DOI: 10.1148/radiol.2443060934
(Radiology 2007;244:852-864.)
© RSNA, 2007
Two- versus Three-dimensional Colon Evaluation with Recently Developed Virtual Dissection Software for CT Colonography1
Se Hyung Kim, MD,
Jeong Min Lee, MD,
Hyo Won Eun, MD,
Min Woo Lee, MD,
Joon Koo Han, MD,
Jae Young Lee, MD, and
Byung Ihn Choi, MD
1 From the Department of Radiology (S.H.K., J.M.L., M.W.L., J.K.H., J.Y.L., B.I.C.) and Institute of Radiation Medicine (J.M.L., J.K.H., B.I.C.), Seoul National University Hospital, Seoul National University College of Medicine, 28 Yongon-dong, Chongno-gu, Seoul 110-744, Korea; and Ewha Woman's University Hospital, Seoul, Korea (H.W.E.). Received May 29, 2006; revision requested August 1; revision received September 8; accepted October 12; final version accepted December 8. Supported by grant 0412-M100-0401-0007 from the Korea Health 21 R&D project, Ministry of Health & Welfare, Republic of Korea.
Address correspondence to J.M.L. (e-mail: leejm{at}radcom.snu.ac.kr).
 |
ABSTRACT
|
|---|
This retrospective study was institutional review board approved; the requirement for informed patient consent was waived. The purpose of this study was to retrospectively compare a two-dimensional (2D) data interpretation technique with a three-dimensional (3D) colon dissection technique in terms of interpretation time and sensitivity for colonic polyp detection, with colonoscopy as the reference standard. Ninety-six patients (56 men, 40 women; mean age, 54.8 years) underwent colonoscopy and multidetector computed tomographic (CT) colonography on the same day. Two radiologists independently analyzed the data on a per-polyp and per-patient basis. The sensitivity of both approaches was compared by using the McNemar test. The time required to interpret CT colonographic data with each technique was also assessed. Compared with the conventional 2D colonic polyp detection method, primary 3D interpretation with use of virtual dissection software for CT colonography revealed comparable per-polyp (77% and 69% for two readers) and per-patient (77% and 73% for two readers) sensitivities and comparable per-patient specificity (99% and 89% for two readers) for the detection of polyps 6 mm in diameter or larger and involved a shorter interpretation time.
© RSNA, 2007
|
INTRODUCTION
|
|---|
Offsetting the promising results of the clinical practicality of computed tomographic (CT) colonography are technical issues related to the mechanics of interpreting the data. There are two strategies for interpreting CT colonographic data sets: a primary two-dimensional (2D) approach and a primary three-dimensional (3D) approach. A recent study (1) revealed that most (80%) experienced interpreters use the primary 2D approach, reserving the use of 3D endoluminal views for problem solving. Pickhardt et al (2) found the performance of CT colonography encouraging when they used a primary 3D approach to detect polyps, and a growing minority of radiologists now prefer this method (3–5). Nevertheless, the 3D approach, a virtual imitation of colonoscopy, has limitations in that it does not enable the inspection of "blind spots," such as the parts of the mucosa that are hidden by colon folds. This is true even when four fly-through passes of the colon are performed with a markedly increased interpretation time, unless the 3D workstation includes a special function to display the blind areas to the reviewer after bidirectional 3D navigation is performed (6–8). Therefore, a major limitation of conventional 3D endoluminal techniques is the time-consuming data interpretation process.
To overcome the current limitations of 3D endoluminal imaging, 3D panoramic display techniques have been developed as an effective way to inspect the inner colonic surface by virtually unfolding and dissecting the colon along its longitudinal axis (9–14). However, Hoppe et al (11) found that the 3D virtual colon dissection method had lower sensitivity for polyp detection and was more time-consuming than axial interpretation.
As an extrapolation of the techniques presented by Hoppe et al (11), 3D colon dissection software (Perspective Filet View; Philips Medical Systems, Cleveland, Ohio) that prevents the blind spots that are sometimes created when all of the projection rays are perpendicular to the centerline (eye point) has been developed (15). This colon dissection software not only unrolls the colon but also bends the projection rays from the eye point to show both sides of the folds and the areas between tight folds. By bending the projection rays as a function of the distance from the central location, the view adds perspective to the previously flat dissection view. This phenomenon enables the user to move through the colon and see inside traditional blind spots behind the folds and thereby makes visualization of the entire colonic mucosa possible in a unidirectional navigation. Therefore, the purpose of our study was to retrospectively compare a conventional 2D interpretation technique with a recently developed 3D virtual colon dissection technique in terms of interpretation time and sensitivity for colonic polyp detection, with colonoscopy as the reference standard.
|
MATERIALS AND METHODS
|
|---|
Patients
Between August 2004 and February 2005, a total of 232 candidate subjects were recruited to participate in a CT colonography study for colorectal cancer screening. That study was conducted with institutional review board approval and informed patient consent that allowed future retrospective analysis of patient data and images. Thirty-two of the 232 candidates were excluded from the study for the following reasons: prior colorectal surgery, inflammatory bowel disease, iron deficiency–related anemia or positive fecal occult blood test results within the previous 6 months, age younger than 40 years, history of familial adenomatous polyposis, and/or history of polypectomy within the previous year. For our current study, we retrospectively collected 96 consecutive CT colonographic cases—of 96 patients—from the cases of the remaining 200 patients who underwent CT colonography. These cases were those of the first 96 of the 200 patients. Fifty-six men and 40 women (age range, 37–76 years; mean age, 54.8 years) were included. Our current retrospective study was institutional review board approved, and the requirement for additional informed patient consent was waived.
CT Colonography
Colon cleansing was performed by means of oral administration of 4 L of polyethylene glycol solution (Colyte-F; TaeJoon Pharmaceuticals, Seoul, Korea), which was tolerated by all candidates. All subjects underwent CT colonography and successive conventional colonoscopy within less than 2 hours of each other. CT colonography was performed by using a 16–detector row scanner (Sensation 16; Siemens Medical Systems, Forchheim, Germany). The colon was gently insufflated with approximately 2 L of room air according to the patient's maximal tolerance by using a mechanical insufflator. No spasmolytic agent was administered.
CT scanning parameters were as follows: 16 x 0.75-mm detector configuration, 2-mm section thickness, 12 mm/sec table feed, 0.5-second rotation time, 1-mm reconstruction interval, 512 x 512 matrix, 120 kVp, and 50 and 200 effective mAs for the prone and supine positions, respectively. The average scanning time required to cover the entire colon was 15–18 seconds for each position. Supine images were acquired after the intravenous administration of 150 mL of iopromide (Ultravist 370; Schering, Berlin, Germany) at a rate of 3 mL/sec after a 70-second delay. Contrast enhancement was generated especially to evaluate the extracolonic findings.
Conventional Colonoscopy and Histopathologic Analysis
In all patients, conscious sedation was induced during colonoscopy with an intravenous sedative (midazolam; Bukwang Pharmaceutical, Seoul, Korea). All colonoscopic examinations were performed by one of five board-certified gastroenterologists, each with 7–15 years of pertinent clinical experience, with use of a segmental unblinding technique (2). Polyps of any size that were detected at CT colonography but not at colonoscopy were subjected to segmental unblinding. Polyp size was estimated by using 10-mm-long biopsy forceps. The gastroenterologists measured the long axis of the base or head of the polyp, excluding the stalk if one was present.
The morphology of a given polyp was described as flat, sessile, or pedunculated. A flat polyp was defined as a shallow mucosal elevation with a height of less than one half its width (16,17). With the exception of some larger lesions, the flat polyps were generally 3 mm or less in height. Biopsy and subsequent histopathologic examination of all lesions found at colonoscopy were performed by one pathologist with 6 years of experience. These examinations were performed exclusively for this study, because in clinical practice, colonoscopists often ignore or cauterize diminutive polyps instead of obtaining histopathologic specimens of them.
Image Processing
During real-time initial CT colonographic data interpretation, one of two radiologists (with previous experience of 100 colonoscopy-proved CT colonographic examinations) interpreted the CT colonographic data by using the primary 2D method. For the primary 2D interpretation method used in both the initial and subsequent retrospective data interpretations, the colon was surveyed on transverse images of the entire volumetric CT data set at a window width and window level of 1500 and –200 HU, respectively, by using dedicated 3D visualization software (Rapidia; INFINITT, Seoul, Korea). When a suspected polyp was found on 2D transverse images, 2D coronal and sagittal images and 3D endoluminal views were evaluated to better characterize the lesion. To further characterize the lesion, the observers examined its internal attenuation by changing the window width and window level settings to 400 and 10 HU, respectively. Lesions with small air attenuation were considered to be stool. Lesion enhancement after contrast material administration increased the possibility that the lesion was a true polyp.
For the 3D interpretation method, the reconstructed supine and prone data sets were transferred to the personal computer (Extended Brilliance Workstation; Philips Medical Systems, Best, the Netherlands) installed with the Perspective Filet View colon dissection package. The main display of the 3D virtual dissection, which is presented in a panoramic way, represents an approximately 380° circular view (360° plus overlapping 20°) of the colon at any given point along an automatically calculated centerline through all colonic segments. This display shows the internal surface of the colon in a linear fashion—unfolded like a pathologic specimen, filmstrip, or roll of tape (18). Unlike with other dissection techniques, with this 3D virtual dissection technique, the 3D view of the dissected colon is viewed as if the viewing area is pushed across a tube, rounding the center of the viewing area (Fig 1). This allows one to see both the proximal and the distal sides of the colonic folds and therefore see around and between the folds of the colon without having to manipulate the image (Fig 1).
View larger version (30K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1a: Principles of 3D virtual colon dissection viewing. (a) The dissected colon is viewed as if the viewing area is pushed across a tube, rounding the center of the viewing area. (b) Top: As a result of this viewing algorithm, the colon dissection software allows the user to see both the retrograde and the antegrade sides of the fold. Bottom: In contrast, with other dissection methods, the view of the colon is flat and shows only the top of the folds. Consequently, these methods do not allow the user to see around and between the colon folds; thus, a lesion or polyp on a fold could be easily missed. (c) Generated supine (top) and prone (bottom) 3D virtual dissection images. Each 10° area at the top and bottom of the image is added to the full 360° view of the colonic surface and displayed with a transparently shaded color so as not to miss any lesion. The colonic surface can be observed quickly by using the top scroll bars. Note that the polyp (arrows) remains unchanged on both images, but the lump of feces (arrowheads) changes in position. A 5-mm adenomatous polyp was confirmed at both colonoscopy and biopsy.
|
|
View larger version (73K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1b: Principles of 3D virtual colon dissection viewing. (a) The dissected colon is viewed as if the viewing area is pushed across a tube, rounding the center of the viewing area. (b) Top: As a result of this viewing algorithm, the colon dissection software allows the user to see both the retrograde and the antegrade sides of the fold. Bottom: In contrast, with other dissection methods, the view of the colon is flat and shows only the top of the folds. Consequently, these methods do not allow the user to see around and between the colon folds; thus, a lesion or polyp on a fold could be easily missed. (c) Generated supine (top) and prone (bottom) 3D virtual dissection images. Each 10° area at the top and bottom of the image is added to the full 360° view of the colonic surface and displayed with a transparently shaded color so as not to miss any lesion. The colonic surface can be observed quickly by using the top scroll bars. Note that the polyp (arrows) remains unchanged on both images, but the lump of feces (arrowheads) changes in position. A 5-mm adenomatous polyp was confirmed at both colonoscopy and biopsy.
|
|
View larger version (61K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1c: Principles of 3D virtual colon dissection viewing. (a) The dissected colon is viewed as if the viewing area is pushed across a tube, rounding the center of the viewing area. (b) Top: As a result of this viewing algorithm, the colon dissection software allows the user to see both the retrograde and the antegrade sides of the fold. Bottom: In contrast, with other dissection methods, the view of the colon is flat and shows only the top of the folds. Consequently, these methods do not allow the user to see around and between the colon folds; thus, a lesion or polyp on a fold could be easily missed. (c) Generated supine (top) and prone (bottom) 3D virtual dissection images. Each 10° area at the top and bottom of the image is added to the full 360° view of the colonic surface and displayed with a transparently shaded color so as not to miss any lesion. The colonic surface can be observed quickly by using the top scroll bars. Note that the polyp (arrows) remains unchanged on both images, but the lump of feces (arrowheads) changes in position. A 5-mm adenomatous polyp was confirmed at both colonoscopy and biopsy.
|
|
The first step of the 3D data interpretation is to match the same position of the colon on each data set on both the longitudinal axis and the transverse axis so that the radiologist(s) can review both images as they pass by. This matching function is useful in enabling one to see both the supine and the prone 3D images simultaneously. It also helps in determining whether a lesion has moved during patient position changes. After the matching process, radiologists should synchronize the two data sets manually by clicking on the "Lock" button. Once the data sets are synchronized, radiologists should click on the "Play" button to interpret the findings seen through the entire colon of the two data sets. In most cases, only one matching at the beginning of the interpretation is sufficient to maintain the matching of the entire colon. However, in some cases—when anatomic landmarks are misarranged between the data sets or when there is a discrepancy between the two data sets due to deformation or change of the colonic arrangement during the patient's position change—radiologists will need to rematch the data sets. In our experience, two to three rematchings were needed per patient in such cases. When a lesion was found, 2D transverse or multiplanar reconstruction images could then be used as problem-solving tools.
Image and Statistical Analyses and the Reference Standard
Two abdominal radiologists (H.W.E., S.H.K., previous experience of 500 and 300 CT colonographic examinations, respectively) who did not participate in the initial CT colonographic data interpretations and who were blinded to the colonoscopy results and the initial CT colonographic data interpretations independently and retrospectively reviewed the CT colonographic images by using the two interpretation methods, with 2 months separating the two interpretations. The radiologists were asked to report polyps of all sizes found on the CT colonographic images. For half the examinations, the primary 2D interpretations preceded the 3D virtual dissection interpretations; for the other half of the examinations, the order of the interpretations was reversed.
The adequacy of bowel preparation and distention was evaluated according to the segment for each supine and prone CT colonographic image data set by using a four-point scale (19). Residual fluid was measured relative to the maximum anteroposterior diameter of the segment: A grade of 1 indicated no residual fluid; grade 2, a largest air-fluid level of less than 25% of the maximal anteroposterior diameter of the segment; grade 3, a largest air-fluid level of between 25% and 50% of the maximal anteroposterior diameter; and grade 4, a largest air-fluid level of greater than 50% of the maximal anteroposterior diameter. For residual feces, grade 1 indicated no feces; grade 2, fewer than five fecal lumps; grade 3, five to 10 fecal lumps; and grade 4, more than 10 fecal lumps. Only residual feces that were 6 mm or greater in diameter were considered in the assessment of bowel preparation quality because it is generally accepted that polypoid lesions smaller than 6 mm can be ignored at CT colonography. For colonic distention, grade 1 indicated distention of more than 75% of the expected maximal luminal dimension; grade 2, distention of 51%–75% of the expected maximal luminal dimension; grade 3, distention of 25%–50% of the expected maximal luminal dimension; and grade 4, distention of less than 25% of the expected maximal luminal dimension.
The colon was divided into six segments: the cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum. For per-polyp analysis, each radiologist recorded the presence of all lesions during each interpretation session. To measure the polyp size, 3D endoluminal measurements obtained by using electronic calipers were used. The radiologists were instructed to obtain measurements along the largest available linear dimension but to avoid including the stalks of pedunculated lesions. The sensitivities of each observer and each method were calculated, and the McNemar test was used to assess the statistical significance.
For per-patient analysis, after finishing the image interpretations for each patient, the radiologist recorded the presence of polyps in that particular subject. The per-patient sensitivities and specificities of each observer and each method were also calculated, and again the McNemar test was used to assess the statistical significance. Analysis based on standard polyp size categories of 5 mm or smaller, 6 mm or larger, 7 mm or larger, 8 mm or larger, 9 mm or larger, and 10 mm or larger was performed (20).
The number of false-positive (FP) polyps detected at each interpretation session by each reader was determined and compared between the two interpretation methods by using the paired Student t test. The readers further analyzed the causes of the clinically important false-negative lesions—those polyps 6 mm or larger that were missed during retrospective unblinded review by the same two radiologists (H.W.E., S.H.K.) in consensus.
The final colonoscopy results after segmental unblinding to the CT colonography results served as the reference standard in our study. Lesions detected during CT colonography were matched to the colonoscopy findings on the basis of the colonoscopy findings and the video registration performed by two other radiologists (J.M.L., J.Y.L.). A polyp detected at CT colonography was considered to be true-positive if it matched a finding at colonoscopy in terms of size, shape, colonic segment, and anatomic interrelation to the haustral folds. For a given lesion to be considered true-positive, it had to be within the same segment or in an adjacent colonic segment, the two recorded sizes had to be the same within a 30% margin of error, and the lesion had to have similar morphologic features (eg, flat, sessile, or pedunculated) at both examinations. In addition, the relationship between a lesion and thick colonic folds—that is, whether the polyp was on a fold or between folds—was also considered in determining a true-positive match. Polyps were categorized by size on the basis of the colonoscopic results, while FP lesions were categorized according to the size measured on the 3D CT colonographic images.
During the image-reading session for each patient, which was defined as the time from the start of the image evaluation at the 3D workstation to the completion of the interpretation for each detected lesion in each patient, the time was documented to compare the time efficiency of the two interpretation methods by using the Wilcoxon signed rank test. Statistical analyses were performed by using SPSS 2003, version 12.0.1, for Windows software (SPSS, Chicago, Ill). P < .05 was considered to indicate statistical significance.
|
RESULTS
|
|---|
Colonoscopic and Histopathologic Findings
A total of 134 polyps (70 flat, 53 sessile, 11 pedunculated) 2–30 mm in diameter were found in 55 (57%) of the 96 patients at conventional colonoscopy. Ninety-nine polyps were 5 mm or smaller, 23 were 6–9 mm, and 12 were 10 mm or larger (Table 1). There were nine advanced adenomas—seven of which were tubulovillous adenomas 6, 7, 8, 10, 10, 15, and 18 mm in diameter—and two villous adenomas 15 and 30 mm in diameter. There were no cancerous lesions.
Fifty-four patients (56%) underwent immediate segmental repeat endoscopic examination after the segmental CT colonography results were revealed. Thirteen (9.7%) additional polyps 2–8 mm in diameter were detected in eight (15%) of these 54 patients. Of these 13 polyps, seven (54%) were adenomatous and four (31%), in three patients, were 6 mm in diameter or larger.
Bowel Preparation and Distention
With regard to colonic preparation and distention, 916 (79.5%) of the 1152 colonic segments were assigned a score of 1 (n = 366, 31.8%) or 2 (n = 550, 47.7%) for residual fluid, 1135 (98.5%) segments were assigned a score of 1 (n = 1059, 91.9%) or 2 (n = 76, 6.6%) for residual feces, and 1081 (93.8%) segments were assigned a score of 1 (n = 901, 78.2%) or 2 (n = 180, 15.6%) for colonic distention.
Sensitivity of CT Colonography with Two Data Interpretation Methods
The sensitivities of both radiologists using each interpretation method to detect clinically relevant (ie, 
6 mm) polyps ranged from 63% (22 of 35 polyps) to 77% (27 of 35 polyps) (Table 2). On a per-polyp basis, the sensitivity of sessile morphology tended to be increased compared with that of flat morphology, and the sensitivity of pedunculated morphology tended to be increased compared with that of sessile morphology. On a per-polyp basis, the sensitivity for the detection of polyps of all sizes and morphologies did not significantly differ between the two interpretation methods for either reader (P > .05); nor did it differ significantly between the two readers with either method.
A flow chart of the study profile based on recommended standards for reporting diagnostic accuracy is presented in Figure 2 (21). Per-patient sensitivities and specificities for the detection of polyps 6 mm or larger ranged from 59% (13 of 22 patients) to 77% (17 of 22 patients) and from 89% (66 of 74 patients) to 99% (73 of 74 patients), respectively (Table 3). Differences in per-patient sensitivity and specificity between the two interpretation methods, regardless of polyp size category, were not significant for either reader (P > .05). Differences in per-patient sensitivity and specificity between the two readers also were not significant with either method (Figs 3–6).
View this table:
[in this window]
[in a new window]
|
Table 3. Sensitivity and Specificity of CT Colonography with Two Data Interpretation Methods Performed on a Per-Patient Basis
|
|
View larger version (65K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3a: True colonic polyp in ascending colon of 63-year-old man. (a) Virtual 3D view of supine (top) and prone (bottom) data sets shows a 22-mm sessile polyp (solid arrows) at the same site relative to the ileocecal valve (arrowheads) and the small diverticuli (open arrows). This viewing mode allows the reader to interactively compare the prone and supine images. (b, c) Transverse 2D CT colonographic images of (b) supine and (c) prone data sets also show the polyp (arrow) in the ascending colon. (d) Conventional colonoscopic findings confirm the presence of the lesion (arrow). Biopsy was subsequently performed, and the final histopathologic diagnosis was hyperplastic polyp.
|
|
View larger version (100K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3b: True colonic polyp in ascending colon of 63-year-old man. (a) Virtual 3D view of supine (top) and prone (bottom) data sets shows a 22-mm sessile polyp (solid arrows) at the same site relative to the ileocecal valve (arrowheads) and the small diverticuli (open arrows). This viewing mode allows the reader to interactively compare the prone and supine images. (b, c) Transverse 2D CT colonographic images of (b) supine and (c) prone data sets also show the polyp (arrow) in the ascending colon. (d) Conventional colonoscopic findings confirm the presence of the lesion (arrow). Biopsy was subsequently performed, and the final histopathologic diagnosis was hyperplastic polyp.
|
|
View larger version (110K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3c: True colonic polyp in ascending colon of 63-year-old man. (a) Virtual 3D view of supine (top) and prone (bottom) data sets shows a 22-mm sessile polyp (solid arrows) at the same site relative to the ileocecal valve (arrowheads) and the small diverticuli (open arrows). This viewing mode allows the reader to interactively compare the prone and supine images. (b, c) Transverse 2D CT colonographic images of (b) supine and (c) prone data sets also show the polyp (arrow) in the ascending colon. (d) Conventional colonoscopic findings confirm the presence of the lesion (arrow). Biopsy was subsequently performed, and the final histopathologic diagnosis was hyperplastic polyp.
|
|
View larger version (108K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3d: True colonic polyp in ascending colon of 63-year-old man. (a) Virtual 3D view of supine (top) and prone (bottom) data sets shows a 22-mm sessile polyp (solid arrows) at the same site relative to the ileocecal valve (arrowheads) and the small diverticuli (open arrows). This viewing mode allows the reader to interactively compare the prone and supine images. (b, c) Transverse 2D CT colonographic images of (b) supine and (c) prone data sets also show the polyp (arrow) in the ascending colon. (d) Conventional colonoscopic findings confirm the presence of the lesion (arrow). Biopsy was subsequently performed, and the final histopathologic diagnosis was hyperplastic polyp.
|
|
View larger version (59K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4a: Sessile polyp (7 mm) in transverse colon of 66-year-old man. (a) Virtual 3D supine (top) and prone (bottom) views show sessile polyp (arrows) closely attached to a fold in transverse colon. (b, c) On transverse 2D CT colonographic (b) supine and (c) prone images, lesion (arrow) was initially missed by reader 2 at 2D interpretation but identified at 3D interpretation. Because this lesion could be identified on the 3D view only, one advantage of 3D view over transverse interpretation may be the detection of such lesions closely attached to a fold. Conventional colonoscopic findings confirmed presence of lesion, and biopsy revealed an adenomatous polyp.
|
|
View larger version (100K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4b: Sessile polyp (7 mm) in transverse colon of 66-year-old man. (a) Virtual 3D supine (top) and prone (bottom) views show sessile polyp (arrows) closely attached to a fold in transverse colon. (b, c) On transverse 2D CT colonographic (b) supine and (c) prone images, lesion (arrow) was initially missed by reader 2 at 2D interpretation but identified at 3D interpretation. Because this lesion could be identified on the 3D view only, one advantage of 3D view over transverse interpretation may be the detection of such lesions closely attached to a fold. Conventional colonoscopic findings confirmed presence of lesion, and biopsy revealed an adenomatous polyp.
|
|
View larger version (106K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4c: Sessile polyp (7 mm) in transverse colon of 66-year-old man. (a) Virtual 3D supine (top) and prone (bottom) views show sessile polyp (arrows) closely attached to a fold in transverse colon. (b, c) On transverse 2D CT colonographic (b) supine and (c) prone images, lesion (arrow) was initially missed by reader 2 at 2D interpretation but identified at 3D interpretation. Because this lesion could be identified on the 3D view only, one advantage of 3D view over transverse interpretation may be the detection of such lesions closely attached to a fold. Conventional colonoscopic findings confirmed presence of lesion, and biopsy revealed an adenomatous polyp.
|
|
View larger version (47K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5a: Flat polyp (4 mm) in sigmoid colon of 70-year-old woman. (a) Virtual 3D view of supine (top) and prone (bottom) data sets shows a small flat polyp (arrows) in the sigmoid colon. (b, c) On transverse 2D CT colonographic images of (b) supine and (c) prone data sets, this small lesion (arrow) was initially missed by both readers at 2D interpretation, but it was identified at 3D interpretation. (d) Conventional colonoscopic findings confirm the presence of the lesion (arrow). Biopsy revealed an adenomatous polyp.
|
|
View larger version (132K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5b: Flat polyp (4 mm) in sigmoid colon of 70-year-old woman. (a) Virtual 3D view of supine (top) and prone (bottom) data sets shows a small flat polyp (arrows) in the sigmoid colon. (b, c) On transverse 2D CT colonographic images of (b) supine and (c) prone data sets, this small lesion (arrow) was initially missed by both readers at 2D interpretation, but it was identified at 3D interpretation. (d) Conventional colonoscopic findings confirm the presence of the lesion (arrow). Biopsy revealed an adenomatous polyp.
|
|
View larger version (129K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5c: Flat polyp (4 mm) in sigmoid colon of 70-year-old woman. (a) Virtual 3D view of supine (top) and prone (bottom) data sets shows a small flat polyp (arrows) in the sigmoid colon. (b, c) On transverse 2D CT colonographic images of (b) supine and (c) prone data sets, this small lesion (arrow) was initially missed by both readers at 2D interpretation, but it was identified at 3D interpretation. (d) Conventional colonoscopic findings confirm the presence of the lesion (arrow). Biopsy revealed an adenomatous polyp.
|
|
View larger version (85K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5d: Flat polyp (4 mm) in sigmoid colon of 70-year-old woman. (a) Virtual 3D view of supine (top) and prone (bottom) data sets shows a small flat polyp (arrows) in the sigmoid colon. (b, c) On transverse 2D CT colonographic images of (b) supine and (c) prone data sets, this small lesion (arrow) was initially missed by both readers at 2D interpretation, but it was identified at 3D interpretation. (d) Conventional colonoscopic findings confirm the presence of the lesion (arrow). Biopsy revealed an adenomatous polyp.
|
|
View larger version (53K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6: Colonic feces in sigmoid colon of 60-year-old woman. Virtual 3D view of supine (top) and prone (bottom) data sets depicts a small polypoid lesion (arrow) in the supine data set only. Depiction in only the supine data set indicates that the lesion can change position and therefore represents feces.
|
|
For reader 1, the average number of FP lesions per patient was 0.53 with the 2D interpretation method and 0.68 with the 3D virtual dissection method. For reader 2, the average number of FP lesions per patient was 0.63 with the 2D method and 0.78 with the 3D method. For both radiologists, the average number of FP lesions identified with the 3D method tended to be greater than that identified with the 2D method, but the difference was not significant (P > .05). For reader 1, 22 FP lesions 5 mm in diameter or smaller, three 6–9-mm FP lesions, and one 13-mm FP lesion were visible on the 3D virtual dissection view only, whereas 10 FP lesions 5 mm or smaller and two 6–9-mm FP lesions were visible at 2D interpretation only. For reader 2, 30 FP lesions 5 mm or smaller, three 6–9-mm FP lesions, and two FP lesions 10 mm or larger were visible on the 3D view only, whereas 16 FP lesions 5 mm or smaller and four 6–9-mm FP lesions were visible at 2D interpretation only.
Regarding the causes of the false-negative polyps 6 mm in diameter or larger (Table 4), three lesions were not identified in retrospect. One lesion was missed by both radiologists with both interpretation methods owing to poor bowel distention. Four flat lesions were missed by both radiologists with both methods. Reader 1 identified two lesions and reader 2 identified three lesions only with 3D virtual dissection owing to their close attachment to a fold (Fig 4). One or two lesions were misinterpreted as feces. However, differences in the causes of the false-negative polyps between the readers and interpretation methods were not significant (P > .05, Fisher exact test).
Interpretation Time
With the 3D virtual dissection interpretation method, the median interpretation time was 9.4 minutes (range, 4.5–24.0 minutes) for reader 1 and 9.6 minutes (range, 4.3–26.7 minutes) for reader 2. With the 2D interpretation method, the median interpretation time was 14.1 minutes (range, 5.8–35.5 minutes) for reader 1 and 14.4 minutes (range, 5.1–38.3 minutes) for reader 2. The difference in interpretation time between the two methods was significant for both readers (P < .05).
|
DISCUSSION
|
|---|
The findings in our study demonstrate that use of the described 3D virtual dissection technique, compared with use of the primary 2D interpretation algorithm, significantly shortens the interpretation time in the detection of colonic polyps. It allows supine and prone images to be displayed adjacent to each other in a flat form and thereby facilitates intuitive and direct perception (18). Theoretically, use of this 3D method, compared with use of the conventional 2D method, can shorten the interpretation time by half because it enables radiologists to simultaneously interpret two sets of CT colonographic data with 3D virtual dissection, whereas the supine and prone data sets must be interpreted separately with the 2D method. Furthermore, application of the 3D virtual dissection technique that we used does not require the antegrade and retrograde navigation that is mandatory with 3D endoluminal views, because it ensures visualization of nearly the entire colonic surface in a single direction mode (12,17). Paik et al (12) reported achieving approximately 75% visibility with either antegrade or retrograde viewing only, 95% visibility with both antegrade and retrograde viewing, and 98% visibility with use of the dissection technique described herein. Such bidirectional navigation inevitably lengthens the interpretation time.
In our study, the 3D method had sensitivity and specificity comparable to those of the conventional 2D method for the detection of colonic polyps. Per-patient specificity in particular was in the high range compared with values published in the literature (2,22–24). The high per-patient specificity in our study can be explained by the fact that we used a wet bowel preparation. It is generally accepted that with wet preparation, a larger amount of fluid is retained, but feces, which is the main cause of the FP findings, is retained in a smaller amount than with dry preparation. Most colonic segments (98.5%) in our study received scores of 1 or 2 for residual feces. In addition, we used a segmental unblinding technique to compensate for the colonoscopy shortcoming of high missed polyp rates (2,25). In our study, after unblinding of the CT colonography results, colonoscopy depicted 13 additional polyps (9.7%) in eight patients, and four of these lesions were 6 mm in diameter or larger. The use of a segmental unblinding technique inevitably reduced the number of FP CT colonographic findings—that is, it increased the overall specificity of the 3D interpretation method.
Although there was no significant difference in the average number of FP polyps detected with either interpretation method for either reader, the average number of FP lesions detected at 3D interpretation was larger than that detected at 2D interpretation. The larger number of FP polyps detected with the 3D virtual dissection method can be explained by the greater sensitivity of this technique for the detection of colonic lesions compared with that of the 2D method. Both the per-polyp and the per-patient sensitivities were greater with 3D interpretation than with 2D interpretation, even though the differences between the two methods were not significant for either reader.
Furthermore, in our study, the initial interpretation of CT colonographic data was performed by using the primary 2D interpretation method with a 3D problem-solving technique. Therefore, among the 26 and 35 FP polyps identified only on the 3D view by readers 1 and 2, respectively, and which were not subjected to segmental unblinding, we cannot be sure how many true-positive lesions were missed at both initial CT colonographic interpretation and colonoscopy. If we had used a more sensitive 3D virtual dissection view at the time of the initial interpretation, the sensitivity of the 3D view might have been higher. This means that the number of FP polyps detected on the 3D view might have been lower.
Our study results differ from those of the previous investigation by Hoppe et al (11), in which a similar 3D colon dissection method was used for data interpretation. In that study, the 3D colon dissection method had lower sensitivity for polyp detection and was more time-consuming than axial interpretation. The results of the Hoppe et al study were mainly attributable to unsuccessful rendering of insufficiently distended and poorly prepared colonic segments and failed depiction of the areas surrounding and between the folds of the colon. Furthermore, in our study, the median interpretation time for 3D viewing was 9.4 or 9.6 minutes, which is less than one-third the 3D virtual dissection viewing time in the Hoppe et al study (11). In fact, the 3D interpretation method that we used differed slightly from that described in the Hoppe et al study. The 3D virtual dissection technique that we used enables the practitioner to see both the proximal and the distal sides of a colonic fold and therefore to see around and between colonic folds. In addition, compared with the method used by Hoppe et al, which involved the use of a 4 x 90° view, our technique involves the use of a full 360° view with an additional overlap of 10° at the top and bottom, for a total view of 380°, which might also account for the increased speed of our CT colonographic data interpretation.
We experienced neither failure in finding the path of the central colonic axis nor software breakdowns, which were the main contributors to the prolonged interpretation times in the Hoppe et al study (11). In addition, although the image distortion away from the equatorial region, a generally recognized drawback of the projection we used, was evident on the virtual 3D view, the area of image distortion was stretched out while it was near the center of the viewing area. Such image deformation causes polyps to be missed and consequently restricts the sensitivity of CT colonography in areas where the colon is strongly curved, such as colonic flexures. However, as observed in another recent study of CT colonography with 360° virtual dissection (26), the sensitivity of this examination for polyp detection is not compromised, even if the colon anatomy is altered. In our study, only one 6-mm lesion was seemingly missed owing to image distortion.
A limitation of our study was the relatively small numbers of patients and large polyps included. Because the patients were asymptomatic and were undergoing first-time colonic screening with colonoscopy, we were likely to find only a few clinically important polyps. Therefore, prospective studies with large patient populations are warranted to determine the time efficiency of 3D virtual dissection viewing. Another limitation was that we did not analyze the rendering time of the virtual 3D and conventional 2D interpretation methods. Poor rendering performance results in a long rendering time and the need for additional time to evaluate the CT colonographic data. However, with recent advances in 3D rendering hardware and software, shortened rendering times are guaranteed. Finally, we did not compare the diagnostic performance of 3D virtual dissection views with that of conventional 3D endoluminal views. At our institution, we do not have a 3D workstation with a special function to display the blind areas to the reviewer after 3D navigation is performed. According to previous literature reports, in which conventional 3D endoluminal display resulted in 93.8% visibility, with use of 3D endoluminal views without display of the blind areas after 3D navigation, one risks leaving parts of the colon unseen.
In conclusion, compared with the conventional 2D colonographic data interpretation method, primary 3D interpretation performed by using virtual dissection software had comparable sensitivity and specificity and facilitated increased time efficiency in the detection of colonic polyps.
|
ADVANCES IN KNOWLEDGE
|
|---|
- Three-dimensional (3D) panoramic display techniques, by enabling the virtual unfolding and dissecting of the colon along its longitudinal axis, represent an effective method of inspecting the inner colonic surface.
- Compared with the conventional two-dimensional colonic polyp detection method, primary 3D interpretation performed by using virtual dissection software for CT colonography enables one to interpret CT colonographic data with comparable per-polyp (77% and 69% for two readers) and per-patient (77% and 73% for two readers) sensitivities, with comparable per-patient specificity (99% and 89% for two readers), and within a shorter interpretation time (9.4 or 9.6 minutes).
|
IMPLICATION FOR PATIENT CARE
|
|---|
- By using three-dimensional virtual dissection software for CT colonography, radiologists can interpret CT colonographic data in a time-efficient manner without sacrificing diagnostic performance.
|
FOOTNOTES
|
|---|
Abbreviations: FP = false-positive • 3D = three-dimensional • 2D = two-dimensional
Authors stated no financial relationship to disclose.
See also Science to Practice in this issue.
Author contributions:Guarantors of integrity of entire study, S.H.K., J.M.L., B.I.C.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, S.H.K.; clinical studies, S.H.K., H.W.E., J.Y.L.; statistical analysis, S.H.K., M.W.L.; and manuscript editing, all authors
|
References
|
|---|
- Barish MA, Soto JA, Ferrucci JT. Consensus on current clinical practice of virtual colonoscopy. AJR Am J Roentgenol 2005;184:786–792.[Abstract/Free Full Text]
- 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]
- Taylor SA, Halligan S, Slater A, et al. Polyp detection with CT colonography: primary 3D endoluminal analysis versus primary 2D transverse analysis with computer-assisted reader software. Radiology 2006;239:759–767.[Abstract/Free Full Text]
- Yasumoto T, Murakami T, Yamamoto H, et al. Assessment of two 3D MDCT colonography protocols for observation of colorectal polyps. AJR Am J Roentgenol 2006;186:85–89.[Abstract/Free Full Text]
- Sorstedt E, Persson A, Noren B, et al. Computed tomographic colonography: comparison of two workstations. Acta Radiol 2005;46:671–678.[CrossRef][Medline]
- Beaulieu CF, Jeffrey RB Jr, Karadi C, Paik DS, Napel S. Display modes for CT colonography. II. Blinded comparison of axial CT and virtual endoscopic and panoramic endoscopic volume-rendered studies. Radiology 1999;212:203–212.
- Fletcher JG, Luboldt W. CT colonography and MR colonography: current status, research directions and comparison. Eur Radiol 2000;10:786–801.[CrossRef][Medline]
- 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. AJR Am J Roentgenol 2000;174:1543–1549.[Abstract/Free Full Text]
- Dave SB, Wang G, Brown BP, McFarland EG, Zhang Z, Vannier MW. Straightening the colon with curved cross sections: an approach to CT colonography. Acad Radiol 1999;6:398–410.[CrossRef][Medline]
- Fletcher JG, Johnson CD, Reed JE, Garry J. Feasibility of planar virtual pathology: a new paradigm in volume-rendered CT colonography. J Comput Assist Tomogr 2001;25:864–869.[CrossRef][Medline]
- Hoppe H, Quattropani C, Spreng A, Mattich J, Netzer P, Dinkel HP. Virtual colon dissection with CT colonography compared with axial interpretation and conventional colonoscopy: preliminary results. AJR Am J Roentgenol 2004;182:1151–1158.[Abstract/Free Full Text]
- Paik DS, Beaulieu CF, Jeffrey RB Jr, Karadi CA, Napel S. Visualization modes for CT colonography using cylindrical and planar map projections. J Comput Assist Tomogr 2000;24:179–188.[CrossRef][Medline]
- Rottgen R, Fischbach F, Plotkin M, et al. CT colonography using different reconstruction modi. Clin Imaging 2005;29:195–199.[CrossRef][Medline]
- Rottgen R, Fischbach F, Plotkin M, et al. Colon dissection: a new three-dimensional reconstruction tool for computed tomography colonography. Acta Radiol 2005;46:222–226.[Medline]
- Dekel D, Durgan J, Fleiter T. Virtual endoscopy (patent pending). Publication no. 2006/000925. Geneva, Switzerland: World Intellectual Property Organization, 2006.
- Park SH, Ha HK, Kim MJ, et al. False-negative results at multi-detector row CT colonography: multivariate analysis of causes for missed lesions. Radiology 2005;235:495–502.[Abstract/Free Full Text]
- Pickhardt PJ, Nugent PA, Choi JR, Schindler WR. Flat colorectal lesions in asymptomatic adults: implications for screening with CT virtual colonoscopy. AJR Am J Roentgenol 2004;183:1343–1347.[Abstract/Free Full Text]
- Juchems MS, Fleiter TR, Pauls S, Schmidt SA, Brambs HJ, Aschoff AJ. CT colonography: comparison of a colon dissection display versus 3D endoluminal view for the detection of polyps. Eur Radiol 2006;16:68–72.[CrossRef][Medline]
- Macari M, Lavelle M, Pedrosa I, et al. Effect of different bowel preparations on residual fluid at CT colonography. Radiology 2001;218:274–277.[Abstract/Free Full Text]
- Dachman AH, Zalis ME. Quality and consistency in CT colonography and research reporting. Radiology 2004;230:319–323.[Free Full Text]
- Bossuyt PM, Reitsma JB, Bruns DE, et al. Towards complete and accurate reporting of studies of diagnostic accuracy: the STARD initiative. Radiology 2003;226:24–28.[Abstract/Free Full Text]
- Rockey DC, Paulson E, Niedzwiecki D, et al. Analysis of air contrast barium enema, computed tomographic colonography, and colonoscopy: prospective comparison. Lancet 2005;365:305–311.[Medline]
- Mulhall BP, Veerappan GR, Jackson JL. Meta-analysis: computed tomographic colonography. Ann Intern Med 2005;142:635–650.[Abstract/Free Full Text]
- Iannaccone R, Catalano C, Mangiapane F, et al. Colorectal polyps: detection with low-dose multi–detector row helical CT colonography versus two sequential colonoscopies. Radiology 2005;237:927–937.[Abstract/Free Full Text]
- Rex DK, Cutler CS, Lemmel GT, et al. Colonoscopic miss rates of adenomas determined by back-to-back colonoscopies. Gastroenterology 1997;112:24–28.[CrossRef][Medline]
- Johnson KT, Johnson CD, Fletcher JG, MacCarty RL, Summers RL. CT colonography using 360-degree virtual dissection: a feasibility study. AJR Am J Roentgenol 2006;186:90–95.[Abstract/Free Full Text]
Related Article
-
Will 3D Virtual Dissection Display Prove to Be the Display Format for Reviewing CT Colonography Examinations?
- Seth Glick
Radiology 2007 244: 629-630.
[Full Text]
[PDF]
This article has been cited by other articles:
|
|
|
|
 
U.S. Preventive Services Task Force
Screening for Colorectal Cancer: U.S. Preventive Services Task Force Recommendation Statement
Ann Intern Med,
November 4, 2008;
149(9):
627 - 637.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
E. P. Whitlock, J. S. Lin, E. Liles, T. L. Beil, and R. Fu
Screening for Colorectal Cancer: A Targeted, Updated Systematic Review for the U.S. Preventive Services Task Force
Ann Intern Med,
November 4, 2008;
149(9):
638 - 658.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
D. Hock, R. Ouhadi, R. Materne, A.-S. Aouchria, I. Mancini, T. Broussaud, P. Magotteaux, and A. Nchimi
Virtual Dissection CT Colonography: Evaluation of Learning Curves and Reading Times with and without Computer-aided Detection
Radiology,
September 1, 2008;
248(3):
860 - 868.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
S. Glick
Will 3D Virtual Dissection Display Prove to Be the Display Format for Reviewing CT Colonography Examinations?
Radiology,
September 1, 2007;
244(3):
629 - 630.
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
