| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Innovation |
1
Department of Radiology, University Hospital Tuebingen, Hoppe-Seylerstr. 3,
D-72076 Tuebingen, Germany.
2
Present address: Department of Radiology, University Hospital Frankfurt,
Theodor-Stern-Kai 7, D-60590 Frankfurt, Germany.
3
Department of Medical Biometry, University Hospital Tuebingen, D-72076
Tuebingen, Germany.
4
Department of Surgery, University Hospital Tuebingen, D-72076 Tuebingen,
Germany.
Received July 10, 2000;
accepted after revision October 6, 2000.
Supported by a Fortune grant.
Introduction
 Top Introduction Subjects and MethodsResultsDiscussionReferences |
|---|
|
TopIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
For study enrollment, patients needed to be able and willing to provide written consent as set forth by the institutional human research review committee, which had approved this study. Exclusion criteria included the presence of a pacemaker, severe claustrophobia, pregnancy, and the clinical diagnosis of an acute abdomen.
After standard oral colonoscopic preparation (GoLytely; Braintee Laboratories, Braintee, MA), MR colonography was performed (Vision 1.5-T scanner; Siemens, Erlangen, Germany). To permit coverage of the entire colon and homogeneous signal reception, the body coil was used. After a disposable rectal enema tube (92MR; E-Z-EM, Westbury, NY) was in place, we asked the patients to assume the prone position on the scanner table. The enema tip was blocked with an insufflated balloon. To minimize peristalsis and alleviate colonic spasm, 20 mg of scopolaminbutylbromid (Buscopan; Boehringer Ingelheim, Ingelheim, Germany) was IV-injected. Subsequently, the enema, consisting of 3 L of water with 60 mL of 0.5 molar gadopentetate dimeglumine (Magnevist; Schering, Berlin, Germany) added, was administered via the rectal tube using up to 80 cm of hydrostatic pressure. We sought to ensure safe and optimal filling of the colon by monitoring the filling process with a nonslice select two-dimensional—spoiled gradient-echo sequence (TR/TE, 4.9/1.3; flip angle, 60°), which automatically provided an updated image every 2 sec. Once we had established that complete filling with adequate distention of the entire colon had been achieved, we performed a 3D spoiled gradient-echo sequence (4/1.6; flip angle, 45°) with the patient in the prone position during a breath-hold lasting 30 sec or less. The imaging parameters, field of view (34-42 cm), section-thickness (2-3 mm), and matrix size (384-256 x 192-160), were chosen to maximize spatial resolution within the confines of breath-holding and coverage of the entire colon.
Commercially available hardware (ultra SPARC 2; Sun Microsystems, Mountain View, CA) and software (Easy Vision 4.2; Philips, Best, The Netherlands) were used to render the 3D virtual double-contrast display. To achieve this display, the volume-rendering algorithm was used in a gradient mode with a lower threshold of 500 H, an opacity of 50%, and a translucency of 3%. The gradient mode calculates the differences between the signal intensities of neighboring voxels and renders the interface between voxels with different signal intensities opaque while keeping neighboring voxels with similar signal intensities translucent.
For the analysis, the colon was divided into eight segments including the six anatomic segments (rectum, sigmoid, descending colon, transverse colon, ascending colon, and cecum) plus the two flexures. On a computer screen, 3D double-contrast MR colonography was evaluated for each segment with regard to the diagnostic value (segment completely assessable or not) and diagnostic confidence (1 = segment definitely negative for a mass, 2 = probably negative for a mass, 3 = not certain, 4 = probably positive for a mass, 5 = definitely positive for a mass) in a coronal projection and subsequently in projections that could be interactively rotated in steps of 15°.
The analysis was independently performed by three radiologists experienced in analyzing double-contrast barium enemas who had no knowledge of any patient's history or colonoscopic findings.
The influence of the reviewer, viewing mode, type of segment, and two-way interaction between these factors on the diagnostic value were analyzed by logistic regression. Defining the ratings of 1 and 2 as a negative finding and from 3 to 5 as a positive finding, the interobserver agreement was analyzed by kappa statistics. The diagnostic confidence in the detection of colorectal mass lesions was evaluated for each reviewer and for every different viewing mode by receiver operator characteristic analysis using colonoscopy as the standard of reference.
|
TopIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
All polyps 10 mm or larger (n = 6) and all carcinomas (n = 8) were visualized on 3D double-contrast MR colonography (Fig. 2A,2B,2C,2D). The innumerable small polyps (<<10 mm) seen at colonoscopy in the patient with the attenuated polyposis coli and a 5-mm polyp in another patient were not visualized on 3D double-contrast MR colonography. Polyps could be clearly differentiated from diverticula.
|
|
|
|
Based on the segments commonly rated as completely assessable by all
reviewers in the coronal (n = 50) and rotated projections (n
= 95), there was moderate agreement in terms of positive and negative findings
for the coronal projection (for R1 vs. R2, 
=
0.52 ± 0.14; for R2 vs. R3, = 0.63
± 0.13; for R1 vs. R3, = 0.58 ±
0.13). The agreement was increased by rotating the 3D double-contrast MR
colonography (for R1 vs. R2, = 0.77 ±
0.07; for R2 vs. R3, = 0.76 ± 0.07; for
R1 vs. R3, = 0.68 ± 0.08).
Rotation of the 3D double-contrast MR colonography increased the diagnostic confidence and accuracy as reflected by the values of the area under the receiver operating curve (coronal: for R1, R2, and R3, Ai = 1-3 = 0.65, 0.83, and 0.76, respectively; rotation: Ai = 1-3 = 0.95, 0.91, and 0.85, respectively).
|
TopIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
10 mm) could be detected even with the
relatively low-voxel resolution of 2.4 x 1.6 x 1.2 mm3
used in this study (Fig.
2A,2B,2C,2D). The technique of selectively imaging colonic contrast filling allows performance of 3D double-contrast MR colonography with no prior segmentation. Therefore, images can be displayed immediately after the acquisition of data. With the use of adequate hardware and software, rotation of the 3D double-contrast MR colonography can be interactively performed with real-time image display.
In contrast to the projection imaging of the double-contrast barium enema, the 3D double-contrast MR colonography is based on a 3D data set, which allows a variety of interpretation methods, such as arbitrary rotation of 3D double-contrast MR colonography or variations in the slab thickness, that can provide successive building up of the 3D double-contrast MR colonography from the back to the front for tomographic views of interest. Because these methods appear to be as time-consuming as virtual colonoscopy, we tested only the impact of a standardized rotation on the image performance of the 3D double-contrast MR colonography. Rotation improved the diagnostic value (Fig. 1), confidence, and accuracy of 3D double-contrast MR colonography in the detection of mass lesions. For example, a carcinoma in the cecum could only be diagnosed when the 3D double-contrast MR colonography was rotated. However, the rotation has different consequences for the individual segments. The right flexure benefits the most from the rotation (Fig. 1), whereas the rectum often remains not completely assessable because of an inadequately chosen field of view or residual air. When used immediately after the data acquisition, the 3D double-contrast MR colonography allows the timely recognition of these problems so that imaging can be repeated if necessary with a larger field of view or with the patient in a different position to shift the air.
Besides controlling the quality of the acquired data set, 3D double-contrast MR colonography can be used simultaneously to analyze the data. All polyps 10 mm or larger (n = 6) and, in particular, all carcinomas (n = 8) were well visualized. Small polyps, however, were not displayed. The clinical relevance of missing these small polyps is still unclear. Because small polyps may physiologically occur in a patient over age 50 and are almost always benign [5], identifying them may increase the number of unnecessary polypectomies and the risk of colonic perforation rather than reduce mortality from colorectal cancer [6]. In addition, given the long precancerous polyp dwelling time of 5-10 years [5] and the correlation between polyp size and the risk of malignant transformation, an interval screening program likely justifies the introduction of a cutoff size for recommending therapeutic colonoscopy. Therefore, when applying a cutoff size of 10 mm, completely assessable and normal findings on 3D double-contrast MR colonography can help the physician or technician to finish the examination earlier without changing the position of the patient or performing subsequent data analysis. Patients with incompletely assessable segments on the double-contrast display require a change in position in cases of residual air and further examination via virtual colonoscopy in cases of redundant bowel loops. For obvious findings, the 3D double-contrast MR colonography currently appears to be the easiest and fastest way to allow first-view diagnosis. Analyzing the data by means of multiplanar reconstruction or virtual endoscopy is more time-consuming for diagnosing obvious findings than analyzing data by looking at the 3D double-contrast MR colonography.
In addition to allowing first-view diagnosis, 3D double-contrast MR colonography can be used for identifying suspicious findings that call for subsequent imaging or more detailed analysis. Data gleaned from double-contrast colonography or from CT can also be used for correcting an automatically calculated center line of a fly-through [7], for showing the results of an automated polyp detector [8], or for providing comprehensive communication of findings.
In conclusion, a 3D double-contrast MR colonography is feasible. It is helpful in deciding if further imaging is needed, and it is promising for quickly filtering out patients with obvious colorectal-mass lesions and for identifying healthy patients as long as their colon is completely assessable. Thus, 3D double-contrast MR colonography should be further enhanced (i.e., unraveled), automatically displayed immediately after the data acquisition, and further tested for its usefulness in rendering MR colonography less time-consuming and thus more cost-effective.
|
TopIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  R. W. F. Geenen, S. M. Hussain, F. Cademartiri, J.-W. Poley, P. D. Siersema, and G. P. Krestin CT and MR Colonography: Scanning Techniques, Postprocessing, and Emphasis on Polyp Detection RadioGraphics, January 1, 2004; 24(1): e18 - e18. [Abstract] [Full Text]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
