DOI: 10.1148/radiol.2361040231
(Radiology 2005;236:118-124.)
© RSNA, 2005
Polyp Size at CT Colonography after Electronic Subtraction Cleansing in an Anthropomorphic Colon Phantom1
Michael E. Zalis, MD,
James J. Perumpillichira, MD,
John Y. Kim, MD, MPH,
Chiara Del Frate, MD2,
Cordula Magee, PhD and
Peter F. Hahn, MD, PhD
1 From the Division of Abdominal Imaging and Intervention, Department of Radiology, Massachusetts General Hospital, 55 Fruit St, White 270, Boston, MA 02114. Received February 16, 2004; revision requested April 21; final revision received August 3; accepted September 20.
Address correspondence to M.E.Z. (e-mail: mzalis{at}partners.org).
 |
ABSTRACT
|
|---|
PURPOSE: To evaluate the effect of various bowel contrast material concentrations and subtraction software on size measurements of well-defined polyp lesions in a colon phantom at CT colonography.
MATERIALS AND METHODS: Repeated scanning and a precise reference standard required the use of a colon phantom in which 21 polyps were randomly distributed. Two readers who had each reviewed computed tomographic (CT) colonographic images from more than 100 cases evaluated polyp size on images obtained when the phantom was partially filled with varying concentrations of contrast material, scanned by using CT colonography, and subjected to electronic subtraction cleansing. The single largest dimension was recorded for each reader for a randomized series of polyps. These measurements were compared with a reference standard that was based on a combination of the manufacturer's polyp size specifications and the subsequent verification of these sizes by an independent consensus panel. Six weeks after initial observations, readers evaluated images of the phantom scanned without the presence of contrast material. Polyp size estimations for the two readers for each series were compared with the reference standard to obtain a mean absolute measurement error for each reader for each series. Data for each reader were compared by using a nonparametric Kruskal-Wallis analysis of variance test. A pair-wise comparison of the experimental and control series was then performed by using the Dunn post hoc test.
RESULTS: Contrast material dilutions resulting in an average attenuation of less than 500 HU resulted in complete subtraction and the absence of streak artifacts. There was no statistically significant difference between the average measurement error for contrast attenuations between 300 and 500 HU when compared with that of control. Streak artifact was noticeable for the highest dilution (mean, 840 HU). No statistically significant differences were observed for series in which cleansing software was used in the absence of bowel contrast material.
CONCLUSION: The combination of electronic cleansing and bowel contrast enhancement in the range of 300–500 HU results in no substantial change in readers' estimations of polyp size at CT colonography.
© RSNA, 2005
|
INTRODUCTION
|
|---|
Previous data support the use of computed tomographic (CT) colonography as a noninvasive means to perform colon cancer screening (1). Current implementation of CT colonography, however, requires that patients undergo a demanding preexamination cathartic bowel preparation (2). The discomfort and embarrassment associated with this preparation contribute to the negative impressions of all structural colon examinations, including CT colonography, and likely affect compliance with colon cancer screening recommendations (3–5). As a result, multiple investigators have attempted to develop a modified CT colonography protocol that reduces the requirement for complete purgation of the bowel prior to examination (6–8).
One proposed modification is the combination of preexamination tagging of ingested foodstuffs with postacquisition electronic cleansing of the colonic contents on resulting CT images, a process dubbed electronic cleansing or digital subtraction bowel cleansing (8). With this technique, the tagging agents are incorporated into ingested materials only and are not appreciably absorbed by the bowel wall or target polyps. Hence, at cross-sectional imaging, polyps and ingested material have markedly different internal attenuations and are distinctly recognizable. Cleansing software selectively removes the high-attenuation tagged materials and leaves the soft-tissue-attenuation structures of the bowel, such as polyps and haustral folds, untouched (8).
The potential advantage of this approach is the simultaneous ability to reduce the duress of preexamination bowel cleansing yet preserve the ability to perform three-dimensional endoluminal evaluations of colon features. Findings from prior studies have demonstrated that the performance of CT colonography is enhanced when these three-dimensional techniques are available (9).
The implementation of minimal preparation techniques, such as electronic cleansing, requires validation that these techniques do not markedly change the imaging characteristics of target polyps. Of these characteristics, polyp size remains the single most important criterion by which to assign risk of carcinoma (10). In addition, polyp size is an important factor in planning clinical management of detected lesions (11). In the imaging scenario proposed for minimal preparation CT colonography, at least two factors could conceivably alter the perceived size of polyp lesions: pseudoenhancement of polyps submerged in high-attenuation contrast material and subtraction artifacts resulting from computerized image processing (12).
The concept of tagging colonic contents, as well as of electronic cleansing, has been demonstrated in preliminary studies by using both barium-based and iodinated contrast agents (6,8). Barium-based contrast agents require emulsifiers in order to remain in aqueous solution and appear to be the most effective agents for solid fecal tagging only; iodinated contrast agents readily dissolve in aqueous solution and are effective as single agents for both solid fecal and fluid tagging (7,8,13). Our experience suggests that iodinated contrast agents provide a more homogeneous bowel preparation and are deliverable to patients in a smaller volume.
We hypothesized that there exists a range of iodinated contrast material concentrations for which the measurements of polyp size obtained following fluid and/or fecal tagging and electronic cleansing are equivalent to those obtained without the use of tagging and electronic cleansing. Thus, the purpose of the present study was to evaluate the effect of bowel contrast material concentration and subtraction software on the size measurements of well-defined polyp lesions in a colon phantom.
|
MATERIALS AND METHODS
|
|---|
Reference Standard and Phantom Design
The precise evaluation of polyp sizes for this study required a reference standard simultaneously more precise and more flexible than that obtained with conventional colonoscopy. Measurement of polyp sizes at conventional colonoscopy is known to be both user and technique dependent and is subject to error (14). In addition, because of ethical and logistic considerations, it was not feasible to conduct repeated CT examinations on human subjects by using differing bowel contrast material preparations. Consequently, we commissioned an anthropomorphic 35-cm colon phantom (Phantom Laboratory, Salem, NY) that included polyps of precisely determined size, shape, and composition.
The reference standard for polyp sizes in this study consisted of a combination of the manufacturer's specifications of polyp size, as commissioned originally by the authors, and the subsequent verification of these sizes by a two-member consensus panel. The consensus panel cross-referenced the manufacturer's measurements for each polyp with the sizes obtained after scanning the phantom in a clean (ie, no bowel contrast material) state; the consensus panel then evaluated the resulting images with a dedicated colonography software reading platform (Vitrea 2, version 3.2; Vital Images, Plymouth, Minn). In instances in which there was a discrepancy between the manufacturer's specifications and the observed measurements, repeat measurements were made. If the discrepancy persisted, the consensus panel's measurements were used as the reference standard. For all measurements in this study, including those of the consensus panel, we used the single largest dimension measurements that were observed on multiplanar views because use of these measurements most closely resembled the typical clinical practice for polyp measurement at CT colonography (14). The members of the consensus panel were not otherwise involved in the evaluation of polyp sizes; the experience of each member of the consensus panel included review of more than 100 cases by using CT colonography and the Vitrea 2 workstation.
The phantom was designed so that the dimensions and attenuation of its constituent materials, including the colonic polyps and folds, closely resembled the features observed in human data sets for CT colonography. In particular, the mean attenuation of the polyps and the colonic wall at CT was 10 HU ± 4. The mean diameter of the colon in the phantom was 3.5 cm, and the colon was U-shaped, which facilitated an even distribution of polyps and features in all three dimensions. The phantom also included extracolonic material to simulate fat (mean attenuation at CT, –25 HU ± 8), and the overall attenuation of the phantom was designed to simulate that of the human body. In addition, the phantom was designed with two ports that permitted the partial filling of the colon lumen with bowel contrast material solutions. These solutions could be subsequently drained from the phantom so that repeated measurements of the same polyps could be performed with differing concentrations of bowel contrast material. In this way, the phantom could be rescanned in multiple contrast material conditions, permitting comparison between the effects of contrast material on the same validated polyp lesions.
The phantom contained 21 soft-tissue-attenuation polyps that ranged from 8 to 25 mm in diameter and included pedunculated and sessile morphologic characteristics. Smaller lesions were not included in the phantom because of manufacturer limitations and the relatively low clinical importance generally assigned to diminutive polyps (1,15). There were seven polyps in each of three (ie, 8–14 mm, 15–20 mm, and 21–25 mm) size ranges. Polyps were randomly distributed on the inner surface of the phantom. There were a total of nine pedunculated lesions that were spherical, but because of manufacturer limitations, these lesions did not have separate stalks. There were three pedunculated lesions and four sessile lesions in each of the three size ranges previously listed. Flat lesions were not included in the construction of the phantom. The polyps were distributed randomly on the inner surface of the phantom so that at least one polyp of each morphologic category and size range was oriented in each of the three orthogonal spatial planes. The phantom included eight folds that were distributed evenly throughout the length of the phantom. All of the polyps in the phantom were attached primarily to the inner surface of the colon. Because of manufacturing limitations, there were no polyp lesions attached to only a colonic fold. Images of the phantom that demonstrate the polyps and folds are shown in Figure 1.
View larger version (108K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1a. Images of 35-cm phantom in native (ie, unopacified) state prior to instillation of contrast material or implementation of subtraction software. (a) CT topogram shows air-filled colon lumen with visible 10-mm sessile polyp (black arrow) and simulated haustral fold (gray arrow). White arrows indicate presence of ports used to fill phantom with contrast material for measurement experiments. (b) Corresponding transverse view of 10-mm sessile polyp (arrow) obtained at CT colonography. (c) Volume-rendered endoluminal CT colonographic view of 10-mm sessile polyp (black arrow) and simulated haustral fold (gray arrow).
|
|
View larger version (63K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1b. Images of 35-cm phantom in native (ie, unopacified) state prior to instillation of contrast material or implementation of subtraction software. (a) CT topogram shows air-filled colon lumen with visible 10-mm sessile polyp (black arrow) and simulated haustral fold (gray arrow). White arrows indicate presence of ports used to fill phantom with contrast material for measurement experiments. (b) Corresponding transverse view of 10-mm sessile polyp (arrow) obtained at CT colonography. (c) Volume-rendered endoluminal CT colonographic view of 10-mm sessile polyp (black arrow) and simulated haustral fold (gray arrow).
|
|
View larger version (108K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1c. Images of 35-cm phantom in native (ie, unopacified) state prior to instillation of contrast material or implementation of subtraction software. (a) CT topogram shows air-filled colon lumen with visible 10-mm sessile polyp (black arrow) and simulated haustral fold (gray arrow). White arrows indicate presence of ports used to fill phantom with contrast material for measurement experiments. (b) Corresponding transverse view of 10-mm sessile polyp (arrow) obtained at CT colonography. (c) Volume-rendered endoluminal CT colonographic view of 10-mm sessile polyp (black arrow) and simulated haustral fold (gray arrow).
|
|
Bowel Contrast Enhancement and Scanning Parameters
The contrast material that was employed for the experiments consisted of 300 mg of nonionic organically bound iodine per milliliter (ioxilan, Oxilan; Guerbet, Bloomingtom, Ind), which was diluted in a contrast material to normal saline ratio of 1:5, 1:10, 1:20, and 1:30. These dilutions were chosen to recreate the attenuation range observed in our clinical experience with fecal and fluid tagging during CT colonography. For each imaging experiment, 700 mL of the appropriate dilution was instilled into the phantom through the built-in ports. Between experiments, the phantom was drained and thoroughly flushed with normal saline to ensure that no residual contrast material remained from prior acquisitions.
For all series, the phantom was scanned by using a four-channel multi–detector row CT scanner (LightSpeed; GE Medical Systems, Milwaukee, Wis) with the same helical scanning parameters employed at our institution for screening CT colonography: 140 kVp, 50 mA, scanning time of 0.5 second, interleaved sections with a pitch of 1:6, table speed of 15 mm/sec, and beam collimation of 3.75 mm. Sections were reconstructed every 1.8 mm. These parameters correspond to those used contemporaneously at our institution for clinical CT colonography. The phantom was oriented in a superior-inferior, transverse orientation for all imaging procedures. Scanning was first performed with no contrast material instilled into the phantom; scanning was then repeated with each of the four contrast material dilutions previously described.
Reader Evaluations
Two readers (M.E.Z., J.J.P.) who had each reviewed more than 100 cases by using CT colonography and the Vitrea 2 workstation were asked to make independent measurements of the partially and completely submerged lesions in the phantom. The two readers were blinded to the attenuation of the contrast material concentration for each series. The contrast material dilution series were presented to the readers in random order with respect to contrast material concentration. Readers were also unaware of the reference measurements determined by the consensus panel. After a delay of 6 weeks, the two readers were then presented two series of phantom images in which no bowel contrast material was used. The first of these series served as the control and consisted of the phantom in its native (ie, unopacified) state; no cleansing software was used. In the second series, the phantom was subjected to the electronic cleansing software, but no contrast material was used. This last series permitted evaluation of the effect of the subtraction software, independent of the effect of the bowel contrast material. The order in which these two unenhanced series were presented to the readers was also randomized.
The two readers made measurements of the polyps by using a combination of coordinated multiplanar reconstructions and volume-rendered endoluminal views, as available on the Vitrea 2 workstation. In all cases, the actual measurements were obtained in the multiplanar views; the endoluminal view was not directly used for measurement because this view can be subject to field-of-view distortions and errors in caliper placement (14). We specifically did not record which of the multiplanar views were used for measurement to avoid influencing the readers and to more closely simulate actual clinical evaluations. Given the nonisotropic voxel scanning parameters used, we expected that the measurements made in the sagittal and coronal planes would be likely to demonstrate greater absolute measurement errors. We deduced, however, that this potential inaccuracy would be constant across the experimental imaging conditions given that the orientation of the phantom in the scanner, the measured polyps, and the scanning parameters were all held constant. Also, to simulate clinical reading conditions, we did not specify or limit readers with respect to zooming images on the display system.
In simulating actual clinical experience, we recorded single measurements for each reader for each polyp. In accordance with standard practice at our institution for clinical CT colonography, readers were instructed to evaluate and record the single largest dimension observed. Readers were given an opportunity to comment freely on the perceived quality of each image series, and these comments were recorded for each dilution series.
Electronic Cleansing Software
To perform the required electronic subtraction, we employed a custom-designed software program that was developed in-house. The software implemented initial global threshold cleansing and was designed to operate only on regions that demonstrated attenuation greater than 200 HU. In addition to threshold methods, the software combined two additional steps that we have observed to be important for clinical evaluation. The first of these, morphologic dilation and linear filtration, was a set of mathematic operations used in image processing to address the scanning-related volume-averaging artifacts inherent to opacified colon data sets (16). In addition, the software implemented a regional nonlinear mucosal reconstruction to recreate the natural, feathered contour of the colon mucosa after the subtraction of contrast material was performed (13). Results from previous experiments have demonstrated that the contour of the bowel wall following this reconstruction filtering was statistically equivalent to the contour of native (ie, unsubtracted) colon (16). Without these additional steps, subtraction cleansing leads to distracting visual artifacts that detract from the readability of the altered data sets. In the present study, for all series in which the software was used, the operating parameters of the software were held constant.
Image Quality
Readers' comments concerning artifacts and image quality were summarized by an independent study coordinator who was not aware of the concentration of contrast material that was associated with the reader comments. After the comments were recorded, they were compared with contrast material concentrations and summarized.
Statistical Analysis
After all measurements were recorded, we compared each reader's size measurement for each polyp with the corresponding consensus panel reference value; the absolute difference of these values was the absolute error. We then calculated the mean absolute error of measurement for each reader for each series. The process was repeated for both the control series and the unenhanced subtraction series. We then compared the mean absolute errors obtained for each reader for the control series with those obtained for each reader for the dilution series. Because the absolute error is constrained to be zero or positive, we made these comparisons by using a nonparametric Kruskal-Wallis analysis of variance test. For each reader, a post hoc, pair-wise comparison between the absolute errors obtained for the dilution series and those obtained for the control series was performed by using the Dunn post hoc test.
|
RESULTS
|
|---|
The 700-mL volume of contrast material resulted in the complete or partial submerging of 11 of the 21 polyps. We observed that the same polyps were either submerged or partially submerged for each dilution series. Only the 11 lesions that were partially or completely submerged were included for measurement. The same 11 polyps were used for comparison in both the contrast material–enhanced and unenhanced series. Five of the 11 submerged lesions were 8–14 mm, four were 15–20 mm, and two were 21–25 mm. With regard to the morphologic features of submerged lesions, three of the 8–14-mm lesions, two of the 15–20-mm lesions, and one of the 21–25-mm lesions were sessile; the rest were pedunculated.
Image Quality
The contrast material dilutions resulted in homogeneous partial opacification of the phantom bowel lumen. The mean attenuation of the bowel contrast material in the phantom for each dilution is presented in Table 1. Readers reported a streak artifact with the 1:10 and 1:5 dilutions but not with the 1:20 or 1:30 dilutions (Fig 2). Images of the 1:30 dilution elicited comments from both readers of incomplete subtraction. The mean attenuation of the 1:30 dilution represented the intended threshold of action for the software, which was designed to operate only in opacified regions that had attenuation greater than 200 HU (13). No comments of incomplete subtraction were reported for the more concentrated dilution series. No comments of either incomplete subtraction or streak artifact were reported for the control series. No comments on contrast material heterogeneity were reported for any of the dilution series.
View larger version (98K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2a. Images of 10-mm polyp in colon phantom for 1:5 dilution (mean attenuation, 840 HU) before and after electronic cleansing. (a) Transverse CT image of polyp (arrow) before electronic cleansing demonstrates marked streak artifact related to attenuation of contrast material in colon lumen. (b) Pseudoenhancement of polyp (arrow) leads to notable degradation of polyp surface following electronic subtraction. Statistically significant differences in mean size measurements were observed with contrast attenuation shown here.
|
|
View larger version (82K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2b. Images of 10-mm polyp in colon phantom for 1:5 dilution (mean attenuation, 840 HU) before and after electronic cleansing. (a) Transverse CT image of polyp (arrow) before electronic cleansing demonstrates marked streak artifact related to attenuation of contrast material in colon lumen. (b) Pseudoenhancement of polyp (arrow) leads to notable degradation of polyp surface following electronic subtraction. Statistically significant differences in mean size measurements were observed with contrast attenuation shown here.
|
|
Polyp Size
When comparing the mean absolute errors for each reader, we observed statistically significant differences across experimental conditions (Kruskal Wallis, P < .002). Mean absolute errors of size measurements (±standard deviations), as well as P values for pair-wise post hoc testing, are shown in Table 2. Pair-wise comparison between the unenhanced and contrast-enhanced series for each reader revealed statistically significant differences in measurement error for reader 2 in the 1:10 series (P < .05) and for readers 1 and 2 in the 1:5 series (P < .01). These measurement errors were small, averaging 1.2 mm for reader 2 in the 1:10 series and 2.2 and 2.3 mm for readers 1 and 2, respectively, in the 1:5 series.
View this table:
[in this window]
[in a new window]
|
TABLE 2. Comparison of Absolute Error for Each Reader for Unenhanced Series and Contrast-enhanced Dilution Series
|
|
We continued by comparing the mean measurement error for the unenhanced subtraction series (electronic cleansing software alone in the absence of bowel contrast material) for each reader. This comparison again revealed no statistically significant difference in mean measurement error, as shown in Table 2.
Examples of lesions in the phantom prior to and following electronic subtraction are shown in Figure 3. Readers did not report any changes in either the morphologic features or attenuation of lesions following the action of the cleansing software for any of the series. Given the limited number of lesions that were partially or completely submerged, we did not perform a subgroup analysis of the mean error according to polyp size.
View larger version (93K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3a. CT colonographic images of colon phantom demonstrate polyp measuring 15 mm in diameter, according to reference standard. (a) Transverse and (b) corresponding endoluminal view of polyp (arrow in a and b) obtained by using control conditions (no contrast material and no subtraction). (c) Transverse image with 1:20 dilution (mean attenuation, 300 HU) shows submerged polyp (arrow) before subtraction. (d) Transverse and (e) corresponding endoluminal views of polyp (arrow in d and e) with 1:20 dilution following subtraction. Measurements for this particular lesion for control and 1:20 postsubtraction series were 14.6 mm and 14.8 mm for reader 1 and 14.9 mm and 14.7 mm for reader 2.
|
|
View larger version (70K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3b. CT colonographic images of colon phantom demonstrate polyp measuring 15 mm in diameter, according to reference standard. (a) Transverse and (b) corresponding endoluminal view of polyp (arrow in a and b) obtained by using control conditions (no contrast material and no subtraction). (c) Transverse image with 1:20 dilution (mean attenuation, 300 HU) shows submerged polyp (arrow) before subtraction. (d) Transverse and (e) corresponding endoluminal views of polyp (arrow in d and e) with 1:20 dilution following subtraction. Measurements for this particular lesion for control and 1:20 postsubtraction series were 14.6 mm and 14.8 mm for reader 1 and 14.9 mm and 14.7 mm for reader 2.
|
|
View larger version (112K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3c. CT colonographic images of colon phantom demonstrate polyp measuring 15 mm in diameter, according to reference standard. (a) Transverse and (b) corresponding endoluminal view of polyp (arrow in a and b) obtained by using control conditions (no contrast material and no subtraction). (c) Transverse image with 1:20 dilution (mean attenuation, 300 HU) shows submerged polyp (arrow) before subtraction. (d) Transverse and (e) corresponding endoluminal views of polyp (arrow in d and e) with 1:20 dilution following subtraction. Measurements for this particular lesion for control and 1:20 postsubtraction series were 14.6 mm and 14.8 mm for reader 1 and 14.9 mm and 14.7 mm for reader 2.
|
|
View larger version (92K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3d. CT colonographic images of colon phantom demonstrate polyp measuring 15 mm in diameter, according to reference standard. (a) Transverse and (b) corresponding endoluminal view of polyp (arrow in a and b) obtained by using control conditions (no contrast material and no subtraction). (c) Transverse image with 1:20 dilution (mean attenuation, 300 HU) shows submerged polyp (arrow) before subtraction. (d) Transverse and (e) corresponding endoluminal views of polyp (arrow in d and e) with 1:20 dilution following subtraction. Measurements for this particular lesion for control and 1:20 postsubtraction series were 14.6 mm and 14.8 mm for reader 1 and 14.9 mm and 14.7 mm for reader 2.
|
|
View larger version (72K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3e. CT colonographic images of colon phantom demonstrate polyp measuring 15 mm in diameter, according to reference standard. (a) Transverse and (b) corresponding endoluminal view of polyp (arrow in a and b) obtained by using control conditions (no contrast material and no subtraction). (c) Transverse image with 1:20 dilution (mean attenuation, 300 HU) shows submerged polyp (arrow) before subtraction. (d) Transverse and (e) corresponding endoluminal views of polyp (arrow in d and e) with 1:20 dilution following subtraction. Measurements for this particular lesion for control and 1:20 postsubtraction series were 14.6 mm and 14.8 mm for reader 1 and 14.9 mm and 14.7 mm for reader 2.
|
|
|
DISCUSSION
|
|---|
In our experimental data we observed that, within a broad range of contrast attenuations, the combination of subtraction software and bowel contrast material for minimal preparation CT colonography does not appear to significantly affect radiologists' estimations of polyp size. In particular, attenuation greater than 200 HU but less than approximately 560 HU resulted in measurement errors that were indistinguishable from those of the control, with a mean measurement error of approximately 0.6 mm. In this attenuation range, our readers did not observe streak artifacts from areas of dense contrast enhancement and did not report incomplete subtraction-related artifacts. Attenuation greater than 560 HU resulted in small but statistically significant changes in mean size measurements, which were accompanied by the presence of beam-hardening artifacts with increasing attenuation. Our cleansing software was designed to ignore structures with attenuation less than 200 HU, and an attenuation of 200 HU was associated with incomplete subtraction. Furthermore, it appears that the subtraction software, when operating in the absence of contrast material, did not by itself lead to systematic errors of measurement. Measurements obtained following subtraction but without contrast material were also indistinguishable from those of the control. In combination, these observations contribute to the validation of minimal preparation approaches to CT colonography. Several qualifications should be addressed, however.
First, the present study consisted of data obtained by scanning a carefully designed phantom and not by evaluating human CT colonography data sets. As such, our observations will require confirmation in a clinical series. We concluded that logistic and ethical considerations required that the first stages of this validation be performed with a colon phantom. This was because of the need to measure precisely the same lesions that could be scanned by using varied contrast material dilutions and software conditions. In theory, it would be possible to validate our observations by using a broad range of contrast material dilutions in a clinical series, but, given the low prevalence of polyps in even moderate risk cohorts, this evaluation would require an impractically large number of cases. Furthermore, consistent application of the most refined measurement techniques for conventional colonoscopy would put considerable burden on the collaborating endoscopists. For ethical reasons, repeated examination of the same patients was never seriously considered. Nonetheless, our observations and phantom data help to delineate a contrast attenuation range for which electronic subtraction can proceed completely and without the generation of substantial artifacts and related measurement errors.
In addition, the morphologic features of the lesions in our study did not include the full range features that are now known to be of interest. In particular, we did not evaluate the effect of contrast enhancement and subtraction on flat lesions. These lesions were excluded from the phantom because of initial uncertainty on the part of the manufacturer as to their ability to construct homogeneous flat lesions with exact dimensions that could be confirmed. Evaluation of flat lesions in the setting of a minimal preparation technique will be particularly important because these lesions typically demonstrate 2–3 mm of vertical impingement into the colon lumen (14). In the optimal contrast attenuation range that we studied, the mean error associated with measurements was approximately 0.6 mm. Although this was not significantly different from the error observed in the control series, it still represents a substantial fraction of the limited height of flat lesions. It is not clear how detection of flat lesions would be affected by a minimal preparation technique. To our knowledge, no studies have evaluated the performance of minimal preparation CT colonography for detection of flat polyps. Flat lesions, though rarely reported, are observable at CT colonography; prior data suggest that the prevalence of advanced dysplasia and, hence, the clinical importance of these lesions may be lower than originally reported (17).
We did not evaluate artifacts and measurement errors associated with pedunculated lesions less than 8 mm in diameter. For the optimized contrast attenuation range, the average error in measurement, 0.6 mm, was a relatively small fraction of the total size of these lesions. We find it unlikely that pseudoenhancement and subtraction artifacts would operate differently for lesions less than 8 mm in size than for lesions in the size range that we studied. Of note, there has been a trend in colonography reporting concerning the upward revision of threshold sizes for polyps that are considered noteworthy. Pickhardt et al (1) recently reported an acceptable balance of sensitivity and specificity for CT colonography in a screening cohort by using a size threshold of 8 mm. With further experience, this upward trend may continue. Nonetheless, direct observation of any measurement-related effects of minimal preparation CT colonography in the smaller size range would help to clarify the performance of the technique.
The contrast material that we instilled into the phantom was of defined attenuation and homogeneity. In the clinical setting for minimal preparation CT colonography, contrast material will be ingested orally and diluted by succus entericus during passage through the small intestine. Hence, predicting the attenuation of a contrast agent in the colon may be difficult given an oral route of administration. The attenuation range that we chose to model represents the attenuation range that we have observed in the colon in our prior experience with minimal preparation CT colonography. Therefore, the concentration range that we considered to be acceptable might not correspond directly to optimal concentrations given orally to patients.
Minimal preparation CT colonography requires the right combination of consistency, attenuation, and homogeneity of bowel contrast material to permit clean subtraction and accurate measurement of colonic lesions. In the present study, we directly evaluated only the attenuation of tagging and began to establish an operational attenuation range for which tagging does not appear to alter measured polyp sizes. In our experience, we have also found that the homogeneity and semiliquid consistency of tagged material can have a considerable effect in that the uniformly tagged pools of contrast material are associated with fewer subtraction-related artifacts. In our current evaluation, we employed a single contrast agent to simulate the clinical tagging of both colonic fluid and fecal material. Iodinated contrast agents, because of their high aqueous solubility, can simultaneously tag both fluid and feces. Barium-based agents, because of their requirement for emulsifiers and their tendency to precipitate in succus entericus, require the addition of iodinated agents for effective tagging of colonic fluid. Our experience has favored the use of iodinated contrast agents alone, and the design of this study reflects that preference.
Finally, while we evaluated the effect of subtraction software in the absence of contrast material on the accuracy of measurement, we maintained the CT scanning parameters, algorithm, and software parameters constant throughout our experiments. In particular, we did not implement subtraction on structures with attenuation less than 200 HU. This limitation was designed as a safety feature to minimize the likelihood that the software would alter the soft-tissue–attenuation structures that were present in the CT colonographic data sets. In addition, we implemented a CT scanning technique that minimized dose at the expense of contrast resolution; this choice was made to reflect the intended use of CT colonography as a screening examination.
Ultimately, we expect that the development of subtraction software will continue to advance. We also expect that a more robust evaluation of our own software parameters beyond what we have evaluated in this study will assist in this future development. It should be noted that there are multiple approaches to performing the required steps of subtraction and mucosal reconstruction, each with the potential to introduce artifacts. The validation steps we performed apply to the software system used in our current study and may not be applicable to other cleansing software; in our view, each subtraction system intended for clinical use should undergo the same kind of rigorous validation we have attempted here.
Practical application: Implementation of a minimal preparation technique for colonography will require radiologists to interpret electronically modified data sets, a practice that departs somewhat from the current interpretation of cross-sectional images. For this practice to be accepted, it will be important to validate that the measurement errors and artifacts that occur because of contrast enhancement and software cleansing are acceptably small. It will also be necessary to define the ranges of relevant parameters, such as contrast attenuation, for which these effects are minimized. In the present study, we observed stability of polyp lesions by using imaging conditions that resembled those required for minimal preparation CT colonography with electronic cleansing. Eventually, the further refinement and validation of operating parameters will require study in a clinical series.
|
ACKNOWLEDGMENTS
|
|---|
The authors thank Elkan Halpern, PhD, for his review of statistical methods.
|
FOOTNOTES
|
|---|
2 Current address: Istituto di Radiologia, Policlinico Universitario, Udine, Italy 
Authors stated no financial relationship to disclose.
Author contributions: Guarantor of integrity of entire study, M.E.Z.; study concepts, M.E.Z., P.F.H.; study design, M.E.Z., J.J.P., J.Y.K., C.D.F., C.M.; literature research, C.D.F., C.M.; clinical studies, M.E.Z., J.J.P., J.Y.K., C.M., P.F.H.; experimental studies, M.E.Z., J.J.P., J.Y.K., C.D.F., C.M.; data acquisition, M.E.Z., J.J.P., J.Y.K., C.D.F., C.M.; data analysis/interpretation, M.E.Z., J.Y.K., C.M., P.F.H.; statistical analysis, M.E.Z., C.M., P.F.H.; manuscript preparation and definition of intellectual content, M.E.Z.; manuscript editing, revision/review, and final version approval, M.E.Z., P.F.H.
|
References
|
|---|
- Pickhardt PJ, Choi JR, Hwang I, et al. Computed tomographic virtual colonoscopy to screen for colorectal neoplasia in asymptomatic adults. N Engl J Med 2003; 349:2191–2200.[Abstract/Free Full Text]
- Macari M, Lavelle M, Pedrosa I, et al. Effect of different bowel preparations on residual fluid at CT colonography. Radiology 2001; 218:274–277.[Abstract/Free Full Text]
- Byers T, Levin B, Rothenberger D, Dodd GD, Smith RA; for the American Cancer Society Detection and Treatment Advisory Group on Colorectal Cancer. American Cancer Society guidelines for screening and surveillance for early detection of colorectal polyps and cancer: update 1997. CA Cancer J Clin 1997; 47:154–160.[Abstract]
- Weitzman ER, Zapka J, Estabrook B, Goins KV. Risk and reluctance: understanding impediments to colorectal cancer screening. Prev Med 2001; 32:502–513.[CrossRef][Medline]
- Ristvedt SL, McFarland EG, Weinstock LB, Thyssen EP. Patient preferences for CT colonography, conventional colonoscopy, and bowel preparation. Am J Gastroenterol 2003; 98:578–585.[CrossRef][Medline]
- Callstrom MR, Johnson CD, Fletcher JG, et al. CT colonography without cathartic preparation: feasibility study. Radiology 2001; 219:693–698.[Abstract/Free Full Text]
- Lefere PA, Gryspeerdt SS, Dewyspelaere J, Baekelandt M, Van Holsbeeck BG. Dietary fecal tagging as a cleansing method before CT colonography: initial results—polyp detection and patient acceptance. Radiology 2002; 224:393–403.[Abstract/Free Full Text]
- Zalis M, Hahn P. Digital subtraction bowel cleansing in CT colonography. AJR Am J Roentgenol 2001; 176:646–648.[Free Full Text]
- Beaulieu CF, Jeffrey RB Jr, Karadi C, Paik DS, Napel S. Display modes for CT colonography. II. Blinded comparison of axial CT and virtual endoscopic and panoramic endoscopic volume-rendered studies. Radiology 1999; 212:203–212.
- Gore RM. Colorectal cancer: clinical and pathologic features. In: Gore RM, ed. Gastrointestinal cancer: radiologic diangosis and staging. Philadelphia, Pa: Saunders, 1997; 409.
- Winawer SJ, Fletcher RH, Miller L, et al. Colorectal cancer screening: clinical guidelines and rationale. Gastroenterology 1997; 112:594–642. [Published correction appears in Gastroenterology 1997; 112:1060 and in Gastroenterology 1998; 114:625.][CrossRef][Medline]
- Birnbaum BA, Maki DD, Chakraborty DP, Jacobs JE, Babb JS. Renal cyst pseudoenhancement: evaluation with an anthropomorphic body CT phantom. Radiology 2002; 225:83–90.[Abstract/Free Full Text]
- Zalis ME, Perumpillichira J, Del Frate C, Hahn PF. Digital subtraction bowel cleansing with mucosal reconstruction in CT colonography: initial observations. Radiology 2003; 226:911–917.[Abstract/Free Full Text]
- Dachman AH, Zalis ME. Quality and consistency in CT colonography and research reporting. Radiology 2004; 230:319–323.[Free Full Text]
- Bond JH; and the Practice Parameters Committee of the American College of Gastroenterology. Polyp guideline: diagnosis, treatment, and surveillance for patients with colorectal polyps. Am J Gastroenterol 2000; 95:3053–3063.[CrossRef][Medline]
- Zalis ME, Perumpillichira JJ, Hahn PF. Digital subtraction bowel cleansing for CT colonography using morphological and linear filtration methods. IEEE Trans Med Imaging 2004; 23:1335–1343.[CrossRef][Medline]
- Fidler JL, Johnson CD, MacCarty RL, Welch TJ, Hara AK, Harmsen WS. Detection of flat lesions in the colon with CT colonography. Abdom Imaging 2002; 27:292–300.[Medline]
This article has been cited by other articles:
|
|
|
|
 
A. H. de Vries, H. W. Venema, J. Florie, C. Y. Nio, and J. Stoker
Influence of Tagged Fecal Material on Detectability of Colorectal Polyps at CT: Phantom Study
Am. J. Roentgenol.,
October 1, 2008;
191(4):
W181 - W189.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
A. Slater, S. A. Taylor, D. Burling, L. Gartner, J. Scarth, and S. Halligan
Colonic polyps: effect of attenuation of tagged fluid and viewing window on conspicuity and measurement--in vitro experiment with porcine colonic specimen.
Radiology,
July 1, 2006;
240(1):
101 - 109.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
M. E. Zalis, J. J. Perumpillichira, C. Magee, G. Kohlberg, and P. F. Hahn
Tagging-based, Electronically Cleansed CT Colonography: Evaluation of Patient Comfort and Image Readability.
Radiology,
April 1, 2006;
239(1):
149 - 159.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
