| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Full Paper |
1 Institute of Diagnostic and Interventional Radiology, Klinikum Offenbach, Starkenburgring 66, 63069 Offenbach, 2 Department of Diagnostic Radiology, University of Ulm, Steinhövelstrasse 9, D-89075 Ulm, 3 Department of Diagnostic Radiology, Kreiskrankenhaus am Plattenwald, Am Plattenwald 1, D-74177 Bad Friedrichshall, 4 Department for Thoracic and Vascular Surgery, University of Ulm, Steinhövelstrasse 9, D-89075 Ulm and 5 Department of Clinical Radiology, University of Münster, A. Schweitzer Straße 33, D-48129 Münster, Germany
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
For patients undergoing carotid angioplasty the morphologic features of the plaque relate to the outcome after carotid angioplasty [3]. In the Imaging in Carotid Angioplasties and Risk of Stroke (ICAROS) study, duplex plaque morphology was correlated with periprocedural complications of carotid bifurcation angioplasty and stenting with the aim of identifying patients who would not profit from treatment [4].
The pre-operative diagnosis of stenoses of the carotid artery is most commonly made using angiography, duplex sonography, and magnetic resonance angiography.
In recent years CT, in particular the use of three-dimensional reconstruction techniques such as maximum intensity projection (MIP) and increasingly, volume rendering, has also assumed a role in the diagnosis of carotid lesions [5–10]. Virtual intra-arterial angioscopy (VIA) has been used as a new technique of secondary image reconstruction by several working groups for depicting the aorta as well as renal and coronary arteries [11–15].
With the help of this method "endoscopic" images of branching vessels can be produced. Problems of depiction occur in sections of vessels containing structures of varying density, such as hard and soft plaques. Separate computation of the respective geometry of these structures, followed by combining the various individual geometries to a total geometry, allows a differentiation of these structures in virtual angioscopy.
The goal of this study was to demonstrate the potential and limitations of virtual angioscopy of the carotids and to compare this technique with digital subtraction angiography (DSA) in a small group of patients.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
). The points of intersection of a straight line of 9 mm length overlaid on the density profile were used to determine the relative density value used as the threshold for the patient study.
|
Data acquisition
The spiral CT was conducted using a CT Twin (Philips Medical Systems, Best, The Netherlands), within 2 weeks of DSA, a non-contrast data set of the cervical region was obtained and this was followed by intravenous injection of 100 ml of contrast agent (Solutrast 300, BYK Gulden, Konstanz, Germany) at a flow rate of 2 ml s-1 into the antecubital vein. After a delay of 30 s, the contrast enhanced data were acquired. The selected delay was adopted from our standard protocols for CT-angiography. Scan parameters were identical in the phantom study and patient series: tube voltage: 120 kV, tube current: 130 mAs, slice thickness: 2.7 mm, increment: 1.0 mm, pitch factor: 1.5.
Post-processing
The contrast medium density was measured within the carotid artery 1 cm below the bifurcation, in the bifurcation, and 1 cm above it. A mean density was calculated separately for each side. The data sets were transferred to a separate image processing workstation (Silicon Graphics Indigo R10000; Silicon Graphics, Inc., Mountain View, CA) and reconstructed using a threshold value-based procedure (Iris Explorer, Silicon Graphics, Inc.). This program implements a so-called "marching cubes algorithm" for the creation of surface geometry as an isosurface of the object [16]. This algorithm uses a grid network structure based on voxels of an identical density previously specified by the user. The definition of the density values, on which this grid network is based, can be made either via direct input in Hounsfield units or indirectly by comparison of the original CT image with the appropriate binary image developed after threshold value modification. The grid network developed in such a way can be coated with any surface texture desired, so that a uniform surface emerges, and the underlying lattice structure is no longer visible. The surface texture was selected in such a way that a realistic impression of the surface of the vessel lumen was created. Lighting conditions were adapted in such a way that all wall structures were depicted in uniform brightness.
In a further step, surface geometry of calcium-containing plaques from the non-contrast slice images were computed using the same procedure and the same threshold value, but a surface texture was selected which clearly differentiated it from the vessel wall. Subsequently, vessel wall geometry was overlaid with this geometry in such a fashion that neighbouring skeletal structures were congruently depicted. An experienced user requires approximately 45 min for completion of the post processing procedure.
First, the morphology of the inner vessel wall and the plaques was evaluated. Subsequently, in accordance with the evaluation criteria for stenoses of NASCET, the residual vessel diameter was compared with the diameter of the post-stenotic vessel segment.
In keeping with the NASCET study criteria, stenosis grades were classified as low grade (<30%), medium grade (30–70%), and high grade (>70%). In addition, we evaluated completely occluded vessels.
The classification into four categories was done in consensus by two examiners on-screen and on a purely visual basis, as no tools were available to objectively measure vessel diameter, and measurements on-screen, in particular of the post-stenotic vessel segment, were not feasible due to distortion of perspective in the angioscopic image.
Sensitivity, specificity as well as the positive and negative predictive value of virtual angioscopy were computed using DSA as the reference procedure.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
The evaluation of stenoses in the vessel wall reconstruction alone was not possible, as vessel sections containing hard plaques were depicted not as stenoses, but as sacculations in the vessel wall. Only by adding the plaque surface geometry could calcium-containing plaques be detected as such and the stenosis correctly diagnosed. The distribution of the individual stenosis grades is shown in Table 1. Clinically relevant high grade stenoses of more than 70% were diagnosed with a sensitivity of 93.7% and a specificity of 91.3%. The positive predictive value was 88.2%, the negative predictive value 95.5%. VIA correlated with DSA for low grade stenoses in 88% of cases, for medium grade in 93% and for high grade stenoses in 71%. All occlusions were correctly classified, however two patients were identified on VIA as being occluded, when angiography showed these to be high grade stenosis. In two patients with high grade stenosis on angiography, VIA misclassified them as medium grade. The low correlation for high grade stenosis is particularly striking and could be observed especially with short stenotic segments (Figure 2). Long, high grade stenotic segments, however, were correctly depicted (Figure 3).
|
|
|
). Those patients demonstrated "low grade" and "medium grade" stenoses that were correctly classified both with VIA and DSA. In the vessel wall geometry of two vessels, freely floating structures appeared in the vessel lumen in immediate proximity to pronounced calcifications. In one case, these very irregularly configured structures led to an inaccurate estimate of the grade of stenosis by distracting attention away from the actual stenosis.
|
Ulcerations in thrombotic wall deposits, which were detected angiographically in three patients, could be depicted by VIA in only one case.
Our procedure was also unable to demonstrate tandem-stenoses.
An aneurysm of the internal carotid artery was clearly recognized as such in virtual angioscopy. Even the outflow from the aneurysm as well as the distal vessel segments could clearly evaluated.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
For virtual angioscopy of the aorta the working group of Kimura used a threshold of 120–160 HU, which was adjusted manually [13]. Another working group set the threshold value for the virtual depiction of coronary arteries at 120 HU, which corresponded to 57.6% of the maximum enhancement [14]. Dessl and coworkers also described the virtual angioscopy of the carotid artery. For reconstruction they used a threshold value of 160 HU [15]. Thus most research using a surface-based procedure employed a threshold value comparable with that used in the present study (56.4% of maximum enhancement, mean value of 144.2 HU). An important difference in the choice of the threshold value is, however, that in the studies referred to above this choice was made subjectively, was therefore examiner-dependent, and was usually adjusted, whereas the selection of the threshold value in the present work was based on the results of the phantom study.
With short segmental stenoses under approximately 3 mm in length, the relatively large slice thickness caused partial volume effects that led to relatively too high or too low density values of the vessel lumen, and thus to the incorrect depiction of the stenosis [17]. In all four misclassified stenoses in VIA a short segmental stenosis was present. An overlapping acquisition of the slices could not eliminate this source of error, since all layers were entered in equal proportion into the reconstruction. The cases of the two patients diagnosed with high grade stenosis in angiography, but classified as occlusive disease by VIA, demonstrate the clinical relevance of the discrepencies of the two methods. VIA must therefore be seen as an experimental "work in progress" technique – it cannot presently be viewed as a standard method for pre-therapeutic diagnosis.
In spatial proximity to large plaques we encountered bizarre intraluminal artefacts. These structures have been described by other authors as "floaters" and result from beam hardening artefacts and/or partial volume effects and inhomogeneities in blood flow and/or vessel contrast [11, 17, 18]. The size and form of these floaters was heavily dependent on the selected threshold value: with a low threshold value they increased in size, with a high threshold value they were no longer visible. By adjusting the threshold value to the actual contrast density in the vessel, the size of these floaters could be reduced to a minimum, so that the evaluation of the stenosis was not impaired in either of the two cases.
High venous contrast in 3D-reconstruction frequently leads to an overlapping of the arterial vessel segment under study by the neighbouring vein. In virtual angioscopy this vessel can be evaluated without previous segmentation of the vein. However, high enhancement of neighbouring vein leads to the occurrence of "holes" in the vessel wall. Rapid cerebral transit time led to early internal jugular vein contrast enhancement and thus to "piercing" artefacts in the wall of the carotid artery. With threshold values set too low and/or an excessive delay of scanning after beginning the contrast medium injection these piercing artefacts occur more frequently [18]. By using a relatively short delay of 30 s and by employing a relatively high threshold value these artefacts could be avoided to a large extent. We observed them in only three patients, in whom these piercing artefacts were located in the common carotid artery and/or in the internal carotid artery distal to the bifurcation and thus in no case impairing the evaluation of the carotid bifurcation and/or the grading of stenoses.
In relation to other spiral CT procedures for vascular imaging, such as maximum intensity projection (MIP) or three-dimensional surface rendering (shaded surface display, SSD), VIA offers the advantage of hard plaques not affecting the evaluation of the residual lumen. In a study presented by Leclerc in 1999 10% of the carotids could not be evaluated by MIP for this reason [10]. In SSD-angiography, calcium-containing plaques were generally removed manually from the data set in order to avoid overlapping of the residual vessel lumen [7, 17, 19]. By introducing a second geometry, however, it is possible in our procedure to reliably differentiate hard plaques from soft wall deposits.
Even in the currently most frequently employed reconstruction procedure, volume rendering, hard plaques must not first be removed from the data set, but are presented transparently, so that an evaluation of the residual lumen becomes possible. However, even here overlapping by neighbouring vessels and other structures of similar density is possible, so that preliminary manual processing of the graphic data sometimes cannot be avoided. For the correct classification of high grade stenoses in volume rendering technique various studies were able to achieve a sensitivity between 92% and 100%, while specificity ranged from 92% to 96%, yet even here high grade stenoses were in some cases misdiagnosed as occlusions [8, 10].
Since patients with an ulcerated plaque surface more frequently suffer cerebral infarction than patients with homogeneous plaque surfaces, an imaging technique should be able to make this differentiation possible. In this study, using virtual angioscopy, we succeeded in only one case of depicting ulceration which had previously been diagnosed with certainty on DSA. This result is consistent with the results of other authors [20, 21]. This is probably due to insufficient spatial resolution, and/or because of partial volume effects, whereby the usually relatively small ulcers are not depicted with sufficient density and thus escape detection.
Another cause could be that plaque ulcers containing calcium deposits are not visualized due to hardening artefacts within the plaques [9]. In this study a structure of soft tissue density situated between a calcium-containing section of the plaque and the vessel lumen could not be depicted, i.e. an ulceration in this location was not visible on virtual angioscopy. VIA is not yet a suitable method for routine clinical use. The lengthy time required for post-processing and the tendency to overestimate short segment, filiform stenosis are limitations of the method. The use of multidetector systems offer the promise of further improvement of this procedure by increasing the spatial resolution of the geometries and thus achieving a more detailed depiction of endoluminal pathology.
Future studies may demonstrate whether this technique can successfully define CT features which might correlate with the risk for cerebral ischaemic events. These may allow more precise assessment of the need for arterectomy, and could, in patients who are candidates for carotid percutaneous transluminal angioplasty, aid in predicting the risk of embolisation from balloon angioplasty.
Received for publication December 2, 2002. Revision received May 22, 2003. Accepted for publication July 24, 2003.
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  D. B. Orbach, B. K. Pramanik, J. Lee, T. S. Maldonado, T. Riles, and R. I. Grossman Carotid Artery Stent Implantation: Evaluation with Multi-Detector Row CT Angiography and Virtual Angioscopy--Initial Experience Radiology, January 1, 2006; 238(1): 309 - 320. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| BJR | DMFR | IMAGING | ALL BIR JOURNALS |
