HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) CME Test (opens in a new window) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Mang, T. Articles by Schima, W. Search for Related Content PubMed PubMed Citation Articles by Mang, T. Articles by Schima, W. Related Collections Computed Tomography Gastrointestinal Radiology

Pitfalls in Multi–Detector Row CT Colonography: A Systematic Approach¹

¹ From the Department of Radiology, Medical University of Vienna, Waehringer Guertel 18-20, A-1090 Vienna, Austria. Presented as an education exhibit at the 2005 RSNA Annual Meeting. Received April 25, 2006; revision requested June 12 and received August 9; accepted August 23. All authors have no financial relationships to disclose. Address correspondence to T.M. (e-mail: thomas.mang{at}meduniwien.ac.at).

	Abstract

Top Abstract LEARNING OBJECTIVES FOR TEST... Introduction CT Colonographic Technique Technical Errors Pitfalls in Evaluation Conclusions References

 
Thin-section multi–detector row computed tomographic (CT) colonography is a powerful tool for the detection and classification of colonic lesions. However, each step in the process of a CT colonographic examination carries the potential for misdiagnosis. Suboptimal patient preparation, CT scanning protocol deficiencies, and perception and interpretation errors can lead to false-positive and false-negative findings, adversely affecting the diagnostic performance of CT colonography. These problems and pitfalls can be overcome with a variety of useful techniques and observations. A relatively clean, dry, and well-distended colon can be achieved with careful patient preparation, thereby avoiding the problem of residual stool and fluid. Knowledge of the morphologic and attenuation characteristics of common colonic lesions and artifacts can help identify bulbous haustral folds, impacted diverticula, an inverted appendiceal stump, or mobile polyps, any of which may pose problems for the radiologist. A combined two-dimensional and three-dimensional imaging approach is recommended for each colonic finding. A thorough knowledge of the various pitfalls and pseudolesions that may be encountered at CT colonography, along with use of dedicated problem-solving techniques, will help the radiologist differentiate between definite colonic lesions and pseudolesions.

© RSNA, 2007

	LEARNING OBJECTIVES FOR TEST 4

Top Abstract LEARNING OBJECTIVES FOR TEST... Introduction CT Colonographic Technique Technical Errors Pitfalls in Evaluation Conclusions References

 
After reading this article and taking the test, the reader will be able to:

• List the spectrum of common and uncommon pitfalls in multi–detector row CT colonography.
  
• Describe multi–detector row CT colonographic signs and criteria that can help distinguish between definite colonic lesions and pseudolesions.
  
• Discuss various evaluation strategies for reducing the number of false-positive and false-negative findings at multi–detector row CT colonography.
  

	Introduction

Top Abstract LEARNING OBJECTIVES FOR TEST... Introduction CT Colonographic Technique Technical Errors Pitfalls in Evaluation Conclusions References

 
Multi–detector row computed tomographic (CT) colonography affords increased opportunities for diagnostic imaging of the large bowel. Isotropic imaging of the colon with thin collimation has become standard and provides high-quality multi-planar reformatted (MPR) images and three-dimensional (3D) assessment of the entire organ. CT colonography provides planar two-dimensional (2D) and virtual 3D endoscopic images of the colon. Although radiologists are traditionally most experienced in 2D CT of the abdomen, the planar 2D approach to the gas-distended colon presents new challenges to "film" readers. The complex intraluminal anatomy of bowel loops, haustral folds, and residual fluid and stool, as well as the degree of distention, may complicate planar evaluation. On the other hand, 3D virtual endoscopy of the large bowel provides an "intraluminal" perspective on CT data that may be unfamiliar to some radiologists. Manual navigation or preprocessing of data and insufficient rendering may have complicated this approach in early versions of virtual endoscopic software.

Published results have indicated a sensitivity for CT colonography of over 90% for polyps 10 mm or larger in diameter (1). To achieve such results, adequate bowel cleansing or "fecal tagging" and reader experience are essential. In clinical practice, inadequate patient preparation and data acquisition impair examination quality. Insufficient image processing (ie, inadequate threshold or rendering) may lead to an increased number of artifacts on both 2D and 3D images. As a result, perception and interpretation errors occur, increasing the number of false-positive and false-negative findings and adversely affecting the diagnostic performance of CT colonography.

The development of multi–detector row CT changed the type and frequency of occurrence of some artifacts that were well known at helical CT. Helical CT artifacts such as respiratory and stair-step artifacts have become rare with multi–detector row CT (2). With improvements in spatial resolution, lesion depiction and differentiation also improved (3–5). Other artifacts, such as image noise, became more important owing to the use of thinner sections and of imaging protocols with dose reduction.

Although at present the most commonly used platform for data interpretation is a primary 2D evaluation with 3D evaluation used for problem solving (6), 3D views are becoming more feasible with recent improvements in 3D evaluation software and hardware that allow fast data handling and better image quality. Thus, primary 3D endoscopic evaluations are being used more frequently for multi–detector row CT data sets than for helical CT data sets (1). However, although the 3D image quality is high for some systems, reconstruction algorithms are still far from yielding perfect "natural" images, and human observers often have problems with navigation and orientation in the colon or with differentiation of depressed structures or artifacts from truly elevated lesions. Radiologists should be familiar with the pitfalls in multi–detector row CT colonography so as to avoid them in daily practice.

In this article, we review pitfalls and pseudolesions at multi–detector row CT colonography. In particular, we discuss and illustrate technical errors related to patient preparation, colonic distention, and image acquisition; various pitfalls of evaluation technique; and perception and interpretation errors that can made on either 2D or 3D images (Table). We also present problem-solving techniques for differentiating between true colonic lesions and pseudolesions.

	CT Colonographic Technique

Top Abstract LEARNING OBJECTIVES FOR TEST... Introduction CT Colonographic Technique Technical Errors Pitfalls in Evaluation Conclusions References

Bowel distention is performed with the patient in the left decubitus or supine position. A thin, flexible rubber catheter (eg, a small-gauge Foley catheter) is placed in the rectum, followed by gentle insufflation (either automated or manual) of carbon dioxide (CO₂) or air to patient tolerance (8). Intravenous administration of antispasmodic agents such as glucagon-hydrochloride or hyoscine-N-butylbromide may reduce discomfort and spasms, but the effect on distention remains controversial (hyoscine-N-butylbromide is not licensed for use in the United States) (9–11).

Thin-section CT is ideally performed with a multi–detector row CT scanner with thin collimation, with the patient first in the supine and then in the prone position. An initial scout view obtained prior to scanning in each position helps ensure adequate distention of the colonic segments, with additional CO₂ or air being insufflated if required (8). Intrinsic colonic lesions can be outlined by tagging residual fecal matter with oral administration of contrast material before the examination. This tagging can be achieved with use of barium or iodinated contrast material (12,13) or a combination of the two (1). In addition, colonic lesions can be enhanced with the intravenous injection of iodinated contrast agent during the second scan. However, intravenous contrast material may be contraindicated in asymptomatic patients due to the increased cost, the need for intravenous access, and the risk of allergic anaphylactic reactions (7). Therefore, the majority of CT colonographic investigators do not routinely administer intravenous contrast material (6). On the other hand, several studies have shown intravenous contrast material to be beneficial in patients with symptoms of colorectal cancer, in the evaluation of stenosis and the prestenotic colon, and in the detection of local recurrence or distant metastases in patients with a history of colorectal cancer. Thus, the administration of intravenous contrast material may be limited to these indications (14–17). Each scan is acquired during a single breath hold. Useful exposure settings are 120 kVp and 50–100 mAs for both prone and supine images (3,18). With 16- to 64-section CT scanners, the use of submillimeter collimations is feasible. Data can be reconstructed as 1-mm-thick sections (7).

Image processing and interpretation is feasible on commercially available CT colonography systems that allow an interactive, manual, mouse-driven, virtual fly-through of the volume- or surface-rendered intraluminal 3D images, as well as evaluation of 2D axial and MPR images in a cine mode. Simultaneously available 2D and 3D image displays allow rapid correlation for the evaluation of any suspected finding.

	Technical Errors

Top Abstract LEARNING OBJECTIVES FOR TEST... Introduction CT Colonographic Technique Technical Errors Pitfalls in Evaluation Conclusions References

 
For optimal image quality, the colon should be clean, dry, and completely distended. Scanning parameters should be chosen that will improve spatial resolution with regard to dose reduction and minimize artifacts.

Patient Preparation
Bowel preparation is one of the most important steps in CT colonography. A well-prepared colon will facilitate lesion detection and minimize false-positive findings, whereas residual matter in the lumen (eg, stool, fluid) may simulate or obscure colonic lesions.

Residual Stool.— Residual stool is a common finding at CT colonography and represents an important source of pitfalls. Stool can either simulate or obscure colonic lesions and lead to interpretation or perception errors. The morphologic features of stool are variable. On 3D views, stool often appears in bizarre geometric configurations with angled borders (19). However, residual fecal material may be pseudopolypoid with a round, oval, or lobulated configuration and may be mistaken for a true polyp on endoluminal views, whereas collections of stool with bizarre configurations are easier to identify as such. Therefore, careful correlation with corresponding 2D images is mandatory (Fig 1). It is especially helpful on 2D images that residual fecal material typically contains trapped gas, which appears as areas of low attenuation, as well as food particles, which appear as areas of high attenuation (19).

View larger version (126K): [in this window] [in a new window] [Download PPT slide]   Figure 1a.  Residual stool in a 47-year-old man with a high familial predisposition for colorectal cancer. (a) Three-dimensional endoluminal CT image shows a broad-based, polypoid filling defect (arrow). (b) On a supine contrast material–enhanced CT scan, the pseudolesion (arrow) is unenhanced, has inhomogeneous attenuation, and contains trapped gas.

View larger version (107K): [in this window] [in a new window] [Download PPT slide]   Figure 1b.  Residual stool in a 47-year-old man with a high familial predisposition for colorectal cancer. (a) Three-dimensional endoluminal CT image shows a broad-based, polypoid filling defect (arrow). (b) On a supine contrast material–enhanced CT scan, the pseudolesion (arrow) is unenhanced, has inhomogeneous attenuation, and contains trapped gas.

In the tagging of fecal material with orally ingested barium or iodinated contrast material prior to the examination (fecal tagging) (12), only residual stool and fluid will take up contrast material, whereas the bowel wall and colonic lesions will not. Similarly, if intravenous contrast material is administered, polyps and masses will enhance, but stool will not (21,22). Both strategies have been reported as helpful in identifying residual fecal matter in the colon (12,23). However, the majority of CT colonographic investigators do not routinely administer intravenous contrast material (6).

Residual Fluid.— Residual fluid is a common finding at CT colonography, depending on the type of bowel preparation. It appears as a horizontal level in the colon on both 2D and 3D images, with homogeneous attenuation equivalent to that of water on 2D images. Residual fluid obscures colonic lesions and leads to perception errors. Because of gravity, residual fluid is always found in the dependent part of a colonic segment. Therefore, fluid is found more commonly in the descending colon and rectum with the patient supine, whereas it moves to the transverse colon with the patient prone. Consequently, performing CT colonography with the patient in both the prone and supine positions will shift retained fluid into other colonic segments, and hidden lesions will become visible (24). However, if large amounts of fluid are present, visualization of the entire mucosa may not be guaranteed at prone and supine imaging.

In general, cathartic colon preparations, such as the "Fleet kit" (Fleet Pharmaceuticals, Lynchburg, Va), which consists of phosphosoda and four bisacodyl tablets, or the "LoSo kit" (E-Z-EM, Westbury, NY), which consists of magnesium citrate and four bisacodyl tablets, render the colon drier than do colonic lavage solutions like polyethylene glycol preparations and are, therefore, preferable (25).

Distention
Underdistention.— Optimal colonic distention is a necessary prerequisite to accurate CT colonographic data interpretation. Underdistention leads to luminal narrowing or colonic segment collapse, which results in perception and interpretation errors (26).

Unlike residual fluid, gas tends to move to the highest point of the colon. Commonly, the left colon, rectum, sigmoid colon, and parts of the descending colon are collapsed when the patient is supine, whereas the transverse colon is often collapsed when the patient is prone. Scanning with the patient in both positions leads to a diagnostically important redistribution of gas during repositioning (27–29). Each segment should be distended on at least one scan and, preferably, on both scans. Adequate distention is recognized by obtaining a scout view after the insufflation of gas into the colon, and additional gas can be insufflated to distend collapsed segments. Bowel distention may be improved with automated insufflation of CO₂, which has been reported to confer the greatest benefit for left colonic distention when the patient is supine (30,31). Fixed gas amounts are impractical because of different anatomic constitutions.

Spasm.— Segmental colonic spasm is a physiologic luminal narrowing that results from peristaltic muscular contraction of the colon. Persistent focal spasm of the colonic wall can simulate a stenotic tumorlike lesion. Segmental colonic spasm may result in focal, circular, and, in many cases, smooth wall thickening, sometimes accompanied by simulated shoulder formation. Areas of spasm also show contrast enhancement of the wall. However, the pericolic fat tissue always appears normal and the borders of the wall are smooth (Fig 2). In contrast, stenotic cancers typically manifest as extensive wall thickening, especially with shoulder formation, as well as stranding, indistinct boundaries, and nodular protrusions into the pericolic fat tissue (32). Pericolic lymph nodes and distant metastases are signs of progression of the disease and can be evaluated with 2D image review.

View larger version (100K): [in this window] [in a new window] [Download PPT slide]   Figure 2a.  Segmental colonic spasm in the descending colon in a 65-year-old man with symptoms of colorectal cancer. (a) Three-dimensional endoluminal CT image shows an irregular, circular narrowing of the colonic lumen (arrow), a finding that simulates a stenosis. (b) Supine coronal contrast-enhanced CT scan shows focal, irregular circular wall thickening with shoulder formation (arrow). The apparent lesion demonstrates enhancement. (c) Prone coronal CT scan shows a normal smooth colonic wall without signs of stenosis or wall thickening (arrow), findings that indicate that the spasm has relaxed.

View larger version (180K): [in this window] [in a new window] [Download PPT slide]   Figure 2b.  Segmental colonic spasm in the descending colon in a 65-year-old man with symptoms of colorectal cancer. (a) Three-dimensional endoluminal CT image shows an irregular, circular narrowing of the colonic lumen (arrow), a finding that simulates a stenosis. (b) Supine coronal contrast-enhanced CT scan shows focal, irregular circular wall thickening with shoulder formation (arrow). The apparent lesion demonstrates enhancement. (c) Prone coronal CT scan shows a normal smooth colonic wall without signs of stenosis or wall thickening (arrow), findings that indicate that the spasm has relaxed.

View larger version (149K): [in this window] [in a new window] [Download PPT slide]   Figure 2c.  Segmental colonic spasm in the descending colon in a 65-year-old man with symptoms of colorectal cancer. (a) Three-dimensional endoluminal CT image shows an irregular, circular narrowing of the colonic lumen (arrow), a finding that simulates a stenosis. (b) Supine coronal contrast-enhanced CT scan shows focal, irregular circular wall thickening with shoulder formation (arrow). The apparent lesion demonstrates enhancement. (c) Prone coronal CT scan shows a normal smooth colonic wall without signs of stenosis or wall thickening (arrow), findings that indicate that the spasm has relaxed.

Although controversial, the administration of antispasmodic drugs, such as glucagon-hydrochloride or hyoscine-N-butylbromide, may reduce the frequency of occurrence of spasms (9,10).

Perforation.— CT colonography has been reported to be a safe, noninvasive method for examining the entire colon. In rare cases, however, bowel distention may lead to perforation of the bowel (8,35). Signs of colonic perforation include focal areas of pericolic gas as well as free intra-peritoneal gas. Wide window settings are optimal for the detection of small amounts of pericolic or intraabdominal gas (Fig 3). Most of the reported cases of perforation caused by colonic distention involved acute inflammation or stenosis of the colon and the use of a rectal balloon catheter (35–37). In cases of inflammatory bowel disease (IBD), especially in acute IBD or in obstructive stenosis of the colon, CT colonography is relatively contraindicated or at least should be performed with caution without the use of a rectal balloon catheter.

View larger version (99K): [in this window] [in a new window] [Download PPT slide]   Figure 3.  Colonic perforation in a 35-year-old woman with acute ulcerative colitis. Supine coronal CT scan shows focal pericolic air formations (arrow) around the transverse colon related to the perforation. Total flattening and disappearance of the haustra is also seen (arrowheads).

View larger version (101K): [in this window] [in a new window] [Download PPT slide]   Figure 4a.  Respiratory artifact in the lower abdomen in a 71-year-old asymptomatic man. (a) Three-dimensional endoluminal CT image shows abrupt wall defects or polypoid structures (arrows) on opposite luminal walls of the colon. (b) Supine coronal CT scan shows linear artifacts of the colonic wall (straight arrows) and wavelike irregularities of the outer abdominal wall (wavy arrow) caused by respiratory motion.

View larger version (130K): [in this window] [in a new window] [Download PPT slide]   Figure 4b.  Respiratory artifact in the lower abdomen in a 71-year-old asymptomatic man. (a) Three-dimensional endoluminal CT image shows abrupt wall defects or polypoid structures (arrows) on opposite luminal walls of the colon. (b) Supine coronal CT scan shows linear artifacts of the colonic wall (straight arrows) and wavelike irregularities of the outer abdominal wall (wavy arrow) caused by respiratory motion.

View larger version (110K): [in this window] [in a new window] [Download PPT slide]   Figure 5.  Stair-step artifacts in the rectum in a 58-year-old asymptomatic woman. Three-dimensional endoluminal CT image (3-mm section thickness) shows a series of concentric rings (arrows) around the colonic lumen.

On 3D images, increased image noise appears as a diffusely granular or nodular surface pattern, whereas the colonic wall retains its normal thickness and anatomic features on 2D images. These "pseudochanges" should not be confused with an IBD, in which granular wall patterns are also present but the wall is thickened in affected segments and, in advanced disease, haustral folds are absent (44). If one or both series (ie, prone and supine) are performed with a standard dose, image noise is easily recognized as such (Fig 6).

View larger version (103K): [in this window] [in a new window] [Download PPT slide]   Figure 6a.  Image noise in a 70-year-old asymptomatic man. (a) Three-dimensional endoluminal CT image (50 mAs) shows a smooth bowel wall with a 10-mm polyp (arrow). (b) Three-dimensional endoluminal CT image (10 mAs) shows the bowel wall with a continuous granular-nodular surface pattern due to increasing noise. Note that the polyp (arrow) is still visible with the decreased dose.

View larger version (101K): [in this window] [in a new window] [Download PPT slide]   Figure 6b.  Image noise in a 70-year-old asymptomatic man. (a) Three-dimensional endoluminal CT image (50 mAs) shows a smooth bowel wall with a 10-mm polyp (arrow). (b) Three-dimensional endoluminal CT image (10 mAs) shows the bowel wall with a continuous granular-nodular surface pattern due to increasing noise. Note that the polyp (arrow) is still visible with the decreased dose.

Metallic Artifacts.— Metallic structures cause beam-hardening artifacts that significantly degrade axial and endoluminal image quality and obscure large portions of the bowel wall. Metallic artifacts are particularly common in the pelvis in patients who have undergone total hip replacement. These artifacts appear as large streaks on 2D images and as bizarre linear intraluminal structures on 3D images and may obscure wall disease and preclude accurate assessment of the bowel segment (Fig 7) (24,39). In patients with a hip prosthesis, evaluation of the rectosigmoid colon is limited.

View larger version (118K): [in this window] [in a new window] [Download PPT slide]   Figure 7.  Metallic artifacts in a 70-year-old woman with bilateral hip prostheses. Prone CT scan shows prominent streaks obscuring large portions of the rectal bowel wall. A 3D endoluminal CT image (inset) demonstrates bizarre luminal artifacts and granular wall irregularity.

	Pitfalls in Evaluation

Top Abstract LEARNING OBJECTIVES FOR TEST... Introduction CT Colonographic Technique Technical Errors Pitfalls in Evaluation Conclusions References

Evaluation Technique
Threshold Values for 2D Imaging.— Two window width and level settings are typically used for 2D data evaluation: a wide "polyp window" setting (width, 1500 HU; level, –200 HU) and a narrow soft-tissue window setting (width, 400 HU; level, 10 HU) (19). Wide window settings (eg, lung or bone windowing) impart a high-contrast interface to help detect intraluminal colorectal lesions and display more structures of the bowel wall, such as the thin haustral folds that might not be seen with narrow window settings. However, this benefit is compromised by the loss of differentiation with regard to the internal structure of the colon.

Conversely, narrow window settings (eg, soft-tissue windowing) permit better interpretation of the findings with regard to attenuation and internal structure (eg, polyp vs lipoma). However, this benefit is reduced by the loss of detail concerning bowel wall structure: Thin haustral folds or interfaces where the colonic wall is adjacent to other bowel segments, as well as pericolic gas as a sign of perforation, may not be displayed properly (Fig 8). Results of a recent study indicated significant differences in results obtained with various reconstruction algorithms, with "soft" algorithms being preferred to the standard algorithm (45).

View larger version (107K): [in this window] [in a new window] [Download PPT slide]   Figure 8a.  Effect of varying the window width and level. (a) Supine CT scan (window width, 1500 HU; window level, –200 HU) clearly depicts a thin haustral fold (arrowhead), with loss of differentiation of the fat attenuation of a lipoma (arrow). (b) The same supine CT scan as seen with narrower window settings (width, 400 HU; level, 140 HU) shows better differentiation of the fat attenuation of the lipoma (arrow), but the thin haustral fold (arrowhead) is now displayed incompletely.

View larger version (121K): [in this window] [in a new window] [Download PPT slide]   Figure 8b.  Effect of varying the window width and level. (a) Supine CT scan (window width, 1500 HU; window level, –200 HU) clearly depicts a thin haustral fold (arrowhead), with loss of differentiation of the fat attenuation of a lipoma (arrow). (b) The same supine CT scan as seen with narrower window settings (width, 400 HU; level, 140 HU) shows better differentiation of the fat attenuation of the lipoma (arrow), but the thin haustral fold (arrowhead) is now displayed incompletely.

Three-dimensional Shine-Through Artifacts.— Modern 3D workstations offer reconstruction presettings for virtually reconstructed images. Although uncommon in recent software releases, suboptimal 3D reconstruction settings (eg, opacity setting for volume-rendered images or perspective shaded-surface-display [SSD] threshold) can lead to shine-through artifacts at virtual endoscopy that simulate ulceration and wall defects and degrade image quality (Fig 9) (39). These artifacts often appear in areas where the colon is not directly surrounded by pericolic tissues and the colonic wall is adjacent to other bowel segments, or perhaps in haustral folds. On corresponding 2D images, no wall defects corresponding to 3D features are present. Adjustment of opacity settings for volume-rendered images or an increased-perspective SSD threshold is helpful in overcoming these pitfalls.

View larger version (108K): [in this window] [in a new window] [Download PPT slide]   Figure 9.  Shine-through artifact of the colonic wall in a 59-year-old asymptomatic woman. Three-dimensional endoluminal CT image obtained with suboptimal reconstruction settings (reduced-perspective SSD threshold) shows pseudodefects of the colonic wall (arrows).

View larger version (135K): [in this window] [in a new window] [Download PPT slide]   Figure 10a.  Effect of different rendering techniques and evaluation software on lesion conspicuity. (a) Three-dimensional endoluminal CT image of an anthropomorphic colon phantom only faintly shows a small (4-mm), simulated flat lesion (crosshairs). (b) Three-dimensional endoluminal CT image obtained with a more recent version of different software shows improved conspicuity of the simulated lesion (arrow).

View larger version (98K): [in this window] [in a new window] [Download PPT slide]   Figure 10b.  Effect of different rendering techniques and evaluation software on lesion conspicuity. (a) Three-dimensional endoluminal CT image of an anthropomorphic colon phantom only faintly shows a small (4-mm), simulated flat lesion (crosshairs). (b) Three-dimensional endoluminal CT image obtained with a more recent version of different software shows improved conspicuity of the simulated lesion (arrow).

There are two primary techniques for data interpretation: a primary 2D approach and a primary 3D approach. With either technique, the alternative viewing technique must be available for problem solving and to aid in differentiating folds, fecal matter, and polyps. Investigation of only 2D or 3D projections has been shown to be less sensitive and less specific than the combined use of both visualization techniques for data evaluation (50,51).

The 3D fly-through as a primary search process makes use of 3D evaluation for lesion detection and axial 2D evaluation for characterization. Primary 3D evaluation has been shown to be sensitive for polyp detection (1). Both the conspicuity (especially of small and medium-sized polyps) and the duration of visualization are increased by using 3D endoluminal fly-through, thereby facilitating the initial search process (1). However, primary 3D evaluation must be performed with the patient in the prone and supine positions in both antegrade and retrograde directions, which is time consuming. If colonic segments are collapsed, axial 2D evaluation must be used as an alternative. Primary 3D evaluation performed in only one direction will result in a poorer performance because of missed lesions behind the haustral folds (52).

Primary axial 2D evaluation makes use of 2D views for lesion detection and 3D views for characterization. Primary 2D evaluation is based on "lumen tracking," as described by Royster et al (50), which interactively tracks through the image volume and focuses on only the air-distended colonic lumen from one end to the other, with a special focus on the cross-section of only one colonic segment at a time. Primary 2D evaluation is more time efficient (51,53) and provides information about the attenuation of findings during the search process. In addition, it has been suggested that flat or annular lesions are best seen on transverse 2D images displayed with narrow window width and level settings (19). Previous reports showed lower interobserver agreement when only 2D images were used (54).

"Overview evaluations" of multiple colonic segments, as with coronal images without a focus on any one segment, will result in a poorer performance. Even an advanced adenoma 1 cm or larger is a diminutive structure to search for in a whole patient data set and may be missed more easily if there is no focus on a particular segment.

In summary, optimal evaluation of CT colonographic data is facilitated by easy access to supine and prone 2D and 3D images. Evaluation of only supine or only prone images or evaluation in only one direction at virtual endoscopy, and restriction to only 2D or 3D evaluation, may lead to either perception or interpretation errors. At present, there is no general consensus on an optimal primary 2D or 3D approach.

Underdistention.— Underdistention of the colon leads to collapsed colonic segments. A colonic lesion located in a collapsed segment is not visible at either 2D or 3D imaging and will be missed. The acquisition of both supine and prone scans improves overall colonic distention and leads to a diagnostically important redistribution of gas, residual fluid, and stool (27–29). A lesion hidden in a collapsed segment with the patient in one position will most likely be visible with the patient in the other position, and even more so if the corresponding segment has been redistended with rectal insufflation of additional gas (Fig 11).

View larger version (121K): [in this window] [in a new window] [Download PPT slide]   Figure 11a.  Underdistention causing perception error in a 65-year-old man with a stenotic sigmoid cancer. (a) Supine coronal CT scan shows collapse of the descending colon (arrow). No lesion is depicted in this segment. (b) Prone coronal CT scan of the distended colonic segment shows a 9-mm synchronous polyp (arrow).

View larger version (101K): [in this window] [in a new window] [Download PPT slide]   Figure 11b.  Underdistention causing perception error in a 65-year-old man with a stenotic sigmoid cancer. (a) Supine coronal CT scan shows collapse of the descending colon (arrow). No lesion is depicted in this segment. (b) Prone coronal CT scan of the distended colonic segment shows a 9-mm synchronous polyp (arrow).

Residual Fluid.— Residual fluid can obscure colonic lesions and lead to perception errors. Lesions obscured by fluid with the patient in one position are easily detected with the patient in the other position (Fig 12). In addition, in cases of excessive retained fluid—for example, in under-distended segments—2D MPR images with narrow window settings should be used to evaluate these segments for intraluminal structures with soft-tissue attenuation. Intravenous contrast material has been reported to be helpful in enhancing submerged polyps in colonic segments that are completely filled with fluid (23). Differentiation of lesions hidden by residual fluid may be possible by tagging residual fluid (fluid tagging) with orally ingested barium or iodinated contrast material prior to the investigation (12,55).

View larger version (162K): [in this window] [in a new window] [Download PPT slide]   Figure 12a.  Residual fluid obscuring a polyp in the descending colon in a 67-year-old man. Colonoscopic examination could not be completed because of a suspected stenosis. (a) Prone CT scan shows a fluid level (arrow) in the ventral dependent portion of the descending colon. (b) On a supine CT scan, the fluid level has shifted to the dorsal dependent portion of the colon, revealing a 9-mm polyp (arrow). The polyp was not seen on the prone CT scan because the lesion was submerged in residual fluid.

View larger version (165K): [in this window] [in a new window] [Download PPT slide]   Figure 12b.  Residual fluid obscuring a polyp in the descending colon in a 67-year-old man. Colonoscopic examination could not be completed because of a suspected stenosis. (a) Prone CT scan shows a fluid level (arrow) in the ventral dependent portion of the descending colon. (b) On a supine CT scan, the fluid level has shifted to the dorsal dependent portion of the colon, revealing a 9-mm polyp (arrow). The polyp was not seen on the prone CT scan because the lesion was submerged in residual fluid.

Bidirectional evaluation from the rectum to the cecum and back helps to overcome such pitfalls (Fig 13). If 3D virtual colonoscopy is used for the primary search process, the evaluation must be performed in both directions; otherwise, significant disease may be missed. Developments like virtual dissection or unfolded cube projections are designed to overcome blind spots and may allow unidirectional 3D evaluation. Recent reports did not show a significant improvement in sensitivity with this approach, but did show a significant reduction in examination time compared with 3D endoluminal views (56–58).

View larger version (107K): [in this window] [in a new window] [Download PPT slide]   Figure 13a.  Advantage of performing primary 3D virtual endoscopy in both the antegrade and retrograde directions. The patient was a 49-year-old asymptomatic man. (a) Antegrade 3D endoluminal CT image shows crowded haustral folds (arrowheads) in the hepatic flexure. No lesion is depicted. (b) Retrograde 3D endoluminal CT image of the same region shows a sessile 9-mm polyp (arrow) located behind a haustral fold (arrowhead).

View larger version (115K): [in this window] [in a new window] [Download PPT slide]   Figure 13b.  Advantage of performing primary 3D virtual endoscopy in both the antegrade and retrograde directions. The patient was a 49-year-old asymptomatic man. (a) Antegrade 3D endoluminal CT image shows crowded haustral folds (arrowheads) in the hepatic flexure. No lesion is depicted. (b) Retrograde 3D endoluminal CT image of the same region shows a sessile 9-mm polyp (arrow) located behind a haustral fold (arrowhead).

Several criteria exist for correct classification of filling defects and polypoid lesions of the colon at CT colonography. Colonic filling defects are characterized on the basis of their morphologic features or shape, attenuation characteristics, and mobility (19,60). On the basis of these criteria, polyps are defined as sessile or stalked, round, oval, or lobulated intraluminal filling defects with homogeneous soft-tissue attenuation that maintain their position with respect to the bowel surface. If intravenous contrast material is administered, polyps will enhance (21,22). Colonic lipomas can be differentiated from adenomatous polyps because of their fat attenuation.

Intraluminal Content.— Residual stool may simulate a polyp if it has a polypoid shape at 3D virtual colonoscopy. The 2D imaging features of stool are helpful in overcoming this pitfall (Figs 1, 14)—in particular, the fact that residual stool has inhomogeneous attenuation because it typically contains trapped gas or food particles, whereas polyps generally have homogeneous soft-tissue attenuation (19). Stool will likely move to the dependent part of the colon when the patient’s position changes from prone to supine because it is not attached to the bowel wall (24,27).

View larger version (111K): [in this window] [in a new window] [Download PPT slide]   Figure 14a.  Polypoid stool simulating a filling defect in a 73-year-old woman. Colonoscopic examination could not be completed because of diverticulosis. (a) Supine 3D endoluminal CT image shows a round, well-circumscribed, polypoid "filling defect" (arrow) in the splenic flexure of the colon. (b) Supine CT scan shows a collection of gas inside the lesion (arrow), which is located on the dorsal dependent wall. (c) Prone 3D endoluminal CT scan shows that the lesion (arrow) has moved to the ventral dependent wall, a finding that is indicative of lesion mobility.

View larger version (140K): [in this window] [in a new window] [Download PPT slide]   Figure 14b.  Polypoid stool simulating a filling defect in a 73-year-old woman. Colonoscopic examination could not be completed because of diverticulosis. (a) Supine 3D endoluminal CT image shows a round, well-circumscribed, polypoid "filling defect" (arrow) in the splenic flexure of the colon. (b) Supine CT scan shows a collection of gas inside the lesion (arrow), which is located on the dorsal dependent wall. (c) Prone 3D endoluminal CT scan shows that the lesion (arrow) has moved to the ventral dependent wall, a finding that is indicative of lesion mobility.

View larger version (147K): [in this window] [in a new window] [Download PPT slide]   Figure 14c.  Polypoid stool simulating a filling defect in a 73-year-old woman. Colonoscopic examination could not be completed because of diverticulosis. (a) Supine 3D endoluminal CT image shows a round, well-circumscribed, polypoid "filling defect" (arrow) in the splenic flexure of the colon. (b) Supine CT scan shows a collection of gas inside the lesion (arrow), which is located on the dorsal dependent wall. (c) Prone 3D endoluminal CT scan shows that the lesion (arrow) has moved to the ventral dependent wall, a finding that is indicative of lesion mobility.

View larger version (88K): [in this window] [in a new window] [Download PPT slide]   Figure 15a.  Solid collections of retained barium simulating a polypoid lesion in a 77-year-old patient with diverticulosis. (a) Prone 3D endoluminal CT image shows a round, well-circumscribed, polypoid filling defect in the sigmoid colon (arrow). (b) Prone coronal CT scan shows the filling defect with homogeneous high attenuation (arrow), a finding that represents barium-tagged fecal material.

View larger version (133K): [in this window] [in a new window] [Download PPT slide]   Figure 15b.  Solid collections of retained barium simulating a polypoid lesion in a 77-year-old patient with diverticulosis. (a) Prone 3D endoluminal CT image shows a round, well-circumscribed, polypoid filling defect in the sigmoid colon (arrow). (b) Prone coronal CT scan shows the filling defect with homogeneous high attenuation (arrow), a finding that represents barium-tagged fecal material.

Depressed bowel wall structures can mimic polypoid lesions on 3D images. Gas bubbles sometimes appear on the surface of fluid levels. These gas bubbles tend to be localized and adjacent to haustral folds or corners, and are often multiple. The bubble layer cannot usually be seen on 2D or 3D images. The margin of a bubble on the surface of a fluid level is caused by adhesive forces and appears as a round depressed structure that can be confused with a polypoid or diverticular lesion on 3D images. Corresponding 2D images are helpful in identifying a slightly depressed structure in a fluid level (Fig 16).

View larger version (56K): [in this window] [in a new window] [Download PPT slide]   Figure 16a.  Gas bubble on an intraluminal fluid level simulating a polypoid lesion in a 64-year-old asymptomatic man. (a) Drawings illustrate how a gas bubble (top) appears at CT (bottom). (b) Prone 3D endoluminal CT image shows a round, "polypoid" abnormality (arrow) in a horizontal fluid level in the colon. (c) Prone CT scan shows the abnormality to be a round, depressed structure caused by a gas bubble (arrow). Note that the bubble layer cannot be visualized at CT.

View larger version (77K): [in this window] [in a new window] [Download PPT slide]   Figure 16b.  Gas bubble on an intraluminal fluid level simulating a polypoid lesion in a 64-year-old asymptomatic man. (a) Drawings illustrate how a gas bubble (top) appears at CT (bottom). (b) Prone 3D endoluminal CT image shows a round, "polypoid" abnormality (arrow) in a horizontal fluid level in the colon. (c) Prone CT scan shows the abnormality to be a round, depressed structure caused by a gas bubble (arrow). Note that the bubble layer cannot be visualized at CT.

View larger version (84K): [in this window] [in a new window] [Download PPT slide]   Figure 16c.  Gas bubble on an intraluminal fluid level simulating a polypoid lesion in a 64-year-old asymptomatic man. (a) Drawings illustrate how a gas bubble (top) appears at CT (bottom). (b) Prone 3D endoluminal CT image shows a round, "polypoid" abnormality (arrow) in a horizontal fluid level in the colon. (c) Prone CT scan shows the abnormality to be a round, depressed structure caused by a gas bubble (arrow). Note that the bubble layer cannot be visualized at CT.

View larger version (123K): [in this window] [in a new window] [Download PPT slide]   Figure 17a.  Bulbous haustral fold simulating a sessile polyp in a 60-year-old asymptomatic woman. (a) Coronal CT scan shows a polypoid filling defect with soft-tissue attenuation in the cecum (arrow). (b) Three-dimensional endoluminal CT image shows the filling defect (arrow) with linear morphologic features, a finding that is indicative of a haustral fold.

View larger version (98K): [in this window] [in a new window] [Download PPT slide]   Figure 17b.  Bulbous haustral fold simulating a sessile polyp in a 60-year-old asymptomatic woman. (a) Coronal CT scan shows a polypoid filling defect with soft-tissue attenuation in the cecum (arrow). (b) Three-dimensional endoluminal CT image shows the filling defect (arrow) with linear morphologic features, a finding that is indicative of a haustral fold.

Scans obtained in patients with collapsed colonic segments should be inspected carefully for signs of malignancy, including asymmetric or circumferential bowel wall thickening, pericolic fat stranding, lymphadenopathy, and metastatic disease (Fig 18) (34). Moreover, a collapsed segment should be reevaluated by scrutinizing it on the corresponding supine or prone scan, after it has most likely become distended due to redistribution of gas during repositioning.

View larger version (168K): [in this window] [in a new window] [Download PPT slide]   Figure 18a.  Collapsed colonic segments in two patients with suspected colorectal cancer. (a) Supine sagittal contrast-enhanced CT scan obtained in a 72-year-old man shows a collapsed rectum (arrow) with only mild pseudothickening of the colonic wall, no shoulder formations, and normal pericolic fat tissue. (b) Supine sagittal contrast-enhanced CT scan obtained in a 61-year-old man shows a collapsed rectum (arrow) with circumferential wall thickening, stranding of the pericolic fat, and contrast enhancement, findings that are indicative of malignancy. Histologic analysis revealed a rectal cancer.

View larger version (147K): [in this window] [in a new window] [Download PPT slide]   Figure 18b.  Collapsed colonic segments in two patients with suspected colorectal cancer. (a) Supine sagittal contrast-enhanced CT scan obtained in a 72-year-old man shows a collapsed rectum (arrow) with only mild pseudothickening of the colonic wall, no shoulder formations, and normal pericolic fat tissue. (b) Supine sagittal contrast-enhanced CT scan obtained in a 61-year-old man shows a collapsed rectum (arrow) with circumferential wall thickening, stranding of the pericolic fat, and contrast enhancement, findings that are indicative of malignancy. Histologic analysis revealed a rectal cancer.

Pedunculated polyps, wit