HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Serlie, I. W. O. Articles by Vos, F. M. Search for Related Content PubMed PubMed Citation Articles by Serlie, I. W. O. Articles by Vos, F. M. Social Bookmarking                      What's this? Hotlight (NEW!) What's Hotlight?

Lesion Conspicuity and Efficiency of CT Colonography with Electronic Cleansing Based on a Three-Material Transition Model

¹ Present address: Clinical Science and Advanced Development, Healthcare Informatics, Philips Medical Systems Nederland, Veenpluis 4-6, 5684 PC, Best, The Netherlands.
² Department of Biomedical Imaging, Technical University Eindhoven, Eindhoven, The Netherlands.
³ Quantitative Imaging Group, Department of Imaging Science and Technology, Faculty of Applied Sciences, Delft University of Technology, Delft, The Netherlands.
⁴ Department of Radiology, Academic Medical Center Amsterdam, Amsterdam, The Netherlands.

Received June 25, 2007; accepted after revision May 10, 2008.

 
Address correspondence to I. W. O. Serlie.

This study was supported by Philips Healthcare, both financially and by providing equipment. The authors had control of the data and information submitted for publication. I. W. O. Serlie and R. Truyen are employees of Philips Healthcare.

The data in this study were provided by J. Richard Choi, Virtual Colonoscopy Center, Walter Reed Army Medical Center.

Abstract

Top Abstract Introduction Materials and Methods Results Discussion References

 
OBJECTIVE. The purpose of this article is to report the effect on lesion conspicuity and the practical efficiency of electronic cleansing for CT colonography (CTC).

MATERIALS AND METHODS. Patients were included from the Walter Reed Army Medical Center public database. All patients had undergone extensive bowel preparation with fecal tagging. A primary 3D display method was used. For study I, the data consisted of all patients with polyps 6 mm. Two experienced CTC observers (observer 1 and observer 2) scored the lesion conspicuity considering supine and prone positions separately. For study II, data consisted of 19 randomly chosen patients from the database. The same observers evaluated the data before and after electronic cleansing. Evaluation time, assessment effort, and observer confidence were recorded.

RESULTS. In study I, there were 59 lesions partly or completely covered by tagged material (to be uncovered by electronic cleansing) and 70 lesions surrounded by air (no electronic cleansing required). The conspicuity did not differ significantly between lesions that were uncovered by electronic cleansing and lesions surrounded by air (observer 1, p < 0.5; observer 2, p < 0.6). In study II, the median evaluation time per patient after electronic cleansing was significantly shorter than for original data (observer 1, 20 reduced to 12 minutes; observer 2, 17 reduced to 12 minutes). Assessment effort was significantly smaller for both observers (p < 0.0000001), and observer confidence was significantly larger (observer 1, p < 0.007; observer 2, p < 0.0002) after electronic cleansing.

CONCLUSION. Lesions uncovered by electronic cleansing have comparable conspicuity with lesions surrounded by air. CTC with electronic cleansing sustains a shorter evaluation time, lower assessment effort, and larger observer confidence than without electronic cleansing.

Keywords: CT colonography • electronic cleansing • lesion conspicuity • material fractions • partial volume effect

Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

 
CT colonography (CTC) is a minimally invasive procedure that is advocated for polyp screening [1]. Residual fecal material and fluid are well known to hinder CTC evaluation, especially in a 3D display mode. Fecal material may cover lesions (preventing detection) or, conversely, mimic polyps (necessitating superfluous 2D verification and possibly reducing specificity). Fecal tagging was introduced to discriminate between tissue and fecal material [2, 3]. It permits a limited bowel preparation, which may contribute to better patient compliance.

Electronic cleansing aims at replacing tagged material (i.e., fecal remains and fluid) with air. This is needed for primary 3D evaluation of data in which there is a large amount of fecal material. Moreover, it may aid 3D problem solving in a primary 2D reading under these conditions.

Several electronic cleansing algorithms based on the increased attenuation of tagged material have been described previously [4–6]. Incomplete processing is still reported to leave artifacts [7]. A specifically noticeable problem is posed by the distracting "ridges" emanating from locations in which air, soft tissue, and tagged material meet [7]. An electronic cleansing algorithm was devised to improve the accuracy at these three-material junctions [8, 9]. The method adapts to patient-specific conditions, such as the local density of the tagged material, and is automated.

The purpose of this study was to assess the electronic cleansing algorithm's effect on lesion conspicuity as well as its practical efficiency. We hypothesize that the studied electronic cleansing method does not affect the conspicuity of lesions and facilitates efficient evaluation regarding time, assessment effort, and observer confidence.

Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

 
Patient Population
Patients were included from the Walter Reed Army Medical Center (WRAMC) public database, made available by the National Cancer Institute [10, 11]. Study I explored the effect of electronic cleansing on lesion conspicuity. All patients were included whose results contained polyps 6 mm in the largest diameter (measured during colonoscopy), irrespective of shape or location (patient group 1). Study II investigated the effect of electronic cleansing on efficiency. Ten randomly selected patients were taken from patient group 1 and nine patients were randomly selected from the public database who did not have polyps (patient group 2) (Table 1). No patients were excluded a priori (apart from those already excluded by WRAMC).

CT Colonography
Patients had undergone standard 24-hour colonic preparation with the oral administration of 90 mL of sodium phosphate (Fleet 1, Fleet Pharmaceuticals) and 10 mg of bisacodyl. Patients consumed 500 mL of barium (2.1% by weight) (Scan C, Lafayette Pharmaceuticals) and 120 mL of diatrizoate meglumine and diatrizoate sodium solution (Gastro grafin, Bracco Diagnostics). Distention was achieved through patient-controlled insufflation of room air.

Colonoscopy
CTC findings were disclosed during colo noscopy using segmental unblinding (at WRAMC). The colonoscopy findings were avail able via a report that included photographs, size measurement, morphology (pedunculated, sessile, and flat), and location (cecum, ascending colon, hepatic flexure, transverse colon, splenic flexure, descending colon, sigmoid, and rectum).

Electronic Cleansing
The electronic cleansing method assumes that the measured density in a voxel arises because of a combination of three materials: soft tissue, air, and tagged material. Initially, the percentage of materials in each voxel is determined. Subsequently, the partial volume of tagged material is replaced by air and the new density is calculated. Finally, a 3D method visualizes the colon from an endoluminal perspective as if there were no fecal remains. Appendix 1 contains an intuitive des cription of the algorithm; an exact (mathematic) explanation has been previously published [8, 9].

Observers
An abdominal radiologist (observer 1) and a research fellow (observer 2) were involved in reading all data for both studies. Observer 1 had previous experience of approximately 1,200 colonoscopy-verified CTC examinations at the start of the study. Observer 2 had previous experience of 350 such examinations. The interval between the two studies was 4 months, during which both observers evaluated approximately 100 additional patients using CTC. This setup was chosen to avoid observer bias (the positive patients from study II are also in study I). Both observers were familiar with pitfalls using primary 3D evaluation with stool tagging and electronic cleansing as described previously [7].

Study I: Image Review
The conspicuity study focused on all polyps from patient group 1 (study I). The prone and supine positions were considered as separate findings. Each finding was assessed by a researcher who had a background in designing electronic cleansing algorithms. This researcher determined whether the finding completely resided in air, was partly covered with tagged material, or was fully covered with tagged material.

Each polyp was marked and presented to the observers on a 3D display. An enhanced 3D dis play (unfolded cube display; ViewForum, Philips Healthcare) was chosen to measure the conspicuity of a polyp without having to turn the camera [12]. The polyps partly or fully covered by fecal mater ial were presented in 3D only after processing by electronic cleansing. The polyps residing in air (not requiring electronic cleansing) were presented directly as such. The cases were presented in random order. The observers were not informed whether the polyp had been uncovered by elec tronic cleansing or not.

The viewing position along the colon's centerline was controlled by the observer. Two-dimensional reformatted views of the original CT data were initialized by clicking on a position in the 3D endoluminal images. The window width and level default of the reformats was width, 1,250 HU; level, –50 HU; however, this could be freely adapted by the observers.

Study I: Assessments
Initially, the observers indicated on a 5-point Likert scale for each case whether it was suspected that the polyp completely resided in air or was uncovered by electronic cleansing. On this scale, 1 corresponded with a polyp definitely considered not uncovered by electronic cleansing and 5 with a polyp strongly suspected to be uncovered. The observers performed this rating based on the 3D display only.

View larger version (86K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 1A —Interface for CT colonography evaluation in 57-year-old man (WRAMC 393). Supine (A) and prone (B) enhanced 3D display images from part 1, in which 3D cine loops (using unfolded cube display; ViewForum, Philips Healthcare) were examined. Note that views are linked.

View larger version (81K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 1B —Interface for CT colonography evaluation in 57-year-old man (WRAMC 393). Supine (A) and prone (B) enhanced 3D display images from part 1, in which 3D cine loops (using unfolded cube display; ViewForum, Philips Healthcare) were examined. Note that views are linked.

View larger version (109K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 1C —Interface for CT colonography evaluation in 57-year-old man (WRAMC 393). Axial CT images from part 2, in which original axial slices were examined to verify whether surface parts were obscured by tagged material while tracking colon's centerline.

View larger version (116K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 1D —Interface for CT colonography evaluation in 57-year-old man (WRAMC 393). Axial CT images from part 2, in which original axial slices were examined to verify whether surface parts were obscured by tagged material while tracking colon's centerline.

The observers performed the conspicuity rating based on the 3D display only. After having provided the conspicuity ratings, each observer had to indicate for the cases scored as "inadequate" or "moderate" whether the electronic cleansing algorithm caused the difficulty or if there was another cause: inappropriate electronic cleansing or some other reason (e.g., CT artifact or a flat lesion). The observers were aware that electronic cleansing algorithms may leave artifacts, specifically clouds of debris and lines of ridges at junctions. The first item was selected if such an electronic cleansing artifact was particularly disturbing, for example, a smoothened polyp surface because of image processing or visible ridges emanating from junctions. In the latter case, the observer was asked to indicate the reason for the difficulty. Both the enhanced 3D display with electronic cleansing and 2D reformats of original data were available to assess the cause of low conspicuity.

Study I: Outcome Parameters and Statistical Analysis
Uncovered by electronic cleansing assessment—The polyps residing in air were compared with those partly or completely covered by tagged material. The difference in the consideration whether a polyp was "uncovered" was tested by means of the Mann-Whitney test (Wilcoxon's rank sum test). A p value 0.05 was considered to indicate a statistically significant difference.

Conspicuity—Differences in conspicuity between polyps were also statistically tested by means of the Mann-Whitney test. The outcomes were stratified by observer, polyp size, and polyp environment (i.e., residing in air or partly or fully covered by tagged material). Two-sided tests were used. Again, a p value 0.05 corresponded to a significant difference.

Study II: Image Review
The data not processed by electronic cleansing from patient group 2 (study II) were reviewed as follows. In part 1, 3D cine loops (using the unfolded cube display, ViewForum) were examined in the prone and supine positions [12]. The two positions were electronically linked. After clicking in one unfolded cube image, the corresponding unfolded cube image of the other position was displayed [13]. The frame rate was controlled by the observers. Two-dimensional reformatted views of the original CT data became accessible by clicking on a position in the 3D endoluminal images. Lesion size was measured by means of electronic calipers on the 3D view. In part 2, after reviewing the unfolded cube images, the original axial CT slices were inspected to verify whether surface parts were obscured by tagged material. The window width and level setting of the reformats and axial CT slices was by default width, 1,250 HU; level, –50 HU; however, this could be freely adapted by the observers. The data processed by electronic cleansing were reviewed similarly.

View larger version (146K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 2A —Walter Reed patient with fluid surface bordering on soft tissue. Image shows 3D display rendered without electronic cleansing.

View larger version (145K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 2B —Walter Reed patient with fluid surface bordering on soft tissue. Image shows result as electronic cleansing is applied only to "pure" materials and two material transitions.

View larger version (142K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 2C —Walter Reed patient with fluid surface bordering on soft tissue. Image shows result after applying electronic cleansing to three-material voxels as well (indicated by shaded band (arrow) in image B).

Study II: Assessments
Observer 1 and observer 2 independently evaluated the original data and the electronically cleansed data from patient group 2. To maintain objectivity, both were blinded to findings during colonoscopy, findings by themselves for the same patient (without and with electronic cleansing), and each other's findings. In addition, they were unaware of the prevalence of polyps. The evaluations using the original data preceded those with the electronically cleansed data. The interval period between evaluating original and electronically cleansed data of the same patient was at least 4 weeks to avoid biased results. The cases were presented in random order. Each surmised polyp was scored with respect to size, morphology, location (colon segment), and lesion confidence. The lesion confidence was qualified on a 5-point Likert scale: zero, not a lesion; and 4, absolute confidence of a lesion. The zero score was used to annotate a potential polyp initially detected through the 3D display but subsequently discarded as "not a lesion" after consulting the 2D reformats or axial CT slices. The observers were aware that in a clinical setting a confidence of 2 or more would indicate a relevant lesion.

The observers rated the assessment effort per colon segment on a 4-point Likert scale: 1, extremely easy; 2, good; 3, difficult; 4, extremely difficult. In addition, they rated their confidence in the reading per patient on a 3-point Likert scale: 1, confident; 2, in doubt; 3, extremely doubtful. The assessment effort and confidence in the reading were recorded after complete evaluation of a patient.

Study II: Reference Standard
Lesions detected during CTC evaluation were matched with the findings in the colonoscopy report (as previously discussed). This was done by a re search nurse (with experience of > 500 such match ings) under the supervision of a research fellow (with experience of > 350 CTC examinations). Neither was involved in reading the CTC studies.

A lesion detected during CTC in this study was considered true-positive if it matched a colonoscopy finding with regard to size, morphology, and loc ation (i.e., same or adjacent segment). A deviation in size by at most 5 mm was accepted to accommodate the inherent inaccuracy of colonoscopic size measurement. It should be noted that in this study the matching strategy from a previous study [14] is adopted, which differs slightly from some others [10, 11]. A false-negative finding was defined as a polyp detected during colonoscopy that did not match any CTC finding. A CTC finding with a lesion confidence of at least 2, a dia meter 6 mm, and not matching a polyp detected with colonoscopy was considered to be false-positive. All detections with a diameter < 6 mm or scored with an obser ver confidence lower than 2 were not considered relevant and discarded.

Study II: Outcome Parameters and Statistical Analysis
Processed volume fraction and processed surface fraction—In this study, the terms "processed volume fraction" and "processed sur face fraction" are intro duced to assess the extent to which electronic cleansing had an effect. The colon volume not processed by electronic cleansing was identified by thresholding the original data (before electronic cleansing) at –650 HU. All voxels with a density below –650 HU after electronic cleansing (notice that electronic cleansing replaced tagged material with air) approximate the colon volume after processing. The "processed volume" is constituted by all those voxels that are part of the colon volume after electronic cleansing and not part of the colon volume before electronic cleansing. The mean processed volume fraction is defined as the average processed volume divided by the average colon volume after electronic cleansing. It is taken to re present the fraction of additionally exposed volume.

View larger version (10K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3A —Histograms of polyp conspicuity for both observers stratified by polyp size. Light gray indicates air and dark gray indicates tagged material. Histograms for observer 1, all data (A); observer 2, all data (B); observer 1, size < 10 mm (C); observer 2, size < 10 mm (D); observer 1, size 10 mm (E); and observer 2, size 10 mm (F).

View larger version (10K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3B —Histograms of polyp conspicuity for both observers stratified by polyp size. Light gray indicates air and dark gray indicates tagged material. Histograms for observer 1, all data (A); observer 2, all data (B); observer 1, size < 10 mm (C); observer 2, size < 10 mm (D); observer 1, size 10 mm (E); and observer 2, size 10 mm (F).

View larger version (10K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3C —Histograms of polyp conspicuity for both observers stratified by polyp size. Light gray indicates air and dark gray indicates tagged material. Histograms for observer 1, all data (A); observer 2, all data (B); observer 1, size < 10 mm (C); observer 2, size < 10 mm (D); observer 1, size 10 mm (E); and observer 2, size 10 mm (F).

View larger version (10K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3D —Histograms of polyp conspicuity for both observers stratified by polyp size. Light gray indicates air and dark gray indicates tagged material. Histograms for observer 1, all data (A); observer 2, all data (B); observer 1, size < 10 mm (C); observer 2, size < 10 mm (D); observer 1, size 10 mm (E); and observer 2, size 10 mm (F).

View larger version (10K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3E —Histograms of polyp conspicuity for both observers stratified by polyp size. Light gray indicates air and dark gray indicates tagged material. Histograms for observer 1, all data (A); observer 2, all data (B); observer 1, size < 10 mm (C); observer 2, size < 10 mm (D); observer 1, size 10 mm (E); and observer 2, size 10 mm (F).

View larger version (10K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 3F —Histograms of polyp conspicuity for both observers stratified by polyp size. Light gray indicates air and dark gray indicates tagged material. Histograms for observer 1, all data (A); observer 2, all data (B); observer 1, size < 10 mm (C); observer 2, size < 10 mm (D); observer 1, size 10 mm (E); and observer 2, size 10 mm (F).

View larger version (10K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 4A —Histograms of polyp conspicuity for both observers stratified by polyp environment (completely surrounded by air [white], fully submerged [dark gray], or partly submerged [light gray]). Histograms for observer 1, all data (A) and observer 2, all data (B).

View larger version (10K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 4B —Histograms of polyp conspicuity for both observers stratified by polyp environment (completely surrounded by air [white], fully submerged [dark gray], or partly submerged [light gray]). Histograms for observer 1, all data (A) and observer 2, all data (B).

All evaluation times for the original data were compared with the corresponding times for the elec tronically cleansed data using the Wilcoxon's matched pairs signed rank test. Differences were con sidered significant at p 0.05. The initialization time (e.g., loading the data) was disregarded in the analysis.

Assessment effort and observer confidence— Differences in assessment effort and observer confidence between the original data and the electronically cleansed data were statistically tested per segment by means of the Wilcoxon's matched pairs signed rank test. Two-tailed p values were used and p 0.05 was considered statistically significant.

Sensitivity and specificity—We do not consider a statistical comparison of sensitivity and specificity between the two methods on the basis of the set of 19 patients possible. Consequently, the measures of sensitivity and specificity by themselves merely serve as descrip tive statistics for polyps 6 mm. The sensitivity of both display methods was determined on a per-polyp basis. The specificity of both display methods was determined by the false-positive rate.

Results

Top Abstract Introduction Materials and Methods Results Discussion References

 
Study I
Four patients from the WRAMC database (WRAMC 26, 98, 220, and 256) were excluded because of incomplete data. In the remaining data, there were 36 polyps between 6 and 10 mm in largest diameter in 32 patients and 29 polyps equal to or larger than 10 mm in 24 patients, amounting to 65 polyps in total. There were 129 polyp findings (regarding supine and prone as separate cases); one polyp remained retrospectively invisible in one position (WRAMC 213p). Eighteen of 129 findings (14%) concerned polyps completely covered and 41 of 129 findings (32%) related to polyps partly covered by tagged material (all uncovered by electronic cleansing); 70 of 129 findings (54%) referred to polyps completely surrounded by air (no electronic cleansing action needed). The median age of the patients was 57 years (age range, 48–76 years); 36 were men, and 15 were women.

Uncovered by electronic cleansing assessment—Both observers rated polyps residing in air significantly lower on the scale reflecting whether they suspected that a polyp partly or fully resided in tagged material (p < 0.05). The readers reported a slight noise suppression by electronic cleansing that was apparent by a slightly smoothened surface (Fig 2A, 2B, 2C).

Conspicuity—Both observers rated the conspicuity of all polyps residing in air as not significantly different from all those partly or fully residing in tagged material (p = 0.5 for observer 1 and p = 0.6 for observer 2). None of the substratifications yielded a significant difference. The numeric data, that is, the distribution of the ratings on conspicuity, are shown in the graphs in Figures 3A, 3B, 3C, 3D, 3E, 3F and 4A, 4B. These figures contain the normalized histograms of the conspicuity stratified by polyp size and polyp environment, respectively. Figure 5 shows examples of 3D views.

View larger version (102K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 5 —Examples of polyps (arrows) with varying conspicuity before (first two columns and last column) and after (third column) electronic cleansing from 52-year-old man with rating of excellent conspicuity (top row), 58-year-old man with rating of good conspicuity (second row), 53-year-old woman with rating of average conspicuity (third row), 53-year-old woman with rating of questionable conspicuity (fourth row), 54-year-old woman with rating of inadequate conspicuity (bottom row). Rating of conspicuity was given by both observers to electronically cleansed data. Circles indicate approximate polyp location. Asterisks and calipers indicate orientation of volume.

Study II
One patient (WRAMC 32) was excluded from the study because of inadequate diagnostic quality: both observers considered the colonic distention in the prone and supine positions insufficient for evaluation. In the remaining data, there were nine polyps between 6 and 10 mm in diameter in seven of 19 patients and three polyps equal to or greater than 10 mm in three of 19 patients, amounting to 12 polyps in total. The median age of the patients was 55 years (age range, 50–78 years); 13 were men, and six were women.

Processed surface fraction and processed volume fraction—The mean processed surface fraction was 269,640 of 975,378 voxels = 0.27 (range, 0.16–0.46) and the mean processed volume fraction was 0.36 of 1.84 L = 0.20 (range, 0.13–0.45). Figure 6A, 6B, 6C, 6D illustrates the effect of electronic cleansing for the cases with the largest and smallest processed volume fractions, respectively.

View larger version (40K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 6A —Visualization of the segmented colonic volume before and after electronic cleansing. Images from patient with highest processed volume fraction in both scans (WRAMC 9) show segment ed colonic volume before (A) and after (B) electronic cleansing (processed volume fraction: supine, 0.44 L / 0.98 L = 0.45; prone, 0.44 L / 1.33 L = 0.33).

View larger version (51K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 6B —Visualization of the segmented colonic volume before and after electronic cleansing. Images from patient with highest processed volume fraction in both scans (WRAMC 9) show segment ed colonic volume before (A) and after (B) electronic cleansing (processed volume fraction: supine, 0.44 L / 0.98 L = 0.45; prone, 0.44 L / 1.33 L = 0.33).

View larger version (56K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 6C —Visualization of the segmented colonic volume before and after electronic cleansing. Images from patient with lowest processed volume fraction in both scans (WRAMC 408) show segmented colonic volume before (C) and after (D) electronic cleansing (processed volume fraction: supine, 0.31 L / 2.33 L = 0.13; prone, 0.30 L / 2.33 L = 0.13).

View larger version (56K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 6D —Visualization of the segmented colonic volume before and after electronic cleansing. Images from patient with lowest processed volume fraction in both scans (WRAMC 408) show segmented colonic volume before (C) and after (D) electronic cleansing (processed volume fraction: supine, 0.31 L / 2.33 L = 0.13; prone, 0.30 L / 2.33 L = 0.13).

Assessment effort and observer confidence—Both observers rated the assessment effort for inspecting the original data significantly larger than for the electronically cleansed data in each segment. The sum of the signed ranks for observer 1 was 4,095 and for observer 2 was 2,045.5 (for both p < 0.0000001) (Fig. 7A, 7B, 7C, 7D). The observer confidence over all patients was rated significantly smaller for the original data than for the electronically cleansed data. The sum of the signed ranks in this respect for observer 1 was 91 (p < 0.007) and for observer 2 was 120 (p < 0.0002) (Fig. 7A, 7B, 7C, 7D). Both reviewers reported a "cloud of debris" artifact in one location emanating from acquisition artifacts. It was not considered to preclude polyp detection. No distracting lines or ridges at junctions were encountered.

View larger version (8K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 7A —Histograms of observer effort and confidence level. Original data = light gray, electronically cleansed data = dark gray. Histograms of assessment effort of CT colonography inspection for observer 1 (A) and observer 2 (B) over all segments and patients.

View larger version (9K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 7B —Histograms of observer effort and confidence level. Original data = light gray, electronically cleansed data = dark gray. Histograms of assessment effort of CT colonography inspection for observer 1 (A) and observer 2 (B) over all segments and patients.

View larger version (8K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 7C —Histograms of observer effort and confidence level. Original data = light gray, electronically cleansed data = dark gray. Histograms of observer confidence in reading using original and electronically cleansed data for observer 1 (C) and observer 2 (D) over all patients.

View larger version (8K): [in this window] [in a new window] [as a PowerPoint slide]   Fig. 7D —Histograms of observer effort and confidence level. Original data = light gray, electronically cleansed data = dark gray. Histograms of observer confidence in reading using original and electronically cleansed data for observer 1 (C) and observer 2 (D) over all patients.

Observer 1 had, in total, five false-positive detections in the original data and four false-positive detections in the electronically cleansed data. Observer 2 had, in total, two false-positive detections in the original data and four in the electronically cleansed data.

Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

 
In study I, no significant difference was found in the conspicuity of polyps that were uncovered by electronic cleansing and polyps surrounded by air. Study II showed that electronic cleansing enabled comprehensive visibility (mean processed surface fraction, 27%) and time-efficient inspection (approximately 12 minutes per patient, saving 40% in inspection time). Moreover, the observers indicated a low assessment effort and a high confidence while reading the electronically cleansed data. For polyps 6 mm, the sensitivity was high (10/12) and the false-positive rate was low: four false-positive findings, in total, in 18 patients.

Several previous articles have described technical innovations regarding electronic cleansing [4–7]. Zalis et al. [4] indicated that it is essential to ascertain how artifacts still present in successfully tagged electronically cleansed data affect observer performance. Pickhardt and Choi [7] identified a reading pitfall caused by an artifact at junctions, that is, locations where air–fluid levels interface with the colon wall. The proposed electronic cleansing method aims at improving the accuracy at such locations. We found that 41 of 129 findings related to polyps partly covered by fecal material, signifying the importance of accurate electronic cleansing in these circumstances. We hypothesize that the insignificant difference of polyp conspicuity (study I), short average 3D evaluation time, high confidence, low assessment effort, and high specificity (study II) indicate that electronic cleansing did not result in artifacts that complicated the reading. The reviewers encountered one artifact in study II, but that was not caused by electronic cleansing.

In study I, it was found that the reviewers could identify whether a lesion was uncovered by electronic cleansing through a slight smoothing of the colon surface. Although smoothing could lead to missing subtle (flat) lesions, we did not experience this in the study. The conspicuity of uncovered polyps did not differ significantly from polyps completely surrounded by air.

It is a complex problem to measure the performance of an electronic cleansing algorithm. Effectively, one should determine the extent to which the algorithm modifies the polyp shape so that it is no longer detected. Study I compared the conspicuity of electronically cleansed polyps and those surrounded by air (not affected by the algorithm). If the electronic cleansing algorithm would change the polyp shape in a destructive manner, this might lead to different distributions of conspicuity rating. In other words, electronically cleansed polyps could become less "conspicuous." We did not find an indication of such an effect.

We did not find previous results on the effectiveness of electronic cleansing regarding processed surface fraction or processed volume fraction, as in study II. A related finding on surface visibility is that, on average, 99.5% of the colon surface becomes visible using the unfolded cube imaging sequences and 95% using forward- and backward-looking imaging sequences [12]. The average processed surface fraction (0.27) by far exceeds the missed parts in those mentioned display modes. Both the display mode and electronic cleansing increase surface visibility such that electronic cleansing should be used with an enhanced 3D display to best benefit from the added surface by electronic cleansing.

The measured reading time per patient (12 minutes) in study II slightly improves on previous work [14]. In that study, a reading time of approximately 14 minutes was measured, with an enhanced 3D display technique. However, fecal tagging was not applied, with the result that additional reading of axial 2D slices for verification of obscured parts was not possible. Previous articles applying 3D displays report evaluation times of 12 [16] and 16 minutes [17].

A limitation of the entire study is that results may not be extrapolated to patients undergoing a limited bowel preparation. It might be expected that a rigorous preparation simplifies electronic cleansing somewhat because there may be less tagged material than in a limited purgation scheme. Moreover, such remains may have a more homogeneous appearance. Still, the complex nature of the problem prompted us to initially study the algorithm's performance regarding extensively prepared patients. It should be noted that we found a considerable processed volume fraction (0.20) despite the extensive preparation. The processed volume and surface fractions might depend on the nature of the preparation (wet or dry). Probably any electronic cleansing algorithm will perform optimally with homogeneously tagged stool [18]. Accordingly, we expect the algorithm to be better suited for removing pools of fluid compared with fecal residue because the former will present a more homogeneous density.

A limitation of study II is the small study population. The sample size of the patient population was calculated to be sufficiently large to meet the primary aim of comparing the original data reading versus electronically cleansed data reading regarding time efficiency, surface visibility, assessment effort, and observer confidence. The reported sensitivity is in the same range as reported in another study [14] in which a similar display method was used. As indicated, the findings on sensitivity and specificity should merely be regarded as descriptive statistics because of the small study population.

We conclude that lesions uncovered by electronic cleansing had similar conspicuity in the 3D display as lesions already exposed to air in extensively prepared patients. In such a population, the electronic cleansing algorithm led to a shorter evaluation time, lower assessment effort, and greater observer confidence than CTC without electronic cleansing. The algorithm may contribute to a practical evaluation strategy involving either a primary 3D display or 3D problem solving in a primary 2D display.

References

Top Abstract Introduction Materials and Methods Results Discussion References

 

1. Vining DJ, Gelfand DW, Bechtold RE, Scharling ES, Grishaw EK, Shifrin RY. Technical feasibility of colon imaging with helical CT and virtual reality. (abstr) AJR 1994;162 [suppl]:104
2. Callstrom MR, Johnson CD, Fletcher JG, et al. CT colonography without cathartic preparation: feasibility study. Radiology 2001;219 : 693–698[Abstract/Free Full Text]
3. Lefere PA, Gryspeerdt SS, Dewyspelaere J, Baekelandt M, Van Holsbeeck BG. Dietary fecal tagging as a cleansing method before CT colonography: initial results—polyp detection and patient acceptance. Radiology 2002;224 : 393–403[Abstract/Free Full Text]
4. Zalis ME, Perumpillichira JJ, Hahn PF. Digital subtraction bowel cleansing for CT colonography using morphological and linear filtration methods. IEEE Trans Med Imaging 2004;23 :1335 –1343[CrossRef][Medline]
5. Chen D, Liang Z, Wax MR, Li L, Li B, Kaufman AE. A novel approach to extract colon lumen from CT images for virtual colonoscopy. IEEE Trans Med Imaging 2000; 19:1220 –1226[CrossRef][Medline]
6. Lakare S, Wan M, Sata M, Kaufman A. 3D digital cleansing using segmentation rays. (abstr) In: Ertl T, Hamann B, Varshney A, eds. IEEE Visualization 2000 (Proceedings of the 2002 IEEE International Conference). New York, NY: ACM Press, 2002:37 –44
7. Pickhardt PJ, Choi JHR. Electronic cleansing and stool tagging in CT colonography: advantages and pitfalls with primary three-dimensional evaluation. AJR 2003;181 : 799–805[Free Full Text]
8. Serlie IWO, Truyen R, Florie J, Post FH, van Vliet LJ, Vos FM. Computed cleansing for virtual colonoscopy using a three-material transition model, MICCAI 2003. Lecture Notes in Computer Science, vol. 2879:175 –183
9. Serlie IWO, Vos FM, Truyen R, Post FH, van Vliet LJ, Classifying CT image data into material fractions by a scale and rotation invariant edge model. IEEE Trans Image Process 2007;16 :2891 –2904[CrossRef][Medline]
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[Abstract/Free Full Text]
11. Walter Reed Army Medical Center, Virtual Colonoscopy Center Website. CT colonography in early detection of colorectal cancer. wramc.vcscreen.com/training/index.html. Accessed March 1, 2006
12. Vos FM, van Gelder RE, Serlie IWO, et al. Three-dimensional display modes for CT colonography: conventional 3D virtual colonoscopy versus unfolded cube projection. Radiology 2003;228 : 878–885[Abstract/Free Full Text]
13. de Vries AH, Truyen R, van der Peijl J, et al. Feasibility of automated matching of supine and prone CT-colonography examinations. Br J Radiol 2006;79 : 740–744[Abstract/Free Full Text]
14. van Gelder RE, Nio CY, Florie J, et al. Computed tomographic colonography compared with colonoscopy in patients at increased risk for colorectal cancer. Gastroenterology 2004;127 : 41–48[CrossRef]
15. Serlie IWO. WRAMC polyps. Delft Technical University Website. www.ist.tudelft.nl/qi. Accessed July 11, 2008
16. Macari M, Milano A, Lavelle M, Berman P, Megibow AJ. Comparison of time-efficient CT colonography with two- and three-dimensional colonic evaluation for detecting colorectal polyps. AJR2000; 174:1543 –1549[Abstract/Free Full Text]
17. Neri E, Vannozzi F, Vagli P, Bardine A, Bartolozzi C. Time efficiency of CT colonography: 2D vs 3D visualization. Comp Med Imag Graph 2006; 30:175 –180[CrossRef][Medline]
18. Zalis ME, Perumpillichira JJ, Magee C, Kohlberg G, Hahn PF. Tagging-based, electronically cleansed CT colonography: evaluation of patient comfort and image readability. Radiology2006; 239:149 –159[Abstract/Free Full Text]

CiteULike

Complore

Connotea

Del.icio.us

Digg

Reddit

Technorati

This Article Abstract Figures Only Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Serlie, I. W. O. Articles by Vos, F. M. Search for Related Content PubMed PubMed Citation Articles by Serlie, I. W. O. Articles by Vos, F. M. Social Bookmarking                      What's this? Hotlight (NEW!) What's Hotlight?