement is a decisive parameter. The accuracy—that is, the exact correspondence between measured and true values—is also important for comparing the results obtained with different scanners and scanner models. Because the use of epidemiologic risk tables requires that the input value—that is, the CAC mass score—be measured accurately and precisely, multiple data sets obtained by using the same scanner and multiple scanner models were acquired to assess the precision and accuracy of our CAC quantification as a function of scanner and scanner model variation (S.U., K.S., C.H.M., S.S.H., other consortium members).

For each data set, the mean Agatston, volume, and calcium mass scores (± standard deviations) for multiple measurements were calculated. To calculate the calcium mass from the CT numbers, the calcium calibration factor for the appropriate scanner model was measured and used. In our simplified phantom model, the segmentation task was completely automated and was based on either CT number or calcium HA density owing to the absence of any other calcified objects in the phantom. Thus, intra- and interuser variation was not assessed. In clinical settings, the segmentation method used (eg, manual vs automated, specific algorithms implementation, or the performance of different observers) may lead to additional variation in quantitative CAC scores. However, such segmentation-induced variation probably exists in previously obtained estimates of clinical variation in CAC scores (discussed later in text) (15,21–24).

A comparison of the measured Agatston scores with known values was not possible because the definition of the Agatston score does not correspond to any physical measure of the calcification test objects. Calcium volume and mass scores were compared with the known values given in Table 1. The evaluation software used probably has a nonnegligible influence on accuracy—and perhaps on reproducibility. For example, connectivity requirements (eg, 1 vs 4 pixels, area, or volume criteria), the use and performance of automated vessel detection algorithms, and calculation and rounding techniques can vary between implementations. Hence, all data presented herein were scored by using the same evaluation software. The data sets acquired by using the CAC thoracic phantom, or patient data sets, can be used with different analysis software to determine the degree to which the software affects the accuracy and precision of the reported scores. Such analysis requires that software manufacturers modify the existing algorithms used to calculate calcium mass scores by using the methods detailed in this report. Various software systems will be validated as such software packages become available. (At least two manufacturers have incorporated the consortium recommendations into their software.)


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 References
 
Variation in Calibration Factor
For a given scanner model, the calibration factor derived from multiple scan acquisitions in the small phantom (with either the same scanner or other scanners of the same model at different institutions) demonstrated coefficients of variation (ie, standard deviations as percentages of the mean) of 0.13%–1.6% (Tables 3 and 4). Similarly, the day-to-day variations observed with a given scanner were between 0.6% and 1.1% (Table 3). Averaged across the three phantom sizes, the calibration factors had coefficients of variation ranging from 3.8% to 5.1% (Table 3). Thus, for a given scanner model, the variation due to object size exceeded the variation due to scanner-to-scanner differences. The high precision of the calibration factor measurements implies that these values can be measured by the system manufacturer or by the user at the time of system installation and confirmed relatively infrequently (perhaps yearly).


View this table:
[in this window]
[in a new window]

  		Table 3. Calibration Factors Measured for Small, Medium-Size, and Large Patients

 

View this table:
[in this window]
[in a new window]

  		Table 4. Mean Resultant Calcium Scores Obtained by Using Fixed 130-HU Threshold

 
Calcium Score Accuracy and Precision Assessment
Regarding Agatston, volume, and mass scores (with use of 130-HU threshold) for the collected phantom CAC data, the coefficient of variation was largest for volume scores (7.9%) and of comparable magnitude for Agatston (4.0%) and mass (4.7%) scores (Table 4). The known total mass of calcium HA within the thoracic phantom (true measurement) was 168.2 mg. For all scanners tested, the mean measured calcium mass score was 167.5 mg ± 7.9 (coefficient of variation, 4.7%). When the electron-beam CT data were excluded (being notably dissimilar to the other data points), the mean (± standard deviation) calcium mass score for the multi–detector row CT scanners was 165.0 mg ± 4.6 (coefficient of variation, 2.8%). For the five multi–detector row CT scanner systems, the total calcium mass score derived was measured correctly to within ±5 mg of calcium HA (Fig 4, Table 4). An assessment of absolute accuracy could not be made for the Agatston score, because this score represents only a mathematical construct and there is no physical reference standard against which to compare it.


Figure 4
View larger version (29K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4: Bar graph illustrates comparison of mean total Agatston, volume, and mass scores for the CAC thoracic phantom derived for the small patient size category (20-HU noise target) with different CT scanners. Volume scores varied the most, whereas Agatston and mass scores had less variation among scanner types. Numbers in parentheses are numbers of scan acquisitions performed with the given scanner model. EBCT = Imatron electron-beam scanner. Seq = sequential.

 
Dose Consequences of Using a Consistent Noise Target
Use of the recommended tube current (Table 2) to achieve a predetermined noise target enabled image noise to be standardized across scanner models (Fig 5). This standardization enabled the tube current—and hence the radiation dose—for a small patient to be reduced by 10%–79% compared with the values observed with use of the tube current–time product settings originally recommended by the manufacturers (Fig 6). For large patients, the tube current would increase 29%–150%—and on one system decrease 56%—relative to the values originally recommended by the manufacturers. This wide range of tube current changes reflects the poor agreement among original recommendations regarding the extent of tube current increase needed for a large patient. Thus, the dose adjustment needed for large patients is much more consistent across manufacturers when an objective noise target is used to determine the required tube current value. As patient size increases, the increase in effective dose becomes smaller than the respective increase in tube current owing to the absorption of x-rays in the adipose tissue surrounding sensitive organs (25). Thus, the increase in effective dose in large patients (approximately 98 kg) is only about 25% when the tube current–time product setting is adapted to achieve the same noise levels as those in "normal-size" (defined as 80 kg) patients (26).


Figure 5
View larger version (28K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 5: Bar graph illustrates the variation in image noise observed with the original manufacturer-recommended technique factors for the small phantom size, as well as the decrease in noise variation achieved after the recommended tube current was varied to achieve a common noise target of 20 HU.

 

Figure 6
View larger version (21K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 6: Bar graph illustrates radiation dose decreases achieved with different CT scanners. Considerable dose decreases were achieved for the small patient size category with use of technique factors that yielded a standardized noise target (20 HU), as compared with the doses delivered with use of the original manufacturer-recommended technique factors.

 
Effect of Density Threshold
Figure 7 shows the resultant scores for 3- and 5-mm calcifications when the patient size and the peak kilovoltage are varied for when either a fixed CT number segmentation threshold (130 HU) or a fixed density segmentation threshold (CT number adapted according to the measured calibration factor to a value corresponding to 100 mg/cm3 calcium HA) is used. The absolute mass was underestimated in both cases owing to the point-spread function of the CT system, which lowers some calcium-containing voxel values to a value below the necessary threshold. With use of the CT number threshold, the results were comparable across peak kilovoltage values but less consistent across phantom sizes. With use of the density threshold, the results were very consistent across phantom sizes; however, the systematic differences across peak kilovoltage values indicate the need to adjust the density threshold if the peak kilovoltage is varied. Thus, further work to identify optimal density thresholds as a function of peak kilovoltage appears to be warranted. However, since the consortium recommends the standardization of scanning protocols, including peak kilovoltage, the increased variation seen when peak kilovoltage is varied has little influence on the conclusions reported herein. It is more important to eliminate variations in mass score due to patient size variations (the harder to control parameter) than to eliminate those due to peak kilovoltage variations (the more easily controlled parameter).


Figure 7A
View larger version (21K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 7a: Bar graphs illustrate calcium mass score as a function of calcification size and peak kilovoltage when a (a) fixed CT number threshold (130 HU) or (b) fixed calcium HA density threshold (100 mg/cm3) is used for segmentation of the 400 mg/cm3 calcium HA test calcifications. Dotted horizontal lines denote the known masses of the 3- and 5-mm test calcifications: 39.3 and 8.5 mg calcium HA, respectively. Data were acquired by using the Volume Zoom four-channel multi–detector row CT scanner in the spiral mode. Density threshold results were more consistent across phantom sizes than were CT number threshold results, whereas CT number threshold results were more consistent across different peak kilovoltage settings. These results demonstrate the need to adjust the density threshold when the peak kilovoltage is varied.

 

Figure 7B
View larger version (18K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 7b: Bar graphs illustrate calcium mass score as a function of calcification size and peak kilovoltage when a (a) fixed CT number threshold (130 HU) or (b) fixed calcium HA density threshold (100 mg/cm3) is used for segmentation of the 400 mg/cm3 calcium HA test calcifications. Dotted horizontal lines denote the known masses of the 3- and 5-mm test calcifications: 39.3 and 8.5 mg calcium HA, respectively. Data were acquired by using the Volume Zoom four-channel multi–detector row CT scanner in the spiral mode. Density threshold results were more consistent across phantom sizes than were CT number threshold results, whereas CT number threshold results were more consistent across different peak kilovoltage settings. These results demonstrate the need to adjust the density threshold when the peak kilovoltage is varied.

 

   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
References
 
The described results show that in a standardized anthropomorphic phantom, calcium mass score is an accurate and precise measure of calcium content and is more robust with patient size variations when a calcium density threshold is used. These results are consistent with those of others who have compared the accuracy and precision of Agatston scores with those of mass scores, including those studies performed when the calcified object is in motion (7,19,20,27–31). Our data represent the best-case agreement among calcium scores derived by using different scanners because of the absence of motion. Other studies have revealed percentage differences in CAC Agatston scores for the same patient and the same scanner of 19%–37% (15,21–24). Devries et al (32) substratified their subject population according to Agatston score and observed a mean difference in score of 72% ± 54 among patients with scores lower than 10.

In our present work, we observed approximately a 4% difference in Agatston scores. Thus, we conclude that scanner-to-scanner variation will be considerably lower than the other types of variation encountered in patients (eg, variations caused by nonreproducible positioning of the coronary artery between heart beats, motion blurring or artifact, motion due to breathing or the need for multiple breath holds [at electron-beam CT]). The increased accuracy and precision of the mass score, as compared with those of the Agatston and volume scores, and the ability to compare the measured mass score with a known physical standard establish the mass score approach as the preferred method of quantitating CAC.

The measured calcium calibration factors for the various scanner models and the small patient size varied minimally (approximately 1%) among the multiple measurements obtained by using either the same scanner or another scanner of the same model at different institutions. Although the variation in calibration factors for a specific phantom size and scanner model was small (approximately 1%), the change in phantom (ie, patient) size caused a much larger change in the calibration factor (approximately 5%). Thus, the scanner operator will need to accurately assess the patient's size by using a lateral width measurement, use the appropriate tube current value to achieve the correct noise target, and use the appropriate calibration factor. Using these protocol guidelines will enable the acquisition of consistent mass scores across a broad range of patient sizes.

The participation and cooperation of the CT scanner manufacturers involved in this effort were and will remain a key element in the success of CAC quantitation. Using the specific guidelines developed and agreed on, manufacturers have begun implementing the described mass score feature in their cardiac evaluation tools; some of these new software packages have already been released. In addition, the consortium has provided their detailed recommendations to third-party software manufacturers (ie, workstation vendors) so that the users of these systems will also have access to mass score calculations.

Ongoing work of the consortium is currently focused on validating the various software tools by using a standard data set that all manufacturers are requested to "score" by using their software. In this manner, these software tools can be validated so that for a given data set, the same CAC mass score will be calculated, regardless of the software used. The only limitation to this process will be the variation in artery segmentation that is introduced in the manual arterial segmentation process. Because most images are scored by experienced cardiac CT technologists or physicians and advanced vessel segmentation that enables accurate and robust identification of the coronary arteries is now offered with most software packages, we expect this variation to be small. As patient data begin to be scored according to the consortium recommendations, analyses of the magnitude of variation introduced in the arterial segmentation process will begin to be conducted.

We conclude that the calcium mass score is the most reasonable quantitative measure of CAC content, having a variation of less than 5% among five models of multi–detector row CT scanners and one electron-beam CT scanner, all from different manufacturers. These results were achieved by using scanners at six institutions for several months (or years in the case of the 64-channel multi–detector row CT system). When the electron-beam CT data points, which were the farthest from the mean value, were excluded, the coefficient of variation decreased to less than 3%. The standardization efforts responsible for the accuracy and precision of the mass scores across the different CT scanners were focused on (a) standardization of the scan acquisition protocols to achieve comparable noise and spatial resolution across scanners, (b) calibration of attenuation measurements—that is, CT numbers—into an absolute physical measure (herein, calcium HA density), (c) use of a segmentation threshold (calcium HA density of 100 mg/cm3) based on a calibrated physical measure, (d) use of a well-defined continuous metric (calcium mass score) for tabulation of the total extent of CAC (total milligrams of calcium HA), and (e) ability to assess absolute accuracy by using a physical reference standard.

It is important to note that the calcium mass score, as defined herein, may not correspond to the absolute mass of a calcified lesion. This is because the attenuation value (or CT number) of some lesion voxels may be below the segmentation threshold for calcium and thus may not be counted in the mass summation. Thus, the total calcium mass score, summed over all user-defined regions of interest, should be reported in milligrams of calcium HA above the segmentation threshold of 100 mg/cm3.

Use of the proposed standardization algorithm enables quantitative comparison of coronary artery calcifications that have been measured at different times and by using different CT scanner models. This facilitates a meaningful comparison of patient data both over time within a patient and across patients and institutions. Future work of the International Consortium on Standardization in Cardiac CT will be focused on the accumulation of patient risk factors and CAC data into an online database. From this multimanufacturer, multi-institutional international database, risk stratification data can be meaningfully summarized with use of a robust standardized measure, the calcium mass score.


   ACKNOWLEDGMENTS
 
The authors acknowledge the helpful discussions of the members of the International Consortium on Standardization in Cardiac CT, chaired by Richard D. White, MD; the technical and Web assistance from Jonathon Sell of the Cleveland Clinic Foundation; and the support of GE Healthcare, Philips Medical Systems, Siemens Medical Solutions, and Toshiba Medical Systems for this standardization effort. The assistance of Natalie Braun with manuscript preparation is also very much appreciated.


   FOOTNOTES
 

Abbreviations: CAC = coronary artery calcium • HA = hydroxyapatite

2 Current address: Siemens Medical Solutions, Forchheim, Germany. Back

3 Current address: University of Florida, Jacksonville, Fla. Back

See Materials and Methods for pertinent disclosures.

Author contributions: Guarantor of integrity of entire study, C.H.M.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, C.H.M., S.U., S.S.H., K.S., W.A.K.; clinical studies, W.A.K.; experimental studies, all authors; statistical analysis, C.H.M., S.U., S.S.H., K.S., W.A.K.; and manuscript editing, all authors


   References
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 References
 

  1. Marincek B, Ros PR, Reiser M, Baker ME. Multislice CT: a practical guide. New York, NY: Springer-Verlag, 2001.
  2. Fishman EK, Jeffrey RB Jr. Multidetector CT: principles, techniques, and clinical applications. Philadelphia, Pa: Lippincott Williams & Wilkins, 2004.
  3. Flohr T, Ohnesorge B, Bruder H, et al. Image reconstruction and performance evaluation for ECG-gated spiral scanning with a 16-slice CT system. Med Phys 2003;30(10):2650–2662.[CrossRef][Medline]
  4. Kachelriess M, Ulzheimer S, Kalender WA. ECG-correlated image reconstruction from subsecond multi-slice spiral CT scans of the heart. Med Phys 2000;27(8):1881–1902.[CrossRef][Medline]
  5. Raggi P, Callister TQ, Cooil B, et al. Identification of patients at increased risk of first unheralded acute myocardial infarction by electron-beam computed tomography. Circulation 2000;101(8):850–855.[Abstract/Free Full Text]
  6. 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(5):1175–1192.
  7. Ulzheimer S, Kalender WA. Assessment of calcium scoring performance in cardiac computed tomography. Eur Radiol 2003;13(3):484–497.[Medline]
  8. Arad Y, Spadaro LA, Goodman K, Newstein D, Guerci AD. Prediction of coronary events with electron beam computed tomography. J Am Coll Cardiol 2000;36(4):1253–1260.[Abstract/Free Full Text]
  9. Sell JC, Halliburton SS. Calcium scoring database. International Consortium on Standardization in Cardiac CT Web site. Available at: https://clinapps.bio.ri.ccf.org/cascore/. Accessed December 2004.
  10. McCollough CH, Ulzheimer S, Halliburton SS, White RD, Kalender WA. A multi-scanner, multi-manufacturer, international standard for the quantification of coronary artery calcium using cardiac CT [abstr]. In: Radiological Society of North America scientific assembly and annual meeting program. Oak Brook, Ill: Radiological Society of North America, 2003; 630.
  11. Ulzheimer S, Shanneik K, McCollough CH, Halliburton SS, Kalender WA. Advantages of using calcium mass in combination with a calcium density threshold for the quantification of coronary calcium [abstr]. In: Radiological Society of North America scientific assembly and annual meeting program. Oak Brook, Ill: Radiological Society of North America, 2003; 428.
  12. Halliburton SS, Sell JC, McCollough CH, Ulzheimer S, Kalender WA, White RD. A Web-based multi-vendor, multi-institutional database of standardized coronary calcium measurements using cardiac CT [abstr]. In: Radiological Society of North America scientific assembly and annual meeting program. Oak Brook, Ill: Radiological Society of North America, 2003; 812.
  13. Ulzheimer S, Kachelriess M, Kalendar WA. New phantoms for quality assurance in cardiac CT [abstr]. Radiology 1999;213(P):402.
  14. Agatston AS, Janowitz WR, Hildner FJ, Zusmer NR, Viamonte M Jr, Detrano R. Quantification of coronary artery calcium using ultrafast computed tomography. J Am Coll Cardiol 1990;15(4):827–832.[Abstract]
  15. Yoon HC, Greaser LE 3rd, Mather R, Sinha S, McNitt-Gray MF, Goldin JG. Coronary artery calcium: alternate methods for accurate and reproducible quantitation. Acad Radiol 1997;4(10):666–673.[CrossRef][Medline]
  16. Ohnesorge B, Flohr T, Fischbach R, et al. Reproducibility of coronary calcium quantification in repeat examinations with retrospectively ECG-gated multisection spiral CT. Eur Radiol 2002;12(6):1532–1540.[CrossRef][Medline]
  17. Hong C, Bae KT, Pilgram TK, Zhu F. Coronary artery calcium quantification at multi–detector row CT: influence of heart rate and measurement methods on interacquisition variability initial experience. Radiology 2003;228(1):95–100.[Abstract/Free Full Text]
  18. Halliburton SS, Stillman AE, Lieber M, Kasper JM, Kuzmiak SA, White RD. Potential clinical impact of variability in the measurement of coronary artery calcification with sequential MDCT. AJR Am J Roentgenol 2005;184(2):643–648.[Abstract/Free Full Text]
  19. Stanford W, Burns TL, Thompson BH, Witt JD, Lauer RM, Mahoney LT. Influence of body size and section level on calcium phantom measurements at coronary artery calcium CT scanning. Radiology 2004;230(1):198–205.[Abstract/Free Full Text]
  20. McCollough CH, Kaufmann RB, Cameron BM, Katz DJ, Sheedy PF 2nd, Peyser PA. Electron-beam CT: use of a calibration phantom to reduce variability in calcium quantitation. Radiology 1995;196(1):159–165.[Abstract/Free Full Text]
  21. Yoon HC, Goldin JG, Greaser LE 3rd, Sayre J, Fonarow GC. Interscan variation in coronary artery calcium quantification in a large asymptomatic patient population. AJR Am J Roentgenol 2000;174(3):803–809.[Abstract/Free Full Text]
  22. 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):550–558.[CrossRef][Medline]
  23. Callister TQ, Cooil B, Raya SP, Lippolis NJ, Russo DJ, Raggi P. Coronary artery disease:improved reproducibility of calcium scoring with an electron-beam CT volumetric method. Radiology 1998;208(3):807–814.[Abstract/Free Full Text]
  24. Greaser LE 3rd, Yoon HC, Mather RT, McNitt-Gray M, Goldin JG. Electron-beam CT:the effect of using a correction function on coronary artery calcium quantitation. Acad Radiol 1999;6(1):40–48.[CrossRef][Medline]
  25. Schmidt B, Kalender WA. A fast voxel-based Monte Carlo method for scanner- and patient-specific dose calculations in computed tomography. In: Guerra AD, ed. Physica medica. Erlangen, Germany: European Journal of Medical Physics, 2002; 43–53.
  26. Schmidt B. Dose calculations for computed tomography: reports from the Institute of Medical Physics. Erlangen, Germany: Institute of Medical Physics, 2001; 7.
  27. Das M, Martensen J, Zou KH, et al. Coronary calcium screening with low-dose 16-slice multidetector-row CT: which calcium scoring method is most robust? [abstr]. European Radiology Supplements 2004;14:539–656.[CrossRef]
  28. Shemesh J, Evron R, Koren-Morag N, et al. Coronary artery calcium measurement with multi–detector row CT and low radiation dose: comparison between 55 and 165 mAs. Radiology 2005;236(3):810–814.[Abstract/Free Full Text]
  29. Martensen JM, Peldschus K, Yucel E, et al. Influence of patient size and acquisition parameters on coronary calcium scoring: a phantom study with 16-slice multidetector-row CT [abstr]. In: Radiological Society of North America scientific assembly and annual meeting program. Oak Brook, Ill: Radiological Society of North America, 2003; 631.
  30. Furuhashi S, Sato Y, Inoue F, Hori Y, Kanmatsuse K, Takahashi M. Evaluation of the plaque texture of means of multislice spiral computed tomography in patients with acute coronary syndrome and stable angina [abstr]. In: Radiological Society of North America scientific assembly and annual meeting program. Oak Brook, Ill: Radiological Society of North America, 2003; 631.
  31. Hoffman U, Bull-Stewart AA, Achenbach S, Ferencik M, Brady TJ, O'Donnell C. Interscan and interobserver variability of coronary artery calcium measurements in prospectively triggered multislice CT using conventional scoring methods and calibrated mineral mass [abstr]. In: Radiological Society of North America scientific assembly and annual meeting program. Oak Brook, Ill: Radiological Society of North America, 2003; 631.
  32. Devries S, Wolfkiel C, Shah V, Chomka E, Rich S. Reproducibility of the measurement of coronary calcium with ultrafast computed tomography. Am J Cardiol 1995;75(14):973–975.[CrossRef][Medline]



This article has been cited by other articles:


Home page
 Am. J. Roentgenol.Home page
M.-T. Wu, P. Yang, Y.-L. Huang, J.-S. Chen, C.-C. Chuo, C. Yeh, and R.-S. Chang
Coronary Arterial Calcification on Low-Dose Ungated MDCT for Lung Cancer Screening: Concordance Study with Dedicated Cardiac CT
Am. J. Roentgenol., April 1, 2008; 190(4): 923 - 928.
[Abstract] [Full Text] [PDF]

Home page
 J. Clin. Endocrinol. Metab.Home page
R. Shroff, A. Kerchner, M. Maifeld, E. J. R. Van Beek, D. Jagasia, and A. Dokras
Young Obese Women with Polycystic Ovary Syndrome Have Evidence of Early Coronary Atherosclerosis
J. Clin. Endocrinol. Metab., December 1, 2007; 92(12): 4609 - 4614.
[Abstract] [Full Text] [PDF]

Home page
 RadiologyHome page
A. Rutten, I. Isgum, and M. Prokop
Coronary Calcification: Effect of Small Variation of Scan Starting Position on Agatston, Volume, and Mass Scores
Radiology, December 1, 2007; 246(1): 90 - 98.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by McCollough, C. H.
Right arrow Articles by Kalender, W. A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by McCollough, C. H.
Right arrow Articles by Kalender, W. A.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE