| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Gastrointestinal Imaging |
1 From the Department of Radiology, Division of Abdominal Imaging, NYU Medical Center, 560 First Ave, Suite HW 207, New York, NY 10016 (Y.W.L., M.M., G.I., J.B.); Department of Medicine, Division of Gastroenterology, NYU Medical Center, VA Medical Center, New York, NY (E.J.B.); and Department of Biostatistics, Fox Chase Cancer Center, Philadelphia, Pa (H.W.). Received October 24, 2002; revision requested January 9, 2003; revision received February 11; accepted March 13. Address correspondence to M.M. (e-mail: michael.macari@med.nyu.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Twenty-five patients underwent multi–detector row CT colonography prior to colonoscopy. CT images were reconstructed with two methods: 1.25-mm sections reconstructed every 1 mm (thin) and 5-mm sections reconstructed every 2 mm (thick). Two independent readers interpreted thin sections, then waited a minimum of 15 days before interpreting thick sections. With colonoscopy as the reference standard, comparisons were made between interpretation of thin and thick sections, including sensitivity, specificity, and number of false-positive observations. Interpretation times were recorded, and comparisons were made by using repeated measures analysis of variance. For all tests, P < .05 indicated a statistically significant difference.
RESULTS: At colonoscopy, 10 patients had 12 polyps (
5 mm, n = 7; 6–9 mm, n = 2; 
10 mm, n = 3). Sensitivity for polyp detection was statistically indistinguishable for thin and thick sections. Reader 1 had three false-positive findings with thin sections and six with thick sections. Reader 2 had six false-positive findings with thin sections and 11 with thick sections. For both readers, the number of false-positive findings was significantly lower for thin sections than for thick sections (P = .035). Specificity was 93.3% with thin sections and 80.0% with thick sections for reader 1 and 80.0% with thin sections and 73.3% with thick sections for reader 2. Mean interpretation time for reader 1 was significantly longer with thin sections (P < .001). Mean interpretation time for reader 2 was 13.0 minutes for both thin and thick sections.
CONCLUSION: Specificity improved for both readers with thin sections, with no difference in sensitivity.
© RSNA, 2003
Index terms: Colon neoplasms, 758.311 • Colon neoplasms, CT, 75.12111, 75.12117, 75.12118 • Computed tomography (CT), multi–detector row, 75.12111, 75.12117 • Computed tomography (CT), thin-section, 75.12118
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
With the development and implementation of multi–detector row CT scanners, optimal section thicknesses and data acquisition parameters are being investigated (1,8). Early clinical work with single-section CT colonography usually involves 3–5-mm-thick sections with a high degree of image overlap for data acquisition (5,9). Multi–detector row CT scanners can acquire data with thin sections (1 mm), providing near-isotropic voxels.Investigators in two recent studies used multi–detector row CT to evaluate different aspects of this technology: Hara et al (1) used a 4 x 5-mm detector configuration to increase acquisition speed, while Macari et al (8) used a 4 x 1-mm detector configuration in conjunction with a high pitch value to optimize section profile.
Unfortunately, it is not possible to compare different section thicknesses in the same patient by using multi–detector row CT technology because of issues related to patient radiation dose. However, multi–detector row CT scanners allow data to be obtained with a nominal section thickness and subsequently reconstructed into thicker sections if desired. The purpose of this study was to evaluate if differences exist in the interpretation of thin and thick section reconstructions at CT colonography.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Bowel preparation was prescribed by the gastroenterologist (E.J.B.). CT colonography was performed with a Plus 4 Volume Zoom multi–detector row CT system (Siemens; Forchheim, Germany) with the patient in the supine and prone positions to encompass the entire colon. CT parameters included 4 x 1-mm nominal section detector collimation, 120 kV, 0.5-second gantry rotation, and 50 effective mAs, with a variable pitch between 6 and 7 to scan the entire abdomen and pelvis in a single 30-second breath hold. For data interpretation, CT images were reconstructed with two methods: 1.25-mm sections reconstructed every 1 mm (thin) and 5-mm sections reconstructed every 2 mm (thick).
A board-certified gastroenterologist (E.J.B.) performed conventional colonoscopy within 3 hours of CT colonography in all patients. All polyps identified at colonoscopy were photographed, measured, sampled for biopsy, and localized to one of six colonic segments (cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum). Specimens were sent for histologic analysis.
Image Interpretation
Two independent CT readers who were blinded to colonoscopic findings, patient history, risk factors, and demographics interpreted both sets of CT images retrospectively for all patients. The first reader was a fellowship-trained abdominal radiologist (G.I.) with 2 years of experience in CT colonography, and the second reader was a 2nd-year radiology resident (Y.W.L.) with several months of experience in CT colonography. Both readers had experience in CT colonographic image interpretation, with training in at least 50 examinations prior to data interpretation for this study. Image interpretation in these 50 examinations occurred prior to initiation of the current study. The primary display method used was a two-dimensional transverse cine technique with a Vitrea 2 workstation (Vital Images, Minneapolis, Minn). Multiplanar reformatted images, as well as volume-rendered three-dimensional endoluminal images, were used to clarify lesion morphology if an abnormality was detected.
For each patient, thin-section CT images were reviewed first, followed by thick-section images, with a minimum of 15 days (range, 15–27 days) between interpretation sessions. Thick-section interpretation was performed with the images in a different order than those at thin-section interpretation. A third radiologist (M.M.), who coordinated the study, removed patient identifiers from the data sets prior to review. Data sets were evaluated for the presence, location, size, and morphology of polyps. Morphology of detected lesions included two- and three-dimensional analysis of whether the lesion was round, oval, or geometric in configuration. Round and oval lesions were also evaluated for sessile or pedunculated morphology. While round and oval lesions may be polyps or retained fecal material, lesions with geometric (angled) borders are retained fecal material.
For mapping of polyp location, the colon was divided into six segments: cecum, ascending colon, transverse colon, descending colon, sigmoid, and rectum. Polyp location was mapped in an identical fashion to the colonoscopic data, and results were given to the third radiologist (M.M.), who entered findings into a database. Total interpretation time, not including time for data transfer and loading, was recorded.
Statistical Analysis
With colonoscopy as the reference standard, comparison was made between thin- and thick-section reconstructions. Analysis of thin- and thick-section image interpretations and colonoscopic findings was based on segmental location and size and morphologic findings. In this way, a lesion seen at CT colonography could be compared with a lesion seen at colonoscopy. If a lesion detected at CT colonography was similar in size (within 2 mm), had the same morphology, and was located in the same segment as that at colonoscopy, it was considered a true-positive finding. If a lesion detected at CT colonography did not match with a lesion detected at colonoscopy or was not present at colonoscopy, it was considered a false-positive finding.
Overall per polyp sensitivity for each independent reader was calculated for lesions 5 mm or smaller, 6–9 mm, and 10 mm and larger. The total number of false-positive findings was calculated for each reader and for both readers combined. Overall per patient specificity was calculated for each independent reader. Generalized estimating equations based on a binary logistic regression model were used to compare thin and thick sections in terms of sensitivity (per polyp) and specificity (per patient). The bootstrap method with resampling of polyps and patients was used to generate 95% CIs for sensitivity and specificity, respectively. Generalized estimating equations in the context of Poisson regression were used to compare thin and thick section reconstructions in terms of the total number of false-positive findings. Mean interpretation time was calculated for each reader in the evaluation of thin- and thick-section images, and comparisons were made by using repeated measures analysis of variance. For all tests, a P value of less than .05 was considered to indicate a statistically significant difference.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
5 mm, n = 7; 6–9 mm, n = 2; 10 mm, n = 3). Sensitivity was the same for both readers in each size category. For polyps 5 mm or smaller, 6–9 mm, and 10 mm and larger, sensitivity was 43%, 100%, and 100%, respectively, with thin sections. Sensitivity with thick sections was 29%, 100%, and 100%, respectively (Fig 1) (Table 1). Averaged over readers and size categories, overall sensitivity was 66.7% for thin sections and 58.3% for thick sections, but the decline in sensitivity was not statistically significant (P = .299). The 95% CIs of the sensitivity difference between thin and thick sections were 0 and 20.8.
|
|
|
|
|
|
|
For the 15 colons that showed no evidence of polyps at colonoscopy, reader 2 identified 12 (80%) as normal with thin sections and 11 (73%) as normal with thick sections (Tables 3 and 4). With thin sections, there were a total of six false-positive findings for reader 2. These included three lesions (3, 5, and 7 mm) in three of 15 patients with normal findings at colonoscopy and three lesions (3, 3, and 7 mm) in two patients with polyps at colonoscopy. These three lesions were not matched with respect to location, size, or morphology with those seen at colonoscopy.
|
|
|
|
|
|
|
Interpretation Time
Mean interpretation times for reader 1 were 11.0 minutes ± 1.4 (SD) and 8.6 minutes ± 1.2 for thin and thick sections, respectively (Table 5). This was found to be statistically significant (P < .001). Mean interpretation times for reader 2 were 13.0 minutes for both thin and thick sections, with an SD of 5.6 minutes for thin sections and 4.7 minutes for thick sections.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
In a study on the comparison of single– and multi–detector row CT colonography, Hara et al (1) used multi–detector row CT technology to decrease scan time. They used a 4 x 5-mm configuration with 5-mm-thick sections and 3-mm reconstruction interval for both single– and multi–detector row acquisitions. The table speed was three times faster with multi–detector row CT. They found no statistically significant difference in the detection of polyps 10 mm and larger in diameter.
In that study, the specificity for polyps 10 mm and larger at CT colonography was 90% for single-section CT and 93% for multisection CT (1). However, the number of false-positive findings for smaller filling defects 6–9 mm in diameter was not reported. On the other hand, investigators in a recent study (8) used a CT colonography protocol with the aim of maximizing z-axis resolution by using a nominal section thickness of 1.25 mm with polyps 10 mm and larger and found sensitivity of 93% and specificity of 98%.
In the present series on the comparison of thin- and thick-section CT colonography, our findings demonstrate no difference in per polyp sensitivity by using thin versus thick section reconstructions. Most small polyps were missed, and all lesions larger than 6 mm were detected. The size at which a polyp should be considered clinically relevant is debatable (11). There is some evidence that a diminutive polyp (<5 mm) is not clinically important, since many are nonadenomatous, and those that are adenomas usually do not grow or grow very slowly over time (11). These small lesions would likely be detected at follow-up screening if there was substantial growth.
In the present study, when the data from the two readers were combined, the number of false-positive findings was decreased with thin sections when compared with thick sections (P = .035). Reader 1 had three more false-positive findings with thick sections than with thin sections (six vs three), and reader two had five more false-positive findings with thick sections than with thin sections (11 vs six). On a per patient basis, there was a trend toward improved specificity with thin sections, which did not reach statistical significance. Improved specificity and decreased number of false-positive findings with thin-section CT colonography are attributable to less in-plane volume averaging and improved resolution of lesion morphology in the z axis by using multiplanar reformatted and endoluminal images. Since there is less volume averaging with thin-section technique, detection of small gas bubbles or areas of high internal attenuation in a filling defect are more readily apparent than if thick sections were used. Detection of gas or areas of high internal attenuation in a lesion implies that the abnormality is retained fecal material and not a polyp.
Decreasing the number of false-positive findings is extremely important if virtual colonoscopy will ultimately be accepted as a screening tool by our clinical colleagues. Depending on the size of the false-positive finding, the misinterpretation of a fold or residual fecal material as a polyp will lead to unnecessary further diagnostic testing, expense, and patient and clinician dissatisfaction with the procedure.
If we considered a 10-mm polyp clinically important, one patient would have undergone unnecessary colonoscopy on the basis of the interpretation of the thick sections by reader 2. No patient would have undergone unnecessary colonoscopy on the basis of interpretations of thin sections by either reader. If we considered a 6-mm polyp clinically important, then two patients would have undergone unnecessary colonoscopy on the basis of the interpretation of the thick sections by reader 2, and one patient would have undergone unnecessary colonoscopy on the basis of the interpretations of reader 1. With use of thin sections and 6 mm as the cut-off of a clinically important polyp, one patient would have undergone unnecessary colonoscopy on the basis of the interpretation of reader 2, and no patients would have undergone unnecessary colonoscopy on the basis of the interpretations of reader 1.
Thinner sections provide higher resolution when compared with thicker sections. However, there are some limitations of routine acquisition of thin sections, including increased scan time, increased radiation dose, and increased size of CT data sets. Larger data sets result in issues regarding networking, postprocessing, and storage. A typical thin-section CT colonography data set results in 700–800 images. Clearly, data management and workflow issues need to be addressed. Currently, workstations are equipped with extensive hard-drive memory that enables large amounts of data to be stored. Moreover, the processing time of these workstations is continually improving such that currently, a complete CT colonography data set can be loaded and ready for viewing within seconds.
With regard to data acquisition time for a thin-section CT colonography protocol, such as that performed in the present study, each acquisition required 30 seconds to obtain. For most patients, this is a tolerable breath hold. With the advent of eight- and 16-section multi–detector row CT scanners, the acquisition time will be considerably shorter, with the possibility of routinely acquiring sections smaller than a millimeter and further improving z-axis resolution.
With regard to radiation exposure, there is currently a dose penalty when using a four-section multi–detector row CT system with the thinnest possible section collimation (12). However, one can substantially lower the milliampere seconds and hence the radiation dose at CT colonography and still maintain accurate assessment of colonic polyps and neoplasms secondary to the very high contrast between the insufflated gas and the colon wall (8,13). The main reason for the dose penalty with thin sections is related to a penumbra effect of radiation delivered to the patient but not used in image formation (12). With the installation of eight- and 16-section multi–detector row CT systems, this penumbra effect will be reduced, and the dose penalty with multi–detector row CT scanners when acquiring thin-section images will be eliminated or reduced.
Our findings suggest that increased amounts of data increase interpretation time. Reader 1 demonstrated a statistically significant increase in interpretation time when evaluating thin sections. Even for this reader (a 2nd-year abdominal imaging attending radiologist), however, the mean interpretation time was only 11 minutes for thin sections. For reader 2 (a 2nd-year radiology resident), mean interpretation time was 13 minutes. This amount of time compares favorably with previously reported studies of interpretation times and should not limit the utility of CT colonography as an imaging tool to evaluate the colon (9).
There are some limitations to our study. This study comprised a relatively small number of polyps of various sizes in 25 patients. It is possible that if more polyps were present or a larger number of patients were included, sensitivity differences between thin and thick sections may have been found. Second, while efforts were made to decrease recall bias by waiting at least 15 days between interpretations and performing interpretations of the thick section data in a different order than that of the thin sections, it is possible that some recall bias existed. However, we believed that if recall bias existed, it would have impacted the sensitivity of the 5-mm-thick sections positively, since they were interpreted last. It is possible that if a false-positive finding was detected in the initial thin-section data, it may have been recalled when interpreting the thick sections.
Third, while there was a trend toward improved specificity, it did not reach clinical importance. This is likely related to the small number of patients. In this study, specificity calculations were made on a per patient basis. However, there were many more false-positive findings overall for both readers when comparing thick and thin sections. For both readers, the number of false-positive findings with thick sections was almost double that with thin sections. Fourth, a reconstructed section by using thin sections does not necessarily equate to a section obtained with 5-mm beam collimation. However, the reconstructed section likely has a better section profile, since less volume averaging is occurring within the section. Finally, the colonoscopist did not know the CT colonography results when colonoscopy was performed. It is possible that some of the false-positive findings that occurred in this study were real lesions that were not seen at colonoscopy.
In conclusion, when comparing thin and thick sections at CT colonography, our preliminary observations suggest that there is no difference in sensitivity. Specificity was improved for both readers by using thin-section CT, and there were substantially more false-positive findings detected with thick sections than with thin sections. For one of the two readers, however, the interpretation time for thin sections was increased with regard to that for thick sections. We believe that the near-isotropic imaging available with a 4 x 1-mm detector configuration yields improved differentiation of polyps from folds and residual fecal material, which thus decreases unnecessary further testing.
|
FOOTNOTES |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. H. Kim, J. M. Lee, C.-I. Shin, H. C. Kim, J.-G. Lee, J. H. Kim, J. Y. Choi, H. W. Eun, J. K. Han, J. Y. Lee, et al. Effects of Spatial Resolution and Tube Current on Computer-aided Detection of Polyps on CT Colonographic Images: Phantom Study Radiology, August 1, 2008; 248(2): 492 - 503. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Jensch, A. H. de Vries, J. Peringa, S. Bipat, E. Dekker, L. C. Baak, J. F. Bartelsman, A. Heutinck, A. D. Montauban van Swijndregt, and J. Stoker CT Colonography with Limited Bowel Preparation: Performance Characteristics in an Increased-Risk Population Radiology, April 1, 2008; 247(1): 122 - 132. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. T. Johnson, J. G. Fletcher, and C. D. Johnson Computer-Aided Detection (CAD) Using 360{degrees} Virtual Dissection: Can CAD in a First Reviewer Paradigm Be a Reliable Substitute for Primary 2D or 3D Search? Am. J. Roentgenol., October 1, 2007; 189(4): W172 - W176. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. D. Johnson, J. G. Fletcher, R. L. MacCarty, J. N. Mandrekar, W. S. Harmsen, P. J. Limburg, and L. A. Wilson Effect of Slice Thickness and Primary 2D Versus 3D Virtual Dissection on Colorectal Lesion Detection at CT Colonography in 452 Asymptomatic Adults Am. J. Roentgenol., September 1, 2007; 189(3): 672 - 680. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Mang, A. Maier, C. Plank, C. Mueller-Mang, C. Herold, and W. Schima Pitfalls in Multi-Detector Row CT Colonography: A Systematic Approach RadioGraphics, March 1, 2007; 27(2): 431 - 454. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. Yasumoto, T. Murakami, H. Yamamoto, M. Hori, R. Iannaccone, T. Kim, H. Abe, M. Kuwabara, K. Yamasaki, N. Kikkawa, et al. Assessment of Two 3D MDCT Colonography Protocols for Observation of Colorectal Polyps Am. J. Roentgenol., January 1, 2006; 186(1): 85 - 89. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Macari and E. J. Bini CT Colonography: Where Have We Been and Where Are We Going? Radiology, December 1, 2005; 237(3): 819 - 833. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. A. Barish, J. A. Soto, and J. T. Ferrucci Consensus on Current Clinical Practice of Virtual Colonoscopy Am. J. Roentgenol., March 1, 2005; 184(3): 786 - 792. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
