| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Technical Developments |
1 From the Department of Radiology, Stanford University School of Medicine, Calif (N.F., M.P., G.D.R.); and First Department of Internal Medicine, Osaka City University, Japan (Y.K.). From the 1999 RSNA scientific assembly. Received July 17, 2001; revision requested September 10; revision received February 20, 2002; accepted February 15. Address correspondence to N.F., Department of Cardiovascular Science and Medicine, Chiba University Graduate School of Medicine (M4), 1-8-1 Inohana, Chuo-ku, Chiba 260-8670, Japan (e-mail: nobusada@ma.kcom.ne.jp).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
© RSNA, 2002
Index terms: Computed tomography (CT), electron beam, 54.12112, 54.12116, 54.12117, 54.12118, 54.12119 • Computed tomography (CT), image processing • Coronary vessels, CT, 54.1211, 54.12116
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
To date, coronary artery luminal caliber has been qualitatively assessed with three-dimensional visualization techniques such as shaded surface rendering performed with a threshold of 80–100 HU (3,4), volume rendering (5), curved multiplanar reconstruction (6), and coronary CT angiography for the purpose of comparing results with those at cine coronary angiography. Comparisons of results with either different thresholds and rendering techniques or different quantitative analyses of threshold selection have not been performed, to our knowledge. It is reasonable to expect that the optimal threshold level for shaded surface rendering or the enhancement setting for volume rendering would depend on the magnitude of arterial enhancement, which has been documented (7) at CT angiography to vary substantially in other vascular locations. As a result, we hypothesized that a single threshold setting or set of enhancement settings could not be used reliably to measure the luminal diameter of the proximal and middle coronary artery tree with electron-beam CT. The purpose of our study was to determine the most accurate method for measuring the dimensions of the coronary arteries that is independent of inter- and intrapatient variations in arterial enhancement.
|
Materials and Methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
Cine Coronary Angiography
Cine coronary angiography was performed with a percutaneous femoral approach. Sublingual nitroglycerin (Nitrostat [0.4 mg]; Parke-Davis, Morris Plains, NJ) was given 2–3 minutes before the contrast agent (iopamidol [300 mg/mL], Iopamion 300; Schering, Berlin, Germany) was injected to minimize the effect of varying vasomotor tone on vessel lumen diameters. Catheters of known diameters were used for calibration. Multiple projections including cranial and caudal angulated views were obtained for all patients.
The 30° right anterior oblique projection was used to assess the left coronary artery and its branches. The 60° left anterior oblique projection was used to assess the right coronary artery (RCA).
Quantitative coronary angiography was performed off line with a computerized edge-detection program (QCA plus; Sanders Data Systems, Palo Alto, Calif) (8), and the findings were analyzed by one cardiologist (Y.K.), who was blinded to the results of electron-beam CT (Fig 1). Diameters were recorded of the proximal left main coronary artery and its distal end just above the bifurcation into the left anterior descending (LAD) branch, the left circumflex branch, and a point between these two sites. Diameters were also measured for the LAD branch, left circumflex branch, and RCA at each of their proximal ends and at two to four equally spaced points between the proximal end and the first arterial side branch.
|
When vessel boundaries in areas of noise or vessel crossings could not be resolved with the computer algorithm, manual editing of short boundary segments was performed with the light pen to correct the computer-generated boundary. At no time did the length of a manually entered margin exceed 20% of the total length of the quantitated segment. After the starting and ending points of the segment were indicated with the light pen,the mean diameter of the segment was computed from perpendicular lines constructed through the length of a computer-generated centerline. The mean diameter of the segment was then used for analysis. At each measurement point, the distance was measured from the ostium of the left main coronary artery or from that of the RCA.
Electron-Beam CT Measurements
Fast scanning speeds are achieved with electron-beam CT as a result of sweeping of a steered electron beam on a fixed tungsten target ring, which provides a moving x-ray source without mechanical motion. The 100-msec modes were used for high-resolution cross-sectional imaging in a step-volume scan mode with electrocardiographic gating.
Electron-beam CT was coupled with an intravenous injection of 160 mL of the iodinated contrast medium, which was performed during breath holding, that was directed over 60–70 mm of the proximal and middle portions of the coronary arteries (1.5-mm collimation, 1.0-mm table incrementation, and triggering at 80% of the R-R interval). The patients received nitroglycerin before electron-beam CT, as in the cine angiographic studies.
Electron-beam CT data were transferred to a workstation (Advantage-Windows; GE Medical Systems, Milwaukee, Wis), and 18–27 measurement points in the coronary tree for each patient were defined relative to identifiable branches. Diameters were measured perpendicular to the median centerline of the vessel by using one of two thresholds: full width at 80, 100, 120, and 140 HU or full width at 50%, 60%, 70%, and 80% of the maximum nonthreshold line density profile (LDP). The latter is called the adaptive method.
Diameters were analyzed by one cardiologist (N.F.), who was blinded to the results of coronary angiography with the workstation program. For both methods, oblique planar reformations were generated that were parallel to the long axis of the vessel; they approximated the oblique projections analyzed with cine coronary angiography. By using the vessel origins and major branch points as landmarks, the corresponding measurement loci that were identified on the coronary angiograms were identified on the electron-beam CT scans.
The fixed-threshold measurement methods simulate shaded surface rendering. A threshold value was applied to the entire data set to reduce the gray scale to black and white. This was accomplished by setting the display window width to zero and the display window level to 80, 100, 120, or 140 HU. The vessel diameters were measured at each point along the same direction as were measured on the coronary angiograms (Fig 2).
|
|
|
|
|
Thus, nine diameter values were recorded for each measurement point on the electron-beam CT scans: four with the different fixed-threshold methods, four with the different LDP methods, and one with the combined LDP method.
With all methods, the peak CT value at each measurement point and the angle of each vessel diameter measurement relative to the longitudinal (through-plane) scan axis were determined. The angle of the vessel diameter measurement relative to the through-plane scan axis was used as an index for assessment of measurement accuracy relative to vessel orientation (9). The peak CT value was used as an index for assessment of arterial enhancement.
Statistical Analysis
Absolute errors were calculated for each method as the differences in the diameter measurements at electron-beam CT minus the corresponding diameter measurements at coronary angiography. Absolute errors for each measurement method were classified on the basis of vessel type and were further stratified on the basis of distance from the vessel ostium. Also, absolute errors were correlated for vessel orientation relative to the electron-beam CT gantry (the angle of the vessel diameter measurement relative to the through-plane scan axis), the vessel types triggered at end diastole, arterial enhancement (peak CT values), vessel size, and the distance from the vessel ostium. As an index of vessel size, we used both quantitative measurements of diameter at coronary angiography and the averages of diameter measurements at electron-beam CT and coronary angiography, according to the recommendations of Bland and Altman (10) regarding the comparison of two measurement methods.
Statistical significance was assessed by means of (a) the paired t test to compare the means of the absolute errors between the 70% LDP and the combined LDP measurement methods, (b) the unpaired t test to compare the means of the angles of the LAD branch measurements at points less than 30 mm from the ostium and the RCA measurements at distances of 30 mm or more from the ostium, and (c) the Hartley test to compare the variances of the absolute errors among the four fixed-threshold methods, the four LDP methods, and, finally, all eight methods.
By means of the F distribution test, variances in vessel diameter were calculated for the following comparisons between every measurement method: (a) RCA at distances of 30 mm or more from the ostium versus LAD branch at distances of less than 30 mm from the ostium, (b) 70% LDP measurements versus combined LDP measurements for every site along the coronary arteries, and (c) 70% LDP measurements for angles less than 45° versus 45° or more.
When analyzing electron-beam CT data, we adopted 0.07 mm as the limit below which any discrepancy in measurements could be accounted for by errors in clinical judgment; this value was reported to be the accuracy of measurements at coronary angiography (11). Furthermore, we considered this limit to be 1.0 mm, on the basis of reports of use of spatial profile curves for MR angiograms (12); this value represents the minimum 2 SDs required by Bland and Altman (10).
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
Table 1 lists the diameters obtained at coronary angiography with the distances from the ostium of the left main coronary artery or the ostium of the RCA. The diameters ranged from 1.46 to 6.09 mm (mean ± SD, 3.57 mm ± 0.84). Distances from the ostium ranged from 0 to 79.77 mm (27.9 mm ± 19.04).
|
|
Points that represented the peak CT values were derived from measurements made on the LDP curve perpendicular to the long axis of the coronary arteries (LDP method) or along the short axis of images perpendicular to the long axis (threshold method). All peak CT value points were located at approximately the center of the vessel lumen and not at the edge; this finding suggests a lack of calcium in the vessel wall. Peak CT values for each site ranged from 44 to 591 HU (228 HU ± 83) for the LDP methods and from 70 to 588 HU (227 HU ± 80) for the fixed-threshold methods. The peak CT values with fixed-threshold measurements correlated with those with LDP measurements (R2 = 0.90, P < .01) (data not shown).
Mean and SD of Absolute Errors among the Eight Methods
Table 3 summarizes the means and SDs of the absolute errors for each of the eight measurement methods, listed according to vessel type and distance from the vessel ostium.
|
Variances of the absolute errors for all vessels were not significantly different among the fixed-threshold methods (SDs were 1.60, 1.60, 1.62, and 1.59, respectively, for 80, 100, 120, and 140 HU) or among the LDP methods (SDs were 1.07, 1.01, 0.98, and 0.98, respectively, for 50%, 60%, 70%, and 80%). However, significant differences in variances were found in the comparisons of the eight methods.
To determine the influence of vessel orientation, we compared variances in measurements from portions of the RCA at distances of 30 mm or more from the ostium with those from portions of the LAD branch at distances of less than 30 mm for each measurement method. SDs for portions of the RCA at 30 mm or more from the ostium were 0.92, 0.94, 0.96, and 1.09 mm for fixed-threshold measurements with 80, 100, 120, and 140 HU, respectively, and were 0.76, 0.74, 0.76, and 0.74 mm for 50%, 60%, 70%, and 80% LDP measurements, respectively. The SDs of the LAD branch at distances of less than 30 mm were 1.67, 1.76, 1.85, and 1.66 mm for the fixed thresholds, and 1.13, 1.09, 1.09, and 1.11 mm for the LDP measurements. Differences between variances for RCA and LAD branch measurements were statistically significant.
The mean and SD of the absolute errors for all vessels were 0.01 and 0.80 mm, respectively, with the combined LDP measurement method. Based on the means of the absolute errors with 70% LDP measurements, results with the combined LDP method were significantly different from those for the left main coronary artery (-0.46 vs 0.46), portions of the LAD branch at distances of less than 30 mm from the ostium (-0.26 vs -0.06), portions of the LAD branch at distances of 30 mm or more from the ostium (0.62 vs 0.24), and all LAD branch sites (0.26 vs 0.12), even though the means for the absolute errors for all vessels were essentially zero (-0.03 vs 0.01). Variances of the absolute errors with 70% LDP measurements compared with combined LDP measurements were significantly different only when all vessels were compared as a group (SD, 0.98 vs 0.80; P < .01). The SD of the absolute errors with the combined LDP method (0.80) is smaller than the SDs with any of the eight methods used in our study. Three percent and 18% of vessels were measured with greater than 2 mm of error with 70% LDP and 100-HU threshold measurements, respectively (data not shown).
Figure 4 is a scatterplot of the absolute errors versus the angles of measurement with the 70% LDP method; the mean (-0.03) and SD (0.98) of the absolute errors for all vessels were the closest to zero and smallest, respectively, for the eight methods used in our study. We divided all 179 sites into two groups: those with angles of diameter measurement relative to the through plane of less than 45° (103 sites) or of 45° or more (76 sites). The SD of the absolute errors for the less than 45° group (SD of 1.07) was significantly larger than that for the 45° or more group (SD of 0.084) (P < .05).
|
|
|
|
|
|
|
|
|
Discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
We compared quantitatively the luminal diameter of the coronary artery by performing imaging with four fixed-threshold, four relative LDP, and combined LDP methods, and we considered several factors that could influence luminal caliber. These factors included vessel orientation, vessel type triggered at end diastole, arterial enhancement, vessel diameter, and distance from the ostium.
Interpretation of Means and SDs
We classified the 179 sites into four vessel types and stratified them according to distance from the ostium (<30 vs 
30 mm). We were able to identify two measurement techniques with which the means of the absolute errors for all vessels were essentially zero (70% LDP [mean, -0.03] and 100-HU fixed threshold [mean, 0.1]). In each vessel type stratified by distance from the ostium, however, the means of the absolute errors were different. Thus, it was possible for the means of the absolute errors to approximate zero when the site of interest was set. In that case, either an appropriate height was selected for measuring LDP width or an appropriate threshold was selected for use with the fixed-threshold method.
Bland and Altman (10) suggested that differences of 2 SDs or less between measurements with an experimental method and those with an established method would not be important in clinical interpretation and either method could be used interchangeably . In this study, we adopted 0.07 mm as the limit below which any discrepancy in measurements could be accounted for by errors in clinical judgment. This is what we then referred to as the accuracy of coronary angiography measurements. With the 70% LDP, 100-HU fixed threshold, or combined LDP methods, we were able to set the means of the absolute errors for all vessels to approximately zero. We concluded that the fixed-threshold method had significantly greater variation than did either of the LDP methods. By accepting 2 SDs of the LDP measurements (1.96–2.14 mm), however, results with both the fixed-threshold (3.18–3.24 mm) and combined LDP (1.6 mm) methods represented unacceptable errors for clinical purposes.
Influence of Vessel Orientation
Spatial resolution in the transverse plane of the electron-beam CT scanner is seven line pairs per centimeter (13), which results in an in-plane resolution of 0.7. In contrast, longitudinal spatial resolution (z axis) was at best 1.5 mm. Therefore, coronary artery orientation should have a profound influence on the accuracy of vessel sizing.
For coronary angiography measurements, we used the 30° right anterior oblique projection to assess the left coronary artery and the 60° left anterior oblique projection to assess the RCA. Measurements with electron-beam CT were made differently from those with coronary angiography. For these, we created a longitudinal view and an orthogonal view of each vessel and measured the LDP on the orthogonal view along the same axis as the corresponding measurement with coronary angiography. Thus, measurements with electron-beam CT of all portions of the left main coronary artery, the proximal and intermediate portions of the LAD branch, and the proximal portions of the RCA were highly influenced by the longitudinal spatial resolution.
In the scatterplot of the absolute errors with the 70% LDP method with respect to the angles of measurement of vessel diameter relative to the through-plane scan axis, the group with angles of measurement that were less than 45° had significantly larger variances in the absolute errors than did the group with angles that were 45° or more. Therefore, we may conclude that the less influenced the measurements are by through-plane spatial resolution, the smaller the variances of their absolute errors.
With respect to the RCA at distances of 30 mm or more from the ostium, the angles at which the vessel diameter was measured relative to the longitudinal axis were significantly larger than those for the LAD branch at distances of less than 30 mm from the ostium. Thus, measurements of vessel diameter of the RCA at distances of 30 mm or more from the ostium were less influenced by longitudinal spatial resolution than were those of the LAD branch at distances of less than 30 mm. Not surprisingly, variances of the RCA measurements at distances of 30 mm or more were significantly smaller than those of the LAD branch at distances of less than 30 mm with every method, even though atrial movement may result in motion artifact for the RCA.
Influence of Arterial Enhancement
Many parameters that relate to patient physiology and anatomy, type of contrast material, and method of delivery can influence arterial enhancement. In addition to these influences, there may be substantial differences in the degree of arterial enhancement among images acquired in the same patient.
Single-threshold settings or a set of enhancement settings are used to make three-dimensional images of coronary arteries with shaded surface rendering or volume rendering techniques applied to the entire data set. If a fixed threshold or a fixed enhancement curve is adopted, however, vessel diameter in one part may be overestimated and that in another part may be underestimated as a result of differences in arterial enhancement or partial volume effects (13). In contrast to the fixed-threshold measurements, there was no correlation between the absolute errors and the peak CT values with LDP measurements, which essentially eliminated the influence of differences in vessel enhancement with the latter technique.
Influence of Vessel Size
There was a significant correlation between the absolute errors for diameters with the LDP measurements versus those with coronary angiography. We attempted to revise the quantitative evaluation by adopting different settings from background CT values with regard to the percentage of maximum height to the maximum CT value (eg, 70% LDP), depending on measurements at coronary angiography. With the resulting combined LDP method, the mean and SD of the absolute errors for diameters of all vessels were the best in this study. Also, vessel sizes with combined LDP measurements were less influenced by those with coronary angiography than were any of the other LDP methods.
In an MR angiographic study with use of the spatial profile curve (12), the minimum value of 2 SDs was 1.96 (0.98 x 2). In comparison, the minimum value of 2 SDs in our study was 1.0 mm, which is better. In their study, however, assessments were restricted to the proximal and intermediate portions of the LAD branch and the RCA.
Study Limitation
Calcification frequently occurs in atherosclerotic plaque. In none of our study subjects was coronary artery calcium visible. Thus, the effect of coronary calcifications on measurement accuracy could not be assessed.
In conclusion, variation with fixed-threshold methods is significantly greater than that with adaptive methods because of substantial variations in arterial enhancement. Therefore, we conclude that measurements of coronary artery dimensions should not be made with fixed-threshold values. Findings based on the LDP showed reduced influence of vessel enhancement on measurement accuracy. Furthermore, by combining three percentages as cutoffs with the LDP method, on the basis of diameters measured with coronary angiography, we succeeded in reducing the influence of diameters determined with the latter.
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, N.F., G.D.R.; study concepts, N.F.; study design, N.F., G.D.R.; literature research, N.F.; clinical studies, G.D.R.; experimental studies, N.F.; data acquisition, G.D.R.; data analysis/interpretation, N.F., Y.K.; statistical analysis, N.F.; manuscript preparation, N.F.; manuscript definition of intellectual content, G.D.R.; manuscript editing and revision/review, M.P., G.D.R.; manuscript final version approval, all authors.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. E. Clouse, A. Sabir, C.-S. Yam, N. Yoshimura, S. Lin, F. Welty, P. Martinez-Clark, and V. Raptopoulos Measuring Noncalcified Coronary Atherosclerotic Plaque Using Voxel Analysis with MDCT Angiography: A Pilot Clinical Study Am. J. Roentgenol., June 1, 2008; 190(6): 1553 - 1560. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. T. Boll, E. M. Merkle, E. K. Paulson, and T. R. Fleiter Coronary Stent Patency: Dual-Energy Multidetector CT Assessment in a Pilot Study with Anthropomorphic Phantom Radiology, June 1, 2008; 247(3): 687 - 695. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. Hong, G. S. Chrysant, P. K. Woodard, and K. T. Bae Coronary Artery Stent Patency Assessed with In-Stent Contrast Enhancement Measured at Multi-Detector Row CT Angiography: Initial Experience Radiology, October 1, 2004; 233(1): 286 - 291. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
U. J. Schoepf, C. R. Becker, B. M. Ohnesorge, and E. K. Yucel CT of Coronary Artery Disease Radiology, July 1, 2004; 232(1): 18 - 37. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. T. Bae, H. Q. Tran, and J. P. Heiken Uniform Vascular Contrast Enhancement and Reduced Contrast Medium Volume Achieved by Using Exponentially Decelerated Contrast Material Injection Method Radiology, June 1, 2004; 231(3): 732 - 736. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
