Background and Purpose No previous study has compared the reliability of carotid artery measurement provided by axial images, shaded surface display (SSD), and maximum intensity projection (MIP).
Methods Helical CT and conventional angiography were performed prospectively in 20 patients with atherosclerotic stenosis of the internal carotid artery. Stenosis measurement was performed in a blinded fashion on angiography and CT by two independent examiners. Calcifications were segmented when they were located far enough from the vascular lumen. SSD and MIP were systematically performed for each carotid bifurcation. We measured stenosis using conventional angiography as standard and the different CT reconstructions (axial images, SSD, and MIP) by comparing the stenosis diameter at its narrowest point to the normal internal carotid artery. The degree of stenosis was classified into six groups: no stenosis, mild stenosis (<30%), moderate stenosis (30% to 70%), severe stenosis (>70%), near occlusion, and occlusion (100%). No measurement was made in cases of normal artery, near occlusion, and occlusion.
Results Correlations between angiography and the three types of reconstruction were very good. Axial sections correctly classified the carotid arteries in 95% of cases. In 10 carotid arteries, stenosis was not assessable by SSD and MIP because of calcifications. In the remaining carotid arteries, MIP correctly classified the degree of stenosis in 96% of cases, whereas SSD misclassified 21% of cases.
Conclusions Our study showed that axial images provide a reliable evaluation of carotid artery stenosis. Calcifications are limiting factors in SSD or MIP. When atherosclerotic plaques are not calcified, MIP reconstructions provide a more reliable measurement of the vascular lumen than SSD.
The North American Symptomatic Carotid Endarterectomy Trial (NASCET)1 has shown the beneficial effect of endarterectomy in patients receiving optimal medical treatment for symptomatic high-grade stenosis of the ICA (>70% in diameter) with a 17% reduction in the risk of cerebral infarction over a 2-year period compared with patients who did not undergo surgery. Although angiography remains the gold standard to evaluate ICA stenosis, it carries an iatrogenic risk mainly because of the use of arterial catheters.2 Helical acquisition CT is a new technique combining continuous gantry rotation with simultaneous displacement of the examination table throughout the acquisition.3 The advantage of this type of volumetric acquisition relies on its ability to obtain 30 scans during the vascular phase of a contrast-enhanced study, allowing accurate 3-D vascular images in any projection.4 5
Direct comparison with conventional angiography is not possible on transaxial sections. Two kinds of 3-D display techniques are currently available: SSD and MIP.4 6 7 SSDs are generated by selecting a CT value above a defined density threshold. The threshold must be carefully chosen on the basis of contrast material attenuation in the area of interest. Then a surface is calculated as if the structure is illuminated by a light source to achieve the 3-D impression through shading; however, this technique provides no information on densities. MIP is a volume-rendering method that is widely used to create MR angiographic displays. The intensity of each pixel in the resulting image is the maximum intensity encountered along parallel rays traced through the volume and projected in the desired viewing direction. With this technique the depth information is lost, but the density information is retained.
The purpose of this study was to compare axial images, SSD, and MIP with conventional angiography in the evaluation of ICA stenosis.
Subjects and Methods
From September 1993 to January 1995, 20 patients with carotid bifurcation atheroma were evaluated by helical CT and conventional angiography. The study group consisted of 6 women and 14 men aged 39 to 85 years (median age, 63 years) suffering from transient ischemic attacks or minor ischemic stokes who had a hemodynamically significant ICA stenosis on color and duplex Doppler sonography. Patients had no contraindication to intravenous injection of contrast material. Conventional angiography was performed with the use of digital subtraction angiography (Integris V3000, Philips) through a femoral artery approach. Selective carotid angiograms were performed in different planes to evaluate the exact degree of ICA stenosis. Angiography was performed after CT (median delay, 9 days; range, 1 to 16 days) in 10 patients and before CT (median delay, 13 days; range, 1 to 27 days) in 10 patients.
CT angiograms were obtained on a Siemens Somatom Plus S system. Patients were placed in the supine position so that they were as comfortable as possible. The head was tilted back to avoid dental hardware. A lateral scanogram was acquired with the shoulders placed as low as possible. We performed 5-mm sections without contrast injection from C2 to C6 to determine the level of carotid bifurcations and to detect the presence of calcifications. Continuous data were acquired during a scan time of 32 seconds and started 3 cm below the carotid bifurcation. Patients were instructed to breathe quietly without swallowing during the scanning period. A total volume of 90 mL of nonionic contrast material (Omnipaque 300, Nycomed) was injected at 3 mL/s into an antecubital vein with a 20-second scan delay after the start of the contrast bolus, a 2-mm collimation, and a table speed of 3 mm/s (total coverage, 90 mm). Axial images were reconstructed at 1-mm increments (total number of sections, 90).
3-D Reconstructions (SSD and MIP)
CT data were transferred to an independent Siemens workstation. A region of interest was drawn to remove the internal jugular vein, collateral branches of the external carotid artery, and osteocartilaginous structures. Calcifications of the ICA were excluded when the segmentation did not contact the adjacent arterial lumen. SSD and MIP were carried out in both carotid arteries. In SSD, the lower threshold was determined by measuring the attenuation value of the intraluminal contrast material at its narrowest point or immediately above and below it. SSD and MIP reconstructions were obtained in multiple projections to determine sites of maximal stenosis.
The three separate sets of CT images (axial, SSD, and MIP) of each patient and the conventional angiograms were examined in a fully randomized order by two independent examiners (X.L. and J.P.P.). We did not perform any measurements when ICAs were normal or occluded. When the residual lumen was too tight to be measured with accuracy or when the distal diameter of the ICA was smaller than that of the external carotid artery, the stenosis was classified as near occlusion and no measurement was performed. In the other cases, the percentage of stenosis was determined by comparing the narrowest diameter of the residual lumen with the diameter of the ICA beyond the bulb. The measurements were made with the use of a large zoom and a computer caliper. Each carotid artery was then assigned to one of six categories: no stenosis, mild stenosis (<30%), moderate stenosis (30% to 70%), severe stenosis (>70%), near occlusion, and occlusion. Cases leading to disagreement between both observers were reviewed by both observers together to reach a consensus.
We performed all comparisons using percent stenosis and stenosis categorization as dependent variables. The first step of the analysis consisted of an evaluation of the level of interobserver agreement for each set of CT images by means of the Spearman’s rank-order correlation test (percent stenosis) and the κ statistic12 13 (stenosis categorization). The second step consisted of a comparison between the three CT measurements and conventional angiography with the use of the same statistical tests.
Some minor problems due to the scanning technique were encountered but did not prevent measurements of carotid stenosis. Motion artifacts were observed in 3 patients and dental hardware in another 2 patients without any effect on the analysis of the carotid bifurcation. Carotid arteries were completely opacified in 18 of 20 patients by a 20-second delay between the contrast injection and the scan. In the remaining patients, carotid stenosis was assessable on CT angiograms despite mild opacification of the common carotid arteries. The positions of ICAs and external carotid arteries were accurately determined on axial images by the two examiners in all cases. The time required to remove unwanted structures was approximately 10 minutes per carotid artery and 20 minutes in 6 patients with severe calcifications (Fig 1⇓). One carotid artery was excluded from the analysis because the helical CT showed that the angiographic view was not appropriate to define the maximal degree of stenosis (Fig 2⇓). In three cases a consensus reading was performed because differences of 10% or more between the two examiners were observed.
Interobserver agreement for the three sets of CT images and the angiograms was judged as good to very good for percent stenosis (angiography: rs=.987, P<.0001; axial sections: rs=.994, P<.0001; SSD: rs=.965, P<.0001; MIP: rs=.996, P<.0001) as well as stenosis categorization (angiography: κ=1.00, P<.0001; axial sections: κ=1.00, P<.0001; SSD: κ=0.96, P<.0001; MIP: κ=0.96; P<.0001).
Comparisons between the three CT technique measurements are shown in Fig 3⇓ (correlation) and the Table⇓ (categorization). Correlations between angiography and CT reconstructions were good (axial sections: rs=.935, P<.0001; SSD: rs=.809, P<.0001; MIP: rs=.856, P<.0001) (Fig 4⇓). However, SSD led to an underestimation of the stenosis of more than 10% in four cases. Axial images and MIP overestimated the stenosis by more than 10% in six cases and one case, respectively (Fig 3⇓).
All normal and occluded carotid arteries and all stenoses with near occlusion were correctly classified by the three CT reconstructions (axial sections, SSD, and MIP) (Table⇑ and Fig 5⇓). In two cases with near occlusion, the stenosis appeared to be opened up and more elongated on SSDs than on conventional angiograms (Fig 6⇓).
Axial images correctly classified 37 of 39 carotid arteries (95%) (κ=0.94; P<.0001), despite intramural calcification present in all cases with severe and moderate stenosis (Fig 7⇓). Stenosis was not assessable in 10 carotid arteries with calcification with the use of SSD and MIP despite multiple projections. In the remaining 29 carotid arteries, 28 stenoses were correctly classified by MIP (96%) (κ=0.96, P<.0001) and only 23 by SSD technique (79%) (κ=0.77, P<.0001). The poor result of the SSD technique was related to an underestimation of the degree of stenosis (Table⇑).
This study shows that axial images and MIP reconstructions provided a measurement of ICA stenosis close to that of angiography. Arterial wall calcifications, which are very common in atherosclerotic plaques, were the main limiting factor in SSD and MIP. Furthermore, even in the absence of calcification, SSD frequently underestimated the degree of stenosis and did not provide a reliable evaluation of carotid stenosis.
The reliability of carotid stenosis measurement depends on the scanning technique. The protocol we used was very similar to that of previous studies except in the study of Castillo,8 who used a 5-mm collimation, a small amount of contrast material (60 mL), and 3-mm increments; this probably accounted for the poor reliability of Castillo’s results, with only 50% of stenoses being correctly classified. Other studies with thinner sections (2 to 3 mm), higher amounts of contrast material (75 to 120 mL), and lower increments (1 to 2 mm) showed a higher agreement between CT angiography and conventional angiography.9 10 11 14
A comparison of the different CT images for the evaluation of ICA stenosis has not been previously estimated. Axial sections provided the most reliable carotid stenosis measurement for different reasons. First, calcifications were not a limiting factor, and axial sections provided stenosis imaging in all cases without requiring subsequent postprocessing such as calcification subtraction. Second, the stenosis at its narrowest point can be measured with much more accuracy than on angiograms when the residual lumen is elongated in the axial plane. Third, the accuracy of stenosis measurement on axial images depends on the scan plane, which has to be perpendicular to the carotid artery. In the present study it was not a limiting factor because the ICA originated roughly perpendicular to the scan plane in all cases. However, this limitation suggests that axial sections must not be used as the sole means of measuring lesions. Finally, in very severe stenosis the decrease in distal ICA diameter precludes an accurate measurement of the degree of stenosis by the NASCET method.15 This usually leads to a classification of cases with ICA diameter lower than that of the external carotid artery as near occlusion. From this point of view, confusion between ICAs and external carotid arteries may be a theoretical limiting factor, particularly in cases with near occlusion. However, in the present study both arteries were easily identified.
The advantage of SSD is the possibility of moving and adjusting for tilt, thereby providing for the opening of the tortuosities; this may provide more accurate images than conventional angiograms in some cases. However, SSD frequently underestimated the lesions in this study, and this might be due to arbitrary selection of the lower threshold.6 In the study of Dillon et al,9 the determination of the threshold with the use of densities measured in the adjacent soft areas also led to an underestimation of the degree of stenosis. Another limiting factor is the segmentation of calcification, which requires much longer postprocessing time and cannot always be performed when calcification contacts the vascular lumen. Dillon et al9 used a very similar technique of segmentation, whereas Schwartz et al11 obtained optimal calcification exclusion with a Sun workstation and customized segmentation software. The calcification was identified in the region of interest on the unenhanced axial CT scans, and was secondly removed from enhancing arterial structures with the use of a segmentation program. This requires an additional 30 minutes of postprocessing time for each carotid artery, but it improves agreement between SSD and conventional angiography.
MIP provided angiogram-like images and correctly classified most stenoses in our study. Calcifications are easily detected on MIP but can artificially shrink the diameter of the artery. This might theoretically be avoided with multiple view angles separating vascular wall calcification from contrast material. However, stenoses are not assessable whatever the projection in cases with large circumferential calcified plaques.10
Axial sections provide the most reliable evaluation of carotid stenosis but require the determination of the carotid axis by another technique, such as MIP reconstruction. MIP is more reliable than SSD; both require lengthy postprocessing time and are inefficient in cases with circumferential arterial wall calcification. Finally, CT images provide a reliable means of evaluating stenosis, as shown by the good level of interobserver agreement. The potential of helical CT in the exploration of atherosclerotic carotid stenosis remains to be determined. If a severe stenosis has been detected by ultrasound, helical CT could be used as a second recourse, before angiography, to select patients with severe stenosis.
Selected Abbreviations and Acronyms
|ICA||=||internal carotid artery|
|MIP||=||maximum intensity projection|
|r s||=||Spearman’s rank correlation coefficient|
|SSD||=||shaded surface display|
The authors gratefully thank P. Demont, E. D’haese, and J. Saulier for the photographic reproductions, T. Saint-Michel for technical assistance, S. Rotsaert for help in preparing the manuscript, and the technical staff in the CT department.
- Received February 13, 1995.
- Revision received June 5, 1995.
- Accepted June 5, 1995.
- Copyright © 1995 by American Heart Association