| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Gastrointestinal Imaging |
1 From the Division of Abdominal Imaging and Interventional Radiology, Department of Radiology, Massachusetts General Hospital, White 270, 55 Fruit St, Boston, MA 02114. Received July 27, 2004; revision requested October 6; revision received March 19, 2005; accepted April 15. Supported by E-Z-Em, Amersham Health (now GE Healthcare), the General Electric–Association of University Radiologists Radiology Research Academic Fellowship (GERRAF), and NCI 1 K22 CA098422 A 01. Address correspondence to M.E.Z. (e-mail: mzalis{at}mgh.harvard.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Materials and Methods: After institutional review board approval and informed consent were obtained, 68 subjects (41 men (60%) men, 27 (40%) women; mean age, 60 years ± 6 [standard deviation]) with average or moderate risk factors for development of colorectal carcinoma were recruited and placed into three study groups. Group 1 (n = 25) ingested 150-mL aliquots of 2% barium sulfate suspension with meals and snacks for 48 hours prior to imaging, without other diet modification or a cathartic. Group 2 (n = 21) ingested 10-mL aliquots of nonionic iodinated contrast material (iopromide) with a concentration of 300 mg per milliliter with meals and snacks for 2 days before imaging, without diet modification or a cathartic. Group 3 (n = 22) ingested nonionic iodinated contrast material (iohexol) with a concentration of 300 mg per milliliter with meals and snacks for 2 days before imaging and ingested 34 g of magnesium citrate the evening prior to imaging. CT colonography was also performed on 10 control subjects who ingested polyethylene glycol electrolyte solution prior to imaging. Subjective and numerical measures of bowel preparation quality, homogeneity, and patient comfort among the noncathartic and cathartic cohorts were compared with nonparametric analysis of variance, the Fisher exact test, and the F test, as appropriate. The study was HIPAA compliant.
Results: Study subjects who received tagging preparations reported significantly improved discomfort scores when compared with those of the control subjects (P < .05, each comparison). There was no significant difference in discomfort scores among groups 1, 2, and 3. For each reader, scores of subtracted image readability were highest for group 3. Dichotomized rates of preparation "success" were also greatest for group 3.
Conclusion: In this series, the patient discomfort scores were significantly improved with tagging preparations for CT colonography. Nonionic iodinated contrast material in conjunction with a hyperosmotic laxative (magnesium citrate) was associated with the best subjective and numerical indices of readability.
© RSNA, 2006
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
To address this compliance barrier, several investigators have explored the possibility of replacing the standard preexamination cathartic preparation with a regimen of orally ingested, positive contrast material to mark the fecal contents of the bowel and to potentially distinguish fecal contents from soft-tissue lesions of the bowel, such as polyps (7,8). In addition, investigators have combined fecal tagging with an image-processing subtraction step wherein the tagged fecal contents are electronically removed from the source CT images, which leaves unmarked soft-tissue features such as polyps untouched (9). The impetus for using electronic, or digital subtraction, bowel cleansing is to preserve the use of three-dimensional visualization. Tagged fecal contents are distinguishable from polyps on cross-sectional multiplanar views of the colon because of the higher cross-sectional attenuation imparted by the tagging material. However, tagged fecal material can submerge and obscure lesions on three-dimensional endoluminal reconstructions. Endoluminal reconstructions have been shown to improve sensitivity for detection of polyps at CT colonography (10).
Electronic subtraction algorithms are based in part on the detection of the increased attenuation of tagged, ingested material (11). If the tagging of ingested colonic contents is not uniform, or if sufficient air is admixed with the colonic contents and leads to volume averaging, the attenuation of affected regions can fall below the threshold required for computer recognition. As a result, the computer will not remove all of the ingested material, leaving behind visually distracting arcs and whorls of opaque, soft-tissue-attenuation material within the bowel lumen. This remnant material can potentially obscure clinically important colon features, such as polyps (11).
Two types of contrast material compose the bulk of orally ingested agents suitable for bowel tagging for CT colonography. Barium-based formulations similar to those routinely used in clinical CT have an excellent safety profile but must be kept in aqueous suspension by emulsifiers to avoid the precipitation of barium sulfate in succus entericus. Iodine-based agents readily dissolve in aqueous solution and have been employed for evaluation of the integrity of the bowel when perforation is suspected. Ionic iodinated agents are hypertonic, and their use can result in adverse reactions due to fluid shifts into the bowel lumen (12). In our own group, limited experience with a group of 10 subjects suggested an increased likelihood of cramping associated with use of ionic iodinated contrast material (M.E.Z., unpublished data, May 2002). Nonionic iodinated agents are less hypertonic than their ionic counterparts and have an improved safety profile when compared with the safety profile of ionic agents (13). Nonionic agents have also been used for evaluation of the gastrointestinal tract when perforation is suspected. To our knowledge, the performance of nonionic agents has never been compared with the performance of barium-based agents for fecal tagging in CT colonography.
We hypothesized that more homogeneous tagging would be associated with improved reader acceptance and that tagging-based preparations for CT colonography would result in improved patient comfort when compared with patient comfort with the use of standard cathartic preparation regimens. Thus, the purpose of this study was to prospectively compare the homogeneity, adequacy, and patient acceptance of nonionic iodine-based regimens with those of a barium-based regimen for noncathartic CT colonography with employment of electronic subtraction cleansing.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Subject Recruitment
We obtained informed written consent for participation of all subjects recruited into this study, and these consent procedures were reviewed and approved by our institutional review board. Our study was Health Insurance Portability and Accountability Act compliant. We chose individuals according to a random set of numbers from a list of those scheduled to undergo screening video-assisted colonoscopy at our institution from July 2001 until July 2003. Men and women with average or moderate risk factors for development of colorectal carcinoma, including age of 50–70 years and positive findings from fecal occult blood testing, composed our digital subtraction preparation study cohort. We excluded individuals with a history of prior colorectal carcinoma or polyps, inflammatory bowel disease, or prior colon surgery. We recruited 68 subjects into this digital subtraction preparation study cohort.
In addition to the 68 study subjects who were recruited to undergo a preparation regimen before CT colonography, we prospectively recruited 10 control subjects for subjective evaluations of bowel preparation and subject acceptance. These 10 individuals were selected according to a random set of numbers from the same list of individuals scheduled to undergo screening video-assisted colonoscopy at our institution from July 2001 until July 2003 and had the same age and risk factors associated with our study subjects. The inclusion, exclusion, and demographic criteria of these individuals were identical to those of patients recruited to undergo a preparation regimen before CT colonography. We obtained institutional review board approval and specific consent to use the images and surveys of these 10 control subjects for this study. The 10 control subjects received a bowel preparation that was standard for colonoscopy in our institution when this study was performed, namely, ingestion of a fiber-restricted diet 2 days prior to examination, ingestion of a liquid diet the day prior to examination, and ingestion of 4 L of polyethylene glycol electrolyte solution (GoLYTELY; Braintree Laboratories, Braintree, Mass) as bowel cathartic the night prior to CT colonography. The control subjects underwent CT colonography immediately prior to their scheduled colonoscopy. We chose 10 as the size of the control group, as prior experience had revealed relatively little variation in subject experience or preparation quality with the use of polyethylene glycol electrolyte preparation (1).
Our study subject cohort of 68 individuals, with mean age of 60 years ± 6 (standard deviation), was composed of 41 (60%) men and 27 (40%) women. Of the 68 individuals, 23 (34%) reported no prior symptoms or relevant history, which qualified these 23 study subjects as asymptomatic screening patients. The remaining 45 study subjects reported history or symptoms, which included a family history of colorectal carcinoma and a prior positive finding from a fecal occult blood test, which qualified them as having moderate risk for development of colorectal carcinoma (14). We placed 25 study subjects into group 1 (which received barium sulfate suspension), 21 study subjects into group 2 (which received iopromide [Ultravist 300; Berlex Laboratories, Montville, NJ]), and 22 study subjects into group 3 (which received iohexol [Omnipaque 300; Amersham Health]). Of the 10 control subjects, four qualified as asymptomatic screening patients, and six reported moderate risk factors for development of colorectal carcinoma.
Tagging Regimens and Subject Groups
From the cohort of referred colonoscopy patients, we sequentially divided study subjects into three groups. Subjects in group 1 ingested seven 150-mL aliquots of 2% barium sulfate suspension (E-Z-Cat; E-Z-Em) with each meal and snack for 48 hours prior to imaging. The morning of CT colonography at 3 hours prior to imaging, subjects ingested a final 700-mL bolus of the same 2% barium sulfate suspension they had ingested during the prior 48 hours. Prior findings from pilot data (M.E.Z., unpublished data, May 2001) have suggested that without this final preexamination bolus, tagging in the cecum was nonuniform because of the entry of untagged succus entericus during the preceding night. We instructed subjects in group 1 to eat their regular diet, without modification.
Subjects in group 2 ingested iopromide, a nonionic iodinated contrast material with a concentration of 300 mg organically bound iodine per milliliter. Subjects in group 2 ingested the contrast material in seven 10-mL aliquots with meals and snacks for 48 hours prior to CT colonography, also without diet modification. Subjects in group 2 ingested a bolus of 30 mL of contrast material diluted in 700 mL of water at 3 hours prior to imaging.
Subjects in group 3 ingested iohexol, a different nonionic iodinated contrast agent, with exactly the same volume and regimen as were used in group 2. Like the members of group 2, subjects in group 3 ingested one 30-mL aliquot of contrast material diluted in 700 mL of water at 3 hours prior to imaging. However, unlike the other subjects, group 3 study subjects also ingested 34 g of magnesium citrate (LoSo Prep; E-Z-Em) the evening prior to CT colonography. Magnesium citrate is a nonabsorbed salt that increases fluid content in the colon and, in sufficient quantity, acts as a hyperosmotic laxative. It is often administered in conjunction with stimulatory cathartics such as bisacodyl sodium; however, we administered magnesium citrate alone, without bisacodyl, to reduce adverse effects such as diarrhea and cramping associated with the use of bowel cathartics. Thus, the total volume of contrast material ingested by each subject in each group was as follows: 1600 mL of 2% barium sulfate suspension for group 1, 90 mL of iopromide (300 mg of organically bound iodine per milliliter) for group 2, and 90 mL of iohexol (300 mg of organically bound iodine per milliliter) for group 3.
The number of subjects in group 1 was 25, with a mean age of 61 years ± 4 (range, 55–69), in which 14 (56%) were men, and 11 (44%) were women. Group 2 consisted of 21 subjects, with a mean age of 58 years ± 5 (range, 50–68), in which 11 (52%) were men and 10 (48%) were women. Group 3 consisted of 22 subjects, with a mean age of 61 years ± 3 (range, 53–70), in which 13 (59%) were men and nine (41%) were women. The 10 concurrent control subjects had a mean age of 59 years ± 7 (range, 51–68); and five (50%) were men, and five (50%) were women. The differences in mean ages and male-female ratios were not significant across the study groups and the control group (P > .05, all comparisons; analysis of variance [ANOVA] and 
2, respectively).
Subject Responses
All subjects were asked to complete a short questionnaire to ascertain their discomfort during CT colonography. After bowel preparation and prior to CT colonography, study subjects (who were to receive tagging bowel preparations for CT colonography) were administered a questionnaire on which they were asked, "What sort of discomfort did you experience during the preparation for the CT study (virtual colonoscopy with digital subtraction cleansing)?" Study subjects rated their discomfort on a five-point scale, as follows: 1, no discomfort; 2, mild discomfort; 3, moderate discomfort; 4, severe discomfort; 5, extreme pain. Control subjects (who were to receive cathartic bowel preparation) were given the same questionnaire after bowel preparation and prior to CT colonography. Control subjects rated their discomfort on the same scale as study subjects. The survey given to all subjects included space for comments. A study coordinator, who was unaware of the type of contrast material each subject had ingested, summarized the subjects' comments, and these summaries were then collected according to contrast material type.
CT Colonography
We employed the same CT colonographic technique for all study subjects and control subjects, regardless of preparation type or group. For each subject, we acquired CT colonographic images with prone and supine patient positions and a four–detector row helical CT scanner (GE LightSpeed; GE Medical Systems, Milwaukee, Wis), after manual air insufflation of the colon to patient tolerance. Radiology staff performed the air insufflation and did not administer spasmolytic. Radiology staff evaluated the degree of colonic distention on the anterior-posterior scout images and administered additional air, if deemed necessary, to patient tolerance. The staff divided the colon into six segments (rectum; sigmoid, descending, transverse, and ascending colon; cecum) for the purposes of evaluating insufflation and administered additional air if at least one whole segment appeared collapsed on the scout image. The staff performed helical acquisitions by using the following technique: 140 kVp, 50 mA, 0.8-second scanning, 3.75-mm collimation, and 1.8-mm section reconstruction interval. At the time of this study, we employed 140 kVp on all screening CT colonographic examinations performed at our institution as a means to reduce absorbed dose and improve uniformity of x-ray beam at low dose. Identical parameters were used for the prone and the supine series. Automatic tube modulation was not available at the time this study was performed.
Electronic Cleansing (Digital Subtraction Bowel Cleansing)
After image acquisition, we employed custom software to electronically cleanse the tagging material from the colon on the CT colonographic images. The software identified and modified the attenuation of opacified regions of the bowel with pixel values of more than 200 HU to render them translucent. The system was also designed to leave soft-tissue elements of the bowel less than 200 HU in attenuation unmodified. The software contained routines to address volume-averaging artifacts present at the borders of opacified regions, as well as routines to perform separate mucosal reconstruction following basic subtraction (11).
The image-processing software operated with a programming system (MATLAB; The Mathworks, Natick, Mass) on a dual 2-GHz processor personal computer (Dell Precision; Dell, Round Rock, Tex). We kept the algorithms and code for the image-processing software constant throughout the conduct of the study.
Previous investigations in which we used the software we employed for this study demonstrated no discernible subtraction artifacts in the colon of data sets where no bowel contrast material was employed (15). Nonetheless, to isolate preparation type as a variable, we applied the software to images of control subjects, as well as images of study subjects, to minimize the likelihood that the software alone could allow readers to identify the images of control subjects.
Image Interpretation and Reader Assessments
We evaluated all CT colonographic images on a dedicated workstation (Vitrea 2, version 3.2; Vital Images, Plymouth, Minn) that offered coordinated transverse, multiplanar, and volume-rendered endoluminal reformations. The processed, electronically cleansed images, which remained compliant with the Digital Imaging and Communications in Medicine (DICOM) standard, were resubmitted to the workstation. Two readers (M.E.Z., P.F.H.), each working independently and each having read approximately 400 CT colonographic images prior to the initiation of this protocol, made their evaluations by using the workstation package. Each reader was presented a random shuffle of control subject and study subject images and recorded his subjective assessments of colon preparation quality, as described below. Readers were encouraged to make full use of the viewing facilities of the workstation, which included the use of multiplanar views, endoluminal subvolume views, and endoluminal fly-through views. We did not specifically record the frequency with which each reader used each type of viewing modality.
Readers were blinded to identifying information on the images, including the type of bowel preparation. Images of control subjects, having been subjected to the same electronic cleansing software as images of study subjects, were not identified as such to the readers. Our study cohorts were only mildly enriched with respect to risk factors for polyp and colorectal carcinoma, with a prevalence of polyps larger than 8 mm estimated to be less than 10% (16). Given the limited number of polyp lesions expected in our study cohorts, meaningful comparison of performance for polyp detection was not possible among the cohorts we studied. As a result, we limited our evaluations to only subjective assessments of preparation quality from readers and did not record performance data concerning polyp detection.
Readers evaluated the electronically cleansed image data sets and recorded their assessment of the readability of the colon on the modified images. These subjective assessments were made with the following five-point scale: 1, uninterpretable images due to unsubtracted feces and artifact; 2, poor preparation, images contain large amount of unopacified feces and subtraction artifact, but images interpretable with difficulty; 3, moderate preparation, images contain moderate amounts of unopacified feces and subtraction artifact, interpretable images; 4, good preparation, images contain small amounts of unopacified material and subtraction artifact, interpretable images; 5, excellent preparation, little to no discernible unopacified material and artifact, easily interpretable images.
To assess the likelihood of each prepartion regimen to yield interpretable images, we dichotomized reader assessments of the electronically cleansed data sets as "success" or "failure." From the five-point scale described previously, we assigned scores 4 and 5 (where 5 was best response) as success and the rest as failure. We then tabulated the success and failure rates for the different preparations, including those of the control subjects. For each reader's assessments, we compared the success rates for the different study groups with the success rates for the control group by using a contingency table.
Finally, we made an assessment of the consistency of the tagged colonic contents visible on the study data sets. We did this on the basis of our initial observations that semiliquid contents subtract more cleanly than desiccated material (M.E.Z., unpublished data, May 2001). Desiccated material, even if uniformly tagged, forms thin, irregular layers around the inner circumference of the colon mucosa. Because the contour of the colon mucosa is complex, this tagged material takes on complex forms that are difficult to subtract cleanly because of the associated volume averaging at CT. We asked one of our readers (M.E.Z.) to make a binary assessment of the consistency of the tagged colonic contents visible on the images, in which the two responses were semiliquid or not semiliquid, on the basis of the following criteria. The reader judged the contents as semiliquid if the tagged contents (a) formed a horizontal or nearly horizontal meniscus in the dependent portion of the colon and (b) moved completely to the contralateral aspect of the colon on movement from prone to supine acquisition. Colonic contents that failed to simultaneously meet these criteria were deemed not semiliquid.
Numerical Analyses of Preparation Homogeneity
In addition to reader-subjective assessments of preparation quality, we made numerical evaluation of the homogeneity of the bowel preparations by taking the standard deviation of pixel values in regions of interest (ROIs) as a measure of tagging homogeneity. We employed specialized, custom software to assist us in this aspect of the study. To minimize confounding due to human selection of images, the software randomly selected images from the tagged colonographic data sets and then calculated the standard deviation of attenuation in opacified regions of colon. The opacified regions of colon were indicated to the computer by one of our readers (J.J.P.), who was not informed of the subject name or type of contrast material employed for each subject's colonographic examination. For each randomly selected image, the reader used the computer mouse to place a graphical ROI box over the single largest opacified region of colon visible on the image. The ROI box was rectangular and deformable by use of a mouse, so that it could be fit into the region of contrast selected by the reader.
If no opacified colon was visible on an image or if the only opacified segments were too small to accommodate an ROI box, the reader would indicate this to the computer, and the computer would then select another image at random from that examination data set. This process of image selection and ROI placement continued until the reader had selected and successfully marked opacified regions on seven images from each colonographic data set. The standard deviation from the ROIs of the seven images was then averaged, and this mean standard deviation was used as the metric of observation for each case. We assumed that the ROIs from one examination data set would be correlated with other ROIs in the same data set, and no statistical comparison was made among the ROIs of one data set. If multiple colon loops were present on an image, the reader was instructed to place the ROI box over the largest opacified segment of colon on the image. The reader was presented images from the supine series from each examination data set.
The computer calculated the mean standard deviation of pixels from the ROIs obtained from each data set, and we took the latter as a measure of attenuation homogeneity for each data set. We did assume that each data set, and hence the mean ROI standard deviation for each data set, was statistically independent. Having obtained these data without knowledge of contrast material type, we then compared the mean standard deviation of ROI attenuation observed in the data sets for group 1, group 2, and group 3.
To assess reproducibility of observations by using this computer-assisted markup technique, a second reader (G.K.), also blinded to patient-identifying information, performed the same procedure on 10 randomly selected cases from the entire study cohort.
Statistical Analysis
For all statistical comparisons, differences associated with a two-tailed P value of less than .05 were considered significant. We compared demographic data by using ANOVA and 2 test of contingency tables. We compared scalar data such as subject discomfort scores and reader preparation scores with a nonparametric ANOVA. The readers' dichotomized evaluations of preparation quality (as success or failure) were compared in a contingency table by using the Fisher exact test. For evaluation of consistency of tagged colonic contents, we tabulated the rate of semiliquid consistency across the different study groups. We compared these data in a contingency table with the Fisher exact method. For the numerical analysis of preparation quality, we compared the mean standard deviation of ROI pixels for CT colonographic data sets opacified by using barium sulfate suspension with that of data sets opacified with nonionic iodine-based contrast material. We made this comparison with the F test. To assess the reproducibility of the numerical pixel ROIs, we compared the differences in mean standard deviation observed between the matched cases evaluated by the two computer-aided readers. We compared these data by using the t test. Power calculation based on a limited cohort of 15 electronically subtracted cases revealed a 90% likelihood to detect differences in average standard deviation as small as 50 HU. We compared reader scores with a nonparametric ANOVA, and differences in rank associated with P < .05 were considered significant.
All statistical calculations were performed by using software packages (Instat 3, StatMate, Macintosh version, GraphPad Software, San Diego, Calif; and Excel, Microsoft, Redmond, Wash).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
|
Reader Assessments
Readers observed at least partial opacification of ingested colon contents in data sets all of the subjects in groups 1, 2, and 3. However, there was considerable variation in quality and readability, as analyzed on postsubtraction data sets. A summary of reader assessments for electronically cleansed data sets is presented in Table 2. In particular, each of our two readers independently scored the readability of the images from groups 1 and 2 as inferior to the readability of images in group 3 or the control group (for each reader, P < .05 for groups 1 and 2 vs either group 3 or control group). There was no significant difference in scores between groups 1 and 2 (which received tagging alone); the median score for each group, for each reader, was 2. In addition, there was no statistically significant difference in readability for group 3 (which received magnesium citrate to supplement tagging) versus the control group (median score 5 for each group, for each reader, P > .05, all comparisons).
|
|
|
|
|
|
|
|
|
|
|
|
|
Consistency of Colonic Contents
Finally, we observed differences in the consistency of tagged colonic contents among the study groups. The rates of semiliquid versus not semiliquid consistency are summarized in Table 5. Among the groups, group 3 demonstrated a higher percentage of semiliquid consistency when compared with groups 2 and 3 (P < .05, all comparisons). There was no significant difference in consistency rates when we compared group 1 with group 2 (P > .05).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Second, our observations suggest differences in the reader acceptance of the different preparation regimens. In the cohorts we studied, each of our independent readers noted fewer electronic cleansing artifacts on images of groups 2 and 3, whose tagging regimens employed nonionic iodinated contrast material. Groups 2 and 3 demonstrated higher preparation success rates. Iodinated contrast is water miscible, a factor that may permit more uniform tagging of ingested foodstuff and may have contributed to the improved reader scores. The numerical assessments we made of preparation homogeneity correlate with the following: The standard deviations of pixels in opacified ROIs were lower for the regimens that employed iodinated contrast material. At the time of this writing, while nonionic agents are used widely for intravascular opacification, their use as bowel contrast agents for CT is less common.
A practical difference among the preparation regimens was the volume of agent employed. In this series, subjects receiving iodinated contrast material ingested 10-mL aliquots, a volume substantially less than that required for studies with the 2% barium sulfate suspension available at the time we performed these observations. Decreased volume of required tagging agent in theory could affect patient compliance, in that a smaller volume of tagging agent is less likely to adversely affect patients' eating habits. In our measure of subject discomfort, we did not observe significant differences between the preparation regimens—a result that suggests the effect of volume of contrast agent on patients' overall experience may be small. However, our relatively crude survey instrument did not address volume as an independent influence on comfort, and this may merit further investigation if tagging for CT colonography becomes more widely used. Since the initiation of our protocol, higher concentrations of barium sulfate suspension have become commercially available for CT colonography. A potentially important topic for further investigation is whether higher-concentration barium-based preparations are associated with either improved subject experience or improved numerical and subjective assessments of preparation adequacy.
Consistency of tagged contents correlated well with reader preferences for tagging regimen. Group 3, which received contrast material plus magnesium citrate to increase the water content in the colon, yielded the highest subjective scores from each of our readers. The consistency of tagged contents in group 3 was also more likely to be judged semiliquid. These observations likely reflect the interplay between the tagging preparation and the subsequent electronic cleansing mechanism. Electronic cleansing, or digital subtraction, systems must recognize tagged contents as distinct from native soft-tissue elements of the bowel, such as polyps and folds (9). Part of this discrimination is necessarily based on attenuation values, which are subject to volume-averaging artifacts at CT. When the tagged material lies in contiguous pools in the dependent portions of the colon, the simplified morphologic consistency of the tagged material facilitates clean subtraction. In our observations, readers gave the highest subjective scores to the preparation that demonstrated the highest likelihood for semiliquid consistency. While in the present study, that preparation combined the use of iohexol and magnesium citrate, other combinations of tagging and osmotic agents may achieve the same result.
In addition, numerical methods for assessing tagging homogeneity appear feasible and may have a role in the quality control of tagging regimens. We calculated the standard deviation of pixel values in ROIs of the opacified colon as a reflection of tagging homogeneity. We used only the supine series to remain consistent in our procedures and assumed that the homogeneity of tagged fecal material would not substantially change as the patient moved from prone to supine positions. As a metric, the standard deviation appeared acceptably reproducible, as observed in a limited cohort of matched test cases. In particular, we did not observe a statistically significant difference in the standard deviation of the matched test cases. Our power calculation suggested that we would be able to detect differences in mean standard deviation less than the differences in mean standard deviation we observed among the different study groups. Group 1 in particular demonstrated the greatest homogeneity and simultaneously the lowest dichotomized success rate assigned by each reader. Group 1 was also the only preparation group in which the mean standard deviation of opacified ROIs exceeded 100 HU.
Numerical measures of homogeneity alone did not capture reader preference. While the numerical measures for group 2 were statistically equivalent to those of group 3, each reader demonstrated a significant preference for the latter. Likely, this result reflects the fact that while homogeneous in attenuation, the relatively desiccated material observed in the images for group 2 formed thin, morphologically complex layers around the colon mucosa and hence left greater artifact following subtraction. When numerical assessments of homogeneity were combined with a gross assessment of consistency of colonic contents, these values tracked closely with reader preferences. Overall, the highest subjective reader ratings and the highest dichotomized success rates were assigned to group 3, which received the preparation that simultaneously demonstrated high homogeneity and semiliquid consistency.
Our observations suggest that a tagging-based preparation for CT colonography is associated with improved reader acceptance, in addition to improved patient comfort, when the preparation regimen consistently yields homogeneous tagging and semiliquid consistency of colonic contents. In our limited cohort, reader scores for group 3 were the highest among the different tagging regimens. As a group, the scores for group 3 were not statistically different from those assigned to the control group. However, two important caveats should be noted. First, the success rate for the control group remained higher than for group 3, although in our relatively small cohorts, this difference did not approach statistical significance. The difference in success rates nonetheless suggests that before tagging is implemented more widely, steps are required to maximize the reliability of the regimen among likely screening subjects. A dichotomized success rate of only 68% for group 3 by one of our readers suggests that there is room for further improvement in this regard. Perhaps more important, we did not make a performance evaluation in this study for detection of polyps, because of the small cohorts we studied and the relatively low prevalence of polyps we expected to observe. It will be essential to ascertain how the artifacts still present in successfully tagged, electronically cleansed cases will affect performance for detection of polyps, as this remains the primary performance measure for CT colonography. While a minimal preparation form of CT colonography may reduce an important barrier for colon cancer screening, the engineering and validation of that technique are complex tasks that have not yet been completed.
Another limitation of our study design was the use of iopromide and iohexol in different regimens, rather than the conceptually simpler comparison of one agent with and one agent without magnesium citrate. This limitation arose from a supply constraint at our institution, which required us to switch to iohexol. We note that the chemical properties for the forms of each agent we used are similar. In particular, the concentration of organically bound iodine for both the iohexol and the iopromide was 300 milligrams of iodine per milliliter. The osmolarity for the iohexol we used was 465 mOsm/L, and the osmolarity for the iopromide was 428 mOsm/L. In addition, since this study was completed, a more concentrated preparation of barium sulfate suspension has become commercially available. It is unclear what affect more concentrated barium sulfate suspension may have on tagging for electronic subtraction, and this merits further study.
Our use of polyethylene glycol electrolyte solution as cathartic agent for our control subjects may depart from practice at most institutions that perform CT colonography at the time of this writing. At least one study has demonstrated an increase in retained colonic fluid when comparing the use of polyethylene glycol electrolyte solution with the use of other preparations, such as phosphosoda and bisacodyl sodium (17). It is possible that our readers adversely scored the control subjects' images because of the retained fluid, and had we utilized a different cathartic, the control subjects' images might have received a higher score. One consequence of this might have been to increase the differences in preparation score and success rate observed between the study groups and the control group. In our institution, at the time we performed this study, we routinely used polyethylene glycol electrolyte solution for bowel preparation in CT colonography, and non-polyethylene glycol electrolyte solution preparations were introduced subsequently. The cathartic experience with optical colonoscopy is an important barrier to recommended colon cancer screening guidelines. An important feature of CT colonography is its potential to obviate or reduce the need for this full cathartic. In the present study, we have compared the study subject experience of using tagging-based preparations with the control subject experience of using the standard cathartic preparation that is used for optical colonoscopy.
The development of a tagging-based, easily tolerated form of colon screening examination may have positive consequences for compliance with recommended colon cancer screening recommendations. CT colonography may offer a unique advantage in that orally ingested tagging in combination with advanced image-processing techniques may alleviate the need for the standard cathartic currently required for both CT colonography and conventional colonoscopy. The observations we present here suggest that a minimal preparation examination is feasible and is associated with an improved patient experience. Nonetheless, the implementation of this technique will require further technical developments to improve the reliability of the current forms of bowel preparation and the accuracy of the electronic subtraction. Finally, clinical use of minimal preparation CT colonography will require rigorous validation of performance for detection of polyps, a step requiring prospective study in a larger cohort.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Abbreviations: ANOVA = analysis of variance • ROI = region of interest
See Materials and Methods for pertinent disclosures.
Author contributions: Guarantor of integrity of entire study, M.E.Z.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, M.E.Z., C.M.; clinical studies, M.E.Z.; experimental studies, G.K.; statistical analysis, M.E.Z., C.M.; and manuscript editing, M.E.Z.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  D. Regge, C. Laudi, G. Galatola, P. Della Monica, L. Bonelli, G. Angelelli, R. Asnaghi, B. Barbaro, C. Bartolozzi, D. Bielen, et al. Diagnostic Accuracy of Computed Tomographic Colonography for the Detection of Advanced Neoplasia in Individuals at Increased Risk of Colorectal Cancer JAMA, June 17, 2009; 301(23): 2453 - 2461. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 I. W. O. Serlie, A. H. de Vries, L. J. van Vliet, C. Y. Nio, R. Truyen, J. Stoker, and F. M. Vos Lesion Conspicuity and Efficiency of CT Colonography with Electronic Cleansing Based on a Three-Material Transition Model Am. J. Roentgenol., November 1, 2008; 191(5): 1493 - 1502. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
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]
|
|
|
|
|
|
S. Jensch, A. H. de Vries, D. Pot, J. Peringa, S. Bipat, J. Florie, R. E. van Gelder, and J. Stoker Image Quality and Patient Acceptance of Four Regimens with Different Amounts of Mild Laxatives for CT Colonography Am. J. Roentgenol., July 1, 2008; 191(1): 158 - 167. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E J HALL and D J BRENNER Cancer risks from diagnostic radiology Br. J. Radiol., May 1, 2008; 81(965): 362 - 378. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Jensch, A. H. de Vries, J. Peringa, S. Bipat, E. Dekker, L. C. Baak, J. F. Bartelsman, A. Heutinck, A. D. Montauban van Swijndregt, and J. Stoker CT Colonography with Limited Bowel Preparation: Performance Characteristics in an Increased-Risk Population Radiology, April 1, 2008; 247(1): 122 - 132. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. D. Johnson, A. Manduca, J. G. Fletcher, R. L. MacCarty, M. J. Carston, W. S. Harmsen, and J. N. Mandrekar Noncathartic CT Colonography with Stool Tagging: Performance With and Without Electronic Stool Subtraction Am. J. Roentgenol., February 1, 2008; 190(2): 361 - 366. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. S. Lee, S. H. Park, E. K. Choi, S. Y. Kim, M.-J. Kim, K. H. Lee, and Y. H. Kim Colorectal Polyps on Portal Phase Contrast-Enhanced CT Colonography: Lesion Attenuation and Distinction from Tagged Feces Am. J. Roentgenol., July 1, 2007; 189(1): 35 - 40. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. J. Beebe, C. D. Johnson, S. M. Stoner, K. J. Anderson, and P. J. Limburg Assessing Attitudes Toward Laxative Preparation in Colorectal Cancer Screening and Effects on Future Testing: Potential Receptivity to Computed Tomographic Colonography Mayo Clin. Proc., June 1, 2007; 82(6): 666 - 671. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
