Pitfalls and Artifacts
Errors leading to incorrect diagnosis can arise at any stage during CT colonography, so the range of potential pitfalls and pseudolesions is wide. Suboptimal bowel preparation and distension, suboptimal CT protocols, and errors of perception and interpretation can lead to false-positive or false-negative results, compromising the diagnostic accuracy of CT colonography. These pitfalls are avoidable if examination and interpretation techniques are optimized according to current guidelines, a thorough knowledge is acquired of the typical CT morphological criteria of the various pseudolesions seen at CT colonography, and, finally, if the interpreter is familiar with the appropriate problem-solving strategies. Table 4.6 lists the most common pitfalls and artifacts encountered in CT colonography. Pitfalls in bowel preparation, CT scanning, and interpretation techniques are listed only briefly here as they are discussed in detail in the sections on those procedures; similarly, morphological features of normal anatomical structures or pathologies of the colon that may be overlooked or misinterpreted are discussed in the sections on those structures or abnormalities.
Spiral CT protocols with breath-hold times of 35–50 seconds are vulnerable to respiratory artifacts and should be avoided. Respiratory movements during image acquisition cause incorrect data acquisition and substantially diminish image quality. They are more common in the transverse colon and in the flexures because of their proximity to the diaphragm. On 2D and 3D images, respiratory movements produce characteristic linear artifacts which may appear as an abrupt wall defect on one side of the bowel and a polypoid structure on the opposite surface. Respiratory artifacts can be rapidly identified on coronal and sagittal 2D views on the basis of irregularities of the outer abdominal wall along the z-axis of the patient (Fig. 4.121) The significance of respiratory artifacts decreases with shorter acquisition times, such as are achieved with MDCT. With scan times of now less than 10 seconds, most patients have no difficulty in holding their breath, thus eliminating respiratory artifacts. As a general rule, for CT colonography the narrowest collimation should be chosen at which complete image acquisition will be possible in one breath-hold. To minimize respiratory artifacts, CT colonographic scans should be performed in the craniocaudal direction.
Stairstep artifacts occur only on reconstructed endoluminal 3D or multiplanar 2D views. On 3D images, the artifacts appear as a series of concentric rings in the wall of the distended colon, creating a step-like appearance (Fig. 4.122). These artifacts increase with increasing slice thickness. Stairstep artifacts are most severe in regions in which the bowel wall is oblique to the z-axis of the patient, for example at the flexures, parts of the rectum, or cecum. This decreases image quality, especially on 3D images. Stairstep artifacts may be avoided by using thinner slices (<1.5 mm) and cease to occur when scanners with at least 16 slices are used.
Given the extremely large attenuation difference between intraluminal gas and the soft-tissue-attenuation bowel wall, a substantial dose reduction is possible without compromising polyp detection. However, substantially reducing radiation exposure leads to increased noise which diminishes the quality of 2D and 3D representations. Image noise increases additionally with the use of narrower collimation and thin-slice reconstructions such as are are routinely applied in CT colonography.
On 3D images, one can identify noise as a diffuse granular or nodular superficial pattern on the bowel wall. On 2D images, however, the bowel wall has a normal thickness and regular anatomical structures (see Fig. 4.9). Such pseudo-changes are not to be confused with the “cobblestone pattern” changes seen in Crohn disease, where 3D images also show a nodular surface pattern on the colonic wall, but the bowel wall of the affected segment is also thickened on 2D views. In advanced disease loss of haustration may also occur. Image noise-related artifacts are not only observed on the colonic wall: They are also present on the surface of residual fluid or on the rectal tube. If one of the two image series is acquired using a standard dose, low-dose-associated changes may be easily identified by comparing the two series. For a routine examination, a dose reduction down to 50 mAs for one or both series is about right, although good results have been reported at even lower dose settings. Because the subjective perception of image noise is related to the window setting, noise can also be reduced on 2D views by selecting a wider window.
The presence of metallic (high-attenuation) foreign bodies can produce beam-hardening artifacts, reducing the quality of 2D and 3D images and compromising or even entirely preventing evaluation of affected portions of the bowel wall. Metal artifacts are often found in the pelvis of patients with hip prostheses.
Because of metal artifacts, evaluation of the rectosigmoid is limited in patients with hip prostheses. This should be noted in the radiologist′s report (Fig. 4.123).
Pitfalls in Data Evaluation
Errors in data evaluation may be technical or reading errors. In terms of correct evaluation technique, it is important to select the appropriate window setting for 2D views and the correct threshold for 3D reconstructions. Reading errors include errors in detection of lesions (perception errors) and errors of interpretation (interpretation errors).
Correct data evaluation depends on complete and artifact-free 2D and 3D depictions of the colon. On 2D views this requires the use of various window settings, while for 3D views, not only proper 3D rendering but also correct selection of the 3D threshold is needed.
Window Settings for 2D Views
Wide window settings (such as lung or bone window settings, e.g., window width 1500 HU, window level –150 HU) produce a high contrast between the colonic lumen and bowel wall structures. This enables visualization of more details of the bowel wall, such as thin semilunar folds, which would not be entirely shown on narrower window settings. The disadvantage of wider windows is poorer differentiation of the internal structure of detected lesions.
Narrow window settings (soft-tissue windows, width 400 HU, level 10 HU) allow better interpretation of colonic lesions in terms of CT density and internal structure (e.g., polyp versus lipoma). A drawback of narrow window settings is loss of detail of colonic wall structures such as thin semilunar folds. CT colonographic evaluation requires both complete depiction of the bowel wall using wide window settings and detailed representation of the bowel wall structure using narrow window settings (soft-tissue windows) (see Fig. 3.2). For manual measurement of polyps on 2D views, however, narrow window settings cause the lesion to appear smaller.
Three-Dimensional Shine-Through Artifacts
Modern 3D workstations offer default settings for virtual 3D reconstructions which do not require to be optimized by the user. On such workstations—although rarely—suboptimal 3D reconstruction settings (e.g., opacity for volume rendering or perspective shaded surface display [pSSD] threshold) can lead to shine-through artifacts on virtual endoscopic views. These can simulate ulcerations and defects in the bowel wall (Fig. 4.124). When navigating through the colon on 3D endoluminal views, it is possible inadvertently to leave the colonic segment through a “defect” of this kind. Such artifacts are found in areas where two adjacent intestinal loops are in contact. Unusual configurations of semilunar folds can also give rise to this kind of artifact. Corresponding 2D views do not show any correlate of the 3D wall defects. Correct adjustment of the opacity settings for volume rendering images or raising the pSSD threshold can help to avoid these reconstruction errors. The best thing to do if such artifacts occur is to return to the manufacturer′s default settings.
Perception errors (the classic “missing” of a lesion) are a common reason for false-negative diagnoses. The perception of a polyp depends on several factors including morphological features of the lesion such as its size, shape and conspicuity, individual examiner-dependent factors such as expertise and level of concentration/fatigue, technical factors such as bowel preparation, distension, scanning parameters, and, not least, the evaluation technique used.
Inadequate colonic distension results in collapsed bowel segments.
A polyp located in a collapsed colon segment cannot be detected on either 2D or 3D views.
The acquisition of datasets in two positions increases the number of distended bowel segments because gas, residual fluid, and stool are all redistributed. Every bowel segment should be distended in at least one of the two scans, ideally in both. A lesion in a segment that is collapsed in one position is very likely to be visible after the patient changes position. This effect can be reinforced by additional rectal insufflation of gas after the patient changes position. During data analysis, collapsed bowel segments should be identified and very closely evaluated on 2D views, ideally with narrow window settings (softtissue windows). Particular care should be taken to identify any masses or pericolic signs of fat stranding as possible signs of a pathological process. These segments should also be evaluated in the other patient position, after redistension if possible (Fig. 4.125). Thus, for complete evaluation of a CT colonographic examination, interpretation of both prone and supine images is required. Significant increases in sensitivity can be achieved in this way, as has been shown in several studies.
Untagged Residual Fluid
Residual fluid is commonly seen at CT colonography. The amount of fluid depends on patient-specific factors as well as the bowel preparation protocol used. On 2D and 3D views, residual fluid appears as a filling defect in the lumen with a horizontal, linear fluid level. On 2D views untagged residual fluid has a homogeneous structure with fluid attenuation (Fig. 4.126). Fluid follows gravity and always collects in the lowest-lying portions of the intestinal lumen, so in the supine position it is more often seen in the descending colon and rectum, while in the prone position it collects in the transverse colon.
Untagged residual fluid can conceal a colonic lesion, leading to an error of perception.
Lesions that are obscured by untagged fluid in one position are often detectable after the patient changes position (Fig. 4.126). If there are large amounts of residual fluid, however, visualization of the entire colonic mucosal surface is not guaranteed, even with evaluation of both prone and supine scans. When there is a large amount of residual fluid, segments that are filled with fluid in both the supine and prone positions should be evaluated using narrow window settings, which may allow detection of intraluminal soft-tissue density filling defects.
In such cases, intravenous administration of a contrast agent can help to improve detection rates because polyps enhance. However, labeling of residual bowel content using an oral contrast medium (fecal tagging) is recommended as the best method of revealing colonic lesions otherwise hidden by intraluminal fluid (Fig. 4.127).