| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Institute of Diagnostic Radiology, Inselspital, University of Berne,
Freiburgstrasse 10, Berne 3010, Switzerland.
2 Department of Gastroenterology, Inselspital, University of Berne, Berne,
Switzerland.
Received September 29, 2003;
accepted after revision November 12, 2003.
Address correspondence to H.-P. Dinkel
(hans-peter.dinkel{at}insel.ch).
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. CT colonography was performed in 22 patients using 4-MDCT followed by conventional colonoscopy on the same day. The CT colonography data sets were analyzed by virtual colon dissection, which virtually bisects and unfolds the colon along its longitudinal axis to inspect the inner colonic surface for polyps. The same CT data sets were independently evaluated using axial interpretation. All data sets were independently interpreted by two radiologists in a blinded manner.
RESULTS. Conventional colonoscopy revealed 31 colonic lesions in 20 patients. Twentytwo of the lesions were smaller than 10 mm; nine were 10 mm or larger. Two of the original 22 patients were excluded, one because of residual stool and fluid and the other because of an impassable stenosing rectal wall cancer. For virtual colon dissection, the per-lesion sensitivity was 42% for observer 1 and 68% for observer 2; for axial interpretation, the respective sensitivities were 48% and 61%. For polyps 10 mm or larger, the respective sensitivities were 67% and 89% for virtual colon dissection and 89% and 100% for axial interpretation. The average time for reconstruction and analysis of virtual colon dissection was 36.8 min versus 29.2 min for axial images. Virtual colon dissection was feasible in both the supine and the prone positions in 45.5% of colonic segments, in either the supine or the prone position in 24.5%, and in neither position in 30% of segments.
CONCLUSION. Although virtual colon dissection may facilitate detection of colonic polyps in isolated cases, its detection rate is not superior to axial interpretation, which is mainly attributable to failed rendering of insufficiently distended colonic segments or regions with residual feces. Virtual colon dissection is also the more time-consuming of the two procedures. With further improvement of path-finding and image segmentation, however, virtual colon dissection has the potential to be a useful interpretation tool for CT colonography.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
A major drawback to CT colonography, however, is the need to view an enormous amount of imaging data, which can involve as many as 1,000 axial sections for combined supine and prone studies. The debate continues about whether review of axial sections alone is adequate or if forms of 3D display are necessary to increase polyp detection [6, 7]. Primarily, CT colonography data sets are often interpreted by viewing axial images [8, 9]. It has been reported that small polyps, especially those residing on colonic folds, may be easily missed by isolated viewing of axial images [10, 11]. In such cases, 3D problem-solving is often not applied because no suspicious lesions are noted on the axial images. Investigators have reported that the use of 3D fly-through viewing of the whole colon, either manual or semiautomatic, is both tedious and time-consuming [4, 12]. New, more user-friendly 3D visualization techniques for analysis of CT colonography are needed. We chose the new virtual colon dissection visualization technique because it is purported to be an efficient way to inspect the inner colonic surface for polyps by virtually bisecting and unfolding the colon along its longitudinal axis. In this study, we compare the sensitivity and time for reconstruction of virtual colon dissection for the detection of colonic polyps with those of axial interpretation.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CT Colonography
On the day before the examinations, each patient was given a wet bowel
preparation consisting of 4 L of methylcellulose solution as prescribed by the
participating gastroenterologist. CT colonography was performed with an
Asteion 4-MDCT scanner (Toshiba). A flexible rubber catheter with a rectal
balloon was inserted into the rectum by the investigating radiologist. The
patient's colon was then insufflated with air according to the patient's
tolerance. The catheter was clamped and left in the rectum, and a single
supine scout CT image was obtained to verify adequate bowel distention. If
bowel distention was inadequate, additional air was insufflated into the
rectum. Once bowel distention was adequate, CT colonography was performed,
first in the supine position in a craniocaudal direction to image the entire
region of the colon and the rectum. The supine scan was started after power
injection of 120 mL (flow rate, 3 mL/sec; scanning delay, 60 sec) of iopromide
(Ultravist 300, Berlex Laboratories) IV contrast medium containing 300 mg
I/mL. The examination was repeated in the prone position without additional
injection of contrast medium.
CT parameters included 4 x 2 mm detector collimation, 120 kV, 0.75-sec gantry rotation, 200 mAs, and a pitch of 1.375. The entire abdomen and pelvis were scanned during a breath-hold of approximately 30 sec. Axial CT images were reconstructed as 2-mm slices with a 1-mm reconstruction interval.
Image Processing
The reconstructed supine and prone data sets were transferred to an
Advantage Windows workstation (version 4.0, General Electric Medical Systems)
running on an Ultra Sparc 60 hardware (Sun Microsystems) featuring two Sun
Ultra Sparc II 450-MHz central processing units and 2 GB of random access
memory. A single monitor system was used. The supine and prone data sets were
electronically anonymous. Image analysis was performed with commercially
available CT colonography software (Voxtool 3.0.51f, General Electric Medical
Systems) using a combination of 2D and 3D reformatted images.
Three-dimensional virtual endoscopic images were reformatted with a
surface-rendering algorithm with automated threshold values. The same software
package also provides a virtual colon dissection tool that virtually bisects
and unfolds the colon along its longitudinal axis. For this purpose, a central
colon path was first defined in a semiautomatic mode. Various points were then
inserted within the colonic lumen. At least two points for start and end of
the path had to be selected. The virtual camera captured one fourth of the
circumference of the complete colon length at each viewing position with a
90° camera field of view that could be rotated in 45° increments (Fig.
1A,
1B,
1C). Eight contiguous panels
displaying the colonic circumference were generated by rotating the virtual
camera in 45° increments around the path. The software is interactive,
allowing the viewer to switch between 2D and 3D views for confirmation of
polyps found using virtual colon dissection on additional views.
|
|
|
CT Image Analysis
All CT colonographic images were interpreted by two radiologists who were
unaware of patient identity or medical history. Both observers were
experienced in the interpretation of CT colonography data sets with 2D and 3D
images in a clinical practice. Before interpretation of the study cases, each
observer analyzed five data sets with virtual colon dissection to become
familiar with this new 3D visualization technique. The study cases were
assigned in random order. Supine and prone acquisitions for each patient were
interpreted in turns in one interpretation session but were not viewed
simultaneously. In the first interpretation session, the supine and prone data
sets were viewed as continuous 2-mm axial sections with high-contrast window
display settings (width, 2,000 H; level, 0 H). This window display setting was
chosen to provide high contrast for polyp detection. For confirmation of
polyps, this window setting was varied. In a second independent interpretation
session several weeks later, the virtual colon dissection images were reviewed
using this method. For each patient and each interpretation session, both
supine and prone image data sets were evaluated for the presence or absence of
polyps. When an abnormality or a suspicious lesion was detected on axial CT
colonographic images, additional views were evaluated to verify the
morphologic features of the lesion. If no abnormality was seen or suspected at
axial image review, no further image processing was performed.
The presence, location, size, and morphologic features of the colorectal lesions were assessed in six colonic segments (cecum, ascending colon, transverse colon, descending colon, sigmoid, and rectum). The head but not the stalk of each polyp was measured. If an abnormality was detected, the internal attenuation—that is, of gas bubbles, high-attenuating areas, or homogeneous attenuation—was carefully inspected using a variety of window and level settings. To differentiate stool or colonic folds from polyps, the morphologic features of the abnormality were evaluated on multiplanar CT views. Lesions with linear margins were regarded as residual fecal material. Endoluminal CT images and multiplanar reformations were used to differentiate a linear fold from a round polypoid.
A record was made of the evaluation time, defined as the time it took to review the CT data sets and produce a statement regarding the presence or absence of polyps. For virtual colon dissection, reconstruction time was recorded separately. Reconstruction time was the time needed to define a path within the colonic lumen and to reconstruct a model of the inner colonic surface. A record was also made of the number of attempts required for reconstruction and how many segments of the colon could be evaluated with virtual colon dissection. If a segment could not be reconstructed with virtual colon dissection, it was either not displayed or replaced by a so-called bridge connecting the previous and the following segments but devoid of diagnostic content (Fig. 2). The number of bridges within the virtual colon dissection model was recorded for each patient. Also, the number of software breakdowns was recorded. The observers also recorded the presence or absence of residual intraluminal stool or fluid retention for all prone and supine segments (rectum, sigmoid colon, descending colon, transverse colon, ascending colon, and cecum) in all CT views.
|
Conventional Colonoscopy
All polyps identified at colonoscopy were photographed and biopsied for
histopathologic analysis. Polyps were measured in millimeters using the open
biopsy forceps technique [13].
The location of each polyp was confirmed by concomitant fluoroscopy and mapped
according to the same six colonic segments (rectum, sigmoid colon, descending
colon, transverse colon, ascending colon, and cecum) evaluated at CT analysis.
Conventional colonoscopy was performed without knowledge of the CT findings.
In all cases in which a polyp was detected at colonoscopy, a pathologist
reviewed the biopsy material. The pathology report on the biopsy material was
reviewed to determine the histopathologic features of all the polyps for which
biopsy was performed.
Data Comparison
The findings of the axial interpretation session and of the virtually
dissected endoluminal CT images were compared separately from those of
conventional colonoscopy. Qualitative results on the presence of polyps were
classified on a per-polyp basis. The analysis was performed at two colonic
lesion size thresholds (< 10 mm and 
10 mm).
Statistical Analysis
Conventional colonoscopic findings were the reference standard for the
presence of colorectal lesions. Colonic lesion size determined at colonoscopy
was used as the reference for calculation of sensitivity and specificity by
size category [14].
Sensitivity and positive predictive values were calculated using conventional
colonoscopy findings as a reference standard. Specificity and negative
predictive values were not calculated for methodologic reasons. To compare
differences between the observers and methods, individual sensitivity results
were examined for heterogeneity using a chi-square test.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
Virtual Colon Dissection
Using virtual colon dissection, observer 1 was able to evaluate the supine
and prone images in 44% (53/120) of segments, either but not both supine or
prone images in 23% (27/120), and neither supine nor prone images in 33%
(40/120) of segments. The complete colon could not be evaluated in 30% (6/20)
of patients. Stool or fluid retention prevented evaluation of either the prone
or supine view in five segments and of both the prone and supine views in
another five segments. The software identified a false path in five patients.
In 34 segments, a bridge containing no diagnostic information was observed
either in the prone or the supine view. A complete breakdown of the software
spontaneously occurred four times without any apparent reason. The software
had to be restarted and path definition and reconstruction of virtual colon
dissection had to be repeated. In 11 patients, more than two approaches were
needed for reconstruction of the supine and the prone views (average number of
approaches, 3.9; range, 2–11).
Observer 2 was able to evaluate the supine and prone images in 47% (56/120) of segments, either but not both supine or prone images in 26% (31/120), and neither supine nor prone images in 28% (33/120) of segments. The complete colon could not be evaluated in 20% (4/20) of patients. Stool or fluid retention prevented evaluation of either the prone or supine view in 13 segments and of both the prone and supine views in another two segments. The software identified a false path in four patients. In eight segments, a bridge containing no diagnostic information was observed in either the prone or the supine view. Two times the software had to be restarted and path definition and reconstruction of virtual colon dissection had to be done again. In nine patients, more than two approaches were needed for reconstruction of the supine and prone views (average number of approaches, 3.1; range, 2–9).
Observer 1 required an average of 18.5 min (range, 4–48 min) for rendering of virtual colon dissection; observer 2, 13.2 min (range, 5–36 min). The mean rendering time for both observers was 15.9 min (range, 4.5–42 min). Observer 1 needed an average interpretation time of 22.5 min (range, 9–40 min) for analysis of virtual colon dissection, and observer 2 needed 19.3 min (range, 7–33 min). The mean interpretation time for the two observers was 20.9 min (range, 8–36.5 min). The total time for analysis of virtual colon dissection was 41 min (range, 15–61 min) for observer 1 and 32.5 min (range 11–64 min) for observer 2. The mean analysis time for both observers was 36.8 min (range, 13–62.5 min).
Table 1 shows the per-lesion sensitivities for detection of colonic lesions. For lesions smaller than 10 mm, individual per-lesion sensitivity was 32% (7/22) for observer 1 and 59% (13/22) for observer 2 (p = 0.10). For lesions 10 mm and larger, individual per-lesion sensitivity was 67% (6/9) for observer 1 and 89% (8/9) for observer 2 (p = 0.48). Per-lesion sensitivity for colonic lesions of all sizes was 42% (13/31) for observer 1 and 68% (21/31) for observer 2 (p = 0.08). The results of both observers were statistically not significant. Five sessile polyps (one of 4 mm, two of 5 mm, two of 8 mm) were identified on virtual colon dissection by at least one observer but were not depicted on axial interpretations by both observers.
|
Axial Interpretation
Individual per-lesion sensitivity of axial CT interpretations for lesions
smaller than 10 mm was 32% (7/22) for observer 1 and 45% (10/22) for observer
2 (p = 0.34) (Table
1). For lesions 10 mm and larger, individual per-lesion
sensitivity was 89% (8/9) for observer 1 and 100% (9/9) for observer 2
(p = 0.74). Per-lesion sensitivity for detection of lesions of all
sizes was 48% (15/31) for observer 1 and 61% (19/31) for observer 2
(p = 0.36). The results of both observers were statistically not
significant. Furthermore, no statistically significant difference between the
results of axial interpretation and virtual colon dissection
(Table 1) was seen.
For axial slices, the average interpretation time was 30.8 min (range, 18–45 min) for observer 1 and 27.6 min (range, 15–41 min) for observer 2. The mean analysis time for both observers was 29.2 min (range, 16.5–43 min), which was significantly different from that of virtual colonic dissection (p = 0.005).
Viewing axial slices, the observers were not able to evaluate the colon of one (5%) patient (six segments) because of excessive stool and fluid retention. Observer 2 was unable to evaluate four more segments (one each of rectum, descending colon, transverse colon, and ascending colon) in various patients because of excessive stool and fluid retention.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Several important issues regarding the use of CT colonography remain unsettled, including the optimal methods of image rendering and diagnostic interpretation [3, 12, 15, 16]. Image rendering and interpretation pose particular challenges because CT colonography produces a relatively large volume of raw data as a result of its need for supine and prone acquisitions and thin slices [17].
Various computer processing techniques are currently being investigated for their ability to shorten interpretation times and increase diagnostic accuracy, desired features for CT colonography. Among the promising techniques are computer methods for postprocessing volumetric data sets to ensure visualization of the entire colonic mucosal surface [10, 18–22]. We chose the new virtual colon dissection visualization technique because it was purported to be an efficient way to inspect the inner colonic surface for polyps by virtually bisecting and unfolding the colon along its longitudinal axis.
Whereas 3D virtual endoscopy is often thought to present the complete colonic inner surface for polyp detection, in reality polyps located around corners or behind haustral folds may not be within the visible portion of the colon. This problem can be minimized only through forward and reverse viewing of both supine and prone images. This process does not, however, ensure full coverage and is extremely time-consuming [10].
Alternatively, 3D virtual endoscopy of the complete colon may be replaced by interactive interpretation of axial images using manual paging to scan the data set for suspicious lesions with selected 3D luminal views for problem solving, a method reported to be quite time-efficient for initial data screening [4–7].
In our study, the average interpretation time was longer for virtual colon dissection (36.8 min) than for axial interpretations (29.2 min). This difference was mainly due to the relatively long rendering time for the 3D virtual colon dissection model. Failure in path-finding of the central colonic axis and software breakdowns additionally prolonged our interpretation times. With the present software version, the process of obtaining a central axis path for virtual colon dissection may be time-consuming unless the colon is totally clean and distended, which was seen in a minority of our patients. Using virtual colon dissection, complete evaluation of the colon was not possible in 30% of our patients for observer 1 and in 20% for observer 2, rates not acceptable for clinical use or screening.
Previous studies have reported a per-polyp sensitivity for polyps larger than 10 mm of 90%, which decreased to 50% for polyps 5 mm in size [3]. In our study, the average sensitivities for lesions larger than 10 mm were 89% (observer 1) and 100% (observer 2) for axial interpretation and 67% (observer 1) and 89% (observer 2) for virtual colon dissection. The respective sensitivities for lesions smaller than 10 mm were 32% (observer 1) and 45% (observer 2) for axial interpretation and 32% (observer 1) and 59% (observer 2) for virtual colon dissection. Although the sensitivity of observer 2 for lesions smaller than 10 mm was slightly higher for virtual colon dissection than for axial interpretation, the sensitivity of virtual colon dissection for lesions larger than 10 mm was not higher than for axial interpretation. The diagnostic accuracy of virtual colon dissection for clinically relevant polyps of 10 mm or larger does not exceed that of axial interpretation [2, 23]. This could also be related to the fact that IV contrast material was administered in our study, because colorectal polyps and carcinomas show enhancement on axial images but not with virtual colon dissection. This method may help to increase the conspicuity of polyps and to differentiate solid lesions from residual colonic fluid and stool on axial images [24].
In our study, high-risk patients with positive results on fecal occult test, iron deficiency anemia, or suspected polyps were examined with CT colonography. This small group of patients showed a high prevalence of polyps and colorectal carcinoma, which may facilitate a higher detection rate for colonic lesions than in a low-prevalence setting with asymptomatic patients [25]. Accordingly, in a larger low lesion prevalence population reflective of the screening setting, the polyp detection rate of virtual colon dissection may be even lower.
With CT colonography, false-positive results may be found with poor colon distention, retained stool, diverticular disease, or thickened and complex haustral folds [26]. In this study, four of seven patients with false-positive results were judged to have stool and fluid retention by both observers, which rendered polyp identification difficult on both axial and 3D images. Five sessile polyps smaller than 10 mm were missed by both observers on the axial images because they were thought to represent a fold. Because these lesions could be identified on virtual colon dissection, one advantage of virtual colon dissection over axial interpretation may be the detection of smaller lesions (Fig. 4A, 4B, 4C, 4D).
|
|
|
|
Both observers in our study missed five polyps on axial interpretation and virtual colon dissection because of retained stool that could not be differentiated from polyps. Of the two procedures, virtual colon dissection was more likely to miss polyps because the software failed to render colonic segments not sufficiently distended or containing residual fecal material. Both of our observers failed to detect a sigmoid wall cancer, and one observer missed a rectal wall cancer (both lesions > 10 mm). A 15-mm-long stalked polyp sitting on a fold was detected by only one observer because of a rendering failure of the other observer (Fig. 5A, 5B, 5C). On the axial interpretation, both wall cancers and the long-stalked polyp were correctly identified. Use of both the supine and the prone positions for CT colonography is reported to improve evaluation of the colon, especially of collapsed colonic segments, and to increase the sensitivity for polyp detection [27]. Five small sessile lesions (5–8 mm) were missed on virtual colon dissection in our patients because the observers were able to render only the supine or prone views of the segment but not both, which hampered lesion identification within residual fluid.
|
|
|
With curved colonic sections, such as colonic flexures, virtual dissection and straightening of the colon may result in distortion of the generated 3D image of the inner colonic surface [19]. This may cause polyps to be skipped or even duplicated. In our study, a long-stalked polyp residing on a fold showed distortion because of the elongated, flattened rendering of the inner colonic surface, but the conspicuity of the polyp was still maintained (Fig. 5A, 5B, 5C).
If the bowel contains residual fluid, which accumulates in the dependent portions of the bowel, the intraluminal air column is interrupted and the automatic calculation of the navigation path is disturbed. Virtual colon dissection provides a tool, termed "bridging," to overcome this problem, but such bridges do not contain diagnostic information. In our study, bridging proved unreliable and frequently failed. If more than two occlusions or torsions of the colon were present, it was nearly impossible to determine a complete central axis path because of software collapse or repeated display of "path errors." Such errors were corrected by repeating the image segmentation, which took additional time to calculate. A complete breakdown of the software occurred spontaneously several times without any apparant reason, leading to abortion of the interpretation session and requiring rebooting of the software. Path definition and reconstruction of virtual colon dissection had to be done again, which was time-consuming.
In conclusion, although virtual colon dissection did facilitate detection of individual small colon polyps, its detection rate was not superior to that of axial interpretation, which is mainly attributable to rendering failure of insufficiently distended colonic segments, or in regions with residual feces. Virtual colon dissection was also more time-consuming. We therefore cannot fully support its sole use in clinical practice at this time, although it may be applied for problem solving or as a back-up method to detect polyps. With further improvement in path-finding and image segmentation, virtual colon dissection may prove to be a helpful tool for radiologists interpreting CT colonography studies.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  D. Hock, R. Ouhadi, R. Materne, A.-S. Aouchria, I. Mancini, T. Broussaud, P. Magotteaux, and A. Nchimi Virtual Dissection CT Colonography: Evaluation of Learning Curves and Reading Times with and without Computer-aided Detection Radiology, September 1, 2008; 248(3): 860 - 868. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. D. Johnson, J. G. Fletcher, R. L. MacCarty, J. N. Mandrekar, W. S. Harmsen, P. J. Limburg, and L. A. Wilson Effect of Slice Thickness and Primary 2D Versus 3D Virtual Dissection on Colorectal Lesion Detection at CT Colonography in 452 Asymptomatic Adults Am. J. Roentgenol., September 1, 2007; 189(3): 672 - 680. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Glick Will 3D Virtual Dissection Display Prove to Be the Display Format for Reviewing CT Colonography Examinations? Radiology, September 1, 2007; 244(3): 629 - 630. [Full Text] [PDF]
|
|
|
|
|
|
S. H. Kim, J. M. Lee, H. W. Eun, M. W. Lee, J. K. Han, J. Y. Lee, and B. I. Choi Two- versus Three-dimensional Colon Evaluation with Recently Developed Virtual Dissection Software for CT Colonography Radiology, September 1, 2007; 244(3): 852 - 864. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. C. Silva, C. V. Wellnitz, and A. K. Hara Three-dimensional Virtual Dissection at CT Colonography: Unraveling the Colon to Search for Lesions RadioGraphics, November 1, 2006; 26(6): 1669 - 1686. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. Yasumoto, T. Murakami, H. Yamamoto, M. Hori, R. Iannaccone, T. Kim, H. Abe, M. Kuwabara, K. Yamasaki, N. Kikkawa, et al. Assessment of Two 3D MDCT Colonography Protocols for Observation of Colorectal Polyps Am. J. Roentgenol., January 1, 2006; 186(1): 85 - 89. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Macari and E. J. Bini CT Colonography: Where Have We Been and Where Are We Going? Radiology, December 1, 2005; 237(3): 819 - 833. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. C. Silva, A. K. Hara, J. A. Leighton, and J. P. Heppell CT Colonography with Intravenous Contrast Material: Varied Appearances of Colorectal Carcinoma RadioGraphics, September 1, 2005; 25(5): 1321 - 1334. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
