AJR ARRS PQI
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by van Ooijen, P. M. A.
Right arrow Articles by Oudkerk, M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by van Ooijen, P. M. A.
Right arrow Articles by Oudkerk, M.
Social Bookmarking
Add to CiteULike   Add to Complore   Add to Connotea   Add to Del.icio.us   Add to Digg   Add to Reddit   Add to Technorati  
What's this?
Hotlight (NEW!)
Right arrow
What's Hotlight?
AJR 2003; 180:223-226
© American Roentgen Ray Society

Computers in Radiology

Noninvasive Coronary Imaging Using Electron Beam CT: Surface Rendering Versus Volume Rendering

P. M. A. van Ooijen1, R. J. M. van Geuns2, B. J. W. M. Rensing2, A. H. H. Bongaerts1, P. J. de Feyter2 and M. Oudkerk1

1 Department of Radiology, University Hospital Groningen, Hanzeplein 1, 9713 GZ Groningen, The Netherlands.
2 Department of Cardiology, University Hospital Rotterdam, Thoraxcentre, P. O. Box 2040, 3000 CA, Rotterdam, The Netherlands.

Received February 11, 2002; accepted after revision June 20, 2002.

 
Address correspondence to P.M.A. van Ooijen.


Abstract
Top
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 References
 
OBJECTIVE. Three-dimensional data for noninvasive imaging of the coronary arteries are acquired from electron beam CT, multidetector CT, or MR imaging. Most commonly, surface rendering is used for three-dimensional processing, but recent advances in hardware and software technology have made it possible to use volume rendering. Our objective was to compare volume rendering with surface rendering for the visualization of the coronary artery tree.

CONCLUSION. Our experience in the comparison of both techniques shows that because of intrinsic problems associated with surface rendering, volume rendering produces better image quality.


Introduction
Top
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 References
 
Three-dimensional (3D) data for displaying the coronary arteries in a noninvasive way can be acquired using contrast-enhanced electron beam CT or multidetector CT. Three-dimensional imaging of this data can be performed in different ways. Most commonly, surface rendering is used, but recent developments in hardware and software have facilitated the use of newer and more computationally expensive techniques such as volume rendering. In this article, we will describe the two rendering techniques, compare them visually, and demonstrate their ability for the visualization of the coronary arteries.


Materials and Methods
Top
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 References
 
For this study, four data sets of human coronary arteries acquired using noninvasive contrast-enhanced electron beam CT were visualized using both volume rendering and surface rendering. Patients were randomly selected from an ongoing study on noninvasive imaging of the coronary arteries. Both institutional review board approval and informed consent were obtained. Data acquisition was performed on an electron beam CT scanner (Evolution XP; Imatron, San Francisco, CA). The acquisition of the 3D data set began with the injection of 120-180 mL of contrast medium (iopromide, 350 g/L I) at 3-4 mL/sec through an antecubital vein. Scanning commenced just proximal to the left main coronary artery after an ECG trigger at 80% of the R-R interval (diastole). Both slice thickness and table increment were set at 1.5 mm, resulting in contiguous, nonoverlapping slices. In all, 40-60 axial slices with a field of view of 18 cm were acquired during a single breath-hold. Surface renderings were constructed on a Magicview workstation (Siemens Medical Systems, Erlangen, Germany). Volume renderings were constructed using 3D rendering software (VoxelView; Vital Images, Plymouth, MN) running on special purpose graphics hardware (Octane; SGI, Mountain View, CA). A qualitative comparison of the resulting images was performed.

Surface Rendering
Surface rendering is a technique that uses only part of the available 3D data set for the reconstruction of an image. The easiest way to select which part of the data has to be used to reconstruct the image is by defining a threshold. Typically, the voxels with a value below this threshold will be discarded from the data, and the voxels with a value equal to or above the threshold will be selected for the rendering. In the case of the coronary artery tree, a threshold of 80-100 H, resulting in an image containing only bone and contrast-enhanced blood, is advised [1]. After this selection, a simplified approximation of the object can be obtained by subdivision of the object into surface elements (e.g., small triangles). A more detailed representation can be obtained using smaller surface elements, which has the disadvantage of being more computationally expensive.

Volume Rendering
Unlike surface rendering, volume rendering does not make use of a surface representation. When a volume rendering algorithm is used, certain properties are assigned to each voxel on the basis of its value (with CT, this is the density value in Hounsfield units). When properties are assigned a certain value, volume rendering uses the histogram of these values. In a typical CT histogram, the x-axis represents the possible voxel values in the data, and the y-axis, the number of voxels with that specific value. Certain values can be related to specific tissue compositions when the properties of a voxel value are determined (Fig. 1). Each of these tissue compositions has specific properties. Partial rendering enables us to mix two different tissue compositions (totaling 100%) to establish the properties of a border voxel. Thus, a voxel can partially belong to the surface of interest and can have properties based on the percentages of the properties of the two tissues involved.



View larger version (69K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 1. —Opacity curve and histogram for volume rendering of coronary arteries. Typical CT histogram shows number of voxels with specific voxel values (CT number). Each CT number can be assigned certain tissue composition. Default tissue compositions consist of air, fat, soft tissue, and bone, with each tissue represented by specific peak in histogram. Tissue composition is identical for each CT scan. Line representing opacity shows maximum opacity for bone and contrast-enhanced blood and low opacity ({approx}30-40%) for fat and soft tissue. Air is fully transparent.

 

Properties that can be assigned to a specific voxel are, among others, contrast and opacity. The contrast is defined by the window level setting. The opacity defines the rate of transparency of a voxel. When 100% opacity is assigned to a voxel, this voxel is completely nontransparent, and when 0% opacity is assigned, the voxel is completely transparent.

To visualize the coronary arteries, the range of Hounsfield units representing the contrast-enhanced blood will have to be visualized with high opacity to display the lumen of the coronary arteries. Visualization of the coronary arteries will result in a window level of 90 H, a window width of 600 H, and an opacity curve, as shown in Figure 1.

Segmentation
Because of the enhancement of the blood using a contrast medium, the complete heart is enhanced and not just the coronary artery tree. Thus, segmentation of the data representing overlapping structures is required in both surface rendering and volume rendering to obtain a clear view of the coronary arteries.

Manual segmentation was performed by drawing curves around the region of interest on a subset of the original slices. The curves drawn in the slices were combined into a 3D surface description, and these surface descriptions were then used to remove obstructing structures [2]. The resulting (segmented) volume was then stored to perform the rendering.


Results
Top
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 References
 
After the required segmentation, a qualitative comparison of surface rendering and volume rendering for visualization of the coronary artery tree was performed using the settings described previously (Figs. 2A,2B and 3A,3B). In one of the cases, the default settings, which were determined by qualitative comparison of several different settings, had to be adjusted to obtain the optimal result. The loss of detail in surface rendering as well as the rendering of only part of the data using a surface representation can be observed in Figure 2A,2B. For coronary artery bypass grafts to be visualized, the same imaging settings can be used (Fig. 3A,3B). Larger vessels (>3 mm) were well visualized using both techniques, whereas smaller vessels (<2 mm) were not adequately visualized by either technique.



View larger version (117K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 2A. —Three-dimensional electron beam CT scans of coronary artery. Ao = aorta, RVOT = right ventricular outflow tract, LAD = left anterior descending coronary artery, RCX = right coronary artery. Surface rendering shows loss of detail, especially close to left main artery, in proximal part of left anterior descending artery, and in proximal part of intermediate branch (arrows). Area with loss of detail in left anterior descending artery was falsely rated as stenotic region. Right ventricular outflow track was hollowed out in surface rendering because of thresholding, thus showing visualization of only small part of data.

 


View larger version (112K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 2B. —Three-dimensional electron beam CT scans of coronary artery. Ao = aorta, RVOT = right ventricular outflow tract, LAD = left anterior descending coronary artery, RCX = right coronary artery. Volume rendering shows no stenosis (arrowheads).

 


View larger version (132K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 3A. —Electron beam CT scans obtained after venuos bypass graft surgey. Ao = aorta, CABG = coronary artery bypass graft. Surface rendering (A) and volume rendering (B) show coronary artery bypass graft.

 


View larger version (117K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 3B. —Electron beam CT scans obtained after venuos bypass graft surgey. Ao = aorta, CABG = coronary artery bypass graft. Surface rendering (A) and volume rendering (B) show coronary artery bypass graft.

 

When the coronary arteries are evaluated, visualization of calcified plaque depositions plays an important role. Visualization of these calcified plaques can be performed easily using volume rendering by slightly changing the opacity curve. The calcified plaques (that have a much higher voxel value > 250 H) become visible when a lower opacity is assigned to those voxels containing contrast-enhanced blood (100-250 H), and the relationship of the calcified plaques to the vessel and the possible stenotic regions can be evaluated (Fig. 4A,4B).



View larger version (130K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 4A. —Electron beam CT scans obtained after coronary artery bypass graft surgery. Volume rendering shows lumen of coronary arteries and coronary artery bypass graft. Arrows show calcified plaques; arrowheads show surgical clips.

 


View larger version (122K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 4B. —Electron beam CT scans obtained after coronary artery bypass graft surgery. Volume rendering shows calcified plaques within vessel lumen (arrows) and surgical clips (arrowheads). Contrast-enhanced blood was displayed in more transparent setting to preserve view of anatomic position and orientation of calcified plaques and clips.

 

In the only patient with an intracoronary stent, surface rendering did not allow differentiation between the vessel lumen and the coronary artery wall stent. Usually, the attenuation of stents is comparable to a densely calcified plaque, and thus the same settings can be used for visualization of the calcified plaque. Careful fine-tuning of the opacity settings allows the visualization of the stent in more detail with some stent types (Fig. 5A,5B).



View larger version (147K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 5A. —Electron beam CT scans obtained after stent placement. Volume rendering shows coronary arteries.

 


View larger version (135K):
[in this window]
[in a new window]
[as a PowerPoint slide]
  		Fig. 5B. —Electron beam CT scans obtained after stent placement. More transparent volume rendering shows position of stent. Inset shows image using setting that results in more detailed visualization of stent.

 

In all four data sets evaluated by visual comparison, volume rendering produced higher quality images compared with surface rendering. In the case of disagreement between surface rendering and volume rendering concerning a possible stenotic region, stenosis was verified on conventional catheter angiography. Volume rendering proved to be correct in all cases of disagreement, and surface rendering showed one false-positive stenosis and one false-negative stenosis (Fig. 2A,2B). No disagreement was found for the coronary artery bypass graft (one patient).


Discussion
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
References
 
A direct comparison of surface rendering and volume rendering has been conducted for CT of several abnormalities [3,4,5], resulting in a number of advantages and disadvantages for both techniques.

The advantages of surface rendering are fast and interactive manipulation after the determination of the surface representation, good depth perspective and unambiguous 3D images using a virtual light source, accurate clinical measurement of structures because of the distinct surface definitions, simple segmentation using multiple thresholds, and assignment of specific attributes to different thresholds (e.g., different colors).

In addition to the advantages, the following disadvantages are inherent to the surface-rendering algorithm. First, surface rendering uses only a small portion of the available data. Second, determination of the surface representation is time-consuming and has to be performed each time a threshold is changed. Because of this, finding the optimal threshold becomes a difficult and laborious task. Third, surface selection is not adequate in structures without a well-differentiated surface. Fourth, because of the approximation of the surface, possible false-positive and false-negative surface elements can be introduced, and small detail can be lost. Finally, internal structures cannot be seen in combination with the surrounding structure.

Advantages of volume rendering are that all available data can be used, partial rendering is possible (small surface detail is better preserved than when using surface rendering), transparency can be assigned to view internal structures or structures in the background, and three-dimensionally unambiguous images are obtained with good depth cues.

Some disadvantages should also be mentioned. First, volume rendering is computationally expensive, which could result in less interactive rendering. Second, determination of the optimal settings is difficult because of the large number of parameters that can be set, possibly leading to overvisualization. When even slightly wrong settings are used, too much data can be visualized, thus obscuring the structure of interest. Finally, high transparency will decrease the depth information in the resulting image and thus make the 3D relationships difficult to observe.

Previous studies of vessels other than the coronary arteries [3,4] have shown that the percentage of arteries visualized using volume rendering is significantly higher (89-96%) than the percentage of arteries visualized using surface rendering (52-70%). Arteries that have a diameter within the range of 2-3 mm were well seen using volume rendering and incompletely seen using surface rendering. Because of technical limitations, acquisition parameters, and (motion) artifacts, larger vessels were visualized with higher quality than were smaller vessels in both rendering techniques. Increased spatial resolution, as with multidetector CT, could increase in the image quality of smaller vessels. Furthermore, an increase in temporal resolution could decrease motion artifacts and thus facilitate visualization of smaller vessels. Decreased visualization when using surface rendering compared with volume rendering can be explained by the decreased overall quality of the images and the increase of artifacts when using surface rendering.

In conclusion, intrinsic problems associated with surface rendering cause a lower image quality when compared with volume rendering because volume rendering uses all the available data instead of using surface approximation. One main reason to use surface rendering instead of volume rendering is that the computation time for volume rendering is comparatively long. However, with modern workstations, this issue becomes less important because of the availability of faster and better hardware and software. Our experiments corroborate our hypothesis by showing that volume rendering produces images of higher quality than those produced with surface rendering for the visualization of the coronary artery tree.


References
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 References
 

  1. Moshage W, Achenbach S, Seese B, Bachmann K, Kirchgeorg M. Coronary artery stenoses: three-dimensional imaging with electrocardiographically triggered, contrast-agent enhanced, electron-beam CT. Radiology 1995;196:707 -714[Abstract/Free Full Text]
  2. Rensing BJ, Bongaerts AH, van Geuns RJ, van Ooijen PM, Oudkerk M, de Feyter PJ. Intravenous coronary angiography using electron beam computed tomography. Prog Cardiovasc Dis 1999;2:139 -148
  3. Soyer P, Heath D, Bluemke DA, et al. Three-dimensional helical CT of intrahepatic venous structures: comparison of three rendering techniques. J Comput Assist Tomogr 1996;1:122 -127
  4. Hong KC, Freeny PC. Pancreaticoduodenal arcades and dorsal pancreatic artery: comparison of CT angiography with three-dimensional volume rendering, maximum intensity projection, and shaded-surface display. AJR 1999;172:925 -931[Abstract/Free Full Text]
  5. Kuszyk BS, Heath DG, Bliss DF, Fishman EK. Skeletal 3-D CT: advantages of volume rendering over surface rendering. Skeletal Radiol 1996;3:207 -214

Add to CiteULike CiteULike   Add to Complore Complore   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us   Add to Digg Digg   Add to Reddit Reddit   Add to Technorati Technorati    What's this?


This article has been cited by other articles:


Home page
 RadioGraphicsHome page
N. C. Dalrymple, S. R. Prasad, M. W. Freckleton, and K. N. Chintapalli
Informatics in Radiology (infoRAD): Introduction to the Language of Three-dimensional Imaging with Multidetector CT
RadioGraphics, September 1, 2005; 25(5): 1409 - 1428.
[Abstract] [Full Text] [PDF]

Home page
 RadiologyHome page
U. J. Schoepf, C. R. Becker, B. M. Ohnesorge, and E. K. Yucel
CT of Coronary Artery Disease
Radiology, July 1, 2004; 232(1): 18 - 37.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by van Ooijen, P. M. A.
Right arrow Articles by Oudkerk, M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by van Ooijen, P. M. A.
Right arrow Articles by Oudkerk, M.
Social Bookmarking
Add to CiteULike   Add to Complore   Add to Connotea   Add to Del.icio.us   Add to Digg   Add to Reddit   Add to Technorati  
What's this?
Hotlight (NEW!)
Right arrow
What's Hotlight?

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS