Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH
QUICK SEARCH:   [advanced]


     

Published online before print March 1, 2002, 10.1148/radiol.2231010441

(Radiology 2002;223:275.)

A more recent version of this article appeared on April 1, 2002
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Quicktime movie
Right arrow All Versions of this Article:
2231010441v1
223/1/275    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
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 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 Raman, R.
Right arrow Articles by Rubin, G. D.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Raman, R.
Right arrow Articles by Rubin, G. D.

Technical Developments

Automated Generation of Curved Planar Reformations from Volume Data: Method and Evaluation1

Raghav Raman, MD, Sandy Napel, PhD, Christopher F. Beaulieu, MD, PhD, Eric S. Bain, MD, R. Brooke Jeffrey, Jr, MD and Geoffrey D. Rubin, MD

1 From the Department of Radiology, Stanford University School of Medicine, 300 Pasteur Dr, S-072B, Stanford, CA 94305-5105. From the 2000 RSNA scientific assembly. Received February 14, 2001; revision requested March 27; revision received August 23; accepted October 8. Supported by National Institutes of Health grant 5RO1HLO58915-03. Address correspondence to G.D.R. (e-mail: grubin@stanford.edu).


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 Materials and Methods
Results
Discussion
REFERENCES
 
The authors developed and evaluated a method to automatically create interactive vascular curved planar reformations with computed tomographic (CT) angiographic data. The method decreased user interaction time by 86%, from 15 to 2 minutes. Expert reviewers were asked to indicate their confidence in differentiating automatically created images from clinical-quality manually produced images. The area under the receiver operating characteristic curve was 0.45 (95% CI: 0.39, 0.51), and a test of equivalency indicated that reviewers could not distinguish between images. They also graded image quality as equivalent to that with manual methods and found fewer artifacts on automatically created images. Automatic methods rapidly produce curved planar reformations of equivalent quality with reduced time and effort.

Supplemental material: radiology.rsnajnls.org/cgi/content/full/2231010441/DC1.

Index terms: Angiography, technology • Arteries, CT, 17.12115, 17.12116, 9*.129162, 9*.12917 • Computed tomography (CT), helical, 9*.12916 • Images, processing • Veins, CT, 17.12115, 17.12116, 9*.12916, 9*.12917


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
Materials and Methods
Results
Discussion
REFERENCES
 
Computed tomographic (CT) angiography provides a noninvasive alternative to conventional angiography (1). These examinations produce hundreds of transverse images that necessitate a faster alternative to section-by-section viewing. At our institution, dedicated technologists create alternative visualizations that aid in the interpretive process. Unfortunately, the technologist’s time for each patient is quickly reaching unacceptable levels owing to the requirement for multiple reconstruction methods (2). To keep this time to a reasonable level, visualization techniques have to be automated where possible. Curved planar reformations depict the cross-sectional profile of a vessel along its length while preserving the relative x-ray attenuation information (2–4), which allows review of the whole cross section on one two-dimensional image. They enhance communication with the referring physician (5) and facilitate radiologic interpretation, for example, in stent-graft studies (4,6), evaluation of renal arteries (7), assessment of tortuous carotid arteries (8), and visualization of circumferential or eccentric calcifications in vessels (4,5,9).

Manually derived curved planar reformations require 15–30 minutes per patient to be produced by an experienced operator, depending on the vessels processed (8,10). This is because curved planar reformations show one vessel per image; thus, a new path has to be drawn for each vessel (10). Different views of the same vessel require adjustments to the path to ensure that it is in the center of the vessel in all planes being imaged. In addition, manually created curved planar reformations are exceedingly operator dependent (3,4,8,10,11). Small deviations from the true centerline, particularly in narrow arteries, can generate spurious stenoses or cause important lesions to be missed. Therefore, automation of this repetitive activity is desirable and has the potential to increase efficiency. The goal of this study was to evaluate a system we developed for generating curved planar reformations automatically from the volumetric image data produced with transverse scanning.


   Materials and Methods
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
Results
Discussion
REFERENCES
 
Automated Method
We have developed a method for automatically and rapidly producing curved planar reformations from volumetric image data that is capable of interactively displaying curved planar reformations of any vessel in any orientation with minimal user input (Movie, radiology.rsnajnls.org/cgi/content/full/2231010441/DC1.

We used a previously developed algorithm (12) to automatically produce the centerlines for curved planar reformations. It makes use of a simple and fast single-threshold segmentation algorithm to mark the boundaries of the aorta and its branches. User input is required only to select the seed voxel in the aorta for the segmentation and subsequently to select the start and end points of the vessels of interest by positioning a crosshair cursor in the transverse, coronal, or sagittal planes. Alternatively, points can be selected by clicking on the beginning or end of a vessel as shown on a three-dimensional surface display of the segmentation.

A raycasting scheme is used to automatically position the crosshair cursor at the point selected on the surface display. For branching vessels, multiple end points can be specified. User-selected points that fall outside the segmentation are rejected, and the user is prompted to reenter them. Once the segmentation and end-point identification are completed, an initial path on the surface of the segmentation is computed from the start to the end points. The segmentation is then morphologically thinned while the path is iteratively corrected toward the medial centerline. On completion of this process, the centerline is smoothed, and its interpoint distance is set to the minimum voxel dimension to make its accuracy at least equal to the data resolution. The user-specified start and end points are corrected toward the calculated medial axis by means of the thinning and smoothing process.

Subsequently, curved planar reformations are calculated in real time by using a linear interpolation algorithm. The user can interactively rotate the curved planar reformations through 360° (in 1° intervals) about the median axis by using an angle manipulation cursor or can view the rotation as a cine sequence (Fig 1). This allows review of the whole cross-sectional profile of the vessel along its length. To facilitate efficient correlation to confirm or refute abnormalities suspected on the curved planar reformations, our implementation allows the user to click on points of interest on the curved planar reformations to view the transverse and other orthogonal planes centered at that point and on automatically generated oblique reformatted planes (multiplanar reformations) perpendicular and parallel to the direction of the path at that point. In contrast to the orthogonal views, the perpendicular oblique reformation shows the true perpendicular cross section of the vessel at that point, while the parallel reformation shows the plane parallel with the vessel at that point (Fig 2).



View larger version (72K):
[in this window]
[in a new window]
  		Figure 1. Automatically created curved planar reformations through aortoiliac arteries. An angle-manipulation cursor allows interactive rotation of the curved planar reformations. a, As the angle manipulation cursor is moved from the sagittal plane (1) by an angle of 40° to position (2), the curved planar reformation rotates accordingly, which effects a transition from a sagittal (b) to an oblique (c) curved planar reformation. The sagittal view shows a stenotic lesion (solid arrow) in the left common iliac artery. In the oblique view, an ulcerative plaque (open arrow), not shown on the sagittal view, is visible in the plane of the reformation.

 


View larger version (122K):
[in this window]
[in a new window]
  		Figure 2. Automatically created curved planar reformations through renal arteries, magnified views. To facilitate cross-referencing, crosshairs indicate the current viewing position on all images. Clicking on a feature of interest sets the viewing position and translates all views to that position, including the perpendicular and parallel oblique reformations. a, Coronal curved planar reformation. A stenosis is not clearly seen on this view. b, Transverse curved planar reformation illustrates an eccentric stenosis, which demonstrates the need for more than one curved planar reformation through a vessel. c, Nearly transverse oblique reformation (multiplanar reformation) parallel to the vessel, which also intersects the stenosed region. The orientation of this plane is shown in a as line C. d, Nearly sagittal oblique reformation perpendicular to the centerline of the vessel shows the true cross section at the point of maximal stenosis. The orientation of this plane is shown in b as line D.

 
Unlike conventional curved planar reformations (2,8), our implementation maintains correlation with the Cartesian coordinate system to measure distances. To maintain the viewer’s orientation, crosshairs on the curved planar reformations index the position shown in the other views (Fig 2). In addition, the curved planar reformations can be presented in a three-dimensional view, which simulates a "magic carpet," that illustrates the actual shape and position of the curved plane (Fig 3). These automatically calculated additional views are not available in current interfaces. The paths can be saved along with the study volume for future use.



View larger version (110K):
[in this window]
[in a new window]
  		Figure 3. Three-dimensional magic carpet views of the shape of the curved planes through vessels. These views are a perspective rendering of the three-dimensional curved plane, which illustrates the actual curvature of the vessels. The inset windows show the corresponding two-dimensional views, which are standard curved planar reformations created by laying the three-dimensional plane down flat. a, Coronal and b, sagittal aortoiliac curved planar reformations. c, Coronal superior mesenteric curved planar reformations. d, Coronal renal curved planar reformations.

 
We implemented our method with a workstation (Dell, Round Rock, Tex) with two 500-MHz processors (Pentium; Intel, Chandler, Ariz) and 1 Gbyte of random access memory, or RAM, which was connected to our network and allowed access to the CT studies used in our evaluation. Since our method was not specifically optimized for a multiprocessor system, only one processor was used.

Validation
The dependence of automatically generated centerlines on user-selected start and end points was characterized by creating multiple centerlines through a small (superior mesenteric) and large (aortoiliac) artery in each patient, with use of the locus of all possible start and end points at the origin and terminus of each vessel. The usefulness of this system was tested with a limited clinical validation by using curved planar reformations of six arteries each (renal, celiac, superior mesenteric, aortoiliac, inferior mesenteric, and left subclavian) from CT angiographic examinations in three consecutive patients with aortic aneurysms. These patients underwent their examinations as part of their clinical workup.

CT angiographic images were transferred to a computer workstation. Curved planar reformations were then created by using both standard manual methods and our automated method. Manually created curved planar reformations were created by an expert technologist with experience in creating curved planar reformations in more than 1,000 patients by using manufacturer software (Advantage Windows, version 3.1; GE Medical Systems, Milwaukee, Wis).

As in standard practice, the technologist drew an approximate path through the center of the vessel by using the transverse, sagittal, and coronal views to specify the center of the vessel every few millimeters. Standard curved planar reformations were produced from these paths. Two reformations were created 90° apart through each vessel, and each included two of the following three possibilities: curved coronal, curved sagittal, or curved transverse, depending on vessel orientation. Automatically created curved planar reformations were produced through the same vessels. Thus, a total of 36 manually created and 36 automatically created curved planar reformations of the 18 arteries were produced, and the time required to produce the images with each method was recorded. All images were then exported from both applications, and histogram matching was used to set and standardize window and level settings to those routinely used in clinical practice.

The 72 images were printed with a clinical-quality dye sublimation printer, with identifying patient markings removed, as illustrated in Figure 4. Automatically and manually created images through all vessels were subsequently separated and arranged in random order. The images were then presented to four radiologists (G.D.R., C.F.B., E.S.B., R.B.J.) and one imaging scientist (S.N.).



View larger version (66K):
[in this window]
[in a new window]
  		Figure 4. a, Automatically and b, manually generated curved planar reformations through superior mesenteric arteries with identifying patient markers removed.

 
With use of a 10-point scale, reviewers determined the confidence with which images could be identified as being automatically or manually created. They submitted their responses on a form that was filled out by using the World Wide Web. A score of -5 indicated that the reviewer was very confident that the image was manually produced, and a score of +5 indicated that the reviewer was very confident that the image was automatically produced. The reviewers were asked to summarize any features that led them to make a decision on whether images were automatically or manually created. Receiver operating characteristic curves were created for each reviewer and for all reviewers pooled to indicate the reliability with which reviewers separated the two groups of images. Therefore, a reviewer who consistently but incorrectly identified manually created images as automatically created would have the same score as one who consistently and correctly identified manually created images as manually created.

Subjective image quality for each image was also assessed on a 10-point scale, and reviewers were asked to comment on the reasons for their assessments. A score of 10 indicated the reviewer had no concerns about the quality of the image. To quantitatively assess the occurrence of artifacts on the images, two reviewers subsequently performed a side-by-side blinded comparison between automatically and manually created images and reported the number of artifacts seen on each image. This allowed the reviewers to detect differences in the depiction of the vessels between paired images. Apparent stenosis, discontinuities, or other image features that were not present on the paired images could be counted as artifacts. Reviewer scores were compared by using paired t tests to reject the null hypothesis that there was no difference between scores. Because a failure to reject this null hypothesis does not indicate that the scores are the same, we additionally applied a test of equivalence to determine if we could reject the null hypothesis that the scores are different.

To perform the test of equivalence, we calculated the equivalency statistic E as where 1 and 2 are the two means being compared, V1 and V2 are their respective variances, and {delta} is the maximum difference that is clinically unimportant. We set {delta} to 15% of the mean score. To achieve 80% power, a value of -1.96 for E was required to reject the null hypothesis for all tests of equivalence. Additionally, we quantified interobserver variability by means of weighted {kappa} analysis. Differences with a P value of .05 were considered statistically significant.


   Results
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
Results
Discussion
REFERENCES
 
Variation in Automatically Created Centerlines
The analyzed vessels had a diameter range of 2.0–21.3 mm. The maximum variation in the start and end points tested was set to the diameter of the vessel. The mean variation in the origin of the automatically calculated median centerline was 2.1 mm in large vessels with a mean diameter of 19.8 mm and 1.3 mm in small vessels with a mean diameter of 9.0 mm. Mean variation in the termini of the median centerline was 1.6 mm in large vessels with a mean diameter of 12.6 mm and less than 0.5 mm in the termini of small vessels with a mean diameter of 3.9 mm. The maximum variation in the origin and termini occurred when the initial points were given as the edges of the vessel. Even so, the variation was never more than half the radius of the vessel. The centerline distal and proximal to the origin and termini, respectively, consistently had a smaller variation. The variation in the centerline became smaller than the minimum voxel dimension within 4.5–6.2 mm of the origins and termini.

Processing Time
Manual creation of curved planar reformations required a mean 14.8 minutes ± 2.2 (range) for each patient, while the automated method required 4.2 minutes ± 0.6 (2.1 minutes ± 0.4 and 2.1 minutes ± 0.2 for user interaction and computer processing, respectively). Thus, the user input time required for creation of curved planar reformations was reduced by 86%. Tortuous vessels and variant anatomy tended to increase manual processing times and necessitated more correction of the path. The renal arteries took the most time to manually process in this study. On the other hand, as indicated by the smaller SD for both the user interaction and computer processing times for the automated method, vessel anatomy did not adversely affect the time required to automatically generate curved planar reformations.

Discrimination of Manually Created from Automatically Created Curved Planar Reformations
The mean area under the pooled receiver operating characteristic curve was 0.45 (95% CI: 0.39, 0.51). Individual reviewers had areas of 0.42 (95% CI: 0.29, 0.56), 0.46 (95% CI: 0.34, 0.59), 0.49 (95% CI: 0.36, 0.62), 0.38 (95% CI: 0.23, 0.54), and 0.57 (95% CI: 0.38, 0.74). All areas were not significantly different from an area of 0.5, and the shapes of the curves approximated a straight line. The calculated equivalency statistic was -2.00 (P < .001) at 80% power, which allowed us to reject the null hypothesis and support the assertion that human observers cannot distinguish between manually and automatically generated curved planar reformations. The {kappa} values for interobserver agreement were between 0.01 and 0.09. In addition, no consistent trend was noted in reviewers’ comments. In their summaries, two reviewers wrote that they were unable to detect any discriminating features on the images. An unusual vessel choice in one curved planar reformation (internal iliac artery) led one reviewer to correctly identify this as an automatically generated curved planar reformation. A shorter curved planar reformation was used by one reviewer as indicative of automatically created images but by another as indicative of manually created images.

Image Quality
The mean image quality was 8.18 (95% CI: 8.03, 8.33) for automatically created images and 7.97 (95% CI: 7.82, 8.13) for manually created images. The equivalency statistic was -2.56 (P < .01) at 80% power, which indicates that image quality was not different between the two groups. The {kappa} values for interobserver agreement were between 0.4 and 0.6, and 83% of scores differed from each other by two points or less. In their feedback, two reviewers identified one automatically generated curved planar reformation of the renal arteries as being of low quality because depiction of the large superoinferior separation of the renal arteries on the image necessitated a discontinuity in the aorta. However, there was no concern expressed about the depiction of the arteries. Consistency in vessel caliber, depiction of smaller branches, and longer paths were criteria that three of the five reviewers noted. Mean scores from the reviewers who used these criteria for scoring image quality were 8.6 for automatically created and 8.4 for manually created images.

Artifacts
Figure 5 exemplifies the types of artifacts that were seen with higher frequency on manually created images. The two reviewers who scored image artifacts identified a mean 21.5 artifacts in the 38 manually created images compared with a mean 3.5 artifacts in the 38 automatically created images. Of those images with artifacts, reviewers on average identified only one artifact per image. Therefore, an average of 18.5 manually created images (48.7%) had one or more artifacts compared with a mean of 3.5 automatically created images (9.2%) with artifacts, a difference of 39.5%. The difference was statistically significant (P < .001). Reviewers agreed about the number of artifacts seen on 69 of the 76 images, with a {kappa} value for interobserver agreement of 0.81.



View larger version (100K):
[in this window]
[in a new window]
  		Figure 5. a, Manually created curved planar reformation through left iliac artery. The lumen of the artery section is obscured by mural calcification (open arrow), which makes it difficult to assess lumen patency. Placement of manual creation points slightly too close to the vessel wall is all that is required to produce this artifact. The solid arrow indicates an area of sharp change in direction of the vessel that is manifested as a small discontinuity in the reconstruction. b, Automatically created curved planar reformation. Mural calcium does not obscure the lumen, and the much greater number of points used to generate the automatically created reconstruction keeps the image smooth despite the sharp bend in the vessel.

 

   Discussion
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
Results
Discussion
REFERENCES
 
Variation in Centerlines
The median centerline was not heavily dependent on the user selection of start and end points. Even large variations in the initially selected points were corrected, which resulted in a much smaller corresponding variation in the median centerline. In addition, user input affected the shape of the median centerline for only about 5 mm at the origin and termini.

Faster Processing
Curved planar reformations can be created substantially faster with an automated method. With multiprocessor systems, automatic processing can occur in the background while the manual interaction for consecutive studies is completed in the foreground. The time required for computer processing is also bound to decrease as faster hardware and parallel processing techniques are incorporated.

Discrimination of Manually Created from Automatically Created Curved Planar Reformations
The test of equivalence indicated with high power that reviewers could not distinguish between images. The 95% CIs for all areas under the receiver operating characteristic curve included 0.5, and the {kappa} value indicated very low interobserver agreement. These findings support this conclusion. We can conclude that expert reviewers cannot differentiate between curved planar reformations generated manually versus those generated automatically.

Image Quality
The test of equivalence indicated that there was no significant difference between quality scores, but interobserver agreement was only fair. The lower interobserver agreement may be due to the subjective nature of the assessment conducted. Nevertheless, even though scores did not agree exactly, most of the scores did not vary by more than two points. We can conclude that overall image quality was not compromised by the use of the automatic method.

Greater Freedom from Artifacts
Findings in our small experiment indicate that automatically created images may have significantly (P < .001) fewer artifacts. There was good agreement between our reviewers. Many authors (3,4,8,10,11) have commented on the occurrence of operator-dependent artifacts in manually created curved planar reformations. The presence of these artifacts is unavoidable in normal clinical practice, not only because of human error but also because of the limitations of manual pointing devices. In addition, only tens of points are used to define a manually created curved planar reformation. Linear interpolation is used to define the centerline between manually selected points, which leads to the possibility of artifact if the interpolated line nears or cuts the vessel wall.

In automatically generated curved planar reformations, 400–700 points are used to define the centerlines, which may decrease the occurrence of operator-dependent artifacts. In addition, it is important to realize that the quality of the centerline is only one factor that determines image quality and artifacts. Our method does not reduce the effect of patient factors, such as movement artifact, and does not address problems caused by inadequate scanning parameters, misregistration, or incomplete vessel opacification. However, these factors would lead to artifacts and poor image quality that are independent of manual versus automatic creation of curved planar reformations. An in-depth investigation regarding artifacts in curved planar reformations has not been carried out, to our knowledge, and is probably warranted in the future.

Interactive Viewing with 360° Rotation
Reference to the original data for transverse images still remains the standard of reference (2,3). Our software allows interactive correlation with other views and 360° rotation of the curved planar reformations. At least two curved planar reformations of each vessel are recommended to detect eccentric plaque that might be missed on just one view (7). Sagittal and coronal curved reformations ignore much of the volumetric data set, as do maximum intensity projections and shaded surface displays. This occurs because only a small subset of the voxels contributes to the image in these methods. While both these alternative modalities show only a small part of the actual data even with 360° cine loops, 360° curved planar reformations show the actual voxel values of the full circumference of the vessel along its length.

As review of CT angiographic data on a computer monitor is increasingly considered to be superior to hard copy review (13,14), this method can be added to the set of tools available to radiologists at their workstations in a manner analogous to the current practice of prerendering cine loops of maximum intensity projection images. This allows the clinician to review the entire wall of the vessel rather than just the coronal and sagittal reformations.

We have developed a technique that automatically generates curved planar reformations that cannot be differentiated from manually generated ones. Our method leads to a reduced postprocessing time and may decrease artifacts. A simple user interface allows the creation of accurate paths through even tortuous vessels, without the need for expert user input. Interactive features allow quick correlation and confirmation with other views. The ability to rotate the image through 360° facilitates a better understanding of anatomic relationships and the lesion under study. This technique adds value to CT angiography and has the potential to promote its widespread acceptance and use. Although the technique was demonstrated with CT angiographic data, it should be equally applicable to MR angiographic data, but this was not tested.


   ACKNOWLEDGMENTS
 
The authors thank Laura Logan, RT (CT), and Marc Sofilos, RT (Stanford 3D Medical Imaging Laboratory, Stanford, Calif), for all manually created curved planar reformations and the Biomedical Imaging Research Unit, Auckland University Medical School, Auckland, New Zealand, for use of their computing facilities for abstract preparation.


   FOOTNOTES
 
2 9*. Vascular system, location unspecified Back

Author contributions: Guarantor of integrity of entire study, G.D.R.; study concepts and design, G.D.R., S.N., R.R.; literature research, R.R.; clinical studies, G.D.R., C.F.B., S.N., E.S.B., R.B.J.; data acquisition, G.D.R., C.F.B., S.N., E.S.B., R.B.J.; data analysis/interpretation, G.D.R., R.R.; statistical analysis, G.D.R., R.R.; manuscript preparation and definition of intellectual content, G.D.R., R.R.; manuscript editing, revision/review, and final version approval, G.D.R., S.N., R.R.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
Results
Discussion
REFERENCES
 

  1. Rubin GD, Shiau MC, Schmidt AJ, et al. Computed tomographic angiography: historical perspective and new state-of-the-art using multi detector-row helical computed tomography. J Comput Assist Tomogr 1999; 23(suppl 1):S83-S90.
  2. Rubin GD, Silverman SG. Helical (spiral) CT of the retroperitoneum. Radiol Clin North Am 1995; 33:903-932.
  3. Rankin SC. Spiral CT: vascular applications. Eur J Radiol 1998; 28:18-29.
  4. Rubin GD, Dake MD, Semba CP. Current status of three-dimensional spiral CT scanning for imaging the vasculature. Radiol Clin North Am 1995; 33:51-70.
  5. Fielding JR, Silverman SG, Rubin GD. Helical CT of the urinary tract. AJR Am J Roentgenol 1999; 172:1199-1206.
  6. Rubin GD. Helical CT angiography of the thoracic aorta. J Thorac Imaging 1997; 12:128-149.
  7. Rubin GD. Spiral (helical) CT of the renal vasculature. Semin Ultrasound CT MR 1996; 17:374-397.
  8. Ochi T, Shimizu K, Yasuhara Y, Shigesawa T, Mochizuki T, Ikezoe J. Curved planar reformatted CT angiography: usefulness for the evaluation of aneurysms at the carotid siphon. AJNR Am J Neuroradiol 1999; 20:1025-1030.
  9. Van Hoe L, Gryspeerdt S. Helical CT angiography of renal artery stenosis (letter). AJR Am J Roentgenol 1997; 168:1380-1381.
  10. Achenbach S, Moshage W, Ropers D, Bachmann K. Curved multiplanar reconstructions for the evaluation of contrast- enhanced electron beam CT of the coronary arteries. AJR Am J Roentgenol 1998; 170:895-899.
  11. Rubin GD. Three-dimensional helical CT angiography. RadioGraphics 1994; 14:905-912.
  12. Paik DS, Beaulieu CF, Jeffrey RB, Rubin GD, Napel S. Automated flight path planning for virtual endoscopy. Med Phys 1998; 25:629-637.
  13. Kirchgeorg MA, Prokop M. Increasing spiral CT benefits with postprocessing applications. Eur J Radiol 1998; 28:39-54.
  14. Zeman RK, Silverman PM, Vieco PT, Costello P. CT angiography. AJR Am J Roentgenol 1995; 165:1079-1088.



This article has been cited by other articles:


Home page
 RadiologyHome page
D. R. Hadizadeh, J. Gieseke, S. H. Lohmaier, K. Wilhelm, J. Boschewitz, F. Verrel, H. H. Schild, and W. A. Willinek
Peripheral MR Angiography with Blood Pool Contrast Agent: Prospective Intraindividual Comparative Study of High-Spatial-Resolution Steady-State MR Angiography versus Standard-Resolution First-Pass MR Angiography and DSA
Radiology, November 1, 2008; 249(2): 701 - 711.
[Abstract] [Full Text] [PDF]

Home page
 RadiologyHome page
R. Raman, B. Raman, S. Napel, and G. D. Rubin
Semiautomated Quantification of the Mass and Distribution of Vascular Calcification with Multidetector CT: Method and Evaluation
Radiology, April 1, 2008; 247(1): 241 - 250.
[Abstract] [Full Text] [PDF]

Home page
 RadiologyHome page
J. E. Roos, D. Fleischmann, A. Koechl, T. Rakshe, M. Straka, A. Napoli, A. Kanitsar, M. Sramek, and E. Groeller
Multipath Curved Planar Reformation of the Peripheral Arterial Tree in CT Angiography
Radiology, July 1, 2007; 244(1): 281 - 290.
[Abstract] [Full Text] [PDF]

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
 Am. J. Roentgenol.Home page
T. S. Desser, F. G. Sommer, and R. B. Jeffrey Jr.
Value of Curved Planar Reformations in MDCT of Abdominal Pathology
Am. J. Roentgenol., June 1, 2004; 182(6): 1477 - 1484.
[Full Text] [PDF]

Home page
 RadiologyHome page
R. Raman, S. Napel, and G. D. Rubin
Curved-Slab Maximum Intensity Projection: Method and Evaluation
Radiology, October 1, 2003; 229(1): 255 - 260.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Quicktime movie
Right arrow All Versions of this Article:
2231010441v1
223/1/275    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
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 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 Raman, R.
Right arrow Articles by Rubin, G. D.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Raman, R.
Right arrow Articles by Rubin, G. D.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH
RADIOLOGY RADIOGRAPHICS RSNA JOURNALS ONLINE