| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||
ABSTRACT |
Tuesday, May 2, 11:20 AM–12:30 PM
Abstracts 096–101
Moderators: Perry J. Pickhardt, MD and Douglas S. Katz, MD
11:20 AM
096. CT Colonographic Evaluation of Patients with Suspected Colonic Tuberculosis
Vashisht S.*; Jain V.; Gulati M.; Bandhu S.; Garg P.; Radiodiagnosis, All India Institute of Medical Sciences, New Delhi, Delhi, India.
Address correspondence to S. Vashisht (sushvash{at}hotmail.com)
Objective: To evaluate patients with suspected colonic tuberculosis (TB) by CT colonography (CTC) and compare the findings with those of conventional colonoscopy (CC).
Materials and Methods: CTC was performed in 21 patients with suspected colonic TB using 4 slice CT scanner with 3.0/2.5mm effective detector row beam collimation and pitch of 5 following administration of intravenous contrast in both supine and prone positions after standard bowel preparation and per rectal air insufflation. Axial images, multiplanar reconstructions, shaded surface display and volume rendered fly through images of colon were analyzed by 3 radiologists in consensus and findings compared with those of CC. Final diagnosis was based on positive histopathological report and/or a definite positive therapeutic response to anti-tubercular treatment in all patients.
Results: Of 12 patients diagnosed with colonic TB, colonic lumen was decreased in 100% (12/12) on CTC and in 91.66% (11/12) on CC. Wall nodularity was found in 58.33% (7/12) on CTC and in 66.66% (8/12) on CC. Thickened or inflamed ileo-ceacal (IC) valve was found in 100% (12/12) on CTC and in 66.66% (8/12) on CC. Ulcers were identified in 83.33% (10/12) on CC only. Terminal ileum involvement was found in 91.66% (11/12) on CTC, whereas on CC terminal ileum could only be evaluated in 33.33% (4/12) of patients, of which two were diseased. In addition, wall thickening of colon was seen in 100% (12/12) of patients on CTC. Other significant extracolonic findings noted on CTC were pericolonic inflammation (12/12), pericolonic lymph nodes (10/12), mesenteric inflammation/nodes (7/12) and lymphadenopathy at other sites (5/12). Three patients were found to have normal colon on both CTC and CC and alternative diagnosis was suggested in another 3. There were three patients who showed findings suggestive of TB on CTC and CC, but finally proved non-tubercular.
Conclusion: CTC is highly comparable to CC in diagnosing patients with suspected colonic TB. While ulcers are not identified on CTC, it has the additional advantage in demonstrating the colonic mural involvement and extracolonic findings.
097. Incidence of Significant Complications at CT Colonography: Collective Experience of the Working Group on Virtual Colonoscopy
Pickhardt P.J.1,2*; Barish M.A.2; Barlow D.S.2; Choi J.R.2; Dachman A.H.2; Fenlon H.M.2; Ferrucci J.T.2; Laghi A.2; Lefere P.2; Macari M.2; McFarland E.G.2; Morrin M.M.2; Paulson E.K.2; Soto J.2; Stoker J.2; Yee J.2; Zalis M.E.2; 1. Radiology, University of Wisconsin Medical School, Madison, WI; 2. Radiology, Working Group on Virtual Colonoscopy, Boston.
Address correspondence to P.J. Pickhardt (pj.pickhardt{at}hosp.wisc.edu)
Objective: To determine the frequency of significant complications at CT colonography (CTC).
Materials and Methods: A formal survey of the "Working Group on Virtual Colonoscopy" was conducted, with respondents representing 16 medical centers from 5 countries. Primary measurement outcomes included the symptomatic perforation rate and the overall significant complication rate. A "significant complication" was defined asany potentially CTC-related event leading to hospitalization. Data were also collected on the number and type of CTC studies performed, colonic distention technique, and presence of a monitoring physician.
Results: The group performed a total of 21,923 CTC studies dating back to 1997, with 11 of 16 centers performing at least 1,000 examinations. The slight majority of CTC studies (53.4%) were labeled as screening and the remaining 46.6% were diagnostic. Colonic distention was achieved by manual room air insufflation in 59.6% of cases (directly controlled by patient in 12.7%) and by automated carbon dioxide delivery in 40.4%. Direct physician monitoring of CTC studies was reported in 45.8% of cases, with individual center practices ranging from 0–100%. Perforations were recorded in two patients undergoing diagnostic CTC; manual room air insufflation was employed in both cases. One patient was asymptomatic and neither hospitalization nor treatment was required. The second patient was hospitalized and treated but was already symptomatic prior to CTC due to annular sigmoid carcinoma and received only 2–3 puffs of air. The symptomatic perforation rate was therefore 0.0046% (1/21,923). No perforations were recorded in patients undergoing screening CTC or with automated carbon dioxide delivery. Three additional patients were admitted following CTC, two for renal failure and one for chest pain, yielding a significant complication rate of 0.018% (4/21,923).
Conclusion: The safety profile for CTC is extremely favorable, particularly for asymptomatic screening and distention with automated carbon dioxide. Higher complication rates reported by other groups may relate to factors of patient selection and distention technique.
098. Retrospective Analysis of Sources of Error in a Large CTC Clinical Trial
Doshi T.1; Rusinak D.J.1; Halvorsen R.A.2; Rockey D.C.3; Dachman A.H.1*; 1. Department of Radiology, University of Chicago, Chicago, IL; 2. Department of Radiology, Virginia Commonwealth University, Richmond, VA; 3. Department of Internal Medicin, University of Texas Southwestern, Dallas, TX.
Address correspondence to A.H. Dachman (ahdachma{at}uchicago.edu)
Objective: A recent study compared computed tomographic colonography (CTC) and optical colonoscopy (OC) for the detection of colorectal polyps reported a lower sensitivity for CTC than in most previous trials. We categorized CTC false negative (FN) cases to discover the causes for this discrepancy and its impact on CTC interpretation.
Materials and Methods: We retrospectively reviewed CTCs using both 2D and 3D analysis for the 228 polyps found in 152 of the 614 patients available for analysis. Detailed reconciliation with OC reports, photographs and pathology reports was performed. Of the 228 polyps identified, 114 were FN on CTC. We classified errors as observer (measurement or perceptual) or technical (artifact, distension, fluid, excessive stool, undetectable, or scan field of view). Difficult reconciliation or classification cases were decided by consensus of two radiologists experienced in CTC. Statistical analyses were performed on a by-polyp and by-patient basis selecting for adenomas (and carcinomas) and again assuming elimination of correctable technical and observer errors.
Results: In the 228 polyp data set the by-polyp sensitivity of CTC for adenomas and cancers > 10 mm was 63.6% and 61.2% for polyps > 6 mm; the by-patient sensitivity at the 10 mm and 6 mm threshold was 69.6% and 67.0%, respectively. After detailed retrospective reconciliation of individual polyps (so as to exclude to any potentially correctable observer error), the by-polyp sensitivity of CTC for adenomas and cancers > 10 mm increased to 92.7% and by-patient to 91.3%. When accounting for observer and correctable technical errors, 8 of 147 (5.4%) adenomas and cancers > 6 mm were not visible in retrospect. If all correctable technical errors and observer errors were scored as true positives, the sensitivity for adenomas and cancers > 6 mm would have been 94.6% by polyp and 94.8% by patient.
Conclusion: CTC sensitivity increased when considering only malignant and pre-malignant lesions on a by-patient basis. Only 5.4% of adenomatous polyps and cancers were not visible in retrospect when excluding correctable technical errors. Therefore, reduction of observer and technical error would markedly increase the sensitivity of CTC. Future use of computer aided diagnosis, better reader training, and rigorous technique has the potential to further increase CTC sensitivity as a colorectal neoplasia screening modality.
099. Intra-observer Variability of Polyp Volume versus Linear Size Measurement using Supine and Prone Datasets at CT Colonography
Lehman V.T.*; Pickhardt P.J.; Taylor A.J.; Radiology, University of Wisconsin Medical School, Madison, WI.
Address correspondence to V.T. Lehman (vlehman{at}students.wisc.edu)
Objective: Noninvasive surveillance of unresected polyps with CT colonography (CTC) would require reliable detection of small incremental changes. The purpose of this study was to assess the intra-observer reproducibility of linear versus volume measurements of polyp size at CTC. Because polyp size is held constant, the supine and prone series from the same study allow for evaluation of potential factors affecting errors in measurement, including local polyp environment.
Materials and Methods: Measurements of 30 proven colorectal polyps on both supine and prone CTC data sets were performed separately by two experienced abdominal radiologists using the same CTC system (Vitrea2, Vital Images), resulting in a total of 120 linear and volume measurements. Linear size was defined as the longest dimension among the standard orthogonal 2D multiplanar reformatted (MPR) views. Semi-automated volume determination required manual tracing of polyp boundaries on 2D MPR. The relative intra-observer error between supine and prone measurements for a given polyp was defined as 100•(|D1–D2|)/Dave for linear size (D) and as 100•(|V1–V2|)/Vave for volume (V). The relationship between measurement error and polyp morphology (flat, sessile, pedunculated), ease of border determination, luminal distention, and the relationship of the polyp to folds and positive oral contrast material (submerged, partially submerged, coated, or surrounded by air) was determined.
Results: Median linear size and volume of polyps was 9.4 mm and 270 mm3, respectively. The overall mean relative error for linear size and volume measurement was 11.2 ± 9.2% and 19.7 ± 16.9%, respectively. Magnitude of measurement error did not correlate with polyp morphology, border determination, luminal distention, or relationship to folds or contrast.
Conclusion: Because changes in volume are amplified compared to linear dimension for a given change in polyp size, the error associated with volume measurement, although greater in absolute value, should allow for improved detection of incremental changes in polyp size. Measurement error was not dependent on polyp morphology, luminal distention, or relationship to contrast.
100. Characteristics of Advanced Adenomas Detected at CT Colonography Screening
Kim D.H.1*; Pickhardt P.J.1; Taylor A.J.1; Pfau P.R.2; 1. Department of Radiology, University of Wisconsin Medical School, Madison, WI; 2. Section of Gastroenterology and Hepatology, University of Wisconsin Medical School, Madison, WI.
Address correspondence to D.H. Kim (dh.kim{at}hosp.wisc.edu)
Objective: The advanced adenoma is the primary target in colorectal screening since increased detection and removal of these lesions would favorably impact cancer incidence. The purpose of this study is to delineate characteristics of advanced adenomas detected at CT Colonography (CTC) in a screening population.
Materials and Methods: Retrospective review of two CTC databases encompassing 2,966 patients over a 30-month period was performed to identify all pathologically proven adenomas measuring = 6 mm. A group of 75 unresected polyps currently enrolled in a CTC surveillance program was excluded from analysis. Cases of proven adenocarcinoma were also excluded. From the known adenoma group, the advanced adenomas were extracted. Advanced adenomas were defined as lesions that measured = 10 mm, demonstrated a prominent villous component, or exhibited high-grade dysplasia. Various characteristics such as lesion size, morphology, histology, and location were recorded.
Results: Out of a total of 283 adenomas = 6 mm, 102 (36%) were classified as advanced lesions. Mean size of advanced adenomas was 16 ± 9.4 mm. The majority of lesions qualified as advanced by the size criterion alone (97/102; 95%). The 5 lesions measuring 6–9 mm constituted 2.8% of all medium-sized adenomas (5/181). Most advanced adenomas demonstrated tubular histology (57/102; 56%), followed by tubulovillous histology (42/102; 41%), and pure villous histology (3/102; 3%). The majority of advanced adenomas were either sessile (45/102; 44%) or pedunculated (48/102; 47%); a minority were flat (8/102; 8%). The majority of non-advanced adenomas were sessile (143/181; 79%). Advanced adenomas were evenly distributed between the proximal (49/102; 48%) and distal colon (53/102; 52%), demarcated by the splenic flexure.
Conclusion: In this series, advanced adenomas were rarely under 10 mm in size, tended to be sessile or pedunculated in morphology; and were evenly distributed between the proximal and distal colon. Polyp size alone may serve as a useful surrogate marker for advanced adenomas at CTC, suggesting that medium-sized lesions (6–9 mm) may be amenable to CTC surveillance schemes.
101. CT Colonography: Performance of a Multi-template CAD Algorithm for Polyp Detection
Blake M.E.*; Soto J.A.; Ferrucci J.T.; Radiology, Boston University Medical Center, Boston, MA.
Address correspondence to M.E. Blake (meghan.blake{at}bmc.org)
Objective: Computer-aided detection (CAD) tools can be valuable adjuncts to screening CT colonography interpretation, by identifying suspicious regions of interest in large datasets for final review by the radiologist. In this study, we evaluated the perfor-mance of a specific CAD tool in a group of colonoscopically-proven CT colonography data sets.
Materials and Methods: For detecting suspicious lesions, the CAD algorithm we tested compares calculated texture and surface feature values within the CT dataset to reference values embodied in "templates". For this study, 95 cases with 126 colonoscopically-proven polyps were collected from five international sites. These were randomly separated into two cohorts, 50 cases for training and 45 cases for testing, so that performance of the CAD algorithm would be measured with fresh new cases. Multiple iterations of this process in the 50 training datasets resulted in a multi-template algorithm, which was then tested with the remaining 45 datasets. The outcomes measured were sensitivity at various size thresholds and number of false-positive findings were acquisition.
Results: When considering only lesions that were identified at colonoscopy and were visible at retrospective review of the CT datasets, the sensitivity of the CAD tool was 92% for polyps 0.55 to 0.89 mm in maximum diameter, 100% for polyps 0.9 to 2.0 cm in diameter, and 100% for polyps sized 2.0 cm or greater. The sensitivity for all colonoscopically-identified polyps (irrespective of whether or not they were found at the retrospective review of the CT images) was 61%, 95%, and 100%, respectively, for the three size categories. Mean number of false positive findings per acquisition was 6.3.
Conclusion: Through the use of different datasets for CAD software training and testing, we have been able to achieve acceptable sensitivity for polyps, producing relatively few false positives per acquisition. However, detection by CAD tools may be limited to those lesions that are visible on the imaging data set, which ultimately relies on a well cleansed and well distended colon.
* Will present paper
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
