点击显示 收起
1 From the Departments of Radiology (K.D.H.) and Health Evaluation Sciences (D.T.M.), Penn State College of Medicine, 500 University Dr, PO Box 850, Hershey, PA 17033; and the Department of Radiology, University of Pittsburgh Medical Center Health System, Pa (D.C.S.). Received February 22, 2000; revision requested April 8; final revision received August 6, 2001; accepted August 21. Supported in part by an unrestricted grant from Marconi Medical Systems.
ABSTRACT |
---|
Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES |
---|
MATERIALS AND METHODS: A working heart phantom simulating normal cardiac motion and providing attenuation equal to that of an adult thorax was used. Thirty tubes with a 3-mm inner diameter were internally coated with pulverized human cortical bone mixed with epoxy glue to simulate minimal (n = 10), mild (n = 10), or severe (n = 10) calcified plaques. Ten additional tubes were not coated and served as normal controls. The tubes were attached to the same location on the phantom heart and scanned with electron-beam CT and helical CT in horizontal and vertical planes. Actual plaque calcium content was subsequently quantified with atopic spectroscopy. Two blinded experienced radiologic imaging teams, one for each CT system, separately measured calcium content in the model vessels by using a Hounsfield unit threshold of 130 or greater.
RESULTS: The sensitivity and specificity of electron-beam CT in detecting CAC were 66.1% and 80.0%, respectively. The sensitivity and specificity of helical CT were 96.4% and 95.0%, respectively. Electron-beam CT was less reliable when vessels were oriented vertically (sensitivity and specificity, 71.4% and 70%; 95% CI: 39.0%, 75.0%) versus horizontally (sensitivity and specificity, 60.7% and 90.0%; 95% CI: 48.0%, 82.0%). When a correction factor was applied, the volume of calcified plaque was statistically better quantified with helical CT than with electron-beam CT (P = .004).
CONCLUSION: Ungated helical CT depicts coronary arterial calcium better than does gated electron-beam CT. When appropriate correction factors are applied, helical CT is superior to electron-beam CT in quantifying coronary arterial calcium. Although further work must be done to optimize helical CT grading systems and scanning protocols, the data of this study demonstrated helical CT’s inherent advantage over currently commercially available electron-beam CT systems in CAC detection and quantification.
Index terms: Computed tomography (CT), comparative studies, 51.12115, 51.12119 • Computed tomography (CT), electron beam, 51.12119 • Computed tomography (CT), helical, 51.12115 • Coronary vessels, calcification, 54.81 • Heart, calcification, 544.12115, 544.12119 • Heart, CT, 542.12115, 544.12115, 544.12119 • Phantoms
INTRODUCTION |
---|
Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES |
---|
Helical CT has also been used to quantify CAC (19–24). Helical CT uses a standard CT tube, acquires a volume of data without cardiac gating, can be used to reconstruct sections at many desired intervals, and is far more available worldwide than electron-beam CT. The attributes of ungated helical CT (ie, improved z-axis resolution, unlimited ability to overlap reconstructed sections, and widespread availability) make it an ideal method to quantify CAC.
At the time this article was written, to our knowledge no investigators had evaluated the sensitivity and specificity of electron-beam CT or helical CT in detecting and quantifying CAC by using phantom vessels with known calcium content. The purpose of this study was to determine the sensitivities and specificities of electron-beam CT and ungated helical CT in detecting and quantifying small amounts of calcium, by using a functional heart phantom and tubes with known quantities of calcium.
MATERIALS AND METHODS |
---|
Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES |
---|
|
Each tube was first oriented parallel (long axis) and then perpendicular (short axis) to the x-y plane to simulate the appearance of the left main and left anterior descending coronary arteries at transverse CT. The line pair and longitudinal resolution with the electron-beam CT scanner were 7.0 and 3.0 mm versus 21.0 and 0.5 mm with the helical CT scanner. A 35-cm field of view was used with both scanners to approximate the field of view commonly used for chest CT. The tubes were positioned at the isocenter of the CT gantry by elevating or lowering the CT table. The 10 normal tubes were scanned three times with each scanner in both vessel orientations to serve as a baseline of 30 normal tubes. In the 120 examinations performed per CT scanner, 30 of the scanned tubes were normal, and 30 contained calcified plaque. A random number table was used to determine the sequence of the 120 examinations. Each tube received a number from 1 to 40 and was labeled at one end with an indelible circle to ensure consistent placement and orientation of the tube on the phantom heart.
CT scanning was performed by lead CT technologists experienced with each system. Data acquisition and detection and measurement of calcium-containing pixels for both electron-beam and helical CT were performed by the CT technologists and confirmed by two experienced radiologists (K.D.H., D.C.S.). The following imaging protocols were used to provide optimal scanner performance and replicate parameters commonly used for clinical chest CT:
With the electron-beam scanner, 62.5 mAs, 130 kV, 3-mm section thickness, electrocardiographic gating, 0% section overlap, 1.37-mm pixel size, and a bone edge-enhancing algorithm were used.
With the helical scanner, 275 mAs, 120 kV, 3-mm section thickness, no electrocardiographic gating, 50% section overlap, 0.7-mm pixel size, an extra-sharp interpolator, a bone algorithm, and a pitch of 1.5 were used.
Routine helical CT of the chest requires a higher mAs than does electron-beam CT because of improved low contrast resolution and accounts for the disparity in this parameter. The CT dose index and surface radiation dose for electron-beam CT were 0.0041 and 0.0078 Gy and for helical CT were 0.0042 and 0.0092 Gy, respectively.
System operators and physicians were blinded to the presence, absence, and/or amount of calcium in any given tube. Phantom vessels were mounted on the heart model by a research assistant working independently of the blinded imaging teams. The characteristics of each tube were known by only the biostatistician (D.T.M.). At completion of the 120 examinations with each CT scanner, the number of pixels containing calcium, with a Hounsfield unit threshold of 130 or greater, was measured for all images of each examination. When the tube was vertically oriented (parallel to the x-y plane), the entire phantom vessel was analyzed. When horizontally oriented (perpendicular to the x-y plane), the proximal 1.5 cm of tube closest to the CT gantry (away from the metallic valve) was evaluated to avoid streak artifact. The number of pixels with calcium was converted into square millimeters on the basis of pixel size. After the first set of measurements was acquired with both systems (for electron-beam CT, D.L.S.; for helical CT, K.D.H.), a period of 3 weeks lapsed, and a second set of measurements was obtained in reverse order by the same two individuals in reverse order.
At study completion, the plaque of each phantom coronary artery was measured (K.D.H.) by using a millimeter ruler for thickness, circumference in degrees, and length. These measurements were converted into a plaque volume by using the formula for the volume of a hollow cylinder: Volume = length x {PI x [1.552 - (1.55 - thickness)2] x degree/360}, where PI is pulsatility index. Calcium weight was converted to measured calcium volume by using the known constant 1.55 mg calcium = 1 mL calcium. The entire tube was cut open, and the glue and calcium plaque was carefully removed by using a scalpel. All portions of the removed plaque were then placed in a sealed vial and marked with the number for the corresponding tube. The vials were then sent to a national clinical laboratory, the Mayo Clinic, which measured specimen weights and total calcium weight and physical density by using atomic spectroscopy.
Statistical comparison of helical CT and electron-beam CT with respect to estimated plaque volume was performed with the paired t test. Agreement between findings of atomic spectroscopy and each method was quantified by using the weighted statistic after categorizing the amount of calcium detected with each method. The estimated calcium values obtained with each method were sorted and assigned a value of zero if less than the median, a value of one if greater than the median and less than the 75th percentile, or a value of two if equal to or greater than the 75th percentile. The statistic was used to measure agreement between these ordinal scores and was interpreted like a correlation.
Sensitivity and specificity were calculated to evaluate the value of the two methods as diagnostic tools. The presence of calcium as determined with atomic spectroscopy was treated as the reference standard. Cutoff levels for helical CT (16.7 mg) and electron-beam CT (11.6 mg) were chosen from the data to minimize the number of errors (false-positive and false-negative results). For the 30 tubes containing calcium, regression analysis was performed to model the relationship between actual and measured plaque volumes. The results of this analysis were used to calculate correction factors that could be applied to the measured volumes so that they better approximated actual plaque volumes. This analysis included instrumentation (helical CT or electron-beam CT) and orientation (horizontal or vertical) as additional prediction factors, so that separate correction factors were calculated for each instrument in each orientation. Helical CT–estimated volumes less than 16.7 were considered normal (CAC not present), whereas those equal to or greater than 16.7 were considered abnormal (CAC present). Receiver operating characteristic analysis was performed to determine the optimal cutoff levels for helical CT (16.7) and electron-beam CT (11.6) to minimize the number of errors (false-positive and false-negative findings). Statistical comparison of helical CT and electron-beam CT with respect to sensitivity and specificity was performed with the Fisher exact test. All analyses were performed with SAS statistical software (SAS, Cary, NC).
RESULTS |
---|
Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES |
---|
|
There was high variability between helical and electron-beam CT depictions of calcified plaque volume. Helical CT of the tubes with calcium depicted higher amounts of calcium than did electron-beam CT (Table 2), whereas only electron-beam CT erroneously depicted calcium on both projections in many tubes without calcium. This increased false-positive rate of electron-beam CT resulted in decreased sensitivity (Fig 2). When the measured calcium volume (in milliliters), as determined with helical CT and electron-beam CT, was compared with the actual calcium mass determined with atomic spectroscopy (Table 3), electron-beam CT measurements were statistically less significant than those of helical CT for most plaque characteristics, including physical density and volume. There was no statistical difference between electron-beam CT and helical CT in measurement of normal tubes without calcium.
|
|
|
|
Combining data from both tube orientations, the sensitivity and specificity of helical CT were 96.4% and 95.0%, respectively, versus 66.1% and 80.0%, respectively, for electron-beam CT (P < .05). The respective sensitivity and specificity of both modalities was unchanged when assessing calculated and measured plaque volume, calcium mass, or calcium physical density (Table 4). For calculated plaque volume, the statistics for combined projections were 0.84 and 0.52 for helical CT and electron-beam CT, respectively. The statistic for helical CT was statistically significant when compared with that for electron-beam CT (P < .01). The corresponding 95% CIs were 77.0% to 91.0% for helical CT and 39.0% to 65.0% for electron-beam CT. For measured plaque volume, calcium mass, and calcium physical density, statistics differed slightly. A statistic of 0.9 (CI: 84.0%, 96.0%) for helical CT was statistically significant (P < .01) when compared with the rated statistic of 0.61 (CI: 49.0%, 74.0%) for electron-beam CT.
|
The data for both helical CT and electron-beam CT were then carefully evaluated to see if a correction factor could be applied to better approximate actual calcified plaque volume, as measured by using the absolute error between the corrected plaque volume estimate and the actual calcified plaque volume. For helical CT, this was accomplished with a factor of 0.188 for horizontal data and 0.303 for vertical data. For electron-beam CT, equivalent factors were 0.551 and 0.471. If a single factor was used for both orientations, a multiplier of 0.238 for helical CT and 0.498 for electron-beam CT was optimal (Fig 3). Despite use of correction factors, calcified plaque volumes measured with both helical CT and electron-beam CT were still statistically different from actual calcified plaque volumes. However, when a correction factor was used, helical CT was superior to electron-beam CT (P = .004) in quantifying actual calcified plaque volume.
|
DISCUSSION |
---|
Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES |
---|
Overestimation of calcium with both systems, especially helical CT, enhanced differentiation between tubes with and without calcium (Fig 3). The 50% overlap of reconstructed helical CT sections effectively doubled calcium measurements and accentuated the difference between calcified and noncalcified tubes. The sensitivity of helical CT was much greater than that of electron-beam CT because of its greater use in differentiating between normal and calcified tubes. Specificity of helical CT was increased over that of electron-beam CT because of overmeasurement of plaque-calcified volume coupled with cardiac motion artifact, especially with vertically oriented tubes (ie, the left anterior descending coronary artery). Although both systems equally depicted the large, dense, calcified plaque (Fig 4), electron-beam CT was much less sensitive in differentiating minimally calcified plaque from a normal tube (Figs 5, 6). Compared with helical CT, electron-beam CT depicted more "pseudocalcifications" (false-positive findings) in normal tubes (Fig 7). This significantly decreased the specificity of electron-beam CT and likely was affected by the decreased linear resolution and increased noise of the electron-beam CT scanner. In addition, although section thickness and voxel size were equivalent between systems, the gated electron-beam CT scanner used much lower mAs than did the helical CT scanner.
|
|
|
|
There have been efforts to apply helical CT with cardiac gating to coronary arterial calcium scoring (21–24). Carr et al (22) and Becker et al (21,23) used either retrospective or prospective cardiac gating with helical CT and found a strong correlation with results of electron-beam CT. Investigators in one study (24) even used multisection helical CT with cardiac gating. Becker et al (21) also used gated conventional CT and found that it correlated better with electron-beam CT than with helical CT. None of these investigators, however, used a known quantity of calcium for comparisons. In addition, specialized scanners not widely available at that time were used.
Other authors (25,26) have found variability similar to that in our data for CAC quantification. Yoon et al (25) found variability between electron-beam calcium scoring examinations performed on the same day of up to 28.4% in women and 43.0% in men. This measurement in variability was most pronounced in patients with lower calcium volumes.
In conclusion, a commercial ungated helical CT scanner has a higher sensitivity and specificity than does a gated electron-beam CT scanner in detecting small amounts of coronary arterial calcium. This greater sensitivity is due to volume averaging of calcium over more sections and greater estimated helical CT volumes. In contrast, electron-beam CT has lower resolution, increased noise, and a marked decrease in sensitivity and specificity in calcium detection. In addition, electron-beam CT overreads calcium in normal tubes, probably because of decreased photons available (low mAs) and lower system resolution.
This is the first reported study, to our knowledge, in which calcium detection with electron-beam CT and ungated helical CT with a phantom model have been methodically evaluated. Our results show that helical CT without cardiac gating performs better than gated electron-beam CT in calcium detection in artificial vessels, by using imaging parameters commonly applied to clinical chest CT for both systems. In addition, when data from both systems are multiplied by their appropriate correction factors, helical CT is superior to electron-beam CT in quantifying calcium content (P = .004).Practical applications: These results have important implications for coronary arterial screening with CT, either as a dedicated cardiac examination or as part of routine chest CT for other reasons. Although further work must be done to optimize rating systems and scanning protocols with ungated helical CT, the data of the current study suggest its inherent advantage and accessibility over currently commercially available electron-beam CT systems in CAC detection and quantification.
Helical CT data must be correlated with current electron-beam CT grading systems to provide a practical method to prospectively assess a patient’s risk of a major cardiac event. Different thresholds of predefined CT numbers of calcium-containing pixels need to be evaluated for sensitivity and specificity in detecting calcium and for predicting risk of future cardiac events. Finally, similar studies involving multi–detector row CT technology with cardiac gating may provide important information about a patient’s risk of cardiac ischemia and myocardial infarction.
ACKNOWLEDGMENTS |
---|
REFERENCES |
---|
Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES |
---|