Sensitivity of CT Colonography for Nonpolypoid Colorectal Lesions Interpreted by Human Readers and With Computer-Aided Detection
Abstract
OBJECTIVE. The purpose of our study was to determine the sensitivity of CT colonography (CTC) interpreted by human readers and with computer-aided detection (CAD) for genuinely nonpolypoid colorectal lesions, defined as 2 mm or less in lesion height at colonoscopy.
MATERIALS AND METHODS. A computerized database search for a 33-month period found 21 patients who had undergone both colonoscopy and CTC and who had a total of 23 genuinely nonpolypoid colorectal lesions: eight adenomas (9-30 mm in width), 10 stage Tis or T1 adenocarcinomas (10-25 mm), and five nonadenomatous lesions (8-20 mm). CTC was performed using a cathartic preparation and fecal tagging and was interpreted by experienced readers in a blinded manner using a primary 3D method and with CAD.
RESULTS. The sensitivities of human readers for nonpolypoid adenomatous lesions (i.e., both adenomas and adenocarcinomas), adenocarcinomas, and nonadenomatous lesions were 66.7% (12/18), 90% (9/10), and 0% (0/5), respectively. Sensitivities were 55.6% (10/18), 90% (9/10), and 0% (0/5) for CAD. A 10-mm stage T1 adenocarcinoma was missed by a human reader on blinded review but was detected with CAD. Both human readers and CAD yielded significantly higher sensitivity for adenomatous lesions than for nonadenomatous lesions (p = 0.014 and 0.046, respectively) and for adenocarcinomas than for noncancerous lesions (p = 0.003 and 0.0001, respectively).
CONCLUSION. CTC showed a high sensitivity for nonpolypoid stage Tis and T1 adenocarcinomas 10 mm or greater in width despite the limited overall sensitivity for nonpolypoid adenomatous lesions, when performed using cathartic preparation and fecal tagging.
Introduction
The importance of nonpolypoid (also colloquially referred to as “flat”) colorectal adenomas as precursors of colorectal cancer is increasingly recognized [1, 2]. Some studies suggested a greater risk for malignancy of nonpolypoid adenomas compared with that of polypoid lesions [2, 3], although there are also contradicting results [4].
CT colonography (CTC) has recently been added to the joint guideline for colorectal cancer screening by the American Cancer Society, the U.S. Multi-Society Task Force on Colorectal Cancer, and the American College of Radiology [5], and it is increasingly being used in clinical practice. The high diagnostic performance of CTC shown in major clinical trials [6, 7] was mostly obtained with polypoid lesions. On the other hand, the diagnostic performance of CTC for nonpolypoid lesions is not yet well-established, although nonpolypoid lesions are generally considered to be more difficult to detect than polypoid lesions [8-11]. Nonpolypoid lesions are still rarely encountered in CTC practice, and there is a paucity of data regarding their detection with CTC either by human readers [8-13] or with computer-aided detection (CAD) [14].
There are several different definitions of nonpolypoid colorectal lesions. A commonly used endoscopic definition is mucosal elevation with a lesion height less than half the greatest lesion diameter [2]. However, the endoscopic definition may be too crude to characterize the flatness of a lesion because lesions of various heights will then be grouped into the same nonpolypoid category on the basis of their widths. The recent Paris endoscopic classification [15] proposes a more refined definition. According to that definition, the polypoid and nonpolypoid terminology describes only superficial lesions [15]. A neoplastic lesion is called superficial when the depth of penetration in the colorectal wall is not more than into the submucosa: benign neoplastic lesions (adenoma or dysplasia) or, in the case of adenocarcinoma, stage Tis (limited to the mucosa) or T1 (limited to the submucosa) according to the TNM classification [15, 16]. Lesions that have penetrated into the muscularis propria or the serosa are classified separately as advanced. In the Paris classification, the lesion height of a superficial lesion is measured in comparison with the height of the closed jaws of biopsy forceps (2.5 mm) [15]. Lesions protruding above the level of the closed jaws of biopsy forceps (i.e., 3 mm or higher by rounding up the submillimeter digit) are classified as polypoid, whereas those protruding below this level (i.e., 2 mm or lower) are classified as nonpolypoid. Similarly, a lesion 2 mm or less in height with respect to the adjacent normal mucosa was proposed as the definition of nonpolypoid lesions on CTC [17].
Previous studies regarding the sensitivity of CTC for nonpolypoid lesions primarily used the definition of lesion height less than half the lesion diameter [10-14]. However, considering that large plaquelike lesions with relatively thick heights are easily recognized on CTC, whereas what causes the diagnostic challenge on CTC is genuinely nonpolypoid lesions that protrude only slightly into the lumen, a more refined definition of nonpolypoid lesions must be used to achieve a more clinically relevant evaluation of the sensitivity of CTC for nonpolypoid lesions.
Therefore, the purpose of this study was to determine the diagnostic sensitivity of CTC interpreted by human readers and with CAD for genuinely nonpolypoid colorectal lesions defined as 2 mm or less in lesion height at colonoscopy according to the Paris endoscopic classification [15].
Materials and Methods
Our institutional review board approved this retrospective study and waived patient informed consent.
Study Patients
We performed a search of the computerized database of colonoscopy and radiology at Asan Medical Center, which is a tertiary referral center, to identify patients who fulfilled all of the following criteria: patients who underwent colonoscopy between July 2005 and March 2008 and whose colonoscopy reports included at least one of the following terms: “nonpolypoid,” “flat,” “depressed,” “carpet,” “lateral spreading,” “laterally spreading,” or “LST” (acronym for lateral spreading tumor or laterally spreading tumor); and patients who had undergone CTC with less than a 6-month interval from colonoscopy.
The search found 34 patients. Gastroenterologists who performed the colonoscopies then retrospectively reviewed the colonoscopy images and reports and the pathology reports of these patients and selected those lesions that fulfilled the following criteria according to the Paris endoscopic definition [15] for nonpolypoid colorectal lesions: 2 mm or less in lesion height at colonoscopy and no extension further than the submucosa in the case of adenocarcinoma.
Thirteen patients were excluded for the following reasons. Eight lesions in eight patients were 4-8 mm in height (Fig. 1), four of which were also advanced cancers (i.e., stage T2 or higher stages). Two lesions in two patients presented with a 4-cm-wide cobblestone-like appearance. However, most of the individual nodular protrusions in the lesions were 3 mm or higher. All of the aforementioned 10 lesions had been correctly diagnosed at CTC in our clinical practice. Two patients with familiar adenomatous polyposis were excluded because lesion-by-lesion correlation between colonoscopy and CTC was not possible due to the presence of numerous colonic lesions. Another patient was excluded because CTC was performed after removal of the nonpolypoid lesion.
The remaining 21 patients (16 men and five women; mean age, 63.5 ± 9.4 [SD] years for men and 62.4 ± 10.2 years for women) with 23 pathologically proven genuinely nonpolypoid colorectal lesions were finally included in this study. In all but one patient, the interval between CTC and colonoscopy was less than 3 months. Sixteen of the 21 patients were referred to our center with a known or suspected diagnosis of colorectal cancer. As part of the pretreatment workup, the 16 patients were concurrently examined with both colonoscopy and CTC; therefore, colonoscopy was requested independently of the CTC results. The remaining five patients underwent CTC for asymptomatic screening (n = 4) and surveillance after low anterior resection for colorectal cancer (n = 1); follow-up colonoscopy was requested because of the detection of the nonpolypoid lesions at CTC (n = 3) and the detection of other polypoid lesions at CTC (n = 2).
CT Colonography
Technique—Patients ingested a semifluid low-residue diet the day before CTC. Eleven patients underwent fecal tagging with 200 mL of 5% w/v barium sulfate taken after each meal the day before CTC (i.e., three doses for a total of 600 mL) and subsequent cathartic preparation with the oral administration of 19.2 g of magnesium citrate and 10 mg of bisacodyl the night before CTC. The remaining 10 patients were prepared with 4 L of polyethylene glycol (PEG) followed by fluid tagging with one dose of 100 mL of Gastrografin (meglumine diatrizoate, Bayer Schering Pharma AG).
Colonic distention with carbon dioxide was achieved using an automated insufflator and a small rectal catheter with a retention balloon (PROTOCO2L colon insufflator and administration set, E-Z-EM). Both supine and prone scans were obtained using a 16-MDCT scanner (Somatom Sensation 16, Siemens Medical Solutions). Seven patients underwent both scans without IV contrast injection using the following scanning parameters: beam collimation, 16 × 0.75 mm; reconstructed slice thickness, 1 mm; reconstruction interval, 0.7 mm; beam pitch, 1; gantry rotation time, 0.5 second; field of view to fit; 120 kV; and 50 mAs. In the remaining 14 patients, one scan was obtained after IV contrast enhancement. The reasons for contrast enhancement were suspected or known colorectal cancer (n = 11) and screening of both the colon and extracolic organs as per patients' requests (n = 3). In accordance with the indications of contrast enhancement, scanning was performed during the portal phase [18] with a 72-second delay after the administration of 150 mL of iopromide (Ultravist 370, Bayer Schering Pharma AG) at a rate of 2.5 mL/s through an antecubital vein. Contrast-enhanced scanning was performed using an automated dose reduction system (CARE Dose 4D, Siemens Medical Solutions) with the maximum allowable tube current set at 200 mAs but otherwise with the same scanning parameters as those used for unenhanced scanning.
Blinded interpretation by human readers—The results of the blinded interpretation during our clinical reading were collected and used for this study. Our way of clinical practice was to initially read the CTC cases without referring to any clinical data. When comparable colonoscopic results were available (e.g., colonoscopy at another institution) at the time of interpretation, we referred to them after the initial blinded reading, and we separately recorded lesions, if any, that had been missed in the blinded reading but were found retrospectively. Cases were interpreted by one of the two experienced radiologists. At the beginning of the study period, each radiologist had had experience with approximately 100 CTC cases and was familiar with the disease entity and the macroscopic morphology of nonpolypoid colorectal lesions. By the end of the study period, each reader had performed more than 1,500 CTC cases.
The interpretation was performed using primary 3D endoluminal fly-through, including both antegrade and retrograde navigations, and 2D problem solving using window width and level settings of 2,000 and 0 HU, respectively, with one of the two CTC systems (Syngo Colonography, Siemens Medical Solutions; and Lucion, Infinitt). Areas submerged under tagged fecal matter were subsequently evaluated with 2D review using window width and level settings of 2,000 and 0 HU, respectively. Readers referred to contrast enhancement at their discretion for confirmation of the suspected lesions. Maximal lesion diameter was measured on the 3D endoluminal view because that was known to be more accurate than the measurement on orthogonal 2D planes [19, 20]. Lesion height from the normal mucosal surface was measured on the 2D view using the same window width and level settings. Then only those lesions 6 mm or greater in diameter were reported [21]. The segmental location and morphologic type of the lesion were recorded. We considered lesions having a height of 2 mm or less with respect to the adjacent normal mucosa to be nonpolypoid [17].
CAD interpretation—Stand-alone CAD interpretation was performed retrospectively using a CAD software program (PEV version VC30A, Siemens Medical Solutions). This version was up-graded from previous versions [22] for improved detection of nonpolypoid lesions in addition to general polyp detection. None of the CTC data sets included in this study had been used to develop the CAD system.
Reference Standard and CTC-Colonoscopy Correlation
Colonoscopy was performed by experienced board-certified gastroenterologists who used a video colonoscope (CF 240L, Olympus Optical). Chromoscopic examination by mucosal spraying of methylene blue was used in some patients once a suspected area was identified. Segmental location, lesion size as estimated by comparing it to the size of biopsy forceps, and morphologic type were recorded.
Lesion matching between CTC and colonoscopy was performed in consensus by the two CTC readers and the gastroenterologists who performed the colonoscopy. Lesion matching was done according to a comparison of the segmental location, size (i.e., size discrepancy within 50% of the colonoscopic size), and morphology of the lesion; the lesion location with respect to the locations of other obviously matched structures; and morphology of the colon in the vicinity of the lesion.
Unblinded Examination of CTC
The cases in which the nonpolypoid lesions had been missed by the human readers on the blinded interpretation were reexamined by the two CTC readers in consensus with knowledge of the colonoscopic findings using the same review methods as the blinded interpretation. In the cases in which the unblinded review failed to reveal the missed lesions, an additional 2D review using soft-tissue window settings (width, 400 HU; level, 20 HU) was performed [12].
Causes for the missed nonpolypoid lesions by the human readers at the blinded review of CTC were determined. For the lesions that were detected on either the initial blinded review or the unblinded review, the two readers determined in consensus whether the lesion conspicuity was higher on the 3D view or on the 2D views using either colon or soft-tissue window settings.
Assessment of CAD Results
The CAD results were assessed by one experienced radiologist. A lesion was considered to be correctly detected by CAD if the lesion was marked by CAD in either position. False-positive CAD marks were counted and their identities were determined. When CAD created multiple false-positive marks in the same structure, each CAD mark was counted separately.
Data Analysis
The characteristics of the nonpolypoid lesions, including histopathology, size, colonoscopic morphology according to the Paris classification [15], and the depth of mural extension in cancerous lesions, were recorded. Although the Paris classification of nonpolypoid colorectal lesions at colonoscopy [15] was originally developed for adenomatous lesions, we classified both adenomatous and nonadenomatous lesions ac cording to that classification.
The diagnostic sensitivity of CTC for nonpolypoid colorectal lesions was analyzed in regard to the lesion size thresholds and histopathology. CAD sensitivity was calculated for all the nonpolypoid lesions identified at colonoscopy and for the identified-by-human lesions (i.e., those lesions that were identified on either the blinded review or the unblinded review of CTC by the human readers). The latter is a useful index for distinguishing the performance of CAD from the intrinsic shortcomings of CTC [23]. The sensitivity was compared using Fisher's exact test: adenomatous lesions (i.e., both nonpolypoid adenomas and adenocarcinomas) versus nonadenomatous lesions, nonpolypoid adeno carcinomas versus non cancerous lesions (i.e., nonpolypoid adenomas and nonadenomatous lesions), and the PEG-prepared group versus the magnesium-citrate-plusbisacodyl group.
Conspicuity of the nonpolypoid lesions was compared between the 3D view and the 2D views using the sign test.
Statistical analysis was performed with SPSS version 15 (SPSS). A p value of less than 0.05 was considered to be statistically significant.
Results
Characteristics of Nonpolypoid Colorectal Lesions
Chromoscopic examination was used in eight patients. In total, eight nonpolypoid adenomas measuring 9-30 mm in width (median, 15 mm), 10 nonpolypoid stage Tis or T1 adenocarcinomas measuring 10-25 mm in width (median, 14 mm), and five nonadenomatous lesions measuring 8-20 mm in width (median, 10 mm) were present at colonoscopy. The histopathology, size, and colonoscopic morphology of the lesions are summarized in Table 1.
| Histopathology | No. of Lesions | Size Range (mm) | Colonoscopic Morphology | |
|---|---|---|---|---|
| Adenomatous | Tubular adenoma | 3 | 9-15 | Slightly elevateda |
| Villotubular adenoma | 2 | 15, 20 | Slightly elevateda (n = 1), slightly depressed with slightly elevated rimb (n = 1) | |
| Adenoma with high-grade dysplasia | 3 | 12-30 | Slightly elevateda (n = 2), slightly elevated with central depressiona (n = 1) | |
| Adenocarcinoma confined in mucosa (stage Tis) | 2 | 10 | Slightly elevateda (n = 1), slightly elevated with central depressiona (n = 1) | |
| Adenocarcinoma with extension to submucosa (stage T1) | 8 | 10-25 | Slightly elevateda (n = 3), slightly elevated with central depressiona (n = 5) | |
| Nonadenomatous | Hyperplastic | 3 | 8-20 | Slightly elevateda |
| Inflammatory | 2 | 9, 10 | Slightly elevateda (n = 1), flatc (n = 1) | |
Diagnostic Sensitivity of CTC for Nonpolypoid Colorectal Lesions
The diagnostic sensitivities of CTC for nonpolypoid colorectal lesions are summarized according to the size thresholds and histopathology in Table 2.
| Sensitivity (%) According to Size Thresholda | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| ≥ 8 mm | ≥ 9 mm | ≥ 10 mm | |||||||
| CAD | CAD | CAD | |||||||
| Histopathology | Human | All Lesions | Identified-by-Human Lesionsb | Human | All Lesions | Identified-by-Human Lesionsb | Human | All Lesions | Identified-by-Human Lesionsb |
| Adenomatous (n = 18) | NA | NA | NA | 66.7 (12/18) | 55.6 (10/18) | 76.9 (10/13) | 70.6 (12/17) | 58.8 (10/17) | 76.9 (10/13) |
| Adenocarcinoma (stage Tis or T1) (n = 10) | NA | NA | NA | NA | NA | NA | 90 (9/10) | 90 (9/10) | 90 (9/10) |
| Adenoma (n = 8) | NA | NA | NA | 37.5 (3/8) | 12.5 (1/8) | 33.3 (1/3) | 42.9 (3/7) | 14.3 (1/7) | 33.3 (1/3) |
| Nonadenomatous (n = 5) | 0 (0/5) | 0 (0/5) | 0 (0/3) | 0 (0/4) | 0 (0/4) | 0 (0/2) | 0 (0/3) | 0 (0/3) | 0 (0/2) |
Note—Numbers in parentheses are numbers of lesions. NA = not applicable because all adenomatous lesions were 9 mm or larger in width and all adenocarcinomas were 10 mm or larger in width.
a
Size refers to lesion width.
b
Identified on either blinded or unblinded review of CTC by human readers.
Blinded interpretation by human readers—The sensitivity of the human readers for all nonpolypoid adenomatous lesions (66.7%, 12/18) was significantly higher than that for nonadenomatous lesions (0%, 0/5) (p = 0.014). The sensitivity for nonpolypoid adenocarcinomas (90%, 9/10) was also significantly higher than that for noncancerous lesions (23.1%, 3/13) (p = 0.003). The sensitivity was not significantly different between the PEG-prepared group (63.6%, 7/11) and the magnesium-citrate-plus-bisacodyl group (41.7%, 5/12) (p = 0.414). Examples of nonpolypoid adenomatous lesions detected by the human readers in the blinded review are shown in Figures 2A, 2B, 2C, 2D, 3A, 3B, 3C, 4A, 4B, and 4C.
CAD interpretation—The general tendency was toward lower sensitivity of CAD compared with the human readers except for nonpolypoid adenocarcinomas, for which both the human readers and CAD yielded 90% (9/10) sensitivity (Table 2). The human reader detected a 10-mm-wide stage Tis adenocarcinoma that was not detected by CAD, whereas CAD detected a 10-mm-wide stage T1 adenocarcinoma that was missed by the human reader in the blinded interpretation because of a perception error (Figs. 5A, 5B, and 5C). The number of false-positive CAD marks ranged from one to 32 per patient (mean, 6 per patient ± 7.3). The false-positive CAD marks were easily dismissible: residual feces (n = 72), small bowel (n = 16), ileocecal valve (n = 7), bubbly surface of colonic fluid (n = 7), stomach (n = 6), suboptimally distended colon (n = 5), colonic folds (n = 4), duodenum (n = 4), extrinsic compression by normal anatomic structures (n = 3), and teniae (n = 2) in decreasing frequency.
The sensitivity of CAD for nonpolypoid adenomatous lesions (55.6%, 10/18) was significantly higher than that for nonadenomatous lesions (0%, 0/5) (p = 0.046). The CAD sensitivity for nonpolypoid adenocarcinomas (90%, 9/10) was also significantly higher than that for noncancerous lesions (7.7%, 1/13) (p = 0.0001). The sensitivity was not significantly different between the PEG-prepared group (45.5%, 5/11) and the magnesium citrate plus bisacodyl group (41.7%, 5/12) (p = 1).
Causes for Missed Nonpolypoid Colorectal Lesions on CTC
Six of 18 adenomatous lesions and all five nonadenomatous lesions were missed by the human readers at the blinded interpretation. The causes for the missed lesions are summarized in Table 3.
| Lesion No. | Size (width in mm) | Histopathology | Cause | Unblinded Reviewa | CAD Resultsa |
|---|---|---|---|---|---|
| 1 | 9 | Tubular adenoma | Invisible despite optimal technique | − | − |
| 2 | 10 | Tubular adenoma | Invisible despite optimal technique | − | − |
| 3 | 15 | Villotubular adenoma | Probable obscuration by large amount of tagged fecal matter | − | − |
| 4 | 20 | Villotubular adenoma | Invisible despite optimal technique | − | − |
| 5 | 12 | Tubular adenoma with high-grade dysplasia | Invisible despite optimal technique | − | − |
| 6 | 10 | Adenocarcinoma with extension to submucosa (stage T1) | Reader perception error | + | + |
| 7 | 8 | Hyperplastic | Reader perception error | + | − |
| 8 | 15 | Hyperplastic | Reader perception error | + | − |
| 9 | 20 | Hyperplastic | Invisible despite optimal technique | − | − |
| 10 | 9 | Inflammatory | Invisible despite optimal technique | − | − |
| 11 | 10 | Inflammatory | Reader perception error | + | − |
Note—CTC = CT colonography, CAD = computer-aided detection.
a
Lesion was (+) or was not (−) detected.
Four missed lesions, including a 10-mm-wide stage T1 adenocarcinoma (Figs. 5A, 5B, and 5C), a 10-mm-wide inflammatory lesion, and one 8-mm-wide and one 15-mm-wide hyperplastic lesions, were perception errors and were retrospectively identified. Of those, the 10-mm T1 adenocarcinoma was correctly detected by CAD (Figs. 5A, 5B, and 5C), whereas the other lesions were not.
Seven missed lesions could not be found even on the unblinded review. Six of these, including two tubular adenomas, a villotubular adenoma, a tubular adenoma with high-grade dysplasia (Fig. 6), and two nonadenomatous lesions, could not be found despite optimal cleansing and distention of the colon and the use of additional 2D review with the soft-tissue window setting. The remaining lesion, a 15-mm-wide villotubular adenoma, was located in a well-distended segment with a large amount of homogeneously tagged stool and therefore was not detected because of obscuration by the tagged fecal matter or flatness of the lesion.
Lesion Conspicuity
Of 16 nonpolypoid lesions found on CTC on either the blinded review (n = 12) or the unblinded review (n = 4), 15 lesions were seen on both 3D and 2D views, whereas the remaining lesion was seen only on the 2D view because it was submerged under tagged fecal matter.
The relative lesion conspicuity between the 3D view and the 2D view depended on the window settings. The lesion conspicuity on the 3D view was superior to that on the 2D view using the colon window setting (p = 0.008): higher conspicuity on the 3D view in eight lesions (Figs. 2A, 2B, 2C, 2D, 3A, 3B, 3C, 4A, 4B, 4C, 5A, 5B, and 5C) but no difference between the two views in seven lesions. On the other hand, the lesion conspicuity was not different between the 3D view and the 2D view using the soft-tissue window setting (p = 0.289): higher conspicuity on the 3D view in six lesions, higher conspicuity on the 2D view in two lesions, and no difference in seven lesions (Figs. 2A, 2B, 2C, and 2D).
Discussion
In this study, we investigated the sensitivity of CTC for nonpolypoid colorectal lesions defined as 2 mm or less in lesion height at colonoscopy according to the Paris endoscopic definition [15]. CTC interpreted by experienced human readers showed moderate sensitivities of 66.7% (12/18) and 70.6% (12/17) for nonpolypoid adenomatous lesions 9 mm or greater in width and those 10 mm or greater in width, respectively, largely because of the low sensitivity for nonpolypoid adenomas. This was lower than the 82-92% sensitivity achieved for mostly polypoid adenomatous lesions of the same size range in several major trials [6, 7].
Nevertheless, the sensitivity for nonpolypoid adenomatous lesions in our study was substantially higher than that in one previous study by Park et al. [10], which reported 25% (1/4) sensitivity for nonpolypoid adenomatous lesions 9 mm or larger (the height of the detected lesion was also 4 mm). We believe that the relatively high sensitivity of our study was attributed to the higher-quality bowel preparation and colon distention, the use of primary 3D interpretation, and readers' familiarity with nonpolypoid colorectal lesions.
In the study by Park et al. [10], 42.9% (6/14) of the missed nonpolypoid lesions at CTC were missed because of poor colon cleansing or distention, whereas in our study, in essence, none of the missed nonpolypoid lesions was missed because of such causes. Fecal tagging was not used in the study by Park et al., whereas it was used in our study. The importance of proper technique in achieving high diagnostic accuracy with CTC is evident [7, 18, 24-26]. This is probably even more relevant for detecting nonpolypoid lesions because nonpolypoid lesions cause even more subtle changes in the colonic contour.
Another factor that contributed to the relatively high sensitivity of our study was probably the use of primary 3D interpretation. Lesion conspicuity was significantly higher on the 3D view than on the 2D view using the colon window setting in our study. One experimental study using simulated lesions [27] showed that the 3D view significantly increased CTC sensitivity for nonpolypoid lesions compared with the 2D view at the colon window setting. Pickhardt et al. [13] reported a fairly high CTC sensitivity of 82.8% (24/29) for nonpolypoid adenomas 6 mm or larger in width using primary 3D review. Those authors also found that nonpolypoid lesions were more conspicuous and easier to detect on 3D than on 2D views at the colon window setting. In fact, the sensitivity in the study by Pickhardt et al. was higher than that of our study. The lower sensitivity of our study compared with theirs may be explained in part by the fact that we used a stricter definition for the nonpolypoid lesions (i.e., ≤ 2 mm in height) than they did (i.e., generally ≤ 3 mm in height but higher for some larger lesions [13]). By contrast, Fidler et al. [12] reported that the 3D view was less helpful than the 2D view for detecting nonpolypoid lesions [12]. However, this result may not have practical implications because the study [12] used a 5-mm slice thickness, which is considered inadequate for the 3D display of CT data.
Unlike the 2D view using the colon window setting, a 2D view using the soft-tissue window setting in our study revealed similar conspicuity of the nonpolypoid lesions compared with the conspicuity on the 3D view. This result is concordant with a previous study [12], which suggested higher sensitivity of the soft-tissue window setting in the detection of nonpolypoid lesions than the colon window setting. Review of the colon using 2D images at a soft-tissue window setting is probably not routinely performed in most institutions; it is uncertain whether this should be routinely added to the clinical interpretation of CTC for improved detection of nonpolypoid lesions at the cost of increased interpretation time. In any case, our results suggest that additional 2D review using the soft-tissue window setting may not be necessary when a 3D review is used because none of the lesions that could not be visualized on the 3D view was identifiable on the 2D view using the soft-tissue window setting.
Similar to the results of human readers, the CAD sensitivity for nonpolypoid adenomatous lesions in our study was lower than those reported for polypoid lesions. The same CAD system as used in this study achieved 86.4% (38/44) sensitivity for polypoid lesions 10 mm or larger [28]. Another CAD system yielded 89.3% (25/28) sensitivity for identified-by-human polypoid adenomas 10 mm or larger [23].
One promising observation made in our study was that both human readers and CAD achieved a fairly high sensitivity of 90% (9/10) for nonpolypoid stage Tis and T1 adenocarcinomas 10 mm or greater in width. CAD also detected one 10-mm stage T1 adenocarcinoma that was missed by the human reader because of perception error. As a result, when both interpretations are combined, the sensitivity of CTC could potentially have been 100%. Unfortunately, there is yet a paucity of comparable data in the literature [14], the CAD detection of the human-missed nonpolypoid T1 adenocarcinoma was only a single-case occurrence, and we did not evaluate the CAD-human interaction [29] in interpretation. Therefore, accumulation of more data is required.
Our study does not provide much information regarding the CTC sensitivity for medium-sized 6- to 9-mm nonpolypoid adenomatous lesions because all but one of the adenomatous lesions in our study were 10 mm or greater. Some ambiguity exists regarding the detection and management of medium-sized polypoid lesions. However, considering the seemingly higher risk for malignancy of nonpolypoid lesions [2, 3], the detection of medium-sized nonpolypoid lesions may have clinical relevance. Therefore, further studies on the CTC sensitivity for medium-sized nonpolypoid lesions are warranted.
The prevalence of nonpolypoid colorectal lesions, in addition to the diagnostic sensitivity of CTC for such lesions, is an important factor in estimating the impact of nonpolypoid lesions on colorectal cancer screening using CTC. The prevalence of genuine nonpolypoid colorectal adenomatous lesions in the screening population is not clearly known yet. In a study by Soetikno et al. [2], 5.84% (36/616) and 0.32% (2/616) of their screening patients had nonpolypoid colorectal neoplasms (i.e., both adenomas and adenocarcinomas) and nonpolypoid stage Tis or T1 adenocarcinomas, respectively. However, their definition of nonpolypoid lesions—lesion height less than half the greatest lesion diameter—left some ambiguity regarding the true flatness of the lesions [30]. Another study by O'Brien et al. [4] showed that approximately 30% of adenomas detected on colonoscopy in the National Polyp Study [31] were nonpolypoid according to the definition of histologically measured adenoma thickness 1.3 mm or less. The lesions appear truly nonpolypoid and would also meet the definition by the Paris endoscopic classification (i.e., ≤ 2 mm lesion height at colonoscopy). However, many (probably more than half) of the nonpolypoid adenomas in the study by O'Brien et al. were 5 mm or smaller in diameter because the reported mean size of the nonpolypoid adenomas was 5 mm [4]. These diminutive lesions should not be discussed in the context of clinically relevant nonpolypoid lesions because they are not clinically significant.
Nonadenomatous lesions in our study showed significantly lower sensitivity on CTC than did adenomatous lesions. In fact, none of the five nonadenomatous lesions was detected on the blinded review. Consistent with our results, the lower sensitivity of CTC for detecting nonpolypoid nonadenomatous lesions was observed in a few other studies [12, 13, 32]. One plausible explanation for this finding may be the tendency of nonadenomatous lesions to efface with air distention of the colon [33]. We believe that the low sensitivity of CTC for nonpolypoid nonadenomatous lesions should be viewed as an advantage of CTC rather than a physical limitation of lesion visualization because it will decrease unnecessary referrals for colonoscopy prompted by the detection of nonadenomatous lesions on CTC.
Our study has limitations. First, it included only a moderate number of lesions. However, it is still one of the largest CTC series of nonpolypoid colorectal lesions reported thus far. To our knowledge, our study is also the first CTC study to adopt the more refined definition for nonpolypoid lesions (i.e., ≤ 2 mm in lesion height at colonoscopy [15]). Second, our study was a retrospective study. Although colonoscopy was requested independently of the CTC results in most of the study patients (16/21), some degree of verification bias [34] and overestimation of CTC sensitivity may exist because of the retrospective sampling of clinical data. Closed forceps were not present in the captured colonoscopy images in most patients. The reference lesion height was retrospectively verified using the colonoscopy reports, pathology reports, and the comparison between the lesion height on captured colonoscopy images and the reported lesion width. Third, no completely flat adenomatous lesions (i.e., stage IIb according to the Paris classification [15]) were included in our study. However, completely flat adenomatous lesions are extremely rare [1]. Fourth, some patients (14/21) received IV contrast enhancement. CTC sensitivity might have been affected slightly by contrast enhancement because contrast material influences the characterization of suspected lesions (if not influencing the initial detection of suspected lesions). Finally, because we investigated a single CAD system, the CAD results may not be generalized to other CAD systems.
In conclusion, CTC, when performed using cathartic preparation, fecal tagging, and good colonic distention and interpreted by experienced readers using a primary 3D method or with CAD, showed a high sensitivity for nonpolypoid (defined as lesions of ≤ 2 mm in height at colonoscopy) stage Tis and T1 adenocarcinomas 10 mm or greater in width despite its limited overall sensitivity for nonpolypoid adenomatous lesions.
Footnotes
Supported by the Korea Research Foundation grant funded by the Korean government (MOEHRD) (KRF-2007-331-E00156).
S. H. Park is a consultant for Infinitt, and L. Bogoni is an employee of Siemens Medical Solutions.
Address correspondence to S. H. Park ([email protected]).
CME This article is available for CME credit. See www.arrs.org for more information.
FOR YOUR INFORMATION
This article is available for CME credit. See www.arrs.org for more information.
References
1.
Soetikno R, Friedland S, Kaltenbach T, Chayama K, Tanaka S. Nonpolypoid (flat and depressed) colorectal neoplasms. Gastroenterology 2006; 130:566-576
2.
Soetikno RM, Kaltenbach T, Rouse RV, et al. Prevalence of nonpolypoid (flat and depressed) colorectal neoplasms in asymptomatic and symptomatic adults. JAMA 2008; 299:1027-1035
3.
Hurlstone DP, Cross SS, Adam I, et al. A prospective clinicopathological and endoscopic evaluation of flat and depressed colorectal lesions in the United Kingdom. Am J Gastroenterol 2003; 98:2543-2549
4.
O'Brien MJ, Winawer SJ, Zauber AG, et al. Flat adenomas in the National Polyp Study: is there increased risk for high-grade dysplasia initially or during surveillance? Clin Gastroenterol Hepatol 2004; 2:905-911
5.
Levin B, Lieberman DA, McFarland B, et al. Screening and surveillance for the early detection of colorectal cancer and adenomatous polyps, 2008: a joint guideline from the American Cancer Society, the US Multi-Society Task Force on Colorectal Cancer, and the American College of Radiology. CA Cancer J Clin 2008; 58:130-160
6.
Johnson CD, Chen MH, Toledano AY, et al. Accuracy of CT colonography for detection of large adenomas and cancers. N Engl J Med 2008; 359:1207-1217
7.
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
8.
Park SH, Lee SS, Choi EK, et al. Flat colorectal neoplasms: definition, importance, and visualization on CT colonography. AJR 2007; 188:953-959
9.
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
10.
Park SH, Ha HK, Kim AY, et al. Flat polyps of the colon: detection with 16-MDCT colonography—preliminary results. AJR 2006; 186:1611-1617
11.
Gluecker TM, Fletcher JG, Welch TJ, et al. Characterization of lesions missed on interpretation of CT colonography using a 2D search method. AJR 2004; 182:881-889
12.
Fidler JL, Johnson CD, MacCarty RL, Welch TJ, Hara AK, Harmsen WS. Detection of flat lesions in the colon with CT colonography. Abdom Imaging 2002; 27:292-300
13.
Pickhardt PJ, Nugent PA, Choi JR, Schindler WR. Flat colorectal lesions in asymptomatic adults: implications for screening with CT virtual colonoscopy. AJR 2004; 183:1343-1347
14.
Taylor SA, Iinuma G, Saito Y, Zhang J, Halligan S. CT colonography: computer-aided detection of morphologically flat T1 colonic carcinoma. Eur Radiol 2008; 18:1666-1673
15.
[No authors listed]. The Paris endoscopic classification of superficial neoplastic lesions: esophagus, stomach, and colon: November 30 to December 1, 2002. Gastrointest Endosc 2003; 58:S3-S43
16.
Compton CC, Greene FL. The staging of colorectal cancer: 2004 and beyond. CA Cancer J Clin 2004; 54:295-308
17.
Dachman AH, Zalis ME. Quality and consistency in CT colonography and research reporting. Radiology 2004; 230:319-323
18.
Park SH, Yee J, Kim SH, Kim YH. Fundamental elements for successful performance of CT colonography (virtual colonoscopy). Korean J Radiol 2007; 8:264-275
19.
Pickhardt PJ, Lee AD, McFarland EG, Taylor AJ. Linear polyp measurement at CT colonography: in vitro and in vivo comparison of two-dimensional and three-dimensional displays. Radiology 2005; 236:872-878
20.
Park SH, Choi EK, Lee SS, et al. Linear polyp measurement at CT colonography: 3D endoluminal measurement with optimized surface-rendering threshold value and automated measurement. Radiology 2008; 246:157-167
21.
Zalis ME, Barish MA, Choi JR, et al. CT colonography reporting and data system: a consensus proposal. Radiology 2005; 236:3-9
22.
Fletcher JG, Booya F, Summers RM, et al. Comparative performance of two polyp detection systems on CT colonography. AJR 2007; 189:277-282
23.
Summers RM, Yao J, Pickhardt PJ, et al. Computed tomographic virtual colonoscopy computeraided polyp detection in a screening population. Gastroenterology 2005; 129:1832-1844
24.
Fletcher JG, Booya F, Johnson CD, Ahlquist D. CT colonography: unraveling the twists and turns. Curr Opin Gastroenterol 2005; 21:90-98
25.
Ferrucci JT. Colonoscopy: virtual and optical—another look, another view. Radiology 2005; 235:13-16
26.
Cotton PB, Durkalski VL, Pineau BC, et al. Computed tomographic colonography (virtual colonoscopy): a multicenter comparison with standard colonoscopy for detection of colorectal neoplasia. JAMA 2004; 291:1713-1719
27.
Mang TG, Schaefer-Prokop C, Maier A, Schober E, Lechner G, Prokop M. Detectability of small and flat polyps in MDCT colonography using 2D and 3D imaging tools: results from a phantom study. AJR 2005; 185:1582-1589
28.
Bogoni L, Wolf M, Lakare S, Dundar M, Jerebko A, Salganicoff M. Automated polyp detection in over 1000 CTC cases: clean and tagged performance. Eur Radiol 2007; 17:A153
29.
Halligan S, Altman DG, Mallett S, et al. Computed tomographic colonography: assessment of radiologist performance with and without computer-aided detection. Gastroenterology 2006; 131:1690-1699
30.
Pickhardt PJ, Levin B, Bond JH. Screening for nonpolypoid colorectal neoplasms. JAMA 2008; 299:2743 [author reply 2743-2744]
31.
Winawer SJ, Zauber AG, O'Brien MJ, et al. Randomized comparison of surveillance intervals after colonoscopic removal of newly diagnosed adenomatous polyps. N Engl J Med 1993; 328:901-906
32.
MacCarty RL, Johnson CD, Fletcher JG, Wilson LA. Occult colorectal polyps on CT colonography: implications for surveillance. AJR 2006; 186:1380-1383
33.
Bertoni G, Sassatelli R, Conigliaro R, et al. Visual “disappearing phenomenon” can reliably predict the nonadenomatous nature of rectal and rectosigmoid diminutive polyps at endoscopy. Gastrointest Endosc 1994; 40:588-591
34.
Obuchowski NA. Special topics III: bias. Radiology 2003; 229:617-621
Information & Authors
Information
Published In
Copyright
© American Roentgen Ray Society.
History
Submitted: December 11, 2008
Accepted: February 23, 2009
Authors
Metrics & Citations
Metrics
Citations
Export Citations
To download the citation to this article, select your reference manager software.