| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, University of Wisconsin Medical School, E3/311
Clinical Science Center, 600 Highland Ave., Madison, WI 53792-3252.
2 Department of Radiology, Uniformed Services University of the Health Sciences,
Bethesda, MD.
Received March 20, 2007;
accepted after revision June 6, 2007.
The opinions and assertions contained herein are the private views of the
authors and are not to be construed as official or as reflecting the views of
the Department of the Navy or Defense.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Ten radiologists, blinded to polyp findings, retrospectively interpreted 730 consecutive colonoscopy-proven CTC cases in asymptomatic adults using a primary 2D approach, with 3D reserved for problem solving. Primary 2D CTC performance was compared with the primary 3D CTC results from the original trial of 1,233 asymptomatic adults. The 10 2D reviewers were significantly more experienced in CTC interpretation (> 100 cases interpreted) than the six reviewers from the original 3D trial.
RESULTS. Primary 2D CTC sensitivity for adenomas 
6 mm was 44.1%
(56/127), compared with 85.7% (180/210) at 3D (p < 0.001).
Sensitivity of 2D CTC for adenomas 10 mm was 75.0% (27/36) compared with
92.2% (47/51) at 3D (p = 0.027). Similar sensitivity trends were seen
for the by-patient analysis and for all polyps at the 6-mm and 10-mm
thresholds. By-patient specificity for 2D evaluation at the 10-mm threshold
was 98.1% (676/689), compared with 97.4% (1,131/1,161) at 3D evaluation
(p = 0.336).
CONCLUSION. Primary 2D CTC is less sensitive than primary 3D CTC for polyp detection in low-prevalence screening cohorts. The disappointing 2D sensitivity in this study was very similar to results obtained with primary 2D evaluation in previous CTC trials.
Keywords: colorectal polyps • CT colonography • polyp detection • virtual colonoscopy
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
A total of 1,233 asymptomatic adults (728 men, 505 women; mean age, 57.8 years) underwent same-day CTC and optical colonoscopy over a 14-month period. This group composed the original screening cohort and formed the control group for primary 3D CTC performance [1]. The 3D performance results for adenomas and nonadenomatous polyps have been reported previously [1, 2]. The primary 2D cohort consisted of 730 consecutive CTC cases (415 men, 315 women; mean age, 58.8 years), representing the available Navy cohort from the DoD screening trial. The size of the 2D cohort was therefore larger than the previous 2D CTC trials by Johnson et al. [3] (703 cases), Cotton et al. [4] (600 subjects), and Rockey et al. [5] (614 subjects). Because of the strict, uniform inclusion and exclusion criteria, the polyp prevalence for the 2D subset from the DoD trial was similar and representative of the entire screening cohort. Specifics regarding primary 2D and primary 3D CTC interpretation techniques are described later.
CTC Technique
Study participants underwent laxative preparation with oral contrast
tagging consisting of 90 mL of phospho-soda, 500 mL of dilute barium (2.1%
weight/weight [w/w]), and 120 mL of water-soluble iodinated contrast material
(diatrizoate), as previously described
[1]. The protocols for colonic
distention and CT were detailed in the original 3D CTC study
[1]. To summarize, colonic
distention was achieved with patient-controlled insufflation of room air.
Breath-hold supine and prone CT acquisitions were obtained on 4- and 8-MDCT
LightSpeed scanners (GE Healthcare). CT technique entailed 1.25- to 2.5-mm
collimation, 1-mm reconstruction interval, 100 mAs, and 120 kVp. Note that our
current CTC protocols for preparation, distention, and MDCT scanning have
evolved from this technique
[9].
CTC Interpretation
CTC interpretation was performed using a commercially available software
system (V3D Colon, Viatronix). For the prospective primary 3D interpretation,
software version 1.2 was used. For the retrospective primary 2D
interpretation, an updated release was used (version 2.0), which features
improvements in both 2D and 3D functionality.
Primary 2D CTC interpretation consisted of cine-mode evaluation of the 2D multiplanar displays (transverse, coronal, and sagittal projections) for both the supine and prone data sets, with window levels set to optimize polyp detection (width, 2,000 H; level, 0 H). Soft-tissue windowing and magnification were used as needed to further evaluate focal findings. The 3D endoluminal view was used only for problem solving to further evaluate lesions detected on 2D evaluation [10]. Interpretation times for 2D polyp evaluation were recorded. All 10 of the 2D reviewers were experienced in both primary 2D and primary 3D CTC interpretation, having interpreted at least 100 CTC cases with optical colonoscopy correlation. To ensure adequate 2D skills, CTC training efforts initially focus on primary 2D polyp detection before a new reviewer is allowed to use 3D for polyp detection. None of the 10 reviewers were involved in the original 3D trial or familiar with any of the DoD CTC cases.
Primary 3D CTC interpretation from the published trial entailed a bidirectional fly-through examination (i.e., rectum-to-cecum and cecum-to-rectum) of the 3D endoluminal display with a 90° field of view for both the supine and prone data sets [1]. The 2D display was mainly used for further evaluation of 3D findings. Prior CTC experience for the six 3D reviewers consisted of 25–50 CTC cases for four reviewers and more than 100 cases for the remaining two reviewers. Unlike the 2D reviewer experience, training cases with optical colonoscopy correlation were not available for the 3D reviewer group. Both the 2D and 3D reviewers recorded all CTC-detected polyps measuring 5–6 mm or greater by segmental location.
Optical Colonoscopy
Optical colonoscopy was performed immediately after prospective 3D CTC
interpretation using standard commercial video colonoscopes
[1]. Colonoscopy was performed
by one of 17 experienced gastroenterologists, each with at least several years
of experience. The colonoscope was advanced to the cecum and then sequentially
withdrawn into more distal segments for polyp detection. Polyps were measured
using a calibrated linear probe. Segmental unblinding of CTC results provided
an enhanced reference standard.
Polyp Histology
All polyps retrieved at optical colonoscopy underwent histologic
examination by an experienced pathologist. For the purposes of this study, two
histologic categories were considered for CTC performance analysis:
adenomatous polyps (colorectal neoplasms) and all polyps (including both
adenomas and nonadenomatous lesions, such as hyperplastic polyps).
Performance and Statistical Analysis
Statistical testing using the chi-square test was performed to compare
differences in primary 2D versus primary 3D CTC performance according to polyp
size threshold and histologic category. A p value < 0.05 was
considered to be significant and a p value < 0.01 was considered
to be highly significant. The polyp-matching algorithm required CTC and
optical colonoscopy agreement according to both polyp size and polyp location,
as previously described [1].
CTC sensitivity for both adenomas and all polyps was assessed with analysis
according to individual polyps ("by-polyp" assessment), and
according to the patient as a whole ("by-patient" assessment).
By-patient specificity was calculated for all polyps at the 6-mm and 10-mm
thresholds. Specificity for adenomas was not considered because nonadenomatous
polyps, such as hyperplastic polyps, cannot be reliably distinguished from
adenomas at CTC or colonoscopy and therefore are better classified as true
polyps rather than false-positives. A true-positive result at a given polyp
size threshold for the by-patient assessment required at least one matching
polyp of that size (or larger). The performance characteristics for primary 2D
CTC evaluation were compared with both the primary 3D CTC evaluation from the
original DoD screening population
[1,
2] and with the results from
the three primary 2D CTC trials
[3–5].
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
6 mm, which includes both adenomatous and
nonadenomatous lesions, the sensitivity for the primary 3D approach was more
than double that for primary 2D detection (80.8% vs 37.9%). The prevalence of
disease among the large consecutive subset of patients (n = 730) from
the naval medical centers evaluated by primary 2D was similar to the overall
DoD study group (n = 1,233). For instance, the prevalence of both
adenomas and all polyps was within 1% at both the 6-mm and 10-mm
thresholds.
|
|
|
|
|
|
The by-patient performance characteristics for primary 2D CTC evaluation
versus primary 3D evaluation are shown in
Table 2. Primary 3D polyp
detection was again superior to primary 2D for all four polyp categories under
consideration. Except for large adenomas ( 10 mm), the differences in 2D
versus 3D polyp detection sensitivity were statistically significant and were
highly significant at the 6-mm polyp size threshold. Specificity was
significantly higher for the primary 2D reviewers compared with 3D at the 6-mm
polyp size threshold. At the 10-mm threshold, the difference in by-patient
specificity was not statistically significant. Because specificity reflects a
combined 2D and 3D assessment for lesions detected at CTC regardless of the
means for initial detection, it likely reflects differences in reviewer
experience more than the 2D versus 3D polyp detection approach.
|
Table 3 shows results from
primary 2D evaluation in the 730 CTC cases from the DoD trial compared with
the primary 2D results from the trial with 703 patients by Johnson et al.
[3], the trial with 600
patients by Cotton et al. [4],
and the trial with 614 patients by Rockey et al.
[5]. The primary 3D results for
all polyps from the DoD trial are also included for comparison
[1,
2]. Overall, the primary 2D CTC
performance characteristics are strikingly similar among these four
low-prevalence cohorts. The sensitivity for large polyps ( 10 mm) and the
specificity at both 6-mm and 10-mm thresholds were slightly higher for the DoD
cohort compared with the other trials, whereas sensitivity at the 6-mm
threshold for the DoD cohort was intermediate among the trials. Of note,
sensitivity for adenoma detection in the current study could not be compared
with the prior 2D studies because of the lack of reporting or incomplete
reporting of adenoma sensitivity in those trials.
|
The mean interpretation time for primary 2D evaluation was 6.7 ± 3.4 minutes (median time, 6 minutes). The average number of CTC cases interpreted by each 2D reviewer was 73 (range, 44–134). Unfortunately, direct comparison with primary 3D reviewing time is not possible because the recorded interpretation time from the original trial included both colonic and extracolonic evaluation.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Of the four large CTC trials to date evaluating cohorts with low prevalence of disease, the three (one single-center and two multicenter) trials restricted to a primary 2D approach for polyp detection fared rather poorly [3–5]. Although these three 2D CTC trials did not evaluate true screening populations, the low prevalence of disease was similar to that found in asymptomatic average-risk adults. In comparison, the DoD CTC screening trial used primary 3D polyp detection and showed that CTC sensitivity for clinically relevant polyps was comparable to optical colonoscopy [1]. Our new findings show that primary 2D evaluation of cases from the DoD trial yields a disappointing sensitivity for polyp detection that is strikingly similar to the previous primary 2D trials (Table 3). This provides direct evidence that primary 2D detection alone, with 3D reserved for problem solving, is an inadequate interpretive approach for low-prevalence CTC screening. Moreover, the real benefit of primary 3D polyp detection used in the initial interpretation is confirmed because other CTC variables such as bowel preparation and colonic distention were fixed. The disparity in CTC reviewer experience and expertise should have favored the primary 2D evaluation.
A recent review of the 114 CTC false-negatives from the 2D trial by Rockey
et al. [5] showed that most
missed lesions could be retrospectively identified at CTC
[12]. Theoretically, the
by-polyp sensitivity for adenomas of 10 mm in this trial would have
increased from 63.6% to 94.5% if all correctable errors were scored as
true-positives. On the basis of our experience, we suspect that most of the 2D
misses from this trial might have been detected prospectively with adequate
primary 3D evaluation. As further support for the primary 3D interpretation
approach used in the DoD trial, this method was recently validated in an
independent CTC screening trial
[13].
There are a number of reasons why the primary 2D technique became the standard interpretive approach in the early development of CTC: 2D CT multiplanar display formats were already mature when CTC began to emerge in the mid to late 1990s, the 2D interpretation approach was a natural extension from general cine-mode CT interpretation, and effective time-efficient primary 3D evaluation was not yet widely available [14]. Although a few smaller studies had suggested that 3D displays offered an interpretive advantage over 2D [15–17], most radiologists remained firmly entrenched in the primary 2D paradigm [18], perhaps in part because most CTC software systems remained incapable of effective, time-efficient 3D polyp detection [16, 19]. However, as more and more CTC software systems incorporate a more robust 3D component, attitudes toward primary 3D interpretation appear to be changing. Perhaps the results of this study will further accelerate this process.
Primary 3D polyp detection is so effective because the conspicuity of polyps among the folds is greatly enhanced compared with cross-sectional 2D images in which haustral folds and polyps have a similar appearance (Figs. 1A, 1B, 1C, 1D, and 1E). The relative ease of polyp detection on 3D (assuming a robust CTC software system) translates into in a much easier search pattern and shorter learning curve [9]. Reviewer fatigue and eye strain resulting from the more onerous 2D polyp search are important considerations in high-volume CTC screening. In our experience, radiologists new to CTC interpretation generally begin as primary 2D reviewers because of their familiarity with this display from general CT evaluation but naturally gravitate toward 3D polyp detection over time. From a logical standpoint, the more sensitive but less specific display—the 3D endoluminal view—is best for initial polyp detection, whereas the more specific but less sensitive display—cross-sectional 2D images—is needed for confirmation of suspected lesions. It must be emphasized that secondary 2D evaluation remains necessary because a pure "virtual reality" evaluation is not feasible. Furthermore, although 3D polyp detection is usually comprehensive, supplementary 2D evaluation can be particularly useful in cases with abundant adherent stool or with areas of partial or total luminal collapse [9].
There are several emerging 3D CTC displays beyond the standard endoluminal projection that are not yet validated for screening but may play an important role in the future. In particular, the "virtual dissection" (or filet) view is an appealing concept because it has the potential to significantly reduce 3D interpretation times [20, 21]. In general, the major trade-off with these novel 3D projections is the spatial distortion that is introduced, which may distort polyps beyond recognition. Currently, the impact on accuracy remains uncertain and further investigation is needed. Another effort toward reducing 3D interpretation time involves widening the endoluminal field of view to increase surface visualization enough to eliminate the need for bidirectional fly-through. Finally, a 2D-based solution aimed at increasing sensitivity is to incorporate computer-aided detection (CAD) into a primary 2D interpretation. However, this approach has yet to match the performance of primary 3D evaluation in the setting of screening [1, 22].
Our study is limited somewhat by the retrospective nature of the 2D evaluation. It is also unfortunate that we cannot directly compare 2D and 3D CTC interpretation times. Although the mean interpretation time of 19.6 minutes reported in the original trial (for both colonic and extracolonic evaluation) clearly translates to a longer primary 3D interpretation compared with the mean time of 6.7 minutes for primary 2D evaluation, this disparity is rapidly vanishing. The navigation speed for endoluminal fly-through has doubled from version 1.2 to version 2.0 of the V3D Colon software system, rendering the published 3D interpretation times obsolete. Bidirectional 3D endoluminal interpretation times are currently under 10 minutes for a typical case [9]. In addition, widening the 3D endoluminal field of view angle to 120° appears to allow unidirectional fly-through without sacrificing polyp detection (Schumacher C et al., presented at the 2007 annual meeting of the Radiological Society of North America), which drastically reduces primary 3D interpretation times even further.
There are a number of features that strengthen the results of this study. The size of the screening population evaluated makes this the largest primary 2D CTC study to date. The strict inclusion and exclusion criteria ensure a true screening population; the low prevalence of disease provides a more rigorous and relevant evaluation compared with a polyp-rich cohort. The use of same-day optical colonoscopy with segmental unblinding of CTC results provides an enhanced reference standard. The primary 2D reviewers were significantly more experienced in CTC interpretation than the original 3D reviewers, which should have favored the 2D results. Finally, the results from the prospective 3D interpretation allow a more direct comparison of the two interpretation techniques.
On the basis of our primary 2D interpretation results, it is clear that the polyp detection technique (2D vs 3D) was likely the critical difference between the multicenter trials. However, another notable protocol difference was the use of oral contrast tagging only in the DoD trial [1–5]. Although the presence of fecal tagging did not appear to affect 2D sensitivity relative to the other trials, it may help to explain the increased specificity (Table 3). The similar sensitivity for primary 2D polyp detection among the published CTC trials argues against any unique or inherent advantages for the DoD patient cohort, such as superior bowel cleansing or distention. With regard to direct comparison of 2D versus 3D results from the DoD cases, we believe that the increased experience of the 2D reviewers likely accounts for the improved specificity. CTC specificity, which measures the ability of the test to exclude polyps in those without disease, involves a combined 2D–3D assessment of detected lesions, regardless of the means for initial detection (2D or 3D). To further support this notion, our current positive predictive value for CTC-detected polyps in clinical practice is now around 90%, which is much higher compared with the DoD trial and implies improved specificity [1, 23].
In conclusion, our results provide the first direct evidence that primary 3D polyp detection at CTC is superior to the standard primary 2D approach for low-prevalence screening. These findings do not apply to CTC software systems with inadequate 3D capabilities because polyp detection is largely limited to 2D evaluation in such cases. This point is of particular relevance to the current CTC screening trial run by the American College of Radiology Imaging Network (ACRIN) [24], which intends to compare primary 2D and 3D detection as a secondary aim. Unfortunately, the majority of centers involved in this trial appear to be using primary 2D-based CTC systems with substantial 3D shortcomings, which may yield confusing results that could underestimate 3D detection capabilities. Regardless, we believe that our findings largely explain the differences in CTC sensitivity reported in the major multicenter CTC trials to date.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  S. H. Park, S. S. Lee, J. K. Kim, M.-J. Kim, H. J. Kim, S. Y. Kim, M.-Y. Kim, A. Y. Kim, and H. K. Ha Volume Rendering with Color Coding of Tagged Stool during Endoluminal Fly-through CT Colonography: Effect on Reading Efficiency Radiology, September 1, 2008; 248(3): 1018 - 1027. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. J. Pickhardt and A. H. Dachman Missed Lesions at Primary 2D CT Colonography: Further Support for 3D Polyp Detection Radiology, February 1, 2008; 246(2): 648 - 649. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
