6
Electronic colon-cleansing for CT colonography: diagnostic performance Markus S. Juchems, 1 Andrea Ernst, 1 Peter Johnson, 2 Sunny Virmani, 2 Hans-Juergen Brambs, 1 Andrik J. Aschoff 1 1 Department for Diagnostic and Interventional Radiology, University Hospitals Ulm, Ulm, Germany 2 Philips Medical Systems, Highland Hights, OH, USA Abstract Background: The purpose of this study was to evaluate whether an electronic-colonic-cleansing (ECC) algorithm is beneficial for the diagnostic performance compared to a CT colonography (CTC) evaluation without electronic cleansing in tagged datasets. Methods: Two blinded readers evaluated CTC datasets from 79 patients with 153 colorectal polyps confirmed by optical colonoscopy. Cases were read in a randomized order with and without the use of electronic colon- cleansing software. Per-polyp sensitivity, per-polyp/per- patient specificity and reading times (with and without ECC) have been calculated and reported. Results: Per-polyp sensitivity for polyps >6 mm without using ECC was 60.4% (Reader 1: 59.7%, Reader 2: 61.1%), while polyps >10 mm were detected with a sensitivity of 58.3% (Reader 1: 66.7%, Reader 2: 50%). On electronically cleansed datasets, the sensitivity was 73.6% (Reader 1: 76.4%; Reader 2: 70.8%) for polyps >6 mm and 83.3% (Reader 1: 83.3%; Reader 2: 83.3%), respectively. Per-patient specificity was 75% without using cleansing (Reader 1: 68%, Reader 2: 82%) and 81.5% using ECC (Reader 1: 86%, Reader 2: 77%). Conclusion: Reading CTC cases using ECC software improves sensitivity in detecting clinically relevant colo- rectal polyps. Key words: CT—CT colonography—Colon—Polyps— Electronic cleansing Since its clinical introduction a decade ago, CT colo- nography (CTC) has benefited from major advances in CT scanner technology, including the introduction of multi-slice scanners and sub-millimeter detectors that may help increase sensitivity in detecting smaller polyps [1]. In studies with older scanners and thicker slices [2], lower sensitivities and specificities have been observed in detecting colorectal lesions. In addition to scanner development, a second focus has been on advanced image post processing. Various 3D visualization methods have been developed with the goal of increasing sensitivity and specificity in detecting colorectal polyps [36]. Nevertheless, several studies have shown that residual fluid and stool residuals can affect the performance of CTC. In poorly prepared patients with suboptimal stool/fluid tagging, residuals can simu- late polypoid lesions (false positive) or hide relevant polyps (false negative) [7, 8]. Two major bowel-cleansing strategies to prepare patients for CTC are used. The first, called ‘‘wet- cleansing’’, PEG-based cleansing, uses a cleansing regi- ment similar to cleansing protocols for optical colon- oscopy and tries to achieve a clean colon with as few stool and fluid residuals as possible. The second strat- egy is to enhance fecal residues and residual fluid (‘‘dry cleansing’’) using oral contrast agents that accumulate in the feces. Based on the high density of tagged stool and fluid, digital subtraction of marked feces and fluid for better 3D visualization was developed (Fig. 1). Some authors [9] have already reported good results with this technique, however, others report difficulties using an electronic cleansing algorithm [10, 11] such as artifacts that may occur due to digital subtraction (Fig. 2). To the best of our knowledge, no other study has yet compared reader sensitivity and specificity in detecting colorectal polyps with and without the use of electronic- colon-cleansing in the same patient cohort. Therefore, Correspondence to: Markus S. Juchems; email: markus.juchems@uni- ulm.de ª Springer Science+Business Media, LLC 2008 Published online: 15 March 2008 Abdominal Imaging Abdom Imaging (2009) 34:359–364 DOI: 10.1007/s00261-008-9386-6

Electronic colon-cleansing for CT colonography: diagnostic performance

Embed Size (px)

Citation preview

Electronic colon-cleansing for CTcolonography: diagnostic performance

Markus S. Juchems,1 Andrea Ernst,1 Peter Johnson,2 Sunny Virmani,2

Hans-Juergen Brambs,1 Andrik J. Aschoff1

1Department for Diagnostic and Interventional Radiology, University Hospitals Ulm, Ulm, Germany2Philips Medical Systems, Highland Hights, OH, USA

Abstract

Background: The purpose of this study was to evaluatewhether an electronic-colonic-cleansing (ECC) algorithmis beneficial for the diagnostic performance compared toa CT colonography (CTC) evaluation without electroniccleansing in tagged datasets.Methods: Two blinded readers evaluated CTC datasetsfrom 79 patients with 153 colorectal polyps confirmed byoptical colonoscopy. Cases were read in a randomizedorder with and without the use of electronic colon-cleansing software. Per-polyp sensitivity, per-polyp/per-patient specificity and reading times (with and withoutECC) have been calculated and reported.Results: Per-polyp sensitivity for polyps >6 mm withoutusing ECC was 60.4% (Reader 1: 59.7%, Reader 2:61.1%), while polyps >10 mm were detected with asensitivity of 58.3% (Reader 1: 66.7%, Reader 2: 50%).On electronically cleansed datasets, the sensitivity was73.6% (Reader 1: 76.4%; Reader 2: 70.8%) for polyps>6 mm and 83.3% (Reader 1: 83.3%; Reader 2: 83.3%),respectively. Per-patient specificity was 75% withoutusing cleansing (Reader 1: 68%, Reader 2: 82%) and81.5% using ECC (Reader 1: 86%, Reader 2: 77%).Conclusion: Reading CTC cases using ECC softwareimproves sensitivity in detecting clinically relevant colo-rectal polyps.

Key words: CT—CT colonography—Colon—Polyps—Electronic cleansing

Since its clinical introduction a decade ago, CT colo-nography (CTC) has benefited from major advances inCT scanner technology, including the introduction ofmulti-slice scanners and sub-millimeter detectors thatmay help increase sensitivity in detecting smaller polyps[1]. In studies with older scanners and thicker slices [2],lower sensitivities and specificities have been observed indetecting colorectal lesions.

In addition to scanner development, a second focushas been on advanced image post processing. Various 3Dvisualization methods have been developed with the goalof increasing sensitivity and specificity in detectingcolorectal polyps [3–6]. Nevertheless, several studies haveshown that residual fluid and stool residuals can affectthe performance of CTC. In poorly prepared patientswith suboptimal stool/fluid tagging, residuals can simu-late polypoid lesions (false positive) or hide relevantpolyps (false negative) [7, 8].

Two major bowel-cleansing strategies to preparepatients for CTC are used. The first, called ‘‘wet-cleansing’’, PEG-based cleansing, uses a cleansing regi-ment similar to cleansing protocols for optical colon-oscopy and tries to achieve a clean colon with as fewstool and fluid residuals as possible. The second strat-egy is to enhance fecal residues and residual fluid (‘‘drycleansing’’) using oral contrast agents that accumulatein the feces. Based on the high density of tagged stooland fluid, digital subtraction of marked feces and fluidfor better 3D visualization was developed (Fig. 1).Some authors [9] have already reported good resultswith this technique, however, others report difficultiesusing an electronic cleansing algorithm [10, 11] such asartifacts that may occur due to digital subtraction(Fig. 2).

To the best of our knowledge, no other study has yetcompared reader sensitivity and specificity in detectingcolorectal polyps with and without the use of electronic-colon-cleansing in the same patient cohort. Therefore,

Correspondence to: Markus S. Juchems; email: [email protected]

ª Springer Science+Business Media, LLC 2008

Published online: 15 March 2008AbdominalImaging

Abdom Imaging (2009) 34:359–364

DOI: 10.1007/s00261-008-9386-6

the purpose of this study was to evaluate if an electroniccolonic cleansing algorithm is beneficial compared withCTC evaluation without electronic cleansing.

Method and materials

PCJ and SV are employees of Philips Medical Systems.(All authors had control of and access to the data. Thereare no other disclosures.)

Study group

The CTC datasets of a screening population have beenprovided courtesy of Dr. Richard Choi, Virtual Colon-oscopy Center, Walter Reed Army Medical Center,Washington DC, USA. 80 patient datasets were availablefor the study [12]. One dataset was excluded due totechnical limitations. Optical colonoscopy was consid-ered as gold standard for this study.

Two experienced readers retrospectively evaluated allpatients blinded to optical colonoscopy results usingboth prone and supine series in each case. Overall, 153colorectal polyps were identified by optical colonoscopy.Polyps were divided into 3 categories based on maximumdiameter measurements reported by optical colonoscopy.There were 81 polyps measuring 5 mm or less, 48 polypswere between 6 and 9 mm and 24 polyps were 10 mm orlarger.

CT colonography technique

CT colonography datasets were acquired according tomethods described by Pickhardt et al. [13]. In brief,patients underwent standard 24-h colonic preparationwith oral administration of 90 mL of sodium phosphate(Fleet 1 preparation, Fleet Pharmaceuticals) and 10 mgof bisacodyl. As part of their clear-liquid diet, patientsalso consumed 500 mL of barium (2.1% by weight; Scan

Fig. 1. Example of a 10-mm polyp hidden under the residualfluid that has been exposed by the use of electronic cleansing.A The polyp is obscured in perspective Filet view (upper left)

and endoluminal view (upper right). B After electroniccleansing the polyp is clearly visible in both 3D viewingmethods.

Fig. 2. A Example of a pseudopolyp in 3D mode due to rendering artifacts after electronic cleansing (arrows). B The 2D viewafter cleansing shows a line-like artifact. C 2D view without cleansing reveals no polypoid lesion.

360 M. Juchems et al.: Electronic colon-cleansing for CT colonography

C, Lafayette Pharmaceuticals) for solid-stool tagging and120 mL of diatrizoate meglumine and diatrizoate sodium(Gastrografin, Bracco Diagnostics) for the opacificationof endoluminal fluid. The CT protocol was as follows.On insertion of a small flexible rectal catheter, pneu-mocolon was achieved through patient-controlled insuf-flation of room air immediately before scanning. CTscanning was performed with the breath held in bothsupine and prone positions.

CT scans were performed using a four- or eight-channel CT scanner (GE LightSpeed, General ElectricMedical Systems). A collimation of 1.25–2.5 mm, a tablespeed of 15 mm per second, a reconstruction interval of1 mm, and scanner settings of 100 mAs and 120 kV wereused.

Case evaluation

Image processing and interpretation were performed ona dedicated workstation (Philips Extended BrillianceWorkspace, version 3.0, Philips Medical Systems,Cleveland, OH, USA) that includes an electronic-colon-cleansing option. The two readers were experienced inperforming CTC. Reader 1 had read more than 300 CTCdatasets before this study and Reader 2 more than 100.Prior to review, colon segments were identified and thecolon lumen center-line was calculated automatically.The level of air insufflation was scored on a 3-point scale(1 (collapsed), 2 (partially distended) and 3 (well dis-tended)). The level of residual fluid was also scored usinga 4-point scale (1 (no residual fluid), 2 (minimal), 3(moderate), and 4 (excessive residual fluid)).

All CTC datasets were read twice. First the cases wereevaluated without using electronic colon-cleansing. Thesecond read was performed with electronic colon-cleansing. Cases were evaluated in a randomized orderand with a gap of at least 2 weeks between the 2 reads toreduce recall bias. Dataset evaluation was performed in aprimary 3D workup. A colon dissection display and anendoluminal ‘‘flight-through’’ were simultaneously dis-played. Furthermore it was possible to evaluate suspectlesions by performing additional multiplanar reforma-tions (MPR).

Polyp findings were reported using a standardizedCase Report Form (CRF). The presence, location, size,and morphologic features of potential polyps were as-sessed for each of the six segments (Rectum, Sigmoid,Descending, Transverse, Ascending, Cecum) in the proneand supine positions.

Software loading and processing time (e.g., automaticsegmentation, calculation of colonic lumen center-lineand automatic cleansing if applied) and pure case reading(interpretation) time (defined as end of automatic pro-cessing until the interpretation is finished were assessedfor each case).

The CTC findings were then compared to the OCfindings (provided by Walter Reed Medical Center).True positive findings were identified by matching polyplocation, size, and type. A given polyp was considered atrue positive match between CTC and OC if it appearedin the same segment of the colon and the diameter re-corded by CTC was within 2 mm of the diameter re-ported by OC.

Normal distribution was tested using the Kolmogo-rov–Smirnov test. For non-normally distributed param-eters the Wilcoxon test was used for calculation. Allstatistics were calculated with SPSS 11.0 (SPSS GmbH,Munich, Germany). A P-value < 0.05 was consideredstatistically significant.

Electronic cleansing software

The software used in this study for electronic subtractionof opacified colonic fluid is available as an option withinthe Virtual Colonoscopy application on Extended Bril-liance Workspace, Philips Medical Systems, Cleveland,OH, USA. The cleansing algorithm processes only thoseintra-colonic regions that had been identified during theautomatic segmentation of the colon. These intra-colonicregions may consist either of air, soft-tissue, contrast(opacified fluid), or combinations of the latter. Thealgorithm identifies and classifies regions using a pre-defined threshold and morphological operations such asfilling, erosion, and dilation. Image processing tech-niques and smoothing filters are applied to reducecleansing and partial-voluming related artifacts. Fecalmaterial and other material floating on the fluid isidentified and removed to make sure no residual fluid orfecal material is left behind after electronic subtraction.The application also allows to simply toggle betweenelectronically cleansed and uncleansed images for 3D aswell as MPR visualization.

Results

Image quality scores

Tables 1 and 2 illustrate the grading of colonic segmentsin regard to the amount of insufflation and residual fluid,respectively. Scores were given for each of the six colonicsegments in both patient positions.

Reader 1 graded 7% of the segments to be collapsedin prone patient position (supine: 8%). Reader 2 rated1% of the segments to be collapsed in prone and supinepatient position.

Reader 1 graded 76% and 83% of the segments inprone and supine position, respectively, as having mod-erate or excessive contrast enhanced residual fluid.Reader 2 had a similar rating of 64% and 69% of thesegments in prone and supine position, respectively, ashaving moderate or excessive tagged fluid.

M. Juchems et al.: Electronic colon-cleansing for CT colonography 361

Sensitivity and specificity for detecting colorectallesions

An overview of per-polyp sensitivity without and withelectronic bowel cleansing over a range of polyp diame-ters for both readers is provided in Table 3.

Polyps measuring between 6 and 9 mm were detectedwith a sensitivity of 56.3% (Reader 1) and 66.7%(Reader 2), for polyps larger than 10 mm 66.7% (Reader1) and 50% (Reader 2) without cleansing. Using elec-tronic cleansing, sensitivity was 72.9% (Reader 1) and64.6% (Reader 2) for polyps between 6 and 9 mm and83.3% for both readers for polyps larger than 10 mm.The sensitivity for detecting polyps larger than 6 mm was59.7% (Reader 1) and 61.1% (Reader 2) without usingelectronic cleansing, which improved to a sensitivity of76.4% and 70.8%, respectively, when using electroniccleansing.

Detection of true positive findings improved signifi-cantly for Reader 1, when using electronic cleansing(P = 0.02). Reader 2 did not improve significantly(P = 0.5).

The results of per-patient sensitivities and specificitiesare provided in Table 4. Reader 1 showed an improve-

ment in per-patient sensitivity from 82% (without elec-tronic cleansing) to 91% (using electronic cleansing).Reader 2 did not benefit from electronic cleansing as per-patient sensitivity was 86% regardless of the use ofelectronic cleansing. Per-patient specificity improvedfrom 68% to 86% for Reader 1, while specificity de-creased from 82% to 77% for Reader 2.

Reading time

For Reader 1 software average loading and preparationtime was 3.16 min without and 2.79 min with electroniccleansing, respectively. The difference in time was ob-served because the center-line calculation without elec-tronic cleansing required manual correction due to fluidfilled colonic segments that were not automatically de-tected. The average reading (evaluation) time for Reader1 was 9.84 and 9.58 min with and without electroniccleansing, respectively.

Reader 2 required 10.53 min to evaluate a case whileusing electronic cleansing, as compared to 11.6 min whenreading without electronic cleansing. Case loading andpreparing time was 3.22 min without electronic cleansingcompared to 2.62 min with electronic cleansing forReader 2. This analysis is also summarized in Table 5.

Table 1. Grading of colonic segments in regard to air insufflation

Level of insufflation Air insufflation

Reader 1 Reader 2

Prone Supine Prone Supine

1 (Collapsed) 33 (7%) 36 (8%) 3 (1%) 4 (1%)2 (Partially distended) 261 (55%) 279 (59%) 121 (26%) 149 (31%)3 (Well-distended) 180 (38%) 159 (33%) 350 (74%) 321 (68%)

Table 2. Grading of colonic segments in regard to residual fluid

Level of residual fluids Residual fluids

Reader 1 Reader 2

Prone Supine Prone Supine

0 (No residuals) 12 (3%) 3 (1%) 5 (1%) 6 (1%)1 (Minimal residuals) 101 (21%) 77 (16%) 166 (35%) 144 (30%)2 (Moderate residuals) 270 (57%) 268 (57%) 227 (48%) 236 (50%)3 (Excessive residuals) 91 (19%) 126 (26%) 76 (16%) 88 (19%)

Table 4. Sensitivity and specificity calculated on a per-patient base

Per-patient sensitivity Per-patient specificity

Withoutcleansing (%)

Withcleansing (%)

Withoutcleansing (%)

Withcleansing (%)

Reader 1 82 91 68 86Reader 2 86 86 82 77

Table 3. Sensitivity of CTC in detecting colorectal lesions

Overall (79 cases) Ground truth (OC) Reader 1 Reader 2

Without cleansing With cleansing Without cleansing With cleansing

Diameter range No. of polyps Polypsfound

Sensitivity(%)

Polypsfound

Sensitivity(%)

Polypsfound

Sensitivity(%)

Polypsfound

Sensitivity(%)

£5 mm 81 34 42.0 37 45.7 38 46.9 35 43.26–9 mm 48 27 56.3 35 72.9 32 66.7 31 64.6‡10 mm 24 16 66.7 20 83.3 12 50.0 20 83.3Total 153 77 50.3 92 60.1 82 53.6 86 56.2Polyps ‡6 mm 72 43 59.7 55 76.4 44 61.1 51 70.8

Table 5. Loading, p reparation, and reading time comparison for bothreaders without and with electronic cleansing

Average loading andpreparation time (min)

Average readingtime (min)

Withoutcleansing

Withcleansing

Withoutcleansing

Withcleansing

Reader 1 3.16 2.79 9.84 9.58Reader 2 3.22 2.62 10.53 11.6

362 M. Juchems et al.: Electronic colon-cleansing for CT colonography

Discussion

CT colonography is becoming widely accepted as ascreening technique for colorectal cancer. However, theperformance of this technique in terms of sensitivity andspecificity seems to be variable. This might be related tothe fact that various scanner types and numerous varia-tions of image-post processing algorithms have been usedin the different studies and trials published in the past.To further evaluate software improvements, the aim ofthis study was to assess possible advantages of applyingan electronic colon-cleansing algorithm in CTC datasetsthat were orally contrast enhanced to tag residual fecesand fluids.

In this study, we compared per-polyp sensitivity, per-patient sensitivity and per-patient specificity in detectingcolorectal polyps with and without the use of electroniccleansing. To reduce bias [14], we decided to have thedatasets read by two readers, both experienced withCTC. To our knowledge, this is the first study that di-rectly compares sensitivity and specificity for the detec-tion of colorectal polyps with and without electroniccolon-cleansing on the same set of patient cases. How-ever, there have been studies published that describe theprocedure and its advantages and pitfalls [9, 10, 15–19].Besides showing the feasibility of electronic cleansing,some of these studies have emphasized and evaluated thehypothesis that electronic cleansing could potentiallyreduce cathartic cleansing that compromises patientcompliance and instead use a dietary (‘‘dry’’) cleansing.

There are two major bowel-cleansing strategies toprepare patients for CTC. The first, so-called ‘‘wet-cleansing’’, PEG-based cleansing, uses a cleansing regi-ment similar to cleansing protocols for optical colonos-copy and tries to achieve a clean colon with as few stooland fluid residuals as possible. This technique requiresgood patient compliance, which is, especially in elderlypatients, not always given.

The second strategy is to mark fecal residues andresidual fluid in less aggressively prepared patients (re-ferred to as ‘‘dry cleansing’’ by some authors), using Na-phospho-soda or Mg-citrate and oral contrast agentsthat accumulate in the feces. There are several publica-tions that prefer the latter method, not only because ofimproved sensitivity and specificity, but also because it isbetter tolerated by patients [10, 15, 20, 21].

A threshold size for clinically relevant polyps is stillbeing discussed in various publications. While some au-thors press for performing polypectomy on all polyps>6 mm (or—alternatively—for >3 polyps of any size),others tend to wait and watch these polys [22]. However,there seems to be consensus that polyps >10 mm need tobe taken out because of their potential to develop toinvasive cancer. Per-polyp sensitivity for this size cate-gory (>10 mm) was 58.3% (Reader 1: 66.7%, Reader 2:50%) without cleansing, which improved to 83.3%

(Reader 1: 83.3%; Reader 2: 83.3%) when electroniccleansing was used. Without electronic cleansing polyps>6 mm were detected with a sensitivity of 60.4% (59.7%for Reader 1 and 61.1% for Reader 2). Again, bothreaders benefited from automatic cleansing also for thatpolyp-size category. On electronically cleansed datasets,the sensitivity for polyps >6 mm improved to 73.6%(Reader 1: 76.4%; Reader 2: 70.8%).

Previously reported literature [2, 13, 23–26] compar-ing CTC readings to conventional colonoscopy for polypdetection has shown heterogeneous results with sensi-tivities for polyps larger 10 mm in the range of 55.6% (2)up to 100% [23, 24]. Looking at a large meta-analysis byMulhall et al. of recent CTC studies carried out, ourresults, both with and without electronic cleansing, areplaced well within the wide range of sensitivities given,especially for large polyps.

Although the datasets were prepared to be read withautomatic cleansing software, we were able to scoreconsiderably good results even without using thecleansing algorithm, especially for large polyps. 60.4%sensitivity for polyps 10 mm or larger is within the rangeother studies published [2]. This fact shows that it is,however, possible to read tagged cases without an elec-tronic software system, although it needs to be said thatsensitivity is compromised and both readers observedincreased sensitivity, especially for the large polyps(‡10 mm), using electronic cleansing on these datasets.

On a per-patient base, Reader 1 showed improvementin sensitivity from 82% to 91% when using electroniccleansing, whereas Reader 2 maintained per-patientsensitivity at 86%, regardless of the use of electroniccleansing.

Previously published studies [2, 26–30] have alsoshown very heterogeneous per-patient specificity rangingfrom 70.7% to 97.7%. Per-patient specificity calculationsshowed diverse results for both readers in our study, too.Reader 1 was able to improve specificity from 68% to86%, but Reader 2 scored lower specificity using elec-tronic cleansing. However, this data places our resultsagain well in the range of the literature cited above. Aspreviously reported [11], electronic bowel-cleansing pro-duces artifacts during the cleansing procedure. Some ofthese artifacts can mimic colorectal polyps. This couldprobably be the reason why Reader 2, who was slightlyless experienced with CT Colonography, showed de-creased specificity when the electronic bowel-cleansingalgorithm was used.

It is still in discussion [4, 31], whether a primary 3Ddataset evaluation is superior to primary 2D evaluation.In our study, a primary 3D workup with the possibilityto evaluate suspect lesions with MPR images was per-formed. However, it is unclear if electronic cleansing isalso beneficial to readers using a solely 2D evaluation, aspolyps that are hidden under fluid in a 3D reconstructionshow up as ‘‘negative’’ blankings in tagged fluid.

M. Juchems et al.: Electronic colon-cleansing for CT colonography 363

Differences in patient population, scanning protocols,patient preparation, and analysis techniques are possiblereasons for the high variability in the sensitivities andspecificities achieved in different studies.

Software loading and preparation time decreasedslightly if the electronic bowel-cleansing option was used(Reader 1: 3.16 vs. 2.79 min; Reader 2: 3.22 vs. 2.62 min)because the preparation time included editing time forincorrect automatic center-line, which was less necessaryin the electronically cleaned datasets.

Overall, we were able to show that the cleansingalgorithm used compares very favorably to previouslypublished articles [11, 16] by means of post processingtime. Pure case evaluation time showed no significantdifference (improvement) when using electronic cleansingfor both readers. However, both readers managed toevaluate cases in a time that is in the range or even belowpreviously published data [4, 13, 32, 33].

In conclusion, we could show in this study that bothreaders did benefit from electronic colon-cleansing, notonly in detecting large polyps ‡10 mm, but also smallerpolyps (6–9 mm) that may be potentially clinically rele-vant according to some authors. Furthermore, electroniccleansing helped one reader by increasing specificity.

References

1. Ferrucci J (2005) CT colonography for detection of colon polypsand cancer. Lancet 365:1464–1465 author reply 1465–1466

2. Cotton PB, Durkalski VL, Pineau BC, et al. (2004) Computedtomographic colonography (virtual colonoscopy): a multicentercomparison with standard colonoscopy for detection of colorectalneoplasia. JAMA 291:1713–1719

3. Vos FM, van Gelder RE, Serlie IW, et al. (2003) Three-dimensionaldisplay modes for CT colonography: conventional 3D virtualcolonoscopy versus unfolded cube projection. Radiology 228:878–885

4. Hoppe H, Quattropani C, Spreng A, et al. (2004) Virtual colondissection with CT colonography compared with axial interpreta-tion and conventional colonoscopy: preliminary results. AJR Am JRoentgenol 182:1151–1158

5. Rottgen R, Fischbach F, Plotkin M, et al. (2005) CT colonographyusing different reconstruction modi. Clin Imaging 29:195–199

6. Juchems MS, Ehmann J., Brambs H-J, et al. (2005) A retrospectiveevaluation of patient acceptance of CT colonography (‘‘virtualcolonoscopy’’) in comparison to conventional colonoscopy in anaverage risk screening population. Acta Radiol 46:664–670

7. Fletcher JG, Johnson CD, MacCarty RL, et al. (1999) CT colo-nography: potential pitfalls and problem-solving techniques. AJRAm J Roentgenol 172:1271–1278

8. Pescatore P, Glucker T, Delarive J, et al. (2000) Diagnostic accu-racy and interobserver agreement of CT colonography (virtualcolonoscopy). Gut 47:126–130

9. Zalis ME, Perumpillichira J, Del Frate C, et al. (2003) CT colo-nography: digital subtraction bowel cleansing with mucosalreconstruction initial observations. Radiology 226:911–917

10. Bielen D, Thomeer M, Vanbeckevoort D, et al. (2003) Dry prep-aration for virtual CT colonography with fecal tagging using water-soluble contrast medium: initial results. Eur Radiol 13:453–458

11. Pickhardt PJ, Choi JH (2003) Electronic cleansing and stool tag-ging in CT colonography: advantages and pitfalls with primarythree-dimensional evaluation. AJR Am J Roentgenol 181:799–805

12. CT Colonoscopy Imaging Database. http://virtualcolonoscopy.nci.nih.gov/

13. Pickhardt PJ, Choi JR, Hwang I, et al. (2003) Computed tomo-graphic virtual colonoscopy to screen for colorectal neoplasia inasymptomatic adults. N Engl J Med 349:2191–2200

14. Obuchowski NA (2003) Special topics III: bias. Radiology 229:617–621

15. Lefere PA, Gryspeerdt SS, Dewyspelaere J, et al. (2002) Dietaryfecal tagging as a cleansing method before CT colonography: initialresults polyp detection and patient acceptance. Radiology 224:393–403

16. Zalis ME, Hahn PF (2001) Digital subtraction bowel cleansing inCT colonography. AJR Am J Roentgenol 176:646–648

17. Zalis ME, Perumpillichira JJ, Kim JY, et al. (2005) Polyp size at CTcolonography after electronic subtraction cleansing in an anthro-pomorphic colon phantom. Radiology 236:118–124

18. Callstrom MR, Johnson CD, Fletcher JG, et al. (2001) CT colo-nographywithout cathartic preparation: feasibility study. Radiology219:693–698

19. Chen D, Liang Z, Wax MR, et al. (2000) A novel approach toextract colon lumen from CT images for virtual colonoscopy. IEEETrans Med Imaging 19:1220–1226

20. Gryspeerdt S, Lefere P, Dewyspelaere J, et al. (2002) Optimisationof colon cleansing prior to computed tomographic colonography.JBR-BTR 85:289–296

21. Lefere P, Gryspeerdt S, Marrannes J, et al. (2005) CT colonogra-phy after fecal tagging with a reduced cathartic cleansing and areduced volume of barium. AJR Am J Roentgenol 184:1836–1842

22. Rex DK (2005) PRO: patients with polyps smaller than 1 cm oncomputed tomographic colonography should be offered colonos-copy and polypectomy. Am J Gastroenterol 100:1903–1905 dis-cussion 1907–1908

23. Chung DJ, Huh KC, Choi WJ, et al. (2005) CT colonography using16-MDCT in the evaluation of colorectal cancer. AJR Am JRoentgenol 184:98–103

24. Macari M, Bini EJ, Jacobs SL, et al. (2004) Colorectal polyps andcancers in asymptomatic average-risk patients: evaluation with CTcolonography. Radiology 230:629–636

25. Mulhall BP, Veerappan GR, Jackson JL (2005) Meta-analysis:computed tomographic colonography. Ann Intern Med 142:635–650

26. Iannaccone R, Laghi A, Catalano C, et al. (2003) Feasibility ofultra-low-dose multislice CT colonography for the detection ofcolorectal lesions: preliminary experience. Eur Radiol 13:1297–1302

27. Van Gelder RE, Nio CY, Florie J, et al. (2004) Computed tomo-graphic colonography compared with colonoscopy in patients atincreased risk for colorectal cancer. Gastroenterology 127:41–48

28. Yee J, Akerkar GA, Hung RK, et al. (2001) Colorectal neoplasia:performance characteristics of CT colonography for detection in300 patients. Radiology 219:685–692

29. Pineau BC, Paskett ED, Chen GJ, et al. (2003) Virtual colonoscopyusing oral contrast compared with colonoscopy for the detection ofpatients with colorectal polyps. Gastroenterology 125:304–310

30. Macari M, Bini EJ, Xue X, et al. (2002) Colorectal neoplasms:prospective comparison of thin-section low-dose multi-detector rowCT colonography and conventional colonoscopy for detection.Radiology 224:383–392

31. van Gelder RE, Florie J, Nio CY, et al. (2007) A comparison ofprimary two- and three-dimensional methods to review CT colo-nography. Eur Radiol 17:1181–1192

32. Pickhardt PJ (2003) Three-dimensional endoluminal CT colonog-raphy (virtual colonoscopy): comparison of three commerciallyavailable systems. AJR Am J Roentgenol 181:1599–1606

33. Macari M, Milano A, Lavelle M, et al. (2000) Comparison of time-efficient CT colonography with two- and three-dimensional colonicevaluation for detecting colorectal polyps. AJR Am J Roentgenol174:1543–1549

364 M. Juchems et al.: Electronic colon-cleansing for CT colonography