| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, Mayo Clinic, 200 First St., SW, Rochester, MN
55905.
2 Department of Biostatistics, Mayo Clinic, Rochester, MN.
3 Division of Gastroenterology, Mayo Clinic, Rochester, MN.
Received January 30, 2007;
accepted after revision April 7, 2007.
Supported by National Institutes of Health grant R01 CA75333.
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. Four hundred fifty-two asymptomatic patients underwent CT colonography (CTC) and colonoscopy. Examinations were reconstructed to 1.25- and 2.5-mm slice thicknesses and interpreted using primary 3D search (360° virtual dissection) and primary 2D search. Two of three experienced reviewers were randomly assigned to each case; 1,808 interpretations were performed.
RESULTS. There were 64 adenomas 
6 mm, 26 of which were large
adenomas 1 cm. For adenomas 6–9 mm in diameter, the area under the
receiver operating characteristic curve (AUC) using 2.5-mm data sets was 0.66,
0.62, 0.90 and 0.78, 0.69, 0.67 for reviewers 1, 2, and 3, respectively, using
primary 3D versus 2D search (p = not significant [NS]). For neoplasms
10 mm, the AUC using 2.5-mm data sets was 0.74, 0.85, 0.89 and 0.66,
0.86, 0.92 for reviewers 1, 2, and 3 using primary 3D versus 2D search
(p = NS). There was no significant difference using 1.25-mm
collimation. Double review using both primary 3D and 2D search yielded
sensitivities of 84% (16/19) and 95% (18/19) for large neoplasms ( 1 cm)
using 2.5- and 1.25-mm data sets, respectively. Five of five (100%)
adenocarcinomas were identified. The sensitivity of colonoscopy for large
neoplasms was 77% (20/26) (20% [1/5] for adenocarcinoma).
CONCLUSION. No advantage exists for 1.25- or 2.5-mm slice thickness or for primary 3D versus 2D search at CTC. Double review using primary 3D (virtual dissection) and 2D search reduces interobserver variability and competes with colonoscopy for the detection of large lesions.
Keywords: colonoscopy • colorectal cancer • CT colonography • oncologic imaging • virtual dissection display
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
New techniques such as 360° virtual dissection present the 3D endoluminal data as if the colon has been straightened and unfolded, allowing quick inspection of its mucosal surface (Figs. 1A, 1B, 1C, and 1D). Unlike images that show the traditional perspective, volume-rendered 3D endoluminal images, which require forward and reverse virtual fly-throughs that mimic the movement of the colonoscope, viewed using 3D virtual dissection, permit the "open" colon to be viewed in colonic segments of varying lengths. Primary review involves examining the endoluminal surface of the colon in a series of segments. Findings on individual colonic segments are correlated with 2D findings and perspective volume-rendered 3D endoluminal images. The drawbacks of this technique include a required midline trace and anatomic distortion, which is most marked at flexural regions of the colon. Early reports indicate that the technique is promising for detecting colon lesions [14–16].
|
|
|
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Twenty-four cases were used in a prior study assessing different polyp detection systems [17], but the computer software, aims, and all radiologist interpretations described herein are entirely unique to this study.
CTC
All patients underwent CTC before same-day colonoscopy. Preparation was
ordered at the preference of the referring physician. The preparations for the
study examinations included a polyethylene glyeol electrolyte lavage (CoLyte,
Braintree Laboratories) and bisacodyl tablets (10 mg) in 372 patients (82%),
magnesium citrate (300 mL) and bisacodyl tablets (20 mg) in 13 patients (3%),
and oral saline (Phospho-soda, Fleet) (90 mL) in 67 patients (15%). Four
hundred sixteen patients received glucagon, 1 mg subcutaneously, 10 minutes
before CT acquisition unless contraindicated or refused by the patient; 36
(8%) did not receive glucagon.
Patients were placed in the left lateral decubitus position for enema tip insertion and slow manual insufflation of approximately 2 L of carbon dioxide or until the patient verbally indicated air administration had reached maximal tolerance. After a CT scout image was obtained before each acquisition to confirm complete colon filling, both supine and prone data acquisitions were obtained. Additional carbon dioxide was added as tolerated by the patient if needed to fully distend the colon.
All examinations were performed using an 8-MDCT scanner (LightSpeed Ultra, GE Healthcare). Images were acquired using a 1.25-mm collimation; table speed of 13.5 mm/s (pitch of 1.35); 1.25-mm reconstruction intervals; matrix of 512 x 512; field of view, to fit; 70 mAs; 120 kVp; standard reconstruction algorithm; and 28-second breath-holds. Images from prone and supine acquisitions were reconstructed to slice thicknesses of 1.25 and 2.5 mm with a 1.25-mm reconstruction interval.
The image data were networked to an offline workstation (AW 4.2 workstation, GE Healthcare; colonography software: Voxtool 5.4.46, GE Healthcare). The software allows simultaneous comparison of supine and prone data sets and adjustments of the field of view. The studies were randomly distributed to two of three experienced radiologists to be interpreted independently using either a primary 2D search method (using a conventional 2D image display technique with 3D endoluminal problem solving) or a primary 3D search method using a 360° virtual dissection image display method that can display both multiplanar 2D and 3D perspective volume-rendered images for problem solving (Fig. 2). All three reviewers had interpreted more than 1,000 colonoscopy-verified CTC examinations before this study. Reviewers had trained on at least 50 cases using 360° virtual dissection software before the study. Virtual dissection used "Autodissection," an automated midline trace tool. The incidence of manual tracing being required was not tabulated. Supine and prone comparisons were used.
|
Each interpretation when performed by the same reviewer using a different search method was separated by at least a 6-week interval to reduce recall bias. Lesion location, size, and reviewer confidence were noted for each abnormality. Lesion size was determined from 2D images using the greatest lesion diameter. Lesion problem solving using virtual dissection used both 2D images (including multiplanar reformatted images) and endoluminal 3D images. Reviewer confidence for each lesion was rated on a scale of 1–6 (not possible to definite). Each data set (supine, prone), regardless of the image display, was evaluated sequentially, and suspicious regions were compared between the two position data sets. The interpretations were performed during time allocated for research, without the rush typical of a busy clinical practice.
Results are reported for each individual reviewer (by slice thickness and by primary 3D vs 2D search method) and for double review—that is, the combined reports of the two individual reviewers using both primary 2D and primary 3D interpretations. Reviewers were instructed to record only lesions that were > 5 mm in diameter.
Colonoscopy
Same-day index colonoscopy was performed in the standard fashion to examine
the entire colorectum. Examinations were videotaped during withdrawal of the
instrument. All examinations were directly performed by a staff
gastroenterologist or were supervised by one of approximately 50 staff
gastroenterologists and colorectal surgeons experienced with this
technique.
Colonoscopists were blinded to the results at CTC. Colorectal inspection was complete in 446 of 452 patients (99%), including 18 patients who had a partial colonic resection with visualization of the ileocolonic anastomosis by the endoscopist. Incomplete studies in six examinations were complete to the sigmoid (n= 2), splenic flexure (n= 2), transverse colon (n= 1), and ascending colon (n= 1). The findings in these patients were compared between CTC and colonoscopy only in the segments of the colon that were visualized endoscopically.
Reviewer 1 performed all lesion matching using the annotated slice number recorded by the reviewer and videotaped colonoscopy examination when required. Lesions detected at CTC were matched manually with those found at colonoscopy using both the colonoscopy and the pathology reports. Lesion size was determined on the basis of the pathology report unless the lesion had been removed in pieces. In those cases, the size estimate at colonoscopy, which was based on comparison with the forceps diameter, was used. Nearly all lesions underwent final pathologic analysis within 24 hours of CTC and colonoscopy. If the lesion was within one colonic segment of the colonoscopic location and within ± 50% in size, it was considered a match at CTC.
The colon was considered to have eight segments: cecum, ascending colon,
hepatic segment, transverse colon, splenic segment, descending colon, sigmoid
colon, and rectum. Videotaped examinations were reviewed if a convincing
lesion (> 5 mm) was identified at CTC and was not reported on the
colonoscopy report. Lesions 1 cm at CTC that were not identified at
colonoscopy were reviewed by all colonographers after both blinded radiologic
assessments had been performed. If the lesion was deemed to have a high
likelihood of being a true neoplasm despite negative colonoscopy (n =
6), a letter was sent to the referring physician requesting repeat
colonoscopy. Repeated colonoscopy was performed in all six patients.
Examination quality was assessed for eight colon segments for each examination by subjectively grading the amount of fluid, stool, distention, and breath-hold artifacts on a 4-point scale by the assigned initial reviewer. Fluid and stool were graded as none, mild, moderate (could miss 5-mm polyps), or nondiagnostic (could miss 1-cm polyps). Distention was graded as > 90% (optimal), 75–90%, 50–74%, or < 50% fully distended. The lowest rating among the eight colon segments was tabulated.
Statistical Analysis
The goals of this study included estimation and comparison of the
performance of CTC examinations by slice thickness (1.25 vs 2.5 mm) and also
by display method (2D vs 3D). Given the time involved in reviewing these
images and the primary study goals of performance estimation, we decided to
assign two of three experienced radiologists to review the images of each
patient. Case assignments were made using a randomization algorithm to allow a
direct, within radiologist, comparison of 1.25 versus 2.5 mm in half of the
study cohort and in the other half of the cohort a direct, within radiologist,
comparison of 2D versus 3D. In other words, in half of the study group, one
radiologist was assigned to interpret the two images obtained with a slice
thickness of 1.25 mm on 2D display and of 2.5 mm on 2D display, whereas the
other radiologist was assigned to interpret the two images obtained with a
slice thickness of 1.25 mm on 3D display and of 2.5 mm on 3D display. For such
a patient, a direct assessment of the difference in performance by slice
thickness could be made within the same radiologist. In the other half of the
study group, one radiologist was assigned to interpret the two images obtained
with a slice thickness of 1.25 mm on 2D display and of 1.25 mm on 3D display,
whereas the other radiologist was assigned to interpret the two images
obtained with a slice thickness 2.5 mm on 2D display and 2.5 mm on 3D display.
For such a patient, a direct assessment of the difference in performance by
display method could be made within the same radiologist. In addition, the two
interpretations assigned to a radiologist were ordered in random fashion with
a "washout" period of at least 6 weeks. A total of 1,808
interpretations were performed (Fig.
2).
The reference standard for the presence of an adenomatous lesion or
adenocarcinoma was established by endoscopy, including the same-day index
colonoscopy examination and any subsequent endoscopies performed during the
study period. The analysis looked only at adenomatous polyps. In patients with
both adenomatous and hyperplastic (mixed) lesions, all polyps were considered
to be adenomatous polyps because we could not distinguish the exact nature of
each polyp. CTC sensitivities were compared for slice thickness (1.25 vs 2.5
mm) and image display technique (2D vs 3D endoluminal problem solving) using
the McNemar test. This study had limited power for these comparisons. Among
the 38 adeno-matous polyps that were 6–9 mm, for a two-sided comparison
at an 
value of 0.05, there would be 80% power for a comparison of
sensitivities of 50% versus 84% assuming independence of slice thickness
(or display) results. Only lesions that were > 5 mm in diameter were
considered in the analysis. Double review included all detections from both
independent interpretations of the primary 3D (virtual dissection) and primary
2D interpretations. No discussion of individual findings or consensus
reporting occurred.
Receiver operating characteristic (ROC) curves were generated and the areas
under the ROC curves (AUCs) were calculated separately for each of the three
reviewers and for the double review. In this analysis, the judgment of lesion
presence was graded by the radiologists as not possible, doubtful, somewhat
doubtful, conservatively confident, moderately confident, and highly
confident. For lesions not detected by a radiologist an answer of "not
possible" was assumed. For the double review, the higher of the two
reviewer confidences was used. Specificity was determined on a per-patient
basis and was reported for two groups of patients: those without a proven
lesion that was > 5 mm and those without a proven lesion that was 10
mm. Interpretation times for primary 3D interpretation using virtual
dissection versus primary 2D interpretation were compared using a paired
Student's t test.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
A total of 93 mass lesions 6 mm in diameter were present in 68
patients, 50 lesions 6–9 mm in diameter were present in 36 patients, and
43 lesions 1 cm in diameter were present in 38 patients. Sixty-four of
the 93 lesions (69%) were neoplastic (i.e., adenoma or adenocarcinoma). There
were 26 neoplasms 1 cm in 26 patients and 38 neoplasms 6–9 mm in 30
patients. The prevalence of at least one adenomatous lesion 1 cm in
diameter was 5.8% (26/452) (Table
1).
|
Colorectal lesions 6–9 mm had histologic findings as follows:
adenoma, n = 26 (52%); hyperplastic, n= 9 (18%); mixed,
n = 12 (24%); and other, n= 3 (6%). Mixed (adenomatous and
hyperplastic) lesions were considered to be adenomas. Lesions 1 cm had
the following histologic findings: adenoma, n= 21 (51%);
adenocarcinoma, n= 5 (12%); hyperplastic, n= 10 (24%);
lipoma, n = 1 (2%); and other, n= 4 (10%). The two remaining
lesions had no recorded histology.
Overall examination quality for each of the eight colonic segments was
judged for the presence of excessive fluid, stool, or collapse. The poorest
quality among the eight segments was tabulated. Residual fluid in the colon
was judged to be none in 48 patients (11%), mild in 240 (53%), moderate in 152
(34%), and nondiagnostic in 12 (3%). Residual stool in the colon was judged to
be none in 179 patients (40%), mild in 180 (40%), moderate in 81 (18%), and
nondiagnostic in 12 (3%). Colon distention was judged to be optimal in 135
patients (30%), suboptimal for one segment in 67 (15%), and suboptimal but
diagnostic for more than one segment in 250 (55%). Breath-hold artifacts were
judged as none in 400 patients (88%), mild in 36 (8%), and moderate in 16
(4%). Fifty-nine patients (13%) had a nondiagnostic examination because of
poor preparation quality, suboptimal distention, or both. Of these, 10
patients had a total of 10 lesions: Four were 6–9 mm and six were 1
cm. Because of their small numbers, they were included in the study analysis
to ensure assessment of all patients undergoing CTC.
Per-Lesion Assessment
CTC sensitivity for the detection of neoplastic lesions that were 6–9
mm and CTC sensitivity for those that were 1 cm are shown in Tables
2 and
3 for primary 2D and for
primary 3D virtual dissection using both 2.5- and 1.25-mm collimations.
Overall, there was no significant difference in sensitivity between the search
methods or between the slice thicknesses (p > 0.05). For lesions
6–9 mm in diameter, the sensitivity of the three reviewers varied from
16% (3/19) to 100% (8/8). For two of three reviewers, there tended to be
improved sensitivity for lesions of this size using primary 3D virtual
dissection, but this trend was not universal. For large lesions ( 1 cm),
the sensitivity ranged from 33% to 88%.
|
|
Double review (different reviewers using conventional and virtual dissection image display methods) resulted in improved lesion sensitivity of 84% (16/19) and 95% (18/19) for the detection of large neoplasms using 2.5- and 1.25-mm collimation, respectively (p = NS). Double review was less effective for polyps 6–9 mm in diameter: The sensitivity was 45% and 55% using slice thicknesses of 2.5 and 1.25 mm, respectively.
Four of the five adenocarcinomas (80%) in our study population were identified using the primary 3D search using both slice thicknesses and primary 2D search using 2.5-mm slices. Five of five (100%) were identified using primary 2D search and 1.25-mm slices, and five of five (100%) were identified using any combination of 2D and 3D double review.
Index colonoscopy initially discovered 20 of 26 large lesions (77%). Missed lesions at index endoscopy that were seen on CT data sets and were confirmed by subsequent endoscopy and surgery included four adenocarcinomas, one tubulovillous adenoma, and one tubular adenoma (Figs. 3A, 3B, 3C, and 3D).
|
|
|
|
For adenomas and adenocarcinomas that were 1 cm, the sensitivity on a
per-patient basis ranged from 33% to 88%, with a double review sensitivity of
84% and 95% for a 2.5- and a 1.25-mm slice thickness, respectively. The
specificity for these larger lesions ranged from 95% to 99%, with a double
review specificity of 98%. The AUC values ranged from 0.66 to 0.92 among the
three reviewers. The double-review AUC value was between 0.92 and 0.97 for a
2.5- and a 1.25-mm collimation, respectively (p = 0.001).
Interpretation times for primary 2D and primary 3D using virtual dissection averaged 14.5 minutes (range = 2–56 minutes, SD = 6.2 minutes) and 10.4 minutes (range = 1.5–37.2 minutes, SD = 6 minutes), respectively.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
1
cm. Selection of the correct slice thickness is an important technical parameter that influences the size of the suspected lesion that can be detected, the quality of the reformatted 2D images, the quality of the 3D images, and the image noise and radiation dose. New CT technology enables submillimeter slice acquisition of the entire abdomen and pelvis within a breath-hold, and radiologists tend to use these thinner slices. To maintain constant image noise, the radiation dose must be doubled each time the slice thickness is reduced by one half [18]. Alternatively, radiologists must be willing to accept images with higher noise levels.
Because CTC data are usually acquired using a low-dose technique, further increases in image noise can degrade image quality and can potentially impact diagnostic capability [19]. For example, 3D endoluminal images can develop surface nodularity or endoluminal "floaters" as image noise increases. Lesions on 2D images can appear heterogeneous and resemble stool if excessive image noise is present. Therefore, selecting a slice collimation that is as thick as possible to perceive the attenuation of a filling defect without degrading the spatial resolution required to detect colorectal lesions is optimal. This study indicates that a 2.5-mm slice thickness results in diagnostic performance that is statistically similar to interpretations using a 1.25-mm slice thickness. This trend was true for both conventional 2D image displays and 3D virtual dissection image displays.
Controversy exists about the optimal image display for reviewing CTC data [20]. Advocates of the conventional 2D approach (i.e., primary 2D search) point to the ability to assess the full integrity of the bowel wall, inspect the internal attenuation of filling defects, and inspect both sides of a haustral fold and to the lack of requirements for endoluminal navigation [5, 21]. The 3D proponents argue that many lesions are more conspicuous, especially subcentimeter lesions, and that overall performance is superior compared with 2D techniques. Further, the difficulties in separating normal colon folds from abnormal mucosa are reduced and interobserver variability is lower with this type of display [10, 12].
Virtual dissection is a new 3D display technique requiring a midline trace. The colorectum is "straightened" along the midline trace and "opened," similar to a surgical specimen, exposing all 360° of the lumen [14, 15]. Anatomic distortion does occur, particularly in the flexural regions of the colon. Studies have indicated that this distortion is predictable and that normal folds can be discriminated from abnormal mucosa [14]. The advantages of this technique are that the entire colorectal mucosa can be inspected for each anatomic segment on a single image without concern that a lesion is hidden behind a fold. The limitations of other 3D techniques do exist: Filling detects must be interrogated using 2D image display and can be hidden by fluid or collapsed lumen. As such, 2D images must be integrated by the software to allow rapid problem solving [14, 15]. Our study indicates that this new technique is promising; we found that there was no statistical difference in performance (sensitivity and specificity) when 3D interpretation was compared with conventional 2D interpretation. Interpretation times for virtual dissection were 28% faster than the conventional 2D approach (10.4 vs 14.5 minutes, respectively).
Ideally, interpretation of a colorectal examination would include review of both 2D and 3D displays, but this is not commonly performed because of time constraints and shortage of adequate personnel available for independent review. However, investigators have shown that this strategy is clinically feasible [22], and we use this method of review daily in our practice [22]. Our study results indicate that overall performance is significantly improved for most reviewers when primary 2D and primary 3D virtual dissection are combined. This review strategy may be a viable option using 3D virtual dissection because interpretation time is relatively short. Further, interobserver variability for interpretation of CTC has been well documented, and double review remains a viable method for addressing that issue. Computer-assisted diagnosis may be another option for reducing interobserver variability [23].
For adenomatous lesions that were 1 cm, the performance of double
review using both image display techniques at 1.25-mm collimation resulted in
a per-lesion sensitivity, per-patient sensitivity and specificity, and AUC of
95%, 95% and 98%, and 0.97, respectively. These performance estimates are
essentially equivalent to the expected performance of index colonoscopy.
Importantly, in six patients, a large lesion ( 1 cm) was identified at CT
colonography that was not found at colonoscopy. This translated into a
sensitivity for initial colonoscopy of 77%, lower than the estimates at CTC.
The lesions that were missed on the initial colonoscopy included four
adenocarcinomas (of five total adenocarcinomas) that were located in the cecum
(1.3 cm), sigmoid colon (1.3 cm), rectosigmoid junction (3.2 cm), and
ascending colon (4.0 cm). There was one tubulovillous neoplasm in the
ascending colon (2.0 cm), and a tubular adenoma was present in the cecum (1.3
cm). Although endoscopy tended to perform more favorably for the detection of
subcentimeter lesions, it also tended to miss advanced lesions that were
1 cm.
The overall quality of the examinations was considered nondiagnostic in 59 of 452 patients (13%). This rate is higher than expected and is likely due to the use of manual insufflation of the colon compared with mechanical insufflation. Our practice since 2003 has been to use mechanical insufflation of the colon for clinical examinations, and the potential to obtain overall better distention was known to the radiologists who rated these research cases. These cases were not excluded from analysis. As such, we expect that the results from the present study would be slightly better than reported if mechanical insufflation had been used.
A moderate or severe amount of residual fluid was present within at least one segment of the colon in 36% of cases. This is likely secondary to the lavage bowel preparation that was used for the majority of patients. In most cases, this was not problematic because both prone and supine data sets were obtained, redistributing fluid to other bowel segments. We currently use stool and fluid tagging in our clinical practice to label this residual fluid and to reduce false-positive examinations. We do not believe that the 8-MDCT scanner had any effect on nondiagnostic studies because all acquisitions were obtained within a breath-hold.
Previous studies evaluating slice thickness and lesion detection rates have
been confined to phantom or animal colons with simulated lesions
[3,
24–28].
Those studies showed a very high sensitivity for lesion detection using narrow
slice thicknesses. In their study of 20 pig colons with 60 lesions, Won and
colleagues [28] found no
significant difference in lesion detection using either 1- or 2.5-mm
collimation for lesions 5 mm in diameter. These data support the findings
of our large study.
Pickhardt and colleagues [29] showed that a single-pass 3D endoluminal fly-through left nearly a quarter of the colorectal mucosa nonvisualized. Adding an antegrade fly-through to the retrograde fly-through provided coverage of 94% of the total lumen. The first report of virtual dissection mapping 100% of the colonic mucosa was by Fletcher and coworkers [30] and showed the feasibility of this technique compared with conventional 2D interpretation. Hoppe and coworkers [16] evaluated virtual dissection in 22 patients and found similar detection rates between it and conventional 2D image display.
The limitations of the 3D virtual dissection technique included longer times to reconstruct and interpret images and the inability to evaluate segments that are insufficiently distended. Rottgen and coworkers [15] in 48 patients found virtual dissection techniques to be superior for lesion detection compared with either conventional 2D or 3D endoluminal displays. Johnson and colleagues [14] in a 20-patient cohort found sensitivities and specificities for 360° virtual dissection among three radiologists to be similar to colonoscopy. Our large study confirms the findings of others that 360° virtual dissection performs similar to primary 2D search. Further, when a double-review technique is used, interobserver variation can be markedly reduced, resulting in overall performance that may exceed that of colonoscopy.
A limitation of this study is that it was initiated in 2001 before the introduction of both mechanical colon insufflation and stool and fluid tagging. These techniques are becoming widely accepted today, but they were not used for our study because they were unavailable when the study was designed. Although these techniques could have provided higher estimates of overall performance if they had been used, the study results remain relevant because both compared techniques were performed similarly. Stool tagging would not likely have affected these data because the specificity remains very high. In fact, tagging could have increased interpretation time because retained barium can appear polypoid on 3D virtual dissection images, requiring increased time for additional problem solving. The effect of better colon distention using an automated insufflator is unknown but likely would have translated into a higher sensitivity.
Comparison of primary 2D search and primary 3D search with virtual dissection was not performed against primary 3D search using conventional-perspective volume-rendered images (fly-throughs). In large part, this comparison was not done because of the daunting logistic issues of reinterpreting 452 examinations by three reviewers already using two slice thickness and two search methods. The study by Rottgen and coworkers [15] provides some insight into this question: In a small patient group (n = 48), virtual dissection review performed better than interpretations performed using 3D endoluminal fly-throughs. Because both are 3D techniques, we anticipate that they should have similar diagnostic capability and that virtual dissection may save time by obviating forward and reverse viewing.
A time interval of at least 6 weeks between examinations seems to be a reasonable time period to eliminate recall bias; however, there may have been a few special cases of unique lesions that were recalled by the reviewers. Precise assessment of this was not performed.
High interobserver variability was encountered in this study and has been summarized in the CTC meta-analysis by Mulhall and colleagues [31] and in other radiologic reports [32–37]. The causes are not fully understood but are likely due to many factors, including rushing, distraction, concentration, and cognitive traits. Double review in our practice has been shown to improve overall reviewer performance. Hopefully, computer-aided detection will also provide a similar benefit in the future.
In summary, for CTC detection of adenomatous lesions 6 mm, there was
no performance advantage between reconstruction of 2.5- and 1.25-mm slice
thicknesses or between primary 2D and primary 3D search using virtual
dissection. Decisions regarding slice thickness should include consideration
of patient dose, image noise, and radiologist preference. Combining primary 2D
and primary 3D (dual interpretation) helps reduce interobserver variability
and can result in accuracy rates comparable to colonoscopy. Virtual dissection
360° images can be interpreted approximately 25% faster than conventional
2D displays.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  P. J. Pickhardt, D. H. Kim, and C. Hassan The Effectiveness of Colonoscopy in Reducing Mortality From Colorectal Cancer Ann Intern Med, June 2, 2009; 150(11): 818 - 819. [Full Text] [PDF]
|
|
|
|
|
|
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. K. Rex, T. F. Imperiale, J. Regula, M. Polkowski, D. S. Glaser, L. F. Butterly, H. Pohl, D. H. Kim, and P. J. Pickhardt CT Colonography versus Colonoscopy for the Detection of Advanced Neoplasia N. Engl. J. Med., January 3, 2008; 358(1): 88 - 90. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
