| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Radiology, Teikyo University School of Medicine, 2-11-1, Kaga, Itabashi-ku, Tokyo 173-8605, Japan.
Received May 10, 2004;
accepted after revision August 19, 2004.
Address correspondence to S. Suzuki.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Vascular models with three wall thicknesses (1.0, 0.8, and 0.5 mm) and an inner diameter of approximately 4 mm were filled with contrast material of three different densities (198, 270, and 350 H) and scanned with helical CT. Three sizes of display field of view (10, 15, and 20 cm) were used. We evaluated the measurement error of the automated software, which was defined as the difference between the diameter measurement of the automated software and the true inner diameter of the vascular model. Statistical analysis involved three-way analysis of variance with repeated measures.
RESULTS. There were significant differences in the measurement errors among the three wall thicknesses of the vascular model, three densities of intravascular contrast material, and three sizes of display field of view. The overall measurement errors progressively increased with larger sizes of display field of view (p < 0.01) and with lower densities of intravascular contrast material (p < 0.001). The measurement errors tended to progressively increase as the thickness of the wall of the vascular models increased.
CONCLUSION. The accuracy of the diameter measurements by automated software for CT angiography was affected by the size of the display field of view, intravascular density of the contrast material, and wall thickness of the vessel. It is necessary to consider the effects of these factors on the diameter measurements of small arteries.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In this study, we assessed the effect of the size of the display field of view, intravascular density of contrast material, and wall thickness of the vascular model on the accuracy of in vitro diameter measurements made by automated software for CT angiography.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The physical phantom was made of nine ethylene-vinyl alcohol copolymer cylinders with the three wall thicknesses filled with contrast material of the three densities. We fixed the nine cylinders in a columnar styrene container (diameter, 5 cm) filled with salad oil (mean attenuation value ± SD, -120 ± 4 H) and placed the cylinders parallel to the central axis of the styrene container. We fixed the styrene container in a water-filled columnar polyethylene container (diameter, 10 cm) with their central axes overlapping (Fig. 1).
|
Helical CT
We performed single-detector CT on a HiSpeed Advantage SG unit (GE
Healthcare) with the central axis of the columnar container of the phantom
overlapping the axis of the gantry rotation. The following parameters were
used to obtain the scans: 1-mm collimation, pitch of 1.0, 1-sec gantry
rotation period, 48.0-cm scanning field of view, 200 mA, and 120 kV.
Measurement
Transverse sections were reconstructed every 0.5 mm with 10-, 15-, and
20-cm fields of view and a 512 x 512 pixel matrix using
"standard" as the convolution kernel. The image data were
transferred to a workstation (Advantage Workstation, version 4.0; GE
Healthcare) and analyzed by automated software (Advanced Vessel Analysis, GE
Healthcare). Ten separate mean diameter measurements were performed in all
vascular models with the three sizes of display field of view.
The diameter measurement consisted of three main steps. In the first step, we identified the segment to be analyzed by designating its starting and ending points in planar cross sections. In the second step, the software automatically detected the centerline of the vascular model between the two defined points. In the last step, diameters were automatically computed for the 10 points that we defined along the centerline. At each point, the software measured the area of the vascular model cross section in the plane orthogonal to the centerline. The mean diameter was defined as the diameter of the circle that would have the same area as the vascular model cross section.
The measurement error was defined as the signed difference between the measured diameter and true inner diameter of the vascular model. Figure 2A shows the cross section and CT attenuation profile of the models with a 0.8-mm-thick wall filled with contrast medium of intermediate density in the three sizes of display field of view. Figure 2B shows the cross section and CT attenuation profile of the models with a 0.8-mm-thick wall filled with the three densities of contrast medium in the 15-cm display field of view. Figure 2C shows those of the models with the three wall thicknesses filled with contrast medium of intermediate density in the 20-cm display field of view.
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
First, we showed the results for the effects of the size of display field of view and the intravascular density of contrast material on the overall measurement error without taking into account the wall thickness of the vascular model (Table 1). Second, we showed the results for the effect that the wall thickness of the vascular model has on the measurement error in all combinations of the three densities of intravascular contrast material and the three sizes of display field of view (Figs. 3A, 3B, and 3C).
|
|
|
|
Effect of the Size of the Display Field of View on the Overall Measurement Error
There were significant differences (p < 0.0001) in the overall
measurement errors among the three sizes of display field of view in all three
densities of intravascular contrast material.
The mean overall measurement errors for the vascular models with all three densities had positive values and progressively increased with larger sizes of display field of view (p < 0.01), as shown in Table 1. The mean overall measurement errors were more than 0.5 mm with the 15- and 20-cm fields of view when the density of the contrast material was not more than 270 H.
Effect of the Density of Intravascular Contrast Material on the Overall Measurement Error
There were significant differences (p < 0.0001) in the overall
measurement errors among the three densities of intravascular contrast
material in the three sizes of display field of view.
The mean overall measurement errors in all three sizes of display field of view progressively increased with lower densities of intravascular contrast material (p < 0.0001), as shown in Table 1.
Effect of Wall Thickness of Vascular Model on the Measurement Error
There were significant differences in the measurement errors among the
three wall thicknesses of the vascular models in all combinations of the three
sizes of display field of view and three densities of intravascular contrast
material (p < 0.0001). The measurement errors were a positive
value for all except the models with a 0.5-mm-thick wall filled with contrast
material of 350 H in a 10-cm display field of view (Figs.
3A,
3B, and
3C).
In the 10-cm display field of view, the mean measurement errors for the vascular models with all three densities progressively increased with the wall thickness of the vascular model (p < 0.0001) except for those with a 0.8- or 1.0-mm-thick wall filled with 270-H contrast material, as shown in Figure 3A. However, in the models with 350-H contrast material, there was no significant difference between the absolute values of the mean measurement error for the models with a 0.5- or 0.8-mm-thick wall (p = 0.487), because underestimation occurred for the models with a 0.5-mm-thick wall. The difference in the mean diameter measurements for the models with a 0.5- versus a 1.0-mm-thick wall was less than 0.4 mm for all three densities of intravascular contrast material. The mean measurement error for models of all three wall thicknesses was less than 0.2 mm when the density of the contrast material was 350 H.
In the 15-cm field of view, the mean measurement errors for the vascular models with contrast material of all three densities progressively increased with the wall thickness of the vascular model (p < 0.0005) except for the models with a 0.8- or 1.0-mm-thick wall filled with 350-H intravascular contrast material, as shown in Figure 3B. The differences in the mean measurement diameters between models with a 0.5- or 1.0-mm-thick wall were 0.20 mm for models with 270-H contrast material and 0.60 mm for those with 198-H contrast material. The mean measurement error for models of all three wall thicknesses was approximately 0.3 mm or less when the density of the contrast material was 350 H.
In the 20-cm field of view, the mean measurement errors for the vascular models with contrast material of all three densities progressively increased with the wall thickness of the vascular model (p < 0.0001), as shown in Figure 3C. The differences in the mean diameter measurements between models with a 0.5- or 1.0-mm-thick wall were 0.53 mm for those with 270-H contrast material and 0.86 mm for those with 198-H contrast material.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In this study, the mean overall measurement errors tended to increase progressively with the larger sizes of display field of view and were more than 0.5 mm with 15- and 20-cm fields of view when the density of the contrast material was 198 or 270 H. Such measurement errors may exert a significant practical impact on the estimation of the diameter of small arteries. The measurement error is probably affected by the slope of the profile curves corresponding to the vascular inner contour, whichever measuring method may be used. As shown in Figure 2A, the slope of the profile curve became smoother with a smaller size of display field of view. Therefore, we believe that small size of the display field of view—in other words, small pixel size—is necessary for accurate measurement of vascular diameter.
The overall measurement errors tended to decrease progressively with higher densities of intravascular contrast material in all sizes of display field of view as shown on Table 1. Thus, adequate contrast enhancement is thought to be necessary for accurate measurement of vascular diameter with this automated software. The optimal value for arterial contrast enhancement for CT angiography has been assessed by some investigators [15-18]. The results were consistent with the findings of Suzuki et al. [15] using automated software. They used acryl pipes filled with contrast medium of three different densities (460, 350, and 210 H) as the vascular models. They reported that measurement errors progressively decreased with higher densities of intravascular contrast medium. On the other hand, Claves et al. [17] reported that a density higher than 250 H increased error without using automated software. The difference between the optimal attenuation of contrast material reported by Claves et al. and our results may arise from differences in the vascular models and measurement methods used. The optimal value for arterial contrast enhancement may differ even among automated software depending on the methods used to calculate the centerline and recognize the vascular contour. Therefore, it is necessary to evaluate the optimal attentuation of contrast medium for each method.
In this study, the wall thickness of the vascular models affected the accuracy of the diameter measurements, as shown in Figures 3A, 3B, and 3C. In 20-cm display field of view, the differences in the mean overall measurement diameters between the models with a 0.5- or 1.0-mm-thick wall were 0.53 mm for those with 270-H contrast material and 0.86 mm for those with 198-H contrast material. These results may have a significant impact on the accuracy of stenosis measurement, especially for small arteries, because atherosclerotic stenotic portions of arteries generally have wall thickening with plaque. The difference in measuring methods may affect the result. However, as shown on Figure 2C, wall thickness affected the slope of the profile curves corresponding to the vascular inner contour. Therefore, wall thickness will probably affect the accuracy of diameter measurements of vessels even if other measuring methods are used. On the other hand, the mean measurement error for models of all three wall thicknesses that were filled with contrast material of 350 H were less than 0.2 mm with a 10-cm field of view, and this level of error was thought to be expected and relatively acceptable.
The CT attenuation of the vascular wall will affect the accuracy too. The wall density of the vascular model in this study was consistent with previously reported findings [19, 20]. Becker et al. [19] found that the CT attenuation (mean ± SD) of noncalcified plaque was 76 ± 35 H, whereas Estes et al. [20] reported that the CT attenuation (mean ± SD) of fibrous and lipid plaque was 90 ± 24 H and 39 ± 12 H, respectively.
This study has some limitations. First, we used single-detector CT, although MDCT has begun to replace single-detector CT [19, 21-24]. MDCT is superior to single-detector CT mainly in the spatial resolution in the direction of the z-axis and in the temporal resolution. However, the diameter measurements in this study were performed essentially on the xy plane because the vascular models were placed parallel to the z-axis. We intentionally used single-detector CT, considering that measurement accuracy in the xy plane would be affected by the "corn beam artifact" peculiar to MDCT. If the artifact problem is solved by refinements in the equipment, this result will be applicable also to MDCT. Second, the phantom model cylinder wall had a homogeneous density (76 H), whereas in vivo atherosclerotic plaques are often heterogeneous. That difference may limit the extension of our findings in phantom models to humans. Third, we used fields of views of 10, 15, and 20 cm, although in clinical practice larger ones are used typically. We consider that the measurement error with a larger field of view will become larger than that with a 20-cm field of view.
Additional studies are required to evaluate the effects of the differences in software, noise, tube voltages, and shape of the atherosclerotic plaque on the measurement accuracy.
In conclusion, the accuracy of diameter measurement using automated software for CT angiography is affected by the size of the display field of view, intravascular density of the contrast material, and wall thickness of the vessel. It is necessary to consider the effects of these factors in the diameter measurement, especially for small arteries.
Acknowledgments
We thank TERUMO Co., Ltd., for generously providing tubes made of
ethylene-vinyl alcohol copolymer and measuring the vascular models by the
profile projector, and we thank Fumiaki Harada and Takeshi Kawai for their
advice and expertise.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  J. P. Villablanca, F. J. Rodriguez, T. Stockman, S. Dahliwal, M. Omura, S. Hazany, and J. Sayre MDCT Angiography for Detection and Quantification of Small Intracranial Arteries: Comparison with Conventional Catheter Angiography Am. J. Roentgenol., February 1, 2007; 188(2): 593 - 602. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
