DOI: 10.1148/radiol.2381042106
(Radiology 2006;238:309-320.)
© RSNA, 2006
Carotid Artery Stent Implantation: Evaluation with Multi–Detector Row CT Angiography and Virtual Angioscopy—Initial Experience1
Darren B. Orbach, MD, PhD,
Bidyut K. Pramanik, MD,
Julie Lee, BA,
Thomas S. Maldonado, MD,
Tom Riles, MD and
Robert I. Grossman, MD
1 From the Neuroradiology Section (D.B.O., B.K.P., J.L., R.I.G.) and Vascular Surgery Division (T.M., T.R.), New York University Medical Center, 530 First Ave, New York, NY 10016. Received December 13, 2004; revision requested February 15, 2005; revision received February 28; accepted March 17, 2005.
Address correspondence to D.B.O. (e-mail: darren.orbach{at}med.nyu.edu).
 |
ABSTRACT
|
|---|
Approval for this HIPAA-compliant study was obtained from the institutional review board; informed consent was not required for retrospective review of patient studies that had been performed for clinical evaluation. The purpose of this study was to retrospectively compare the accuracy of intrastent luminal diameter, as measured on transverse computed tomographic (CT) angiograms and virtual angioscopic views, with the manufacturer's specifications for phantom diameter and with digital subtraction angiographic (DSA) measurements of stent diameter obtained in patients. Intrastent diameter was measured by using standard and stent-optimized reconstruction kernels with three window settings. Endoluminal virtual angioscopic views of the stent-containing vessels were also generated. Measurements at CT angiography were compared with known specifications for the phantom and with DSA measurements in patients. Erroneous measurements of intrastent diameter occurred when a standard kernel and nonoptimized window settings were used. A set of parameters that minimized error relative to measurements obtained at DSA was also identified. Virtual angioscopy helped demonstrate morphologic aspects of stenosis that were otherwise difficult to appreciate.
© RSNA, 2006
|
INTRODUCTION
|
|---|
The North American Symptomatic Carotid Endarterectomy Trial and the Asymptomatic Carotid Atherosclerosis Study demonstrated the greater benefit of carotid endarterectomy versus medical management for the treatment of severe carotid artery stenosis (defined as greater than 70% stenosis the former trial and greater than 60% stenosis for the latter trial) (1–3). Although a randomized scientific trial comparing endarterectomy with carotid stent implantation has not yet been completed, endovascular treatment of carotid stenosis with expandable stents is being increasingly proposed as an alternative therapy to endarterectomy, and the respective roles of the two procedures are still being elucidated (4–7). While carotid restenosis after endarterectomy is relatively uncommon (8), the reported incidence of restenosis after stent implantation ranges from 3% to 29% (9–13). Early restenosis (ie, within 3 years after stent implantation) is attributed to intimal hyperplasia, whereas late restenosis is attributed to secondary or recurrent atherosclerosis.
No standard of care has been established for follow-up imaging of patients with carotid artery stents. Most patients, however, are followed-up yearly with Doppler ultrasonography (US), and patients with findings that are suggestive of high-grade stenosis are typically evaluated with digital subtraction angiography (DSA). While Doppler US is a reliable and accurate method for screening patients who have undergone endarterectomy, Doppler US has been shown to be markedly less specific in patients who have undergone stent implantation, with a false-positive rate of more than 30% (14). Attempts to refine Doppler US criteria to improve specificity have proved unsuccessful (15).
Computed tomographic (CT) angiography has been shown to correlate with Doppler US and DSA for the evaluation of carotid stenosis in carotid arteries that do not contain stents (16,17). The sensitivity and specificity of CT angiography have been reported to be as high as 90%. CT angioscopy has also been described for the assessment of coronary arteries (18) and atherosclerotic carotid arteries that do not contain stents (19,20). To the best of our knowledge, however, CT angioscopy has not been used to internally assess carotid stents. The CT angioscopy algorithm relies on differences in CT attenuation between the contrast material–enhanced lumen and its borders, which consist of the vascular endothelium in the native vessel and the interior of the newly implanted stent in cases in which one had been implanted. The software generates a transparency table for which all pixels that are within an attenuation range that is consistent with that of contrast-enhanced blood are rendered transparent; all pixels outside this range are rendered opaque. A color table is then used to assign a color to each opaque pixel on the basis of attenuation. Structures that are not readily apparent on transverse source images may be better appreciated on an endoluminal view, as shown in Figure 1 , which demonstrates that the individual struts of two stents can be seen.
View larger version (80K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1a: (a) Virtual angioscopic views of carotid stents obtained at the proximal stent orifice in a patient (top) and at the midpoint of the stent in a carotid phantom (bottom). Images demonstrate that visualization of the internal surface, including the concentric struts, is possible. (b) Single coronal section from a series of two-dimensional reformatted images obtained through a stent shows the perspective along which the angioscopic view was obtained.
|
|
View larger version (92K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1b: (a) Virtual angioscopic views of carotid stents obtained at the proximal stent orifice in a patient (top) and at the midpoint of the stent in a carotid phantom (bottom). Images demonstrate that visualization of the internal surface, including the concentric struts, is possible. (b) Single coronal section from a series of two-dimensional reformatted images obtained through a stent shows the perspective along which the angioscopic view was obtained.
|
|
The purpose of this study was to retrospectively compare the accuracy of intrastent luminal diameter, as measured on transverse CT angiograms and virtual angioscopic views, with the manufacturer's specifications for phantom diameter and with DSA measurements of stent diameter obtained in patients (reference standard).
|
MATERIALS AND METHODS
|
|---|
Patients
Approval for this study was obtained from our institutional review board. Informed consent was not required for retrospective review of studies that had been obtained for clinical evaluation. This study complied with all relevant Health Insurance Portability and Accountability Act guidelines.
The patients, all with previously diagnosed severe carotid atherosclerotic disease, underwent carotid angioplasty and stent implantation (Wallstent; Boston Scientific, Natick, Mass). It is the routine clinical practice of the vascular surgeons at our institution to obtain follow-up CT angiograms within 30 days after carotid stent placement. Six patients (four women, two men; age range, 59–77 years; mean age, 69 years) with six stents were scanned with CT angiography within a variable period of time after stent placement (Table 1). As shown in Table 1, the stents, which extended past the bifurcation, were typically placed with the proximal end within the distal common carotid artery and the distal end within the proximal internal carotid artery. The only exceptions were one patient in whom the entire stent was located within the common carotid artery and another patient in whom the entire stent was located within the internal carotid artery. During catheter angiography, at least one lateral view of the common carotid bifurcation was obtained both before and after stent implantation. Informed consent was obtained from patients for CT angiography with intravenous contrast material.
Five of the six patients (patients 2–6) had newly placed stents. In these patients, CT angiography was performed less than 20 days after stent placement. Patient 1 was imaged approximately 6 months after stent placement for recurrent severe intrastent stenosis. In this patient, DSA was performed for the purpose of intrastent angioplasty, and in this patient alone, CT angiography preceded DSA by 33 days.
Phantom
A carotid artery phantom (Siemens Medical Solutions, Erlangen, Germany) was used to assess the accuracy of our luminal diameter measurements. The phantom consisted of three parallel polyurethane tubes (inner diameter, 4 mm) that were filled with contrast material. The contrast material was diluted to approximately 1:75 in iso-osmolar saline, as was similarly performed in previous CT studies of stent phantoms (21,22). At this concentration, the mean attenuation of the solution was 275 HU. One of the tubes did not contain a stent (phantom A), while the other two tubes contained one stent each (phantoms B and C). Of the two tubes that contained a stent, one had a soft-tissue-attenuation stenosis located within the stent, which reduced the inner diameter to 2.2 mm (phantom B); the other tube consisted of the stent alone (phantom C). The phantom was imaged in longitudinal and transverse orientations while immersed in a water bath that was warmed to body temperature to simulate the soft-tissue attenuation that surrounds the carotid arteries. Image acquisition parameters were identical to those used in patients (described later).
CT Angiography
CT angiograms were obtained by using a 16–detector row CT scanner (Sensation 16; Siemens Medical Solutions). Imaging was performed with a section thickness of 0.75 mm, a collimation of 0.5 mm, an overlap of 0.25 mm, 120 kVp, a tube rotation time of 500 msec, an image acquisition rate of 32 sections per second, a matrix of 512 x 512, and a field of view of 20 cm. This yielded isotropic voxels of 0.75 mm. Contrast material was injected by way of an 18-gauge antecubital fossa intravenous catheter at an injection rate of 4 mL/sec without a saline flush. Image acquisition was initiated with bolus tracking when the threshold for intravascular enhancement in a region of interest located over the aortic arch was exceeded.
A standard soft-tissue reconstruction kernel (B30; Siemens Medical Solutions) was used for all patient data and for one set of phantom data. A new reconstruction kernel (B46; Siemens Medical Solutions) that was optimized for sharp-edged, high-attenuation structures, such as stents, was made available to us, and the phantom data were reconstructed by using this kernel. Because the raw data from the patient cohort were not saved, the patient data could not be reconstructed by using the new kernel.
Image Analysis
CT angiograms were analyzed on a workstation (Vitrea workstation; Vital Images, Plymouth, Minn) by a board-certified radiologist who had 3 years of experience in interpreting carotid CT angiograms (D.B.O.). Analysis was performed by using transverse images and fly-through virtual angioscopic views of the endoluminal surface. Because patient identification information was removed from the images, the reader was blinded to the results of all prior analyses at all times.
Transverse images.—Intrastent luminal diameter was measured on transverse images by using three different window settings. The first setting consisted of a window width of 500 HU and a window level of 200 HU, which was the standard window setting used for CT angiography. The second setting consisted of a window width of 1500 HU and a window level of 400 HU, which was the recommended window setting for the imaging of stents in prior studies (21). The third setting consisted of a window width of 1500 HU and a window level of 1500 HU, which was the window setting at which the patients' stents and the cortices of bone structures were just visible. In all patients, there was a well-defined border between the luminal region, which contained contrast material, and the surrounding stent region, which had substantially higher attenuation than the lumen.
Although the carotid phantom stents had a uniform diameter outside the region of stenosis, in the patient cohort, the stents did not assume a simple cylindric shape; rather, the diameter typically varied along the length of the stent. We therefore measured the diameter of the lumen at three points along the long axis of the stent—that is, at the proximal and distal ends and in the middle. The proximal end was defined as the portion of the stent that was located three CT sections or 2.25 mm distal to the first section on which the stent circumference was seen in its entirety (ie, the point closest to the aorta), and the distal end was defined as the portion of the stent that was located three CT sections proximal to the last section on which the stent was fully visualized (ie, the point furthest from the aorta). For stents that had an hourglass shape (ie, a narrow center and wider proximal and distal ends), the middle was defined as the narrowest point along the stent. For stents with more complex shapes, the middle was defined as the CT section obtained at a gantry position exactly midway between the proximal and distal sections.
Although the stent diameter typically varied along the long axis of the stent, in the majority of stents, the shape was circular when cut in cross section and thus had a single diameter. In those few cases where the cross section of the stent was ovoid rather than circular, measurements along both the long and the short axes were recorded.
For the carotid phantom images, in addition to measuring the luminal diameter, we compared the transverse images obtained with the two kernels at all three window settings and attempted to determine whether we could resolve the fine spatial details, such as the individual metal struts, of the stents. We also attempted to ascertain whether we could clearly differentiate the contrast-enhanced lumen from the stenotic lesion in the phantom that contained the artificial stenosis.
Virtual angioscopic images.—A software program (3D Fly Through Vessels; Vital Images) was used to generate virtual angioscopic endoluminal images of the stent-containing carotid artery. For the carotid phantoms, we used the default choices of transparency and color tables, as well as the tables that were generated by us. To choose appropriate values for these tables, we measured the range of Houndsfield units for 20 randomly chosen pixels within three regions on each scan for both the phantom and the patients. These regions consisted of (a) the contrast-enhanced lumen, (b) the perimural soft tissue (or in the case of the phantom, the artificial stenosis) because the normal vessel wall itself cannot be resolved at CT, and (c) the stent material itself. The measured values were then used to generate transparency and color tables for the creation of virtual angioscopic views.
To measure vessel diameter on the angioscopic views, the user chooses a point along the wall and then draws a line across the lumen to the other side. Because of the complex changes in visual perspective that are present on angioscopic views, however, there are an infinite number of potential lines of differing lengths that one might draw across, and it is not apparent which is the true diameter (Fig 2).
Our solution is as follows: The user chooses point A along one wall and traverses across the lumen to any point B on the other side of the vessel. The user then travels up and down a short distance from point B perpendicular to the long axis of the vessel. Because the software places the cursor on an opaque point at all times, every point traversed is on the wall rather than inside the lumen, and the sum of the points defines an arc. The point on the arc at which line AB measures a maximum is, by definition, the point directly across the lumen from the point of origin; this point is called B' (Fig 3a).
View larger version (13K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3a: Diagram illustrates the measurement technique for obtaining an accurate luminal diameter on virtual angioscopic views. (a) In the first image, point B' is chosen, which gives the maximal distance from point A along the arc described. As per the geometry of a circle, the point directly across the starting point is, by definition, the furthest point on the circumference and defines the diameter, as shown. (b) In the second image, point P is chosen, which gives the shortest distance from point A along the line described.
|
|
Next, beginning at point B', the user travels up and down the long axis of the vessel along which the shortest distance from point A must lie exactly opposite and measures the true lumen diameter. We label this point P (Fig 3b). In practice, this two-step technique for length determination yields robust values with little variation on repeated measurement. Measurements for stents that had an hourglass shape are seen in Figure 4.
View larger version (19K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3b: Diagram illustrates the measurement technique for obtaining an accurate luminal diameter on virtual angioscopic views. (a) In the first image, point B' is chosen, which gives the maximal distance from point A along the arc described. As per the geometry of a circle, the point directly across the starting point is, by definition, the furthest point on the circumference and defines the diameter, as shown. (b) In the second image, point P is chosen, which gives the shortest distance from point A along the line described.
|
|
View larger version (18K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4a: Schematics of cylindric and noncylindric vessel configurations. (a) Drawing demonstrates that even in a vessel with an hourglass shape, the choice of point P remains accurate. (b) Drawing shows how, with a noncircular vessel cross section, point B', which is furthest from point A, is incorrectly chosen.
|
|
View larger version (12K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4b: Schematics of cylindric and noncylindric vessel configurations. (a) Drawing demonstrates that even in a vessel with an hourglass shape, the choice of point P remains accurate. (b) Drawing shows how, with a noncircular vessel cross section, point B', which is furthest from point A, is incorrectly chosen.
|
|
Because the angioscopic view allows for visualization of the individual struts of the stent (Fig 1), the intrastent diameter will depend on whether one measures across the lumen from strut to strut, from strut to interstrut, or from interstrut to interstrut. We chose as our convention to measure from strut to strut.
DSA Images
DSA images were obtained by the vascular surgeons and were interpreted by a board-certified radiologist who had 3 years of experience in interpreting carotid DSA images (D.B.O.). By removing patient identification information from the images, the reader was blinded to the results of all prior analyses at all times. During placement of the carotid stents, lateral views of the DSA images were obtained intraoperatively. We employed a commonly used technique for vessel calibration by taping a dime to the patient's skin and imaging on the same view as the vessel that contained the stent. This allowed for calibration of the measurements to a known standard (a dime measures 17.9 mm). In one patient, we obtained an image of a calibration catheter inside the aortic arch (23) together with the dime on the patient's face. This allowed us to compensate for the parallax that was induced by using an external length standard (ie, on the body surface) rather than an internal length standard; when not corrected for parallax, the measurement was 16.4 mm rather than 17.9 mm (ie, 91.5% of the true value). We measured the dime on the image by using the calibration catheter as a ruler. From this, we calculated the coefficient needed to arrive at a correct diameter of 17.9 mm; we subsequently used this coefficient to correct all our DSA images.
In all patients, the location of the stent was determined by noting the texture generated by the struts. Intrastent luminal diameter was measured as close to the proximal and distal ends of the stent as possible, as well as at the midpoint or narrowest portion of the stent, in order to best correlate these measurements with those obtained at CT.
Statistical Analysis
We used the manufacturer's specifications for phantom diameter and the measured patient stent diameters at DSA as the reference standards. The three variables of interest included CT reconstruction kernels (standard or stent-optimized), location (proximal, middle, and distal portion of the stent), and window setting. Three measures of error—that is, measurement error, absolute error, and relative error—were used to evaluate and compare these variables. Measurement error, which is an assessment of bias, was calculated as the observed diameter minus the reference standard diameter. Absolute error, which is an estimate of accuracy, was calculated as the absolute value of measurement error. Relative error, which is a measure of accuracy that is independent of the true diameter, was calculated as the absolute error, presented as a percentage, of the reference standard diameter. A mixed-model analysis of variance was used to evaluate differences between the variables for each of the three error types. In each case, the assessment of error constituted the dependent variable, and the model included CT window settings, location, and reconstruction kernel (for phantom data only) as fixed classification factors. Also included was the indicator variable that identified patients as a random effect. Models for patients and phantoms were constructed independently. The Tukey honestly significant difference was used for all pairwise comparisons among the variables, with significant differences determined by maintaining the type I error rate at or below 5%.
|
RESULTS
|
|---|
Transverse CT Angiography
Carotid phantom.—The standard reconstruction kernel was compared with the stent-optimized reconstruction kernel. On the basis of the images generated by using the standard kernel, it was impossible to differentiate between the contrast-enhanced fluid within the stent lumen and the artificial mural stenosis; thus, the measured diameter falsely gave a reading that spanned from stent wall to stent wall, thereby overestimating stent patency. With the stent-optimized kernel, the stenosis was clearly separate from and had a different attenuation than the opacified lumen (Fig 5).
View larger version (58K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5: Comparison of two transverse CT angiograms obtained in the carotid phantom with the same window settings. The top image was reconstructed by using the standard kernel, and the bottom image was reconstructed by using the stent-optimized kernel. The stenosis (arrows) is more subtle in the top image and is much more discretely depicted on the bottom image.
|
|
As is evident from Table 2, the diameter of the phantom that did not contain a stent (phantom A) did not change measurably with changes in CT window settings, regardless of the reconstruction kernel used. The apparent diameter of the phantoms that contained a stent, however, did change dramatically. The true diameter of the carotid phantom is demonstrated in Table 3. In the absence of intrastent stenosis, the intrastent diameter was underestimated by 38%–58% for the standard kernel and by 25%–38% for the stent-optimized kernel.
View this table:
[in this window]
[in a new window]
|
Table 2. Luminal Diameter of Carotid Phantom as Measured by Using Different Reconstruction Kernels and Window Settings
|
|
By using two different window settings (Table 2), there was a statistically significant difference between the standard and stent-optimized kernels with respect to absolute error (P < .001) and relative error (P = .001). No significant difference, however, was found in terms of measurement error (P = .133). Thus, while the two kernels did not exhibit different levels of bias, the stent-optimized kernel provided significantly more accurate diameter assessments than did the standard kernel. The window setting that consisted of a window width of 1500 HU and a window level of 400 HU was found to have a significantly lower mean absolute and relative error than the window setting that consisted of a window width of 500 HU and a window level of 200 HU (P = .033).
Mean bias (ie, measurement error) was significantly smaller at the middle of the stent (ie, in the segment of the stent that contained the artificial stenosis) (P = .018) than at either the distal or proximal end of the stent (ie, in the naked contrast material–filled segments of the stent) (P = .033). There was no significant difference between any of the regions in terms of absolute or relative error.
Patient cohort.—Unlike the stents that were placed in carotid phantoms, most of the stents that were placed in patients assumed a noncylindric configuration, with an hourglass shape being most common. In comparing the diameters derived from transverse images with those derived from DSA images (Tables 4, 5), it is clear that appropriate window settings are essential, as was the case for the carotid phantom (Fig 6). The standard CT angiography window setting (window width, 500 HU; window level, 200 HU) caused substantial artificial luminal narrowing compared with the results of DSA. Extremely wide window settings at which only the struts of the stent and the cortices of bone structures could be seen (ie, at a window width of 1500 HU and window level of 1500 HU) resulted in artificial widening of the measured lumen. This was particularly noticeable in patient 1, who had known severe intrastent stenosis. In this patient, decreased visualization of soft tissue resulted in a misdiagnosis of stent patency at these wide window settings. Window settings with a width of 1500 HU and a level of 400 HU generated images with diameters that were in good agreement with those obtained at DSA. Window settings with a width of 500 HU and a level of 200 HU had significantly higher mean absolute and relative error than did the other window settings (P < .001).
View larger version (30K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6: Transverse CT angiograms demonstrate how the measured luminal diameter changes with window settings. The left image was obtained with a window width of 500 HU and window level of 200 HU, the middle image was obtained with a window width of 1500 HU and window level of 400 HU, and the right image was obtained with a window width of 1500 HU and window level of 1500 HU. All images are from the same transverse section and were obtained in the same patient. The measured diameter ranges from 2.3 to 4.5 mm.
|
|
Virtual Angioscopy
Given the results described earlier, we used the transverse images that were generated by using the stent-optimized kernel for all further analysis of the phantom data. We used the transverse data set to measure the range of the attenuation values for randomly selected pixels in the four regions of interest (ie, contrast-enhanced lumen, perivascular and intravascular soft tissue, calcium, and stent). After excluding calcified perivascular or intravascular foci, there was little overlap in attenuation values for these three regions (Table 6), and it was readily possible to generate transparency and color table values. Pixels between 200 and 500 HU were rendered transparent, and all other pixels were rendered opaque. Pixels with attenuation greater than 500 HU were colored yellow, and pixels with attenuation less than 200 were colored red.
View this table:
[in this window]
[in a new window]
|
Table 6. Range of Attenuation Values for Stent Material, Calcified and Noncalcified Plaques, and Contrast-enhanced Lumen
|
|
In patient 1 (Fig 7), the angioscopic representation suggests that the stenotic lesion is eccentric, with the part of the inner surface of the stent that is located toward the bottom of the image appearing free of stenosis (Fig 7a). This observation is borne out by careful examination of the transverse source image on which the intrastent high attenuation contrast material is seen to collect anteriorly and to the right (Fig 7b) while the lower attenuation of the intrastent stenosis lies posterior and to the left.
View larger version (56K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7a: Patient 1. Severe intrastent stenosis in a 74-year-old woman. (a) Endoluminal views of a stent obtained with same raw data set were used to generate both images. Top image uses software default values for transparency and color tables, which generated the impression of a clean stent. Bottom image uses corrected settings for transparency and color tables; the substantial red intrastent stenosis is clearly visible. (b) Transverse source image demonstrates that, as is shown on the angioscopic views in a, the stenosis is eccentric, with high-attenuation intrastent contrast material collecting anteriorly (arrow) and lower-attenuation stenosis collecting posteriorly.
|
|
View larger version (99K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7b: Patient 1. Severe intrastent stenosis in a 74-year-old woman. (a) Endoluminal views of a stent obtained with same raw data set were used to generate both images. Top image uses software default values for transparency and color tables, which generated the impression of a clean stent. Bottom image uses corrected settings for transparency and color tables; the substantial red intrastent stenosis is clearly visible. (b) Transverse source image demonstrates that, as is shown on the angioscopic views in a, the stenosis is eccentric, with high-attenuation intrastent contrast material collecting anteriorly (arrow) and lower-attenuation stenosis collecting posteriorly.
|
|
In another patient (Fig 8), the distal end of the stent is partially occluded by material that is in continuity with and of a similar attenuation as the stent material, which is suggestive of a possible kinked strut. There is currently no way of independently corroborating this finding, however.
View larger version (75K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 8: Angioscopic images viewed from inside the stent (top) and from outside the stent (bottom) demonstrate partial occlusion at the distal end of the stent (arrow). The attenuation of the occluding material appears to be similar to that of the stent itself. The material also appears to be in contiguity with the stent, which suggests the possibility of a kinked strut.
|
|
The detailed angioscopic views allow the user to choose whether to measure the luminal diameter from strut to strut or from interstrut to interstrut. We found that the difference between these measurements was on the order of 0.5 mm (Fig 9). The actual strut thickness of the stents that were used, as measured with a caliper, was 0.2 mm. This provided an expected 0.4-mm difference between measurements obtained from strut to strut and those obtained from interstrut to interstrut.
View larger version (159K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 9: Angioscopic view of carotid phantom stent demonstrates how the measured intrastent diameter changes depending on whether one measures from strut to strut (approximately 3.7 mm) or from interstrut to interstrut (approximately 4.1 mm). The actual strut thickness, as measured with a caliper, is 0.2 mm. This generates an expected 0.4-mm difference between measuring from strut to strut and measuring from interstrut to interstrut.
|
|
In the absence of a stent, the standard software algorithm overestimated the diameter by approximately 10% (Tables 7, 8). Conversely, by using the corrected color and transparency tables, the diameter of arteries that did not contain a stent was underestimated by approximately 17%. For the stent-containing phantoms, the standard software values generated diameters that did not vary along the length of the stent (ie, the artificial stenosis was invisible), while our modified transparency and color tables accurately measured the stenosis in phantom B as 2.2 mm. The internal stent diameter along the nonstenotic segment, however, was underestimated by approximately 15%.
View this table:
[in this window]
[in a new window]
|
Table 8. Intrastent Diameter as Measured on Virtual Angioscopic Views Generated by Using Corrected Transparency and Color Tables
|
|
The angioscopic views in the patient cohort tended to underestimate the diameters measured at DSA by 13%–50%. In patient 1, who had known intrastent stenosis, agreement between the angioscopic view and the DSA image at the point of maximal stenosis was excellent.
The overall mean and standard deviation of the absolute and relative error assessments was highest for the standard window setting (window width, 500 HU; window level, 200 HU) and lowest for the virtual angioscopic fly-through diameter measurements. The difference between these measurement methods, however, did not achieve statistical significance (P > .12).
|
DISCUSSION
|
|---|
We used transverse source images and reconstructed endoluminal views to measure the intrastent luminal diameter after the placement of carotid artery stents in patients and in a carotid artery phantom. We found that, in our carotid phantom, the reconstruction kernel that was optimized for high-attenuation sharp-edged structures, such as a stent, made the otherwise invisible intrastent stenoses apparent and allowed for more accurate measurement of intrastent diameter. Whether the same kernel can be optimally used in the patient setting for assessing CT angiograms of carotid arteries that either do or do not contain stents remains to be determined.
Even with the preferred kernel and optimal window settings, however, intrastent diameter was still underestimated by a factor of approximately 25%. One potential contributing cause for this is that we measured intrastent diameter from strut to strut, while the measurements of the carotid phantoms and the measurements obtained at DSA are more likely to be made from interstrut to interstrut. The difference between the strut-to-strut and interstrut-to-interstrut distances that we found, which was as much as 0.5 mm, represents a 12% difference in a 4-mm carotid artery and may underlie some of the discrepancy. We believe that the narrowest diameter within the stent (ie, the strut-to-strut distance) is likely to be the most biologically relevant because it defines the narrowest pathway through which blood must flow.
The measured stent diameter varied inversely with the narrowness of the CT window settings. With wide window settings, however, all structures other than bone and metal are rendered invisible, including intrastent soft-tissue stenosis. On the CT angiograms of the carotid phantoms, CT angiography with the stent-optimized kernel generated accurate measurements of the diameter, which were obtained in the region containing the simulated stenosis. Thus, CT angiograms of stents should be interpreted with the understanding that, in regions without substantial mural soft tissue (ie, regions with bare stent material), the CT measurement is likely to underestimate the true measurement by about 25%. In regions containing substantial intrastent soft tissue, measurements at CT angiography are likely to be highly accurate.
On the angioscopic views there was an increase in luminal narrowing artifacts in the patient cohort (13%–50%) relative to the phantom (approximately 25%). Given that the stents in patients were of larger diameter than those in the carotid phantom, which should militate against increased narrowing artifacts, the likeliest explanation is our inability to use the stent-optimized kernel in the patient cohort. In fact, the phantom data show that luminal narrowing artifacts increased from approximately 28% to 43% when the suboptimal filter is used.
Two in vitro studies (21,22) evaluated whether CT angiography was a suitable alternative to conventional angiography in the evaluation of small vessel stents for intracranial angioplasty. These investigators found that the luminal narrowing artifacts increased markedly as the diameter of the stents decreased. By focusing on the larger subset of stents, which approximate the size of our carotid phantom (4 mm), investigators found that CT blooming artifacts impaired visualization of both naked stents and those containing simulated stenoses; artificial luminal narrowing of 39%–67% was seen in naked stents measuring 3–4 mm, and simulated stenoses of 25% that were located inside 4-mm stents had apparent stenotic degrees of 60% at CT angiography. Much of the difference between our findings and theirs is likely the result of technique: they used a two–detector row CT scanner with a collimation of 1 mm, whereas we used a 16–detector row CT scanner with a collimation of 0.5 mm. Despite this, the values we report for the measured diameter with the standard reconstruction kernel are similar in magnitude to their results, and thus, the optimized kernel is the likely explanation for the balance of the difference.
Our carotid phantom was 4 mm in diameter, whereas the diameter of the stent-containing carotid arteries in the patient cohort was larger. Given that luminal narrowing artifacts diminish with increasing stent diameter, our error measurements in the phantom carotid likely represent a ceiling relative to those measurements that will be obtained in clinical practice.
Errors in measurement on two-dimensional images are reported by the software manufacturer to be on the order of plus or minus 10% for structures measuring less than 2 mm and on the order of plus or minus 5% for structures measuring between 2 and 10 mm. For the surface area of three-dimensional structures, the error is reported as plus or minus 20% for objects with a surface area of less than 4 mm2 and as plus or minus 10% for objects with a surface area of between 4 and 100 mm2. Our measurements, which were obtained from transverse images and three-dimensional angioscopic views, were within these ranges, thereby demonstrating that the presence of a high-attenuation stent does not add substantially to the error when appropriately compensated.
Our study had limitations. A potential source of error by using the technique that we outlined earlier for measuring luminal diameter from the angioscopic view lies in the fact that the technique assumes a cylindric vessel. As we demonstrated, unlike the carotid phantom stents, most of the stents that were placed in patients assumed a noncylindric configuration, with an hourglass shape being the most common. This shape could potentially inject error into the assessment of point P as one travels up and down the length of the stent. However, the geometry dictates that, even in the presence of an hourglass configuration, point P is still the shortest point across from point A as one travels up and down the long axis of the stent; in fact, the difference between point P and all other points is magnified rather than reduced so long as the shape of the stent is concave outward, which we always found to be the case.
Alternatively, the technique could fail if the cross section of the vessel is not circular but rather oviod or irregular; in this case, the choice of B' as the furthest from point A along the arc defined by the user would be incorrect. Fortunately, we found this geometry to be uncommon. For cases in which the transverse images demonstrate a noncircular cross section, however, one should interpret luminal diameter measured on the angioscopic views with great care.
Another limitation relates to our assessment of the DSA images. First, because we did not have an internal standardized measurement in every case, we had to calculate the true size of structures by using a surface-based standard (a dime) in an attempt to correct for the inherent parallax. As described earlier, we found the parallax error in comparing the skin surface to the aortic arch to be on the order of 8%. Differences among patients in terms of the distance between the skin surface and the depth of the internal carotid artery may be substantial and would thus skew our measurements.
Second, we usually had only a single intraoperative view of the stent-containing carotid artery from which to make our measurements. Thus, we could not ensure that we visualized the smallest lumen measurement at angiography. Depending on the vessel tortuosity and the angle of the long axis of the vessel relative to the image intensifier, magnification effects may substantially change the perceived size of the vessel.
Finally, because these clinically acquired angiograms were not optimized to demonstrate the fine details of the stents, the struts were not always optimally seen and bone artifacts were not always minimized. The determination of the proximal and distal ends of the stent cannot be taken as precise; in one case, the proximal and distal ends could not be determined at all. Because the stents varied in diameter along their length, it was important to align the CT angiographic section at which the diameter was measured with the spot on the DSA image on which the diameter was measured; this is an impossible task if the ends are not well visualized. Given these caveats, as well as the fact that the CT images of the patients were reconstructed with the suboptimal standard kernel, it is remarkable that the measurements at CT angiography were in as good accord with the DSA measurements as they were.
Two groups have previously performed virtual angioscopy of the carotid artery (19,20). These groups described the relative inaccuracy of virtual angioscopy in assessing the grade of stenosis and the insensitivity of the technique to the presence of plaque ulceration. Because these investigators did not use a multi–detector row CT, however, the transverse source images were thick, thereby causing a loss of detail because of low spatial resolution and volume averaging. Furthermore, these researchers effectively used a binary color table for which every nontransparent voxel was rendered with the same color, thereby blurring the distinction between plaque components and endothelium.
In summary, we have shown that, with the appropriate choice of reconstruction kernel and window parameters, transverse CT angiograms can provide a measurement of intrastent luminal diameter that approximates the diameter measured at DSA in the stent-containing carotid artery. While the generation of endoluminal virtual angioscopic views is, at the present time, labor intensive, these images can accurately depict the morphologic details of the internal stent lining and possibly the vessel endothelium and may have a valuable role to play in the assessment of restenosis. Both of these findings require further substantiation with larger cohorts.
|
ACKNOWLEDGMENTS
|
|---|
We thank Bernhard Schmidt, PhD, of Siemens Medical Solutions, Erlangen, Germany, for use of the carotid artery phantom and for his helpful insights on the subject of CT phantom imaging. We are also grateful to James Babb, PhD, for statistical analysis.
|
FOOTNOTES
|
|---|
Abbreviations: DSA = digital subtraction angiography
Authors stated no financial relationship to disclose.
Author contributions: Guarantors of integrity of entire study, D.B.O., B.K.P., R.I.G.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, D.B.O., B.K.P., T.S.M.; clinical studies, all authors; experimental studies, T.R.; statistical analysis, D.B.O.; and manuscript editing, D.B.O., B.K.P., T.S.M., T.R., R.I.G.
|
References
|
|---|
- North American Symptomatic Carotid Endarterectomy Trial Collaborators. Beneficial effect of carotid endarterectomy in symptomatic patients with high-grade carotid stenosis. N Engl J Med 1991;325:445–453.[Abstract]
- Executive Committee for the Asymptomatic Carotid Atherosclerosis Study. Endarterectomy for asymptomatic carotid artery stenosis. JAMA 1995;273:1421–1428.[Abstract]
- Barnett HJ, Taylor DW, Eliasziw M, et al. Benefit of carotid endarterectomy in patients with symptomatic moderate or severe stenosis: North American Symptomatic Carotid Endarterectomy Trial Collaborators. N Engl J Med 1998;339:1415–1425.[Abstract/Free Full Text]
- Yadav JS, Wholey MH, Kuntz RE, et al. Protected carotid-artery stenting versus endarterectomy in high-risk patients. N Engl J Med 2004;351:1493–1501.[Abstract/Free Full Text]
- Malek AM, Higashida RT, Phatouros CC, et al. Stent angioplasty for cervical carotid artery stenosis in high-risk symptomatic NASCET-ineligible patients. Stroke 2000;31:3029–3033.[Abstract/Free Full Text]
- Ouriel K, Hertzer NR, Beven EG, et al. Preprocedural risk stratification: identifying an appropriate population for carotid stenting. J Vasc Surg 2001;33:728–732.[CrossRef][Medline]
- CaRESS Steering Committee. Carotid revascularization using endarterectomy or stenting systems (CARESS): phase I clinical trial. J Endovasc Ther 2003;10:1021–1030.[CrossRef][Medline]
- Gorlitzer M, Heine B, Mendel H, et al. Redo surgery for carotid artery stenosis: when and how? Cardiovasc Surg 2000;8:366–371.
- Christiaans MH, Ernst JM, Suttorp MJ, et al. Restenosis after carotid angioplasty and stenting: a follow-up study with duplex ultrasonography. Eur J Vasc Endovasc Surg 2003;26:141–144.[CrossRef][Medline]
- CAVATAS Investigators. Endovascular versus surgical treatment in patients with carotid stenosis in the Carotid and Vertebral Artery Transluminal Angioplasty Study (CAVATAS): a randomised trial. Lancet 2001;357:1729–1737.[CrossRef][Medline]
- Wholey MH, Al-Mubarek N. Updated review of the global carotid artery stent registry. Catheter Cardiovasc Interv 2003;60:259–266.[CrossRef][Medline]
- Willfort-Ehringer A, Ahmadi R, Gschwandtner ME, Haumer M, Lang W, Minar E. Single-center experience with carotid stent restenosis. J Endovasc Ther 2002;9:299–307.[CrossRef][Medline]
- Berkefeld J, Turowski B, Dietz A, et al. Recanalization results after carotid stent placement. AJNR Am J Neuroradiol 2002;23:113–120.[Abstract/Free Full Text]
- Ringer AJ, German JW, Guterman LR, Hopkins LN. Follow-up of stented carotid arteries by Doppler ultrasound. Neurosurgery 2002;51:639–643.[CrossRef][Medline]
- Robbin ML, Lockhart ME, Weber TM, et al. Carotid artery stents: early and intermediate follow-up with Doppler US. Radiology 1997;205:749–756.[Abstract/Free Full Text]
- Link J, Brossmann J, Grabener M, et al. Spiral CT angiography and selective DSA of internal carotid artery stenosis. AJNR Am J Neuroradiol 1996;17:89–94.[Abstract]
- Cinat ME, Pham H, Vo D, Gordon I, Wilson SE. Improved imaging of carotid artery bifurcation using helical computed tomographic angiography. Ann Vasc Surg 1999;13:178–183.[CrossRef][Medline]
- Nakanishi T, Kohata M, Miyasaka K, Fukuoka H, Ito K, Imazu M. Virtual endoscopy of coronary arteries using contrast-enhanced ECG-triggered electron beam CT data sets. AJR Am J Roentgenol 2000;174:1345–1347.[Free Full Text]
- Rilinger N, Seifarth H, Sokiranski R, et al. Virtual intra-arterial angioscopy (VIA) of the carotid artery based on helical CT data. Br J Radiol 2003;76:792–797.[Abstract/Free Full Text]
- Dessl A, Giacomuzzi SM, Springer P, et al. Virtual endoscopy with post-processing helical CT data sets [in German]. Aktuelle Radiol 1997;7:216–221.[Medline]
- Hahnel S, Trossbach M, Braun C, et al. Small-vessel stents for intracranial angioplasty: in vitro comparison of different stent designs and sizes by using CT angiography. AJNR Am J Neuroradiol 2003;24:1512–1516.[Abstract/Free Full Text]
- Trossbach M, Hartmann M, Braun C, Sartor K, Hahnel S. Small vessel stents for intracranial angioplasty: in vitro evaluation of in-stent stenoses using CT angiography. Neuroradiology 2004;46:459–463.[Medline]
- Fortin DF, Spero LA, Cusma JT, Santoro L, Burgess R, Bashore TM. Pitfalls in the determination of absolute dimensions using angiographic catheters as calibration devices in quantitative angiography. Am J Cardiol 1991;68:1176–1182.[CrossRef][Medline]
This article has been cited by other articles:
|
|
|
|
 
A.C. Yi, E. Palmer, G.Y. Luh, J.P. Jacobson, and D.C. Smith
Endovascular Treatment of Carotid and Vertebral Pseudoaneurysms with Covered Stents
AJNR Am. J. Neuroradiol.,
May 1, 2008;
29(5):
983 - 987.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
P. Bergeron
Symptomatic carotid artery stenosis: the dilemma for a reasonable selection of carotid stenting or endarterectomy
Eur. Heart J.,
January 1, 2008;
29(1):
7 - 9.
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
