| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, Medical University of Vienna, Waehringer Guertel
18-20, A-1090, Vienna, Austria.
2 Academic Medical Center, Amsterdam, The Netherlands.
3 University Medical Center Utrecht, Utrecht, The Netherlands.
Received September 26, 2004;
accepted after revision December 6, 2004.
Address correspondence to T. G. Mang
(thomas.mang{at}akhwien.at).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Thirty-six polypoid and 39 flat polyps (44 lesions, 2-5 mm; 31 lesions, 6-8 mm) were placed in three explanted segments of a thoroughly cleaned porcine colon (overall length, 4.5 m) that was distended with air and submerged in a water phantom. MDCT data sets with 4 x 1 mm collimation and 6-mm table feed were reconstructed every 0.7 mm with 1.25-mm effective slice width. The data were reviewed by three radiologists using 2D images in all three projections and with 3D volume-rendered images available as an adjunct to the 2D images.
RESULTS. Additional 3D as a problem-solving tool significantly increased the overall sensitivity (96% vs 90%), decreased the total number of false-positive calls (n = 9 vs n = 5), and increased the diagnostic confidence level (p < 0.03) compared with 2D images alone. Small polyps less than or equal to 5 mm (89% vs 95%, p = 0.004) and flat polyps (82% vs 94%, p = 0.001) especially benefited from 3D. Sensitivity was generally higher for polypoid than for flat polyps (99% vs 94%, p = 0.041).
CONCLUSION. Under phantom conditions, simulating an ideal clinical setup, MDCT colonography is not limited by spatial resolution and detects polyps less than or equal to 5 mm in size with high sensitivity and specificity. Additional 3D image tools improve diagnostic accuracy and reviewer confidence, especially for the detection of flat and small polyps.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
There is no general agreement about the lower limit for the size of polyps that should be followed. Although the risk for malignant transformation of adenomatous polyps smaller than 10 mm is low, a small percentage showed high-grade dysplasia when removed after 3 years of observation [4]. As a result, controversy remains regarding the minimum polyp size for which the potential benefits of polypectomy outweigh the risks [5]. Although some favor a threshold of 0.8 cm [6], others advocate total polyp clearance [7].
CT colonography has become a viable imaging tool for evaluating the colon. Its suitability as a screening tool is under lively discussion. It appears, therefore, warranted to assess the diagnostic accuracy of this method with regard to lesions that require immediate therapeutic action and smaller polyps that could be followed in adequate time intervals [6].
Using single detector-row CT scanners and conventional colonoscopy as the gold standard, an average sensitivity between 70% and 100% has been reported for the detection of polyps larger than 10 mm; for polyps smaller than 10 mm, however, the sensitivity dramatically decreased with a quoted detection rate of less than 50% for polyps smaller than 5 mm [8-17]. Other limitations of single detector-row CT colonography include the low sensitivity for depicting flat polyps and the relatively high false-positive rate [18-20].
|
|
|
|
|
The purpose of this experimental study was twofold: to assess the sensitivity and false-positive rate of MDCT colonography for the detection of small and flat polyps under ideal imaging conditions using an anthropomorphic phantom model and to quantify the added diagnostic value of 3D colonography tools (endoluminal views and target views) when used as an adjunct to 2D images (axial and multiplanar reformats) alone.
|
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
The distribution of the lesions in the colon segments was carefully documented because it served as the gold standard for further statistical analysis. To affix the lesions on the colon wall, we used cyanoacrylate glue (Loctite, Henkel) that had a radiologic attenuation similar to soft tissue and did not influence the conspicuity of the lesions. The colonic segments were then turned right side out and the distal end (10 cm) of the specimen was tied air-tight with yarn. A plastic tube was placed in the proximal end (10 cm) of the colon, which was then also firmly fixed with triple-tied yarn. Thus, each colon segment yielded a length of 150 cm for further analysis.
Each of the three colon segments was separately placed in a dedicated 20 x 30-cm acrylic phantom box and distended with air through the insufflation tube (Fig. 4). Within the phantom box, each colon segment was arranged in multiple flexures, each with angulations ranging from 5° to 160°. The phantom box had the diameter and form of an abdominal transverse cross-section. The box was filled with water to simulate abdominal attenuation. Inside the phantom box, an acrylic plate was fixed horizontally in the upper third of the height to keep the air-distended colon segment from buoyancy in the center of the box. In this way, we ensured that the specimen was totally surrounded by water. The phantom box was closed watertight using a cover with two lockable holes. One of the holes was used to fill the phantom case with water, and the second was connected to the tube to distend the colon segment with air from outside the case.
Scan Technique and Image Processing
Helical scanning was performed using an MDCT scanner (Somatom Volume Zoom,
Siemens Medical Solutions). All of the prepared colon phantoms were scanned in
the supine position. An initial scout view was performed to ensure adequate
distention of the colonic segments. Scanning was performed with a 4 x 1
mm collimation and a table feed of 6 mm. Exposure settings were 120 kVp and
140 mAs (CT dose indexvol, 17.4 mGy). Images were reconstructed
every 0.7 mm with an effective slice thickness of 1.25 mm. The transverse
source images were transferred to a 3D computer workstation with commercially
available evaluation software (Vitrea 2, Vital Images) that included
multiplanar reformations, a volume-rendered endoluminal fly-through view, and
volume-rendering techniques for target viewing of cubic subvolumes around
areas of interest (e.g., suspected polyps).
CT Colonographic Data Interpretation
All images were reviewed on the computer workstation by three observers.
All observers were board certified and had more than 10 years of experience in
abdominal CT. Experience in interpreting CT colonographic studies at the time
of the review sessions was more than 2 years. The reviewers were not involved
in the phantom assembly, including the placement of the lesions. They were
informed about the size range and morphology of the lesions to look for but
they were unaware of the number and distribution of lesions. All review
sessions were performed on the same 19-inch CRT monitor in the same room under
similar conditions (subdued ambient light).
Planar cross-sectional images were used for the primary search process. Each observer assessed the cross-sectional images in the transverse, coronal, and sagittal planes. The 2D images were presented in monitor-filling format with wide window settings (width, 1,200 H; level, -100 H). Images were evaluated by interactively tracing through the images of the air-distended colonic lumen from one end to the other. With planar images in all three planes available, all detected abnormalities were digitally marked and rated using a three-point scale of confidence ranging from "equivocal finding" to "lesion probably present" and "definite lesion."
In a second review session, the supplementary value of 3D imaging as a problem-solving tool was assessed for all lesions determined in the first review as "equivocal" or "probable" lesion. For this purpose, volume-rendered endoluminal fly-through views and volume-rendered target views were available for evaluation. This combined 2D/3D evaluation of cross-sectional images, virtual endoscopy, and 3D target views yielded the final ratings for each observer. The time interval between the first and the second review sessions was between 3 and 4 weeks. To minimize the influence of fatigue, the interpreters reviewed only one colon segment (1.5 m, 25 polyps) per day, either 2D or 3D.
Statistical Analysis
We performed descriptive statistics with means of sensitivity and
false-positive rates separately for each reviewer. Sensitivity was defined as
the proportion of correctly localized lesions compared with all lesions, as
defined by the standard. The false-positive rate was defined as the proportion
of false-positive calls compared with all positive calls.
Sensitivities were calculated for all lesions and separately for subgroups of lesions with respect to lesion size and polyp type (polypoid and flat). For calculation of the sensitivity, only definite lesions were considered.
Significance of difference was tested using the McNemar test; a p value of less than 0.05 indicated significance. The Wilcoxon's rank sum test was applied to assess whether the reviewers' diagnostic confidence for correctly detected lesions significantly differed when using 2D images alone versus a combined 2D/3D approach.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
The mean diagnostic confidence (averaged over the three reviewers) for correctly localized lesions significantly increased (p = 0.03, Wilcoxon's rank test).
Impact of Polyp Size
No significant difference was found in the detection rates for subgroups of
different lesion sizes when only 2D images were evaluated: Lesions larger than
5 mm were seen in 91% and lesions less than or equal to 5 mm were seen in 89%
(Table 2).
|
Additional 3D images significantly increased the sensitivity for both subgroups: For lesions larger than 5 mm, detection increased from 91% to 98% (p = 0.031); for lesions less than 5 mm, it increased from 89% to 95% (p = 0.004).
Impact of Polyp Type
For flat polyps, the mean sensitivity with 2D images alone amounted to 82%
and was significantly lower compared with the sensitivity for ovoid polyps
(98%, p = 0). The availability of additional 3D techniques
significantly increased the average detection rate of flat polyps to 94%
(p = 0.001) but did not significantly affect the average detection
rate of polypoid lesions (98% vs 99%; p = 1)
(Table 3).
|
False-Negative Rates
We found a low rate of false-negative calls, which were, in all cases,
related to perception errors. Retrospectively, all missed lesions could be
correctly identified. Three polyps could not be displayed with virtual
colonography because of their location between adjacent haustral folds. In
these cases, however, correct identification was possible with 2D images.
False-Positive Rates
False-positive rates were generally low. With the combined 2D/3D approach,
reviewer 1 saw two false-positive lesions and reviewer 2 saw three
false-positive lesions, amounting to a false-positive rate of 0.9% and 1.4%,
respectively. Reviewer 3 had no false-positive calls. Causes for
false-positive interpretations were haustral folds and wall irregularities due
to extraluminal structures (e.g., wall adherent connective tissue or fat).
The number of misinterpretations decreased with the combined 2D/3D approach (three vs five for folds and two vs four for wall irregularities, respectively).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MDCT scanners offer a higher scanning speed, fewer motion artifacts, and substantially higher spatial resolution in all three spatial dimensions compared with single detector-row CT if MDCT is obtained with a thin-section protocol of approximately 1-mm section collimation. It can be expected that the improved speed and spatial resolution of MDCT will also lead to an improved sensitivity and specificity compared with single detector-row CT, especially for the detection of small and flat polyps.
Using a phantom set-up and ideal scanning parameters, we found a very high mean sensitivity of greater than 90% for flat polyps and polyps smaller than 9 mm. The availability of additional 3D display increased the mean sensitivity even further to 96%. Our results considerably exceeded the numbers that were reported for MDCT colonography in a clinical setting. Macari et al. [20] report a sensitivity of 93% (13/14) for polyps larger than 10 mm, but sensitivity dropped to 70% (19/27) for polyps 6-9 mm in diameter. For polyps of 5 mm or less, the sensitivity was only 12% (11/91).
Other authors who evaluated MDCT colonography reported a significantly reduced number of examinations with disturbing respiratory artifacts (52% vs 19%) or suboptimal colonic distention (61% vs 16%) compared with single-detector CT [22]. The fact that they could not find an increased sensitivity rate despite these improvements with respect to image quality is most likely because single detector-row and MDCT data acquisition were performed with an identical section width of 5 mm and that only polyps of 10 mm in size or greater were studied. The impact of section thickness on the detection rate, especially of small lesions (< 5 mm in size) was underscored by the results of two other studies. Rogalla and coauthors [23] describe a decrease in sensitivity from 96% to 74% when altering the width of the reconstructed sections from 1 to 5 mm. Taylor et al. [5] found that the detection of small polyps (2-4 mm in size) was optimum with a thin collimation of 1.25 mm and a pitch of 3, achieving sensitivity rates between 60% and 70%.
Our results further suggest that a combined 2D/3D approach that includes volume-rendered endoluminal fly-through views and volume-rendered target views is superior to an evaluation of 2D images alone, even when multiplanar reformats of all three projections are available. The 3D display was especially helpful for the evaluation of small wall abnormalities, which could be checked against the findings on cross-sectional images and volume rendering of a coned-down region of interest. Thus, not surprisingly, 3D had a more pronounced effect on flat polyps compared with polypoid polyps. The false-positive rate, although already low with 2D images, could be further decreased by evaluating 3D images. The number of correctly identified lesions and the mean diagnostic confidence significantly increased when both 2D and 3D imaging tools were available. In this aspect, our results are in agreement with results published by Hara et al. [11], who also reported a superiority of the combined 2D/3D approach for single detector helical CT data and with the results from Pickhardt et al. [6], who reported similar high sensitivity rates with MDCT colonography and a combined 2D/3D evaluation.
In Macari et al.'s [20] study that had found a considerably lower detection rate for small lesions, images were also obtained with a thin collimation (4 x 1 mm) MDCT technique, but the reviewers used axial images alone for primary detection of lesions. Further image processing, such as multiplanar reformations and 3D endoscopic views, were only applied when an abnormality was suspected on the axial images.
Our study was designed to assess the supplementary value of 3D as a problem-solving tool, not to compare the performance of a primary 3D versus a primary 2D approach. The reviewers used the 3D images for evaluation of lesions that had been determined as equivocal or indeterminate in the first review session that was based on multiplanar reconstructions. The 3Ds were not restricted to an endoluminal fly-through display but also consisted of volume-rendered target views. All three reviewers regularly used both options.
A recent study showed significant performance differences between different workstations when complete virtual endoscopy is used for the primary evaluation [24]. To our knowledge, no data have been published concerning the influence of different 3D systems on the performance of CT colonography when 3D is used only for problem solving. Even if there was an influence of the 3D system on the performance, we think that our main conclusion with respect to the additional value of 3D is independent of the software used.
All false-negative ratings in our study were caused by perception errors, and all missed lesions could be correctly identified retrospectively. These results indicate that under ideal conditions using optimum scanning parameters, spatial resolution is sufficient for display of such small lesions if no other disturbing factors, such as poor bowel distention or stool residues, are present.
Hara et al. [9] write that 55% of all missed lesions larger than 5 mm occurred because of perception errors. In that study, images were obtained with a 5-mm section thickness, and the authors note that insufficient spatial resolution of the imaging system may have also contributed to the misdiagnosis of small or flat lesions. Other reasons that have been mentioned under in vivo conditions include incomplete bowel distention, residual stool or fluid [14], small lesion size, and flat morphology [20, 25]. Yee et al. [16] report in their study that more than 70% of false-negative diagnoses are related to a poor preparation or inadequate distention.
Analysis of the false-positive ratings in our study identified two sources for misinterpretation: haustral folds and colonic wall impressions from intramural or extraluminal structures. Misinterpretation of haustral folds becomes more likely when only 2D image planes are evaluated. Our results confirm that haustral folds can usually be distinguished from true polyps by relying on a combined imaging approach of 2D and 3D images. It has to be noted that those folds misinterpreted as polyps in our study were small and short in shape (< 5 mm).
False-positive diagnoses resulting from circumscribed wall impressions from outside occur more often when evaluation is based on virtual endoscopy only. These impressions are caused by extraluminal connective tissue or fat tissue from colon loops that are lying tightly adjacent to each other. Three-dimensional endoluminal image evaluation alone precludes visualization of extraluminal structures and is particularly susceptible to this type of interpretive error.
We chose a phantom setup that included a sufficient number of subtle polypoid lesions and provided an absolute standard of truth for statistical analysis. Special care was taken to design an anthropomorphic phantom that simulated realistic anatomic conditions. The water basin was designed to simulate the scatter conditions of the abdomen. Each bowel segment was arranged in several curves to simulate realistic conditions of the colon. The intraluminal architecture of the colon model contained folds and wall irregularities. Other phantoms reported in the literature were considerably smaller than ours, with a bowel length of approximately 20-60 cm, and contained a smaller number of lesions. Lesions were formed using seed, yarn, meat balls, plastic material, or were digitally introduced into the image [26-29].
Another recently published phantom study is noteworthy because the investigators used a human colectomy specimen with polyposis coli containing 117 polyps from 1 to 15 mm [5]. Unlike our study, the bowel was arranged in a natural course of ascending, transverse, and descending sections. A separate analysis of the impact of segment orientation was therefore performed. On the other hand, as the authors write, a potential drawback of this phantom may have been that, unlike artificially produced phantom polyps and most naturally occurring adenomatous polyps, the tiny polyps in polyposis coli tend to form a shallow carpet throughout the entire colon with larger polyps intermittently interspersed. However, this is important because under realistic conditions, the target lesion is a single polyp. Another phantom study by Johnson et al. [29] used a phantom consisting of a boro-silicate tube with simulated soft-tissue polyps (12, 10, 7, and 5 mm in diameter). This phantom, however, differed from ours with respect to the material used and the polyp size.
We formed our lesions from muscle tissue because it shows soft-tissue attenuation and can be easily attached to the bowel wall using cyanoacrylate glue, with soft-tissue-equivalent density. The glue had no influence on the conspicuity of the lesions. Polypoid lesions had a diameter and height of 2 to 8 mm. Since flat polyps must be raised to be detected, we chose to create thin sheets of muscle tissue with a diameter of 2 to 8 mm and a height of only 0.5 to 2 mm to simulate such lesions. The availability of an absolute standard of truth appears to be important, considering that in most previous studies, conventional colonoscopy was chosen for comparison [17]. Conventional colonoscopy, however, has been shown to overlook up to 24% of all adenomas, which calls into question the appropriateness of this technique as a standard [30].
The following points may limit the extent to which our results can be transferred into a clinical environment. Similar to Taylor et al.'s [5] phantom study using a human specimen, examination conditions were ideal in our phantom setup. Factors such as suboptimal distention and fecal residuals or artifacts due to colon movement or respiration, which all have an impact on the diagnostic performance in vivo, were reduced to a minimum or were nonexistent.
We focused on the detectability of rather small polyps to exclude the fact that spatial resolution is the limiting factor for decreased detection rates of small lesions. The clinical importance of small polyps (< 1 cm), however, is still controversial [18]. Their tendency for malignant transformation is known to be low, thus their immediate therapeutic impact is limited [18]. Little is known about their growth rates or whether they should be followed and, if so, at what time intervals [31].
We conclude that MDCT colonography, under ideal conditions, is technically able to provide detection rates for small and flat polyps that are substantially higher than originally reported. When comparing these results against reports from the literature, the prerequisites for such high detection rates appear to be thin-section imaging and ideal examination conditions with optimum bowel cleansing and good bowel distention. The spatial resolution of CT no longer appears to be a limiting factor. A combined 2D/3D approach for image evaluation is optimum because of the complementary role of 2D and 3D images to provide luminal and extraluminal information.
Acknowledgments
We thank Dr. Michael Weber for statistical assistance.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  P. J. Pickhardt, A. D. Lee, A. J. Taylor, S. J. Michel, T. C. Winter, A. Shadid, R. J. Meiners, P. J. Chase, J. L. Hinshaw, J. G. Williams, et al. Primary 2D Versus Primary 3D Polyp Detection at Screening CT Colonography Am. J. Roentgenol., December 1, 2007; 189(6): 1451 - 1456. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
