June 2006, VOLUME 186
NUMBER 6

Recommend & Share

June 2006, Volume 186, Number 6

Gastrointestinal Imaging

Original Research

Polyp Measurement Using CT Colonography: Agreement with Colonoscopy and Effect of Viewing Conditions on Interobserver and Intraobserver Agreement

+ Affiliations:
1Intestinal Imaging Centre, St. Mark's Hospital, Harrow, Middlesex, England HA1 3UJ.

3Medical Statistics Group, Centre for Statistics in Medicine, Institute of Health Sciences, Oxford, England.

4Consultant, Ruislip, Middlesex, England.

Citation: American Journal of Roentgenology. 2006;186: 1597-1604. 10.2214/AJR.05.0171

ABSTRACT
Next section

OBJECTIVE. This article presents inter- and intraobserver agreement for estimates of polyp diameter using CT colonography, including the effects of different visualization displays and prior experience.

MATERIALS AND METHODS. Four observers, three of whom had prior experience with CT colonography, estimated the maximum diameter of 48 polyps using three different visualization displays: 2D colonography window, 2D abdominal window, and 3D surface rendering. Each re-measured a subset of 10 polyps. Polyps measured 2 to 12 mm according to a colonoscopic reference. Inter- and intraobserver agreement and agreement with the reference measurement were determined using the Bland-Altman method, paired Student's t testing, analysis of variance, and analysis of covariance (ANCOVA), and by calculating the components of variance.

RESULTS. CT measurements overestimated polyp diameter, a phenomenon found least using the 2D abdominal display. Generally, 95% limits of agreement encompassed different size categories for individual polyps: the widest spanned 14.6 mm (-4.6 mm to 10.0 mm) for an experienced observer using the 3D display. When using the 2D abdominal display, no significant difference was found between estimates and the reference value for the other two experienced observers (p = 0.83 and 0.23). All the observers' measurements were significantly different from the reference when using the 3D display (p < 0.001). The novice was significantly different from the experienced observers in some analyses. Inter- and intraobserver agreement were poorest for the 3D display.

CONCLUSION. Measurement of polyp diameter from CT colonography is subject to variation contingent on the observer's experience and the viewing display used. Although 3D visualization display is commonly used for polyp detection, it should not be used for measurement.

Keywords: colon, colonoscopy, colorectal cancer, CT colonography

Introduction
Previous sectionNext section

Colorectal cancer is believed to arise from adenomatous polyps. Although most adenomas never become cancerous, malignant transformation occurs in a small percentage. Malignant transformation is multifactorial [1, 2] and is positively correlated with the degree of dysplasia, villous histology, and diameter of the polyp [3, 4]. Only 1% of adenomas with a diameter less than 10 mm are malignant, whereas 50% that have a diameter of 20 mm or more are malignant [3]. In addition, patients with an adenoma measuring 10 mm or more have a higher incidence of both synchronous and metachronous advanced adenomas [5-7]. As a result, individuals with larger polyps are at higher risk for developing colorectal cancer than the general population, and surveillance guidelines advise more frequent examination in this group [8, 9].

The biologic significance of an adenoma is thus heavily influenced by its maximum diameter. Diameter can be determined during colonoscopy, most often by comparing the polyp to adjacent biopsy forceps. This procedure can be followed immediately by polypectomy. CT colonography is increasingly advocated as an alternative screening method because it is less invasive and safer than colonoscopy [10]. However, immediate polypectomy is not possible with this process, so when a polyp is detected, recommendations for future management must be made based on measurements obtained from the CT data set. Large polyps require subsequent endoscopy for polypectomy; small polyps may be safely left in situ because the risk of malignant transformation is outweighed by the small but significant risk of adverse events related to colonoscopy and the cost and inconvenience of a second procedure (Zalis M, presented at the Fifth International Symposium of Virtual Colonoscopy). It is therefore important for measurements of polyp diameter obtained at CT colonography to be both accurate and reproducible. However, although several published studies have focused on the variability of polyp measurements obtained during colonoscopy [11-13], at the time of writing, no study has determined this for CT colonography. A degree of inter- and intraobserver disagreement is inevitable, and this is potentially confounded by the viewing conditions (e.g., window width and level), the method of image rendering used, and the experience of the observer.

figure
View larger version (71K)
Fig. 1A —65-year-old man with sigmoid polyp who underwent CT colonography on 4-MDCT scanner with collimation of 1.25 mm, pitch of 1.5, and 100 mA. NAVG = navigator. Two-dimensional colonography window view, show 6-mm sigmoid polyp.

figure
View larger version (55K)
Fig. 1B —65-year-old man with sigmoid polyp who underwent CT colonography on 4-MDCT scanner with collimation of 1.25 mm, pitch of 1.5, and 100 mA. NAVG = navigator. 2D abdominal window view, show 6-mm sigmoid polyp.

figure
View larger version (129K)
Fig. 1C —65-year-old man with sigmoid polyp who underwent CT colonography on 4-MDCT scanner with collimation of 1.25 mm, pitch of 1.5, and 100 mA. NAVG = navigator. 3D surface-rendered view show 6-mm sigmoid polyp.

We aimed to determine the level of agreement between CT colonography measurements of maximum polyp diameter and their colonoscopic equivalent and, in addition, to determine the level of inter- and intraobserver measurement agreement for CT, including the effects of different viewing conditions and prior experience.

Materials and Methods
Previous sectionNext section
Polyp Data Set

Forty-eight polyps in 24 patients were identified by a study coordinator from a database of CT colonography examinations accumulated as part of an ongoing study comparing CT with optical colonoscopy. The study had full approval from our local ethics committee, and all subjects gave written consent. All polyps were protuberant with no morphologically flat lesions. Their diameters ranged between 2 mm and 12 mm. All sizes were represented (except 11 mm) in 1-mm increments: four were 12 mm, five were 10 mm, one was 9 mm, three were 8 mm, two were 7 mm, seven were 6 mm, 14 were 5 mm, six were 4 mm, three were 3 mm, and three were 2 mm. The reference size was defined as the maximal diameter estimated during colonoscopy, performed by a single expert endoscopist who had performed more than 5,000 colonoscopies. Maximum polyp diameter was estimated by comparing each polyp to adjacent, open biopsy forceps, measuring 7 mm, pushed against the polyp. Twelve patients had one polyp, seven had two, one had three, two had four, one had five, and one had six polyps.

In brief, CT colonography was performed after full bowel preparation and distention using carbon dioxide for insufflation. Prone and supine scanning was performed using a 4-MDCT unit (LightSpeed Plus, GE Healthcare), collimation of 1.25 mm to 2.5 mm, pitch of 1.5, and 50 mA to 100 mA. This technique has been described in detail previously [14]. Same-day optical colonoscopy was performed after CT colonography.

The 24 individual patients were anonymous; after their order was randomized, each was given a unique study number. The study coordinator, who had prior experience interpreting more than 200 CT colonography studies, determined and recorded on a study sheet the segmental location and CT coordinates for each of the known 48 polyps. This was achieved by reference to the previous CT colonography and optical colonoscopy reports, and by viewing the prone and supine CT data sets for each study on a commercially available workstation using proprietary software (Advantage Windows 4.1 and Navigator colon package, GE Healthcare). The axial slice number for the epicenter of each individual polyp was noted for the prone and supine studies individually, if visible on both. This procedure facilitated polyp identification and location for the study interpreters (because polyp detection was not an aim of this study) and allowed them to select one of the paired studies from which to make their measurements.

Polyp Measurement

Four observers interrogated each of the 48 polyps. The observers were unaware of the reference measurement for each polyp and also unaware of the distribution of polyp sizes in the data set. All four observers were familiar with the proprietary software for reporting routine abdominopelvic CT, and the study coordinator also ensured they were familiar with the display and measurement functions required for 2D and 3D blinded assessment (see below). Observers 1, 2, and 3 had previous experience with CT colonography interpretation, with a minimum of 150 examinations each at the time of the study. Observer 4 had no prior experience and had only received instruction on how to use the software, not on how best to measure polyps. Each observer was asked to measure the maximal diameter of each polyp indicated on the study sheet using one of three image display methods: colonography window setting, abdominal window setting, and 3D image. Measurements were made using the same proprietary workstation and software used by the study coordinator to select the polyps. A four-quadrant display was used (4:1), and the first measurement was made from the 2D axial images or 2D multiplanar or oblique reformatted images obtained with the patient in either the prone or supine position, whichever was felt by the observer to better depict the maximal diameter of the polyp. The observers were also able to magnify the images according to their individual preference. The workstation was set to measure from 2D images, and a standard colonography window setting was used for viewing (width, 1,500 H; level, 150 H) (Fig. 1A). The measurement was made by placing software calipers across what was judged to be the maximal diameter of the polyp. To reduce bias, the screen annotation function was disabled during this procedure so the observer was unaware of the value of the measurement made. The observer then changed the display to standard abdominal viewing windows (width, 400 H; level, 40 H) (Fig. 1B) and made a second measurement of maximal polyp diameter in an identical fashion to the first, again unaware of the value of the measurement made. Last, the observer switched to a default endoluminal surface-rendered perspective display (Fig. 1C), repositioned the viewing angle to best depict the polyp, and made a final measurement of maximal diameter after changing the software settings to account for diameter measurement using 3D rendering. Once this measurement was done, the screen annotation function was enabled, and the observer recorded all three individual measurements on the study sheet. The interpreters were unaware of each other's results.

To assess intraobserver agreement, each observer repeated the three measurements on a subset of 10 polyps chosen by the study coordinator from the study data set, which broadly represented all polyp sizes. After the order of the patients was again randomized, each observer remeasured the same subset of polyps.

Once all measurements were completed, the study sheets were collated by the study coordinator and individual observer measurements for each polyp transferred to an Excel worksheet (version 2000, Microsoft). Analysis was performed using Stata (version 7, StataCorp LP) and MLwiN (version 1.10, Institute of Education, University of London).

Statistical Analysis

The Bland-Altman method [15] was used to determine the level of agreement between observers' estimates of maximum polyp diameter and the reference endoscopic diameter. The 95% limits of agreement were used to define the range of discrepancies between the test (CT) and the reference (colonoscopy) value on 95% of occasions. Also, because each observer measured the same polyps, a paired Student's t test was used to determine the presence of a significant difference between each set of measurements and the endoscopic reference size (for each observer). Paired Student's t tests were also used to determine any significant difference between estimates from the novice observer and the experts when using the three different visualization displays. A repeated measures analysis of variance was used to examine the difference in measurements between the four observers. We allowed for the repeated measurements on the same polyp, adjusting for this using the Huynh and Feldt method. Repeated measures analysis of covariance (ANCOVA) was used to determine if the differences between observers varied for polyps of different sizes. This was done by including the endoscopic reference size in the analysis and examining if there was an interaction between this and the observer.

To assess the repeatability of the CT polyp measurements, the variability between interpreters was measured regardless of the reference value by calculating the components of variance. It was possible to break down the data variability into that which resulted from between and from within polyp measurements. A large proportion of variability between polyp measurements implies that the observer had relatively little impact on the measurements. Conversely, if a large proportion of the variability was within polyps, this implies that the observer had more influence on the measurements. Polyp variation was further broken down into those attributable to different observers and those attributable to repeat measurements. The components of variance were calculated using a cross-classified multilevel model, with the multilevel component allowing for the structure of the data and the cross-classification allowing for repeat observations on the same polyp. Any missing observations were excluded from the analysis.

Results
Previous sectionNext section

Two of the experienced observers were unable to measure accurately a total of four polyps when using the abdominal window setting only: Observer 1 was unable to measure one polyp with a reference size of 4 mm, and observer 2 was unable to measure three polyps with reference sizes of 2 mm, 3 mm, and 5 mm. The observers attributed this to window setting effects—the reassignment of gray-scale values according to the change to abdominal windows meant that voxels with lower Hounsfield units appeared black and were therefore indistinguishable from intraluminal gas. Two experienced observers (observers 2 and 3) and the novice also were unable to measure a 2-mm polyp accurately when using the 3D display. All other measurements were complete.

Of the 48 polyps measured using a 2D display, observer 1 used an axial display for 27, an oblique display for 15, a coronal display for 5, and a sagittal display for 1. Only observer 3 used axial views for all polyps, and observers 2 and 4 used a combination of axial, coronal, and sagittal views.

Observer Agreement with Colonoscopic Reference Size

The mean differences between observers' estimates of polyp diameter and the colonoscopic reference size, the SD of these differences, and limits of agreement for each observer are summarized in Table 1 for each of the three visualization displays. In addition, the difference between individual observer measurements and the reference size are displayed graphically as Bland-Altman dot plots (Fig. 2A and 2B). These data show that the CT measurements, irrespective of the display used, on average overestimated the diameter of the polyp studied when compared with the reference value. However, the least amount of measurement error occurred when using the 2D abdominal window setting. The 95% limits of agreement were relatively wide for all observers and certainly had sufficient span to encompass different size categories for individual polyps. The narrowest limits of agreement spanned 7.6 mm for observer 1 using the 2D colonography window setting, whereas the widest spanned 14.6 mm for observer 3 (an experienced observer) using the 3D display. When significant differences between the observers' estimates and the reference values were considered, observers 1 and 2 were not significantly different, but only when using the 2D abdominal window setting; these two observers were significantly different from the reference standard when using the other two displays, and observers 3 and 4 were significantly different when using all three displays (see Table 1). The novice (observer 4) had errors that were significantly larger than expert observer 1 for all three displays (p = 0.001, 0.01, and < 0.001 for 2D colonography, 2D abdominal, and 3D displays, respectively) and significantly larger than observer 2 for two of the displays (p = 0.04 and 0.001 for 2D colonography and 3D displays, respectively). No significant difference was found between the novice and observer 3 for any display. For all displays, measurement error was smallest, on average, for observer 1, followed by observer 2, with observers 3 and 4 having greater errors. Supporting these data—when the absolute error was calculated for each observation and observer and compared with the visualization platform used—no significant difference was seen between the 2D colonography window setting and the 3D display. A significant difference occurred, however, when the 3D display was compared with the 2D abdominal window setting in all but one comparison (Table 2).

TABLE 1: Agreement Between Observer CT Measurements and Colonoscopic Reference

TABLE 2: Comparison of Absolute Errors Between Visualization Displays for Four Observers

Observer Agreement Irrespective of Reference Size

Between-polyp variance (variability when measuring different polyps) and within-polyp variance (variability when measuring the same polyp, further broken down into those attributable to different observers and those attributable to random variation [between-measurement variance]) are shown in Table 3. These data show that measurements obtained using the three visualization displays are relatively similar with very little within-polyp variation resulting from differences between observers; approximately 70% of the total variability in the data was attributable to between-polyp differences. Of the within-polyp variation, most was random variability of the measurements rather than differences between observers. When the differences between observers' measurements were examined using analysis of variance, the most significant difference occurred when using the 3D display (Table 4). No evidence that this difference was significantly influenced by polyp size was seen (2D colon window setting [p = 0.96], 2D abdominal window setting [p = 0.84], 3D display [p = 0.86], analysis of covariance).

TABLE 3: Variability in Measurement Data: Components of Variance

TABLE 4: Differences in Polyp Measurements Between Observers

figure
View larger version (16K)
Fig. 2A —Bland-Altman dot plots showing agreement between CT measurements and colonoscopy reference for each observer grouped according to visualization display (A) and per observer 1 to 4 (B). Effect of visualization display on polyp diameter measurement.

figure
View larger version (19K)
Fig. 2B —Bland-Altman dot plots showing agreement between CT measurements and colonoscopy reference for each observer grouped according to visualization display (A) and per observer 1 to 4 (B). Interobserver agreement. Colon = colonography, abdo = abdominal.

Predictably, assessment of intraobserver agreement revealed that it was superior to interobserver agreement, with the narrowest limits of agreement for the 2D abdominal display and the widest for the 3D display (Table 5). Differences between the two sets of observer measurements are displayed graphically as Bland-Altman dot plots (Fig. 3A and 3B).

TABLE 5: Intraobserver Agreement

Discussion
Previous sectionNext section

Accurate assessment of polyp size during CT colonography is fundamental for determining biologic significance of the polyp and therefore whether the patient should proceed to colonoscopy and polypectomy. Measurement errors during colonoscopy are well established [11-13], and we have shown that errors also occur during CT colonography. The errors depend on the observer's experience and the viewing conditions used to make the measurement. Furthermore, we used a viewing platform from a single manufacturer. Lesion conspicuity may be affected by the type of workstation [16], so it is conceivable that further differences between measurements would occur using platforms from different manufacturers.

Researchers group polyps into three size categories contingent on their maximum diameter (small, 5 mm or less; medium, 6 to 9 mm; large, 10 mm and larger). Expert consensus is that polyps in the small category are clinically insignificant; surveillance scanning should be considered for medium and large polyps, but polypectomy is recommended for those that are large (Zalis M, presented at the Fifth International Symposium of Virtual Colonoscopy). Such grouping is convenient, but measurement error is inevitable and could potentially result in large polyps being inadvertently classified as small and vice versa, especially if the error occurs around a category cut point. Supporting this, we found limits of agreement as wide as 14 mm, raising the possibility that a polyp whose true diameter is 10 mm or more could be assigned to the small category because of measurement error. As a result, the polyp might be assigned to interval surveillance or even be disregarded when the correct course of action would be endoscopic polypectomy. Conversely, by including clinically insignificant polyps (2- to 5-mm polyps) in our data set, we showed that diminutive polyps can be incorrectly categorized as large and therefore inappropriately referred for polypectomy.

Our analysis of variance found evidence of a significant difference between mean polyp measurements for the four observers, especially when the 3D image was used to make the measurement. This might be expected because there are more subjective opportunities for cursor placement using this type of image. The angle and direction from which the polyp is viewed may be continuously varied by the observer, as may be the distance from the polyp, even though the software is calibrated to account for this. Intuitively, these factors suggest that inaccurate measurements most likely occur when using a 3D display, and our results support this hypothesis. Alternatively, it could be argued that 3D visualization offers the best chance of detecting the long axis of the polyp, especially when the polyp is large and its morphology irregular. However, we found no evidence that accuracy was influenced by the size of the polyp being measured.

Predictably, intraobserver agreement was better than interobserver agreement, but, again, the 3D image was the least accurate. This is important because polyps left in situ require reexamination and remeasurement, perhaps after an interval of several years, which makes it less likely that the subsequent measurement will be performed by the same observer who made the first. The clear message is that whereas 3D visualization may be the best method with which to detect polyps [10], it is not the most reliable method for measuring them.

Whether to use colonography or abdominal window settings for measurement is open to debate from examination of our data. Both generally overestimated the diameter of polyps when compared with the reference colonoscopy measurement, but, overall, this discrepancy was least for the abdominal window setting. However, the narrowest limits of agreement were achieved by observer 1 (an expert) when using the colonography setting, and the least interobserver error overall (irrespective of agreement with the reference values) was achieved using the colonography setting. Also, two observers reported that very small polyps “disappeared” when they enabled the abdominal window setting. Intuitively, this latter finding suggests abdominal windows should not be used when there is an appreciable reduction in polyp conspicuity by using these settings. Moreover, in vitro studies of lung nodule measurement indicate that far more accurate measurements are obtained using lung window settings (closer to the colonography setting) compared with using an abdominal window setting [17]. Data also showed that observers differed in their ability to make accurate measurements. Notably, observers 1 and 2 (two experts) were generally more accurate than observers 3 and 4 (an expert and a novice); the novice observer had significantly larger errors than two of the experts but not the third. Although this suggests that prior experience in CT colonography interpretation may facilitate more accurate measurements, this study is not absolutely conclusive.

figure
View larger version (10K)
Fig. 3A —Bland-Altman dot plots showing intraobserver agreements, grouped according to visualization display (A) and per observer 1 to 4 (B). Effect of visualization display on intraobserver agreement.

figure
View larger version (13K)
Fig. 3B —Bland-Altman dot plots showing intraobserver agreements, grouped according to visualization display (A) and per observer 1 to 4 (B). Intraobserver agreement. Colon = colonography, abdo = abdominal.

There are several limitations to our study, and perhaps it raises more questions than it answers, especially regarding the true diameter of a polyp. We used the colonoscopic estimate of polyp diameter as a reference standard because this is conventional, well-established practice. We also used a single expert endoscopist whose experience exceeded 5,000 colonoscopies. Despite this, it is well recognized that colonoscopic measurements are subject to considerable error, even when made by experienced practitioners [11-13]. Endoscopists tend to overestimate polyp diameter during in vivo assessment and underestimate diameter during in vitro assessment [11, 13]. For pragmatic reasons we used open biopsy forceps to determine the reference size because this is the method preferred by our endoscopists. Using a calibrated linear measuring tool may have been preferable, but even this is imperfect because it may not be possible to align the tool adjacent to the maximal diameter of the polyp, leading to a semisubjective assessment [13]. Alternatively, it could be argued that we should have used pathologic estimates as our reference standard as they have been shown to be more reliable than endoscopic measurements [12]. However, even this approach is imperfect because excised polyps tend to significantly decrease in diameter as a result of cauterization and vascular collapse and may either enlarge or shrink with formalin fixation [12, 18]. We did not attempt to analyze our data according to polyp morphology; for example, our data set did not contain flat polyps, which are notoriously difficult to detect and to measure [19, 20]. Also, it is conceivable that elliptic polyps may be more accurately measured on 3D because of better depiction of their shape. Indeed, our study provides some indirect evidence to suggest this is the case. Although observers were free to use oblique 2D reformatted images, only observer 1 actually did so, and observer 1 was closest to the reference standard. In contrast, observer 3 solely used the axial plane and was the least accurate of the experts.

All these factors conspire to thwart a truly reliable estimate of polyp diameter with which comparisons can be made. The Bland-Altman approach is especially relevant to this analysis because it can be argued that the true measurement is not known with certainty from either colonography or colonoscopy. In this case, the best estimate of the true diameter is the mean of the measurements obtained by colonoscopy and colonography, which is the bedrock of the Bland-Altman analysis. We found that absolute agreement between the CT colonography estimate and the colonoscopic reference measurement was variable, evidenced by limits of agreement that were frequently enough to span not only one but two polyp size categories.

We used a single viewing platform (with a variety of visualization displays). The 3D image we used, which was surface-rendered, may be more inherently inaccurate than a volume-rendered alternative [21]. Also for pragmatic purposes, we did not formally assess the effect of display magnification on polyp measurement accuracy, although observers were free to alter magnification according to individual preference. Further research on these topics is required. Our use of a single novice may be criticized because this individual was effectively acting as a proxy for all novices. We were able to show significant differences between our three experts and no doubt there would also be differences between novices. One potential solution to minimize inter- and intraobserver variation of assessment of polyp diameter is to fully automate the measurement process using computer-aided boundary detection, which removes the subjective element of cursor placement. Further development in this area is awaited.

In conclusion, measurement of maximum polyp diameter during CT colonography is subject to inter- and intraobserver variation, the degree of which is contingent on the observers, their level of expertise, and the viewing conditions used. This variation may result in polyps detected by CT being inadvertently assigned to an incorrect size category. In day-to-day clinical practice, the window setting and anatomic plane used for measurement should be documented to facilitate subsequent interval comparisons. Finally, our study suggests that 3D visualization displays, commonly used for polyp detection, should not be used for polyp measurement.

Address correspondence to S. Halligan.

2Present address: Department of Specialist Radiology, Level 2 Podium, University College Hospital, 235 Euston Rd., London, England, NW1 2BU.

References
Previous sectionNext section
1. Terry MB, Neugut AI, Bostick RM, et al. Risk factors for advanced colorectal adenomas: a pooled analysis. Cancer Epidemiol Biomarkers Prev 2002; 11:622-629 [Google Scholar]
2. Giovanunucci E, Ascherio A, Rimm EB, Colditz GA, Stampfer MJ, Willett WC. Physical activity, obesity, and risk for colon cancer and adenoma in men. Ann Intern Med 1995; 122:327-334 [Google Scholar]
3. Muto T, Bussey JR, Morson BC. The evolution of cancer of the colon and rectum. Cancer 1975; 36:2251-2270 [Google Scholar]
4. O'Brien MI, Winawer SJ, Zauber AG, et al. The National Polyp Study: patient and polyp characteristics associated with high-grade dysplasia in colorectal adenomas. Gastroenterology 1990; 98:371-379 [Google Scholar]
5. Read TE, Read JD, Butterfly RN, Butterfly LF. Importance of adenomas 5 mm or less that are detected by sigmoidoscopy. N Engl J Med 2003; 336:8-12 [Google Scholar]
6. Wallace MB, Kemp JA, Trnka YM, Donovan JM, Farraye FA. Is colonoscopy indicated for small adenomas found by screening flexible sigmoidoscopy? Ann Intern Med 1998; 129:273-278 [Google Scholar]
7. Noshirwani KC, Van Stolk RU, Rybicki LA, Beck GJ. Adenoma size and number are predictive of adenoma recurrence: implications for surveillance colonoscopy. Gastrointest Endosc 2000; 51:433-437 [Google Scholar]
8. Atkin WS, Saunders BP. Surveillance guidelines after removal of colorectal adenomatous polyps. Gut 2002; 51(suppl V):V6-V9 [Google Scholar]
9. Winawer S, Fletcher R, Rex D, et al; Gastrointestinal Consortium Panel. Colorectal cancer screening and surveillance: clinical guidelines and rationale—update based on new evidence. Gastroenterology 2003; 124:544-560 [Google Scholar]
10. 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 [Google Scholar]
11. Fennerty MB, Davidson J, Emerson S, Sampliner RE, Hixson LJ, Garewal HS. Are endoscopic measurements of colonic polyps reliable? Am J Gastroenterol 1993; 88:496-500 [Google Scholar]
12. Schoen RE, Gerber LD, Margulies C. The pathologic measurement of polyp size is preferable to the endoscopic estimate. Gastrointest Endosc 1997; 46:492-496 [Google Scholar]
13. Gopalswamy N, Shenoy VN, Choudhry U, et al. Is in vivo measurement of size of polyps during colonoscopy accurate? Gastrointest Endosc 1997; 46:497-502 [Google Scholar]
14. Taylor SA, Halligan S, Saunders BP, et al. Use of multidetector-row CT colonography for detection of colorectal neoplasia in patients referred via the Department of Health “2-Week-wait” initiative. Clin Radiol 2003; 58:855-861 [Google Scholar]
15. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1:307-310 [Google Scholar]
16. Pickhardt PJ. Three-dimensional endoluminal CT colonography (virtual colonoscopy): comparison of three commercially available systems. AJR 2003; 181:1599-1606 [Abstract] [Google Scholar]
17. Harris KM, Adams H, Lloyd DC, Harvey DJ. The effect on apparent size of simulated pulmonary nodules of using three standard CT window settings. Clin Radiol 1993; 47:241-244 [Google Scholar]
18. Morales TG, Sampliner RE, Garewal HS, Fennerty MB, Aickin M. The difference in colon polyp size before and after removal. Gastrointest Endosc 1996; 43:25-28 [Google Scholar]
19. Tsuda S, Veress B, Toth E, et al. Flat and depressed colorectal tumours in a southern Swedish population: a prospective chromoendoscopic and histopathological study. Gut 2002; 51:550-555 [Google Scholar]
20. Macari M, Bini EJ, Jacobs SL, Lange N, Lui YW. Filling defects at CT colonography: pseudo- and diminutive lesions (the good), polyps (the bad), flat lesions, masses, and carcinomas (the ugly). Radio-Graphics 2003; 23:1073-1091 [Google Scholar]
21. Hopper KD, Iyriboz A, Wise SW, Neuman JD, Mauger DT, Kasales CJ. Mucosal detail at CT virtual reality: surface versus volume rendering. Radiology 2000; 214:517-522 [Google Scholar]

Recommended Articles

Polyp Measurement Using CT Colonography: Agreement with Colonoscopy and Effect of Viewing Conditions on Interobserver and Intraobserver Agreement

Full Access, , ,
American Journal of Roentgenology. 2006;186:1792-1796. 10.2214/AJR.05.1315
Abstract | Full Text | PDF (303 KB) | PDF Plus (355 KB) 
Full Access, , , , , , , ,
American Journal of Roentgenology. 2007;188:122-129. 10.2214/AJR.05.1535
Abstract | Full Text | PDF (1536 KB) | PDF Plus (433 KB)