Published online November 11, 2002, 10.1148/rg.e11
(Radiographics. 2003;23:e11-e11.)
© RSNA, 2003
Navigating the Aorta: MR Virtual Vascular Endoscopy1
James F. Glockner, MD, PhD
1 From the Department of Radiology, Mayo Clinic, 200 First Street SW, Rochester, MN 55901. Presented as a scientific exhibit at the 2001 RSNA scientific assembly. Received March 29, 2002, revision requested July 29, revision received and accepted October 3. Address correspondence to J.F.G. (e-mail: glockner.james@mayo.edu).
 |
Abstract
|
|---|
Virtual vascular endoscopy (VVE) uses two- and three-dimensional (3D) data sets from magnetic resonance (MR) or computed tomographic angiography to create endoluminal views of blood vessels. This technique is relatively new and has become practical only recently as cheap and powerful computers have become widely available. MR-generated VVE can produce striking images and may prove useful as an alternative or accessory means of presenting large quantities of data. This exhibit reviews technical aspects of MR VVE, describes common errors and artifacts, and provides several examples of MR VVE along with more traditional presentations of 3D gadolinium-enhanced MR angiographic data.
© RSNA, 2002
|
Introduction
|
|---|
Virtual endoscopy is the computer-generated simulation of endoscopic images derived from three-dimensional (3D) computed tomographic (CT) or magnetic resonance (MR) imaging data sets. This technique allows the viewer to explore the inner surfaces of anatomic structures from a unique perspective. Virtual endoscopy has its most frequent application in CT virtual colonoscopy. The success of virtual colonoscopy can be attributed to several factors: The technique is analogous to a common clinical procedure, endoscopic presentation of data has been shown to be useful in visualization of polyps and cancers, and the endoscopic perspective is useful in distinguishing artifacts from true lesions.
Virtual vascular endoscopy (VVE) has been a less popular application. There is no corresponding clinical technique and no clear evidence that this form of data presentation provides any advantages over traditional methods. Nevertheless, MR-generated VVE can produce striking endovascular images, and its importance as a marketing tool should not be underestimated. The technique is still relatively new, and important clinical applications may yet be demonstrated.
This exhibit briefly reviews technical aspects of MR VVE, describes common errors and artifacts, and provides several examples of MR VVE along with more traditional presentations of 3D gadolinium-enhanced MR angiographic data.
|
Methods
|
|---|
Virtual endoscopic displays are produced with use of surface or volume rendering techniques. Surface rendering segments data on the basis of low and high thresholds. Isointensity surfaces corresponding to the selected thresholds are identified at the edges of the vessel of interest and are transformed into a smooth surface composed of triangles or patches. The resulting vessel lumen can be traversed with a virtual endoscope by surface rendering the interior walls with a 3D ray-casting algorithm that selects visible voxels by tracing rays from the current viewing position (1,2). Disadvantages of surface rendering include the need for complex segmentation as a preprocessing step and the retention of only a small fraction of the data in the final image. Volume rendering assigns a color and opacity to each voxel in the data set with use of a transfer function. Assignments can be made on the basis of any information contained in the voxel (eg, intensity, position, spatial gradient, etc). For MR VVE, voxels above an intensity threshold are rendered transparent, while variable opacity is assigned to voxels below the threshold. This ensures that the lumen is transparent and allows visualization of the vessel wall (3,4). Volume rendering techniques are more flexible, allow greater user manipulation of parameters, and permit visualization of structures outside the vessel wall. The major disadvantage of volume rendering is that it is a much more computer-intensive process than surface rendering, since all of the raw data are included in the endoscopic image, not just a small fraction. However, as increasingly powerful computers continue to become available at lower cost, distinctions such as these will almost certainly become less important. In fact, many of the newest generation of image processing workstations allow volume-rendered virtual endoscopy in near real time.
All images in this exhibit were generated with Advantage Navigator software (GE Medical Systems, Milwaukee Wis), on an Advantage Windows 3.1 workstation (GE Medical Systems). This software uses a dual-threshold surface rendering technique. Viewers can adjust low and high thresholds. In principle, thresholding can be adjusted for every view, although this would be time consuming. Navigation through the vessel lumen can be performed in a fly-through mode (ie, moving the camera within the vessel lumen with the computer mouse) or by movement of the cursor on axial, sagittal, and coronal reformatted images. Navigation can also be performed on a semiautomated basis by asking the computer to select the next "step" (position along the vessel); the computer chooses the direction by finding the longest available open path. Movies are generated by saving the user-defined endoscopic path through the vessel and reconstructing images at arbitrary intervals along the path, typically every 2–3 mm to generate a smooth movie without an inconveniently large number of images. The digital communications in medicine (DICOM) images are then converted to audiovisual interleave (AVI) movie files suitable for viewing on a PC running a homebuilt software program (written by Mayo Clinic software engineers).
Three-dimensional gadolinium-enhanced MR angiography was performed with a standard technique on a 1.5-T Signa CVi system (GE Medical Systems). Gadopenetate dimeglumine (0.1–0.2 mL/kg) was injected at 2–3 mL/sec with an MR-compatible power injector. Typical imaging parameters included a repetition time/echo time of 4.5–5/1.5 msec, 256 x 192 matrix, 35° flip angle, a bandwidth of approximately 62.5 kHz, 30–40 sections with a thickness of 1.4–3 mm reconstructed with 50% overlap, centric phase encoding, and 0.5–1.0 signals acquired. A 1–2 mL test bolus was injected before performance of MR angiography to determine the imaging delay needed to ensure acquisition of the center of k space during peak vessel enhancement.
Processing time needed to generate VVE displays was quite variable, depending primarily on the image quality obtained with MR angiography. Most of the VVE image sequences in this exhibit were produced in 10–15 minutes. The requirements for high-quality VVE images are fairly simple but not always achievable: homogeneous vascular enhancement throughout the region of interest, excellent definition of vessel boundaries with high contrast, and high signal-to-noise ratio (SNR) within the vessel lumen. Spatial resolution should be as high as possible without compromising image SNR and contrast. As a general rule, good source images translate into high-quality VVE images, and poor source images generate low-quality VVE images.
|
Errors and Artifacts
|
|---|
Most errors and artifacts are related to the somewhat arbitrary nature of data thresholding. The goal of VVE is to define a set of thresholds that will perfectly define the edge of the vessel while excluding everything else. This is most likely to occur when the 3D data set is optimal (ie, signal intensity within the vessel is uniformly high and signal intensity outside the vessel uniformly low). The most common error is trying to generate VVE images from suboptimal data sets. This leads to two typical problems: holes in vessel walls, which can be mistaken for the origins of small vessels such as accessory renal arteries, and floating shape artifacts within the vessel lumen. Both of these artifacts are threshold problems. Floating shape artifacts occur when voxels within the lumen with relatively low signal intensity are interpreted as not belonging to the vessel and are rendered as distinct objects within the lumen. Pierced-surface artifacts occur when voxels outside the vessel with relatively high signal intensity are interpreted as belonging within the vessel; therefore, the surface defining the edge of the vessel is disrupted and a hole appears (5). These concepts are illustrated in Movies 1–3.
|
Current Applications
|
|---|
Several case reports and technical notes have appeared in the literature describing applications of VVE that use data obtained at either CT or MR angiography (6–14). Visualization of pulmonary emboli has been described with use of both CT and MR data (6,7). VVE of aneurysms and pseudoaneurysms of the aorta has been performed (8,9). Neri et al investigated the diagnostic accuracy of VVE using CT angiography in detection of accessory renal arteries and found that evaluation of source images and maximum-intensity projection (MIP) images had significantly higher sensitivity and specificity (10). The major problem with VVE in detection of small accessory arteries was generation of segmentation artifacts that could cause both false-positive and false-negative interpretations. Maeder et al applied VVE to the evaluation of patients with intracranial aneurysms by using data obtained with 3D time-of-flight MR angiography and concluded that, compared with MIP images, VVE provided superior definition of the aneurysm neck and the morphology of saccular aneurysms (11). Two studies have shown that MR VVE is capable of demonstrating luminal irregularities of the inner surface of the aorta in patients with familial hypercholesterolemia and in a rabbit model of atherosclerosis (12,13). Both authors concluded that this technique could be useful in early detection of atherosclerotic lesions and in assessment of disease progression and response to therapy.
|
Conclusions
|
|---|
MR VVE offers a unique perspective for visualizing data obtained with 3D gadolinium-enhanced MR angiography. A variety of software packages are available to generate endoscopic images, and data processing speed continues to improve, with near real-time visualization achievable on the most advanced workstations. The clinical applications of this technique are not yet clear but may include preoperative planning for surgical and interventional procedures, detection of atherosclerotic plaque, and monitoring of disease progression or regression; or the technique could simply be an alternative method for data presentation as 3D volumetric data sets continue to increase in size.
|
Footnotes
|
|---|
Abbreviations: MIP= maximum-intensity projection,
SNR= signal-to-noise ratio,
3D= three-dimensional,
VVE= virtual vascular endoscopy.
|
References
|
|---|
- Summers RM. Morphometric methods for virtual endoscopy. In: Bankman IN, eds. Handbook of medical imaging processing and analysis. New York, NY: Academic Press, 2000; 747-755.
- Jolesz FA, Lorensen WE, Shinmoto H, et al. Interactive virtual endoscopy. AJR Am J Radiol 1997; 169:1229-1235.[Free Full Text]
- Rubin GD, Beaulieu CF, Argiro V. Perspective volume rendering of CT and MR images: applications for endoscopic imaging. Radiology 1996; 199:321-330.[Abstract/Free Full Text]
- Smith PA, Heath DG, Fishman EK. Virtual angioscopy using spiral CT and real–time interactive volume–rendering techniques. J Comput Assist Tomogr 1998; 22:212-214.[CrossRef][Medline]
- Neri E, Caramella D, Falaschi F, et al. Virtual CT intravascular endoscopy of the aorta: pierced surface and floating shape thresholding artifacts. Radiology 1999; 212:276-279.[Abstract/Free Full Text]
- Konen E, Rozenman J, Amitai M, Gayer G, Garniek A. Virtual CT angioscopy of pulmonary arteries in a patient with multiple pulmonary emboli. AJR Am J Radiol 1998; 171:399-400.[Free Full Text]
- Ladd ME, Gohde SC, Steiner P, Pfammatter T, McKinnon GC, Debatin JF. Virtual MR angioscopy of the pulmonary artery tree. J Comput Assist Tomogr 1996; 20:782-785.[CrossRef][Medline]
- Davis DP, Ladd ME, Romanowski BJ, Wildermuth S, Knoplioch JF, Debatin JF. Human aorta: preliminary results with virtual endoscopy based on three–dimensional MR imaging data sets. Radiology 1996; 199:37-40.[Abstract/Free Full Text]
- Neri E, Caramella D, Cioni R, Trincavelli F, Vignali C, Bartolozzi C. Pseudoaneurysm of the abdominal aorta: evaluation with virtual angioscopy of spiral CT data sets. Eur Radiol 1999; 9:1227-1230.[CrossRef][Medline]
- Neri E, Caramella D, Bisogni C, et al. Detection of accessory renal arteries with virtual vascular endoscopy of the aorta. Cardiovasc Intervent Radiol 1999; 22:1-6.[CrossRef][Medline]
- Maeder PP, Meuli RA, de Tribolet N. Three–dimensional volume rendering for magnetic resonance angiography in the screening and preoperative workup of intracranial aneurysms. J Neurosurg 1996; 85:1050-1055.[Medline]
- Yoshida K, Endo M, Mori K, Katada K, Ueda M, Toriwaki J, Tateno Y. Virtualized angioscopy of the thoracic aorta in a rabbit model of atherosclerosis. Jpn Circ J 1998; 62:198-200.[CrossRef][Medline]
- Summers RM, Choyke PL, Patronas NJ, et al. MR virtual angioscopy of thoracic aortic atherosclerosis in homozygous familial hypercholesterolemia. J Comput Assist Tomogr 2001; 25:371-377.[CrossRef][Medline]
- Fellner F, Blank M, Fellner C, Bohn-Jurkovic H, Bautz W, Kalender WA. Virtual cisternoscopy of intracranial vessels: a novel visualization technique using virtual reality. Magn Reson Imaging 1998; 16:1013-1022.[CrossRef][Medline]
Top
Abstract
Introduction
