| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Radiology, Mayo Clinic Rochester, 200 First St. SW, Rochester,
MN 55905.
2 Department of Health Sciences Research, Section of Biostatistics, Mayo Clinic
Rochester, Rochester, MN 55905.
Received June 18, 2003;
accepted after revision October 13, 2003.
Supported by National Institutes of Health grant R01–CA7533.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. In this prospective and retrospective study, 500 asymptomatic patients at high risk for colorectal cancer underwent CT colonography and colonoscopy. Each CT data set was interpreted by two independent observers, who were unaware of endoscopic findings, using a method of searching through enlarged axial images to detect intraluminal lesions. Another observer identified and characterized lesions missed at prospective interpretation. Polyps were assessed for size, method of visualization, intrinsic and extrinsic features, and examination quality.
RESULTS. We found 116 polyps at least 5 mm in diameter, 54 (47%) of which were missed by at least one of the prospective observers. Polyps seen in only one position were missed more often than polyps seen in both supine and prone positions (84% vs 50%, p < 0.01). Polyps located in suboptimally prepared colonic segments or along a thickened colonic wall were more frequently missed (p = 0.02 and p = 0.05, respectively). Endoscopic morphology and irregular surface contour were associated with missed lesions of all sizes (p = 0.03 and p = 0.04, respectively). Rounded intraluminal lesions were detected more often than other morphologies on CT (p = 0.04).
CONCLUSION. Factors that influence the likelihood that a polyp may be missed at interpretation of CT colonography include being seen only in one position, having flat endoscopic or CT morphology, having surface irregularity, and being located in a poorly prepared segment or along a thickened colonic wall. Understanding these features should lead to improved polyp detection on CT colonography.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The introduction of MDCT, new rendering software, and computer-aided diagnosis promises to improve the performance of this technique. However, the experience and skill of the interpreting radiologist have been shown to be paramount factors in CT colonography [8–10]. Problem-solving techniques have been previously presented [11, 12], with those descriptive reports focusing primarily on pseudolesions (i.e., false-positive results). Recently, results in large screening trials have shown wide interobserver variability in interpreting CT colonography [13, 14]. The recent trial in CT colonography performed by the American College of Radiology Imaging Network also described similar findings [15]. These data suggest that differences in polyp appearance may be recognized by one observer and dismissed or not recognized by another. Many potential variables affecting polyp detection on CT colonography have not been examined in a systematic fashion.
Factors that influence polyp conspicuity can be divided into those that are intrinsic to polyps themselves and those that are extrinsic to polyps—that is, related to the local milieu of the colonic structures and contents. Intrinsic factors affecting polyp conspicuity are polyp size, morphology, and surface contour [16]. Extrinsic factors include the relationship of a polyp to folds or flexures, distention or collapse of a colonic segment, and the amount of retained fluid or stool. The purposes of this study were to systematically characterize the imaging features of all polyps in a high-risk screening population, including mode of visualization (positioning and 2D and 3D imaging); intrinsic features such as size, morphology, and surface contour; and extrinsic features such as relationship to folds and flexures, colonic segment, quality of preparation (retained stool, fluid, and distention) in a single large series; to compare the features of polyps identified prospectively with those not prospectively detected; and to suggest ways to minimize perceptive error.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Image Acquisition
Ten minutes before scanning, patients received a subcutaneous injection of
1 mg of glucagon (to eliminate cramping), which was followed by gentle rectal
air insufflation using CO2 to patient tolerance (i.e., a feeling of
fullness without pain). Complete colon inflation was confirmed with CT scout
images. After being scanned in the supine position, the patient was turned to
the prone position and additional air insufflation was performed to patient
tolerance. A CT scout image verifying adequate inflation was again obtained.
Prone CT images were then acquired. Each acquisition required an
approximately-20-sec breath-hold.
Technical Parameters
Helical scanning was performed with the following parameters on a
single(HiSpeed Advantage, General Electric Medical Systems) or multidetector
(LightSpeed QX/I or LightSpeed Plus, General Electric Medical Systems)
scanner. Single-detector parameters included a 5-mm collimation, 6.5 mm/sec
table speed (pitch, 1.3), 3-mm reconstruction interval, 70 mAs, 120 kVp, 512
x 512 matrix, standard reconstruction algorithm, and 20-sec
breath-holds. Acquisitions between breathholds were performed with a 3-cm
overlap to avoid not scanning any region of colon. MDCT scans were obtained
with a 5-mm slice thickness, 15 mm/sec table speed, HQ mode, 2.5-mm
reconstruction interval, field of view to fit, 80 mAs, 120 kVp, 512 x
512 matrix, and standard reconstruction algorithm. Conventional colonoscopy
was performed later the same day in all patients.
Prospective Polyp Detection
Image processing was performed on a customized computer workstation
(Ultra-2; Sun Microsystems) that was developed at our institution
[17]. Three radiologists with
experience in interpreting more than 200 CT colonography examinations served
as observers. Two of these three radiologists interpreted each data set and
were unaware of the results of conventional colonoscopy. All data sets were
first interpreted by panning through enlarged supine or prone axial images on
a 19-inch (48-cm) monitor using both lung (level, –500 H; width, 2,000
H) and soft-tissue (level, 0 H; width, 500 H) window settings. Suspicious
filling defects were further characterized with 2D multiplanar reformatted
images plus 3D perspective volume-rendered images, with interactive comparison
of supine and prone data sets. The location (axial slice number and colonic
segment), size (from 2D axial or reformatted images), and visibility of polyps
on supine and prone images were recorded. Radiologists were instructed to
ignore polyps smaller than 5 mm.
CT colonography results were compared with the colonoscopy report and the pathology report. For the location of a lesion, conventional colonoscopy was considered to be the reference standard. For these purposes, the colon was considered to have eight segments (i.e., rectum, sigmoid, descending, splenic flexure, transverse, hepatic flexure, ascending, and cecum). The size of an individual lesion was characterized according to the pathology report unless the lesion was removed in pieces, in which case the endoscopic size estimate was used. All polyps visualized at endoscopy were included in the analysis unless histologic analysis showed normal mucosa (n = 5), focal glandular distortion (n = 1), or an inflammatory pseudopolyp (n = 1). If a CT colonography lesion was reported within one segment of an endoscopic lesion and the reported size was within 50% of the size of the pathologic or endoscopic size, that was considered a true-positive CT colonography result. Polyp morphology was classified according to the colonoscopy report as sessile, flat, or pedunculated. Flat lesions were defined as those with a base at least twice as wide as the height.
One of the two prospective observers also assessed the global quality of the data sets with respect to fluid, stool, bowel distention, and breath-holding artifacts [13]. Fluid was characterized in four categories: none, some fluid without diagnostic compromise, moderate amount of fluid with diagnostic compromise, and a large amount of fluid resulting in a nondiagnostic examination. Stool was characterized as none, some without diagnostic compromise, moderate amount with diagnostic compromise, and a large amount resulting in a nondiagnostic examination. Distention was considered to be either good, suboptimal in one or two segments, or suboptimal in more than two segments. Breath-holding artifacts were classified as none, minor (with no diagnostic compromise), moderate (diagnostic compromise), or severe (resulting in a nondiagnostic examination).
Retrospective Characterization of Polyps
A fourth observer who was not one of the prospective observers performed a
retrospective characterization of all polyps. This was done by reviewing the
CT colonography data sets of all patients with polyps (missed and identified
lesions), the colonoscopy report, and all pathology reports of biopsied
specimens.
Polyps were classified according to the following criteria: detected by one observer, detected by two observers, or not detected by either observer. Lesions that were not prospectively detected by either prospective interpretation were further classified as retrospectively detected or not seen retrospectively. All lesions that were found in the CT colonography data sets were then characterized according to several criteria. Polyps were analyzed by two size groups, those of 5–9 mm and those 10 mm or greater. Lesions smaller than 5 mm were not considered. Retrospective evaluation noted whether polyps were displayed in both supine and prone positions or were displayed in only the supine or only the prone position. Retrospective evaluation also recorded polyps as being shown on only 2D images (axial or reformatted images), on only 3D endoluminal images, or on both 2D and 3D images.
Lesion surface contour was characterized as regular (i.e., smooth) or irregular (i.e., shaggy or rugged) on the basis of its CT appearance on 2D images. Polyp shape on CT described the shape of a colonic lesion on 2D axial or reformatted images and was characterized as round, flat (height one-half the width or less), oblong, or a focally thickened fold.
The relationship of polyps to colonic folds was also classified. Polyps were considered to be either related or not related to colonic folds. If they were related to folds, they were further characterized as being located on a fold, at the base of a fold (or adjacent to it), or at the convergence of two or more folds (Fig. 1). The relationship of polyps to folds was determined by analysis of 2D and 3D endoluminal images. By this classification, polyps located at the convergence of folds were a subgroup of polyps located at the base of folds. Some larger polyps (> 10 mm in diameter) were classified as being located at both the base of a fold and on the fold itself.
|
The colonic wall adjacent to a lesion was characterized as being thickened or not thickened. Colonic wall thickening was present if the colonic wall measured greater than 3 mm and the thickening could be the result of any cause, including collapse, muscular hypertrophy, or neoplasia.
Polyps were also classified according to their relationship with colonic flexures. The term "flexure" generically refers to any curvature in the colon, not only to those adjacent to the liver and spleen. Other commonly encountered flexures are in the sigmoid and transverse colons. When a polyp was located in a tortuous region of the colon, the lesion was classified as being at the inner (or short) or at the outer (or long) aspect of a flexure (Fig. 2). If the lesion was localized to a straight segment of colon, it was considered not related to a flexure.
|
The global quality of the entire CT colonography examination and the local
quality of the CT examination in the region of each polyp were evaluated. One
of the two prospective reviewers described the global quality of the CT
colonography preparation, as reported earlier
[13]. Local quality refers to
the observer's ability to visualize the colonic lumen in the vicinity of each
polyp and was assessed by the retrospective observer. Because of the
simultaneous and complex interrelationship of factors such as collapse, stool,
and fluid that predispose to false-negative examinations, single causes of
suboptimal visualization or preparation were not assigned. Rather, a
comprehensive assessment of local quality was rendered as optimal (excellent
distention and visualization of the colonic wall), moderate (some compromise
due to a combination of collapse, fluid, and stool, but still likely to see
lesions 
1 cm), or poor (compromise resulting in probability of missing
lesions 1 cm, if present). If the segment of the colon under assessment
was considered to be nondiagnostic because of any of the listed factors, it
was also rated as poor preparation.
Statistical Analysis
Modes of visualization and intrinsic and extrinsic features of polyps and
their CT appearance were analyzed in a univariate model to ascertain if a
significant difference existed between polyps visualized by both prospective
observers and those missed by one or both of the prospective observers but
seen at the retrospective assessment. In other words, for tests of
significance, polyps seen by the unblinded observer were combined with those
polyps seen by only one of the two prospective observers. The tables give
descriptive results for these two scenarios separately. Tests of univariate
significance were made using a chi-square or Fisher's exact test, as
appropriate. A multivariate analysis of intrinsic and extrinsic features
associated with missed lesion detection was performed for lesion
characteristics trending toward significance (p 
0.1). In both
univariate and multivariate analyses, the significance level was set at less
than or equal to 0.05.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Visualization
All polyps were characterized according to their ability to be seen in the
supine or the prone data set (or both) and on 2D or 3D images, or both
(Table 1 and Fig.
3A,
3B). Polyps of 5–9 mm
seen only in the supine or only in the prone position were missed more often
than polyps seen in both positions (p < 0.04; 79% vs 45%,
respectively). Results were similar for polyps 10 mm or larger (p =
0.03; 92% vs 56%, respectively).
|
|
Significantly more polyps were missed by an observer when lesions were visible using 2D images only, as opposed to being seen on both 2D and 3D images. This trend was significant for polyps of 5–9 mm (p = 0.01; 100% vs 50%) but not for polyps of 10 mm or larger (p = 0.64; 83% vs 65%).
Intrinsic Features
Polyp size.—Fifty (69%; 95% confidence interval [CI],
57–80%) of 72 polyps that were 5–9 mm in diameter could be
visualized on prospective or retrospective review, and 37 (84%; 95% CI,
70–93%) of 44 polyps that were 10 mm or greater could be seen on
prospective or retrospective review. Of the 72 lesions of 5–9 mm, 43
(60%; 95% CI, 48–71%) were detected by either one or both prospective
observers. Of the 29 missed lesions, only seven could be found retrospectively
(Table 1). Of the 44 lesions of
10 mm or larger, either one or both prospective observers detected 30 (68%;
95% CI, 52–81%). Seven of the 14 missed lesions of 10 mm or larger could
be found retrospectively (Table
1).
Endoscopic polyp morphology.—Most sessile, pedunculated, and flat polyps of 1 cm or more were seen by one or both reviewers (18/25, 72%; 6/9, 67%; 6/10, 60%, respectively; Table 2, Fig. 4). For polyps 5–9 mm, most sessile and pedunculated polyps were detected prospectively. In contradistinction, only a minority of flat lesions 5–9 mm in size could be detected (3/9; 33%). All the large pedunculated lesions that were missed by both prospective observers (3/3) could be seen retrospectively. Conversely, however, only 25% (1/4) of the large flat lesions missed by both prospective observers were detectable on retrospective review. Endoscopic polyp morphology tended to correlate with missed lesions for small and large polyps, but this trend was not significant until all polyps greater than 5 mm were considered (p = 0.03).
|
|
Polyp shape on CT.—The shape of a colonic lesion on CT (i.e., round, flat, oblong, or flat or round focal thickening of a fold) did not correspond to whether a lesion was detectable for polyps 5–9 mm (p = 0.32). However, CT shape did correspond to whether a lesion was missed for larger lesions (p = 0.04), with flat, oblong, and focal thickening being missed more often (Table 2). Colonic lesions that were round were easier to detect than other morphologies. All round polyps of 10 mm or larger (13/13; 100%) were identified by at least one observer, and eight (62%) of 13 were identified by both observers.
Nearly all polyps of 5–9 mm (47/50, 94%) were characterized as round lesions on CT, regardless of their endoscopic characterization. Endoscopically flat lesions (all sizes) were characterized according to CT shape as flat lesions in five of 10 cases, as flat or round regions of focal fold thickening in four of 10 cases, and as a rounded lesion in one of 10 cases.
Surface contour.—Polyps 5–9 mm in diameter, as well as those 10 mm or larger, with an irregular surface were missed by an observer more frequently than those with a smooth regular surface (100% vs 55% and 80% vs 59%, respectively), although in neither size group were these differences statistically significant (p = 0.25 and p = 0.29). Because of greater power in the comparison, when considering all polyps 5 mm or greater, those with an irregular surface were missed by an observer more often than those with a smooth, regular surface (p = 0.04; 83% vs 57%, Table 2).
Extrinsic Features
Relationship to folds.—Most lesions of 1 cm or greater seen
on CT colonography were related to folds (84%; 31/37), whereas only about half
the subcentimeter lesions were related to folds (52%; 26/50)
(Table 3 and Fig.
5A,
5B). No significant association
was seen between a polyp being located on a fold and whether it was missed,
for polyps in either the 5–9 mm or the 10 mm or greater group
(p = 0.09 and p = 1.0, respectively). Of the subset of 57
polyps 5 mm or larger located on a fold, 11 (85%) of 13 located at the
convergence were missed by at least one observer, but this location was not
significantly associated with a missed lesion compared with other locations on
a fold (p = 0.2).
|
|
|
Relationships to colonic segment, flexures, and adjacent colonic wall.—No significant difference was seen in the detection rates of polyps with respect to location in a colonic segment or relationship to a flexure (Table 3). Lesions adjacent to a region of colonic wall thickening were missed more frequently than those adjacent to colonic walls less than 3 mm thick, regardless of lesion size (p = 0.05, 86% vs 58%, Table 3).
Quality of the colon preparation.—Global quality was based on the evaluation given by the observers during the prospective data interpretation and was an evaluation of the quality of the entire data set without considering the local preparation in proximity to the lesion. No significant relationship was seen between the global quality of an examination with respect to fluid, stool, distention, or breath-holding artifacts and a lesion's propensity to be missed by at least one observer.
Local quality was determined retrospectively in the region of each lesion, reflecting the local milieu about each polyp, with results displayed in Table 3. The local preparation (optimal vs moderate or poor) did correlate with whether a polyp was missed for polyps 5–9 mm in diameter and for all polyps greater than 5 mm (p < 0.001, 32% vs 79%; and p = 0.02, 51% vs 76%, respectively). This trend did not reach statistical significance for lesions 10 mm or greater in diameter (p = 0.15 [58% vs 85%]).
Multivariate Analysis
Multivariate models were assessed using lesion characteristics that had
univariate test results at a p value of 0.10 or greater for
association with missed detections for all lesions 5 mm or greater in
diameter. These characteristics included morphology (p = 0.03),
relationship to a fold (p = 0.09), shape (p = 0.04),
adjacent to wall thickening (p = 0.05), surface irregularity
(p = 0.04), and quality of local preparation (p = 0.02).
Three multivariate models were found to have two lesion characteristics that
were independently associated with missed detection. None of the three models
was clearly a "best" model. However, all three included local
preparation quality, indicating the importance of this factor in lesion
detection. Model results (Table
4), in no particular order, were model 1: local preparation
(p = 0.01, odds ratio [OR] = 3.5) and adjacent to thickened colonic
wall (p = 0.04, OR = 5.4); model 2: local preparation (p
< 0.01, OR = 4.2) and relationship to a fold (p = 0.03, OR = 3.2);
model 3: local preparation (p = 0.03, OR = 2.9) and morphology (flat,
relative to pedunculated, p < 0.01, OR = infinite, and sessile,
relative to pedunculated, p = 0.88, OR = 1.1).
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Conventional colonoscopy, which has a high sensitivity for the detection of polyps of 10 mm or larger [18, 19], served as the reference standard for the presence of a polyp. Of the 44 lesions that were 10 mm or larger identified on conventional colonoscopy, seven (16%) could not be found in the CT colonography data set, and 22 (31%) of 72 lesions of 5–9 mm could not be identified. These cases must be considered technical failures. Factors likely to reduce technical failures include routine use of MDCT scanners (to improve distention) [20, 21], thinner slices and reconstruction intervals [7], reduced image noise, the use of IV contrast material [2], computer-aided detection [22], and novel rendering techniques [23]. Further research in these areas may clarify the importance of these factors for improving lesion detection.
All three observers were staff gastrointestinal radiologists, each of whom had interpreted more than 200 CT colonography studies. Most lesions detected by both observers were visible in both the supine and the prone data sets: 11 (92%) of 12 polyps of 10 mm or larger, and 17 (81%) of 21 polyps of 5–9 mm. However, when lesions were seen in only the supine or only the prone position, most were missed by at least one observer, and this trend was significant when all polyps 5 mm or larger were considered (p < 0.01). This observation underscores the previously described importance of data acquisition with the patient in two positions [3]. Suspicious lesions seen in one position cannot be considered to be stool simply because they are not observed in another position because about one third of the polyps in our study were seen in only one position (Fig. 3A, 3B).
To detect polyps, our observers used a 2D search method for colonic lesions supplemented by the use of 2D multiplanar reformations and 3D volume renderings to evaluate suspicious lesions. This method is advocated by many colonographers [4, 7]. Most polyps were seen using both 3D endoluminal imaging and 2D reformatted images; however, a minority of polyps could be seen with 2D, but not with 3D, images. Radiologists may feel most comfortable diagnosing lesions that appear polypoid on 2D and 3D images but should remember that some lesions (e.g., flat lesions) may be occult on 3D endoluminal images alone. Because we used a 2D search method, we were unable to characterize the number of polyps that may be visible on 3D endoluminal views and occult on 2D images. Using software that relies primarily on 3D endoluminal renderings to detect lesions may result in finding additional polyps or in finding a different subset of polyps.
Several intrinsic features of polyps play a significant role in lesion conspicuity. The number of missed and detected lesions varied according to endoscopic polyp morphology, surface contour, and polyp shape on CT. Sessile lesions are most likely to be missed because they are much more common than the other morphologies. On a percentage basis, however, endoscopically flat lesions were more likely to be missed than sessile ones. Flat lesions do not alter the colonic contour as much as sessile or pedunculated polyps. Flat lesions represented only 17% of all polyps in this study but are important to detect because they may have a shortened polyp dwell time and early transition to carcinoma [24, 25]. Endoscopically flat lesions corresponded with flat CT lesions or focal fold thickening in nine of 10 patients (Fig. 4). Although most endoscopically flat lesions of 10 mm or larger could be identified, most endoscopically flat lesions 5–9 mm could not be identified with the 5-mm slice thickness used in our study. We believe flat lesions are best seen on 2D axial or multiplanar reformatted images as focal regions of wall thickening with bone or soft-tissue window settings. Three-dimensional endoluminal images are usually not helpful in the detection of flat lesions.
Pedunculated polyps were missed more frequently than was expected. In this series, even large polyps that were missed could easily be seen retrospectively. Radiologists should be aware that the stalk of a pedunculated polyp can be quite long, resulting in a significant change in the shape and location of a pedunculated polyp with a change in positioning. Both stool and pedunculated polyps can change their location as the patient changes from a prone to a supine position.
Approximately 40% of large polyps will show an irregular surface on CT colonography, and both prospective observers in our series failed to identify about one fourth of them. The irregular surfaces were likely misinterpreted as stool. Larger polyps may have an undulating surface or may have stool adhering to their surface. In contrast, nearly all polyps of 5–9 mm had a smooth surface contour.
Extrinsic factors also affect polyp conspicuity. Abnormal colonic wall thickness was associated with a lesion being missed in both univariate and in one of the three final multivariate analyses. In our experience, abnormal wall thickness is most often a result of poor distention, but we did not quantify this relationship in our study. Other causes of abnormal wall thickness may include muscular hypertrophy, inflammation, and neoplasia. Radiologists need to be vigilant for lesions located near regions of abnormal wall thickness.
No statistically significant relationship was seen between missed polyps and location on a haustral fold in our univariate model. However, in a multivariable model that also included quality of local preparation, location on a fold was independently associated with a greater likelihood of a false-negative examination. Lesions located at the convergence of folds were frequently missed (11/13, 85%). If a polyp is suspected at the convergence of folds, 3D endoluminal views are often helpful to distinguish between folds and a polyp (Fig. 5A, 5B).
We found no relationship between the location of a polyp in a flexure and the likelihood of false-negative results. In addition, we found no significant difference between the overall global bowel preparation and whether a lesion was missed or identified. Obviously, this global interpretation did not consider the colonic segment containing each lesion.
The local quality of the examination at the location of each lesion did show that smaller lesions are more likely to be missed by both observers in regions with less than optimal preparation. Suboptimal local examination quality was independently associated with missed lesions in each of three multivariate models. Careful scrutiny of the colon for full distention is important before an examination is completed. Additional air and rescanning may be required to produce an optimal study.
Our comparison of visualization methods and intrinsic and extrinsic features of missed and identified polyps on CT colonography has several limitations. One of the advantages of CT colonography over barium enema is its ability to examine the internal attenuation of colonic lesions, and we did not examine the effect of lesion attenuation on identification. We thought such characterization would be subjective, necessitating the creation of a scale of lesion attenuation and homogeneity. Additionally, it is difficult to determine the complex interrelationship between many factors (e.g., polyp morphology, surface contour, and quality of local preparation) that contribute in a combined manner to polyp conspicuity. We chose to assess this interdependence of factors by performing a univariate analysis of individual factors and then subsequently performing a multivariate analysis that specifically examined significant univariate variables. Finally, we used a customized computer workstation developed at our institution for the analysis of prone and supine data sets [17]. This workstation was optimized to evaluate for polyps on the basis of a paradigm of 2D search, using 3D endoluminal images to examine suspicious lesions [4, 7]. If we had used software that relies primarily on an endoluminal fly-through, we might have detected some polyps that were visible on 3D endoluminal renderings and difficult to appreciate on 2D images alone.
In conclusion, specific features of polyps on CT colonography contribute to the likelihood that a polyp may escape detection by the interpreting radiologist. Knowledge and recognition of intrinsic polypoid features, such as flat and pedunculated morphology and irregular surface contour, may improve detection of neoplasia on CT colonography. Extrinsic features, such as location at the convergence of folds or adjacent to a thickened colonic wall, or suboptimal local preparation, may also impede polyp detection. Lesions seen in only one position or only on 2D images should not be disregarded unless the corresponding colonic segment in the opposite position is well visualized and normal in appearance.
Acknowledgments
We thank Patricia Bard and Karen Madsen for their help in the preparation
of this manuscript.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  K. T. Johnson, J. G. Fletcher, and C. D. Johnson Computer-Aided Detection (CAD) Using 360{degrees} Virtual Dissection: Can CAD in a First Reviewer Paradigm Be a Reliable Substitute for Primary 2D or 3D Search? Am. J. Roentgenol., October 1, 2007; 189(4): W172 - W176. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. G. Fletcher, F. Booya, R. M. Summers, D. Roy, L. Guendel, B. Schmidt, C. H. McCollough, and J. L. Fidler Comparative Performance of Two Polyp Detection Systems on CT Colonography Am. J. Roentgenol., August 1, 2007; 189(2): 277 - 282. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Mang, A. Maier, C. Plank, C. Mueller-Mang, C. Herold, and W. Schima Pitfalls in Multi-Detector Row CT Colonography: A Systematic Approach RadioGraphics, March 1, 2007; 27(2): 431 - 454. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. H. Park, H. K. Ha, A. Y. Kim, K. W. Kim, M.-G. Lee, P. N. Kim, Y. M. Shin, J.-S. Byeon, S.-K. Yang, J. H. Kim, et al. Flat polyps of the colon: detection with 16-MDCT colonography--preliminary results. Am. J. Roentgenol., June 1, 2006; 186(6): 1611 - 1617. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. L. MacCarty, C. D. Johnson, J. G. Fletcher, and L. A. Wilson Occult Colorectal Polyps on CT Colonography: Implications for Surveillance. Am. J. Roentgenol., May 1, 2006; 186(5): 1380 - 1383. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. T. Johnson, C. D. Johnson, J. G. Fletcher, R. L. MacCarty, and R. L. Summers CT Colonography Using 360{degrees} Virtual Dissection: A Feasibility Study Am. J. Roentgenol., January 1, 2006; 186(1): 90 - 95. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. H. Park, H. K. Ha, M.-J. Kim, K. W. Kim, A. Y. Kim, D. H. Yang, M.-G. Lee, P. N. Kim, Y. M. Shin, S.-K. Yang, et al. False-Negative Results at Multi-Detector Row CT Colonography: Multivariate Analysis of Causes for Missed Lesions Radiology, May 1, 2005; 235(2): 495 - 502. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
