CAD for VC improves reader performance, sensitivity for larger polyps

The addition of computer-aided detection (CAD) systems will help virtual colonoscopy make the transition into a standard-care modality, according to two studies presented this month at the 2004 American Roentgen Ray Society meeting in Miami Beach, FL.

"We believe that CAD will have a major impact in the transition of virtual colonoscopy from a research tool to general clinical practice," said Dr. Abraham Dachman, professor of radiology at the University of Chicago. "We think that CAD may be the single focal point that will help make CT colonography readable with a high degree of confidence by a general radiologist."

Dachman and his colleagues conducted an observer-performance study to evaluate the effect of CAD on CT colonography, or virtual colonoscopy. The group collected 20 datasets consisting of 10 abnormal sets with 11 polyps ranging in size from 5-12 mm, along with 10 normal datasets.  The information was collected retrospectively from CTC examinations of patients who had undergone a standard pre-colonoscopy bowel preparation and room-air colonic distension.

The data were examined using two different displays -- one with 2D reformatted views and the other using a real-time 3D endoluminal-perspective view "with a fly-through mode for 3D problem-solving," Dachman said.  The team set the detection performance of the CAD scheme to a by-polyp sensitivity of 92%, with an average of one false-positive per dataset.

The researchers selected four readers with varying degrees of training to evaluate the datasets and rate their diagnostic confidence levels in terms of the presence of polyps 5 mm in diameter or larger. Data were evaluated both with and without computer-aided detection, with readers rating their confidence level with each reading.

The team conducted a receiver operating characteristic (ROC) analysis, calculating the value of the area under the ROC curve (Az) as a measure of the detection performance.   Differences between readers’ performance with and without CAD were then evaluated using the Student’s t-test.

According to the results, the detection performance increased for all observers when aided by CAD. The average Az value for radiologists with CAD was 0.85 and 0.70 without CAD. Dachman's group found the difference to be statistically significant (p=-0.025). Among the four observers, the use of CAD improved results most dramatically (Az = 0.21) for the gastroenterologist, Dachman said.

CAD and CTC plus contrast

In another presentation, Dr. Ronald Summers from the National Institutes of Health in Bethesda, MD, discussed how CAD works when paired with contrast-enhanced virtual colonoscopy for polyp and mass detection.

For this study, Summers’ team enrolled 29 patients who had a high suspicion of having colonic polyps or masses. All patients underwent VC, and the results were correlated with conventional colonoscopy performed the same day.

Noncontrast prone CT imaging was performed first, followed by intravenous contrast-enhanced supine imaging. Parameters for scanning were 3-mm slice thickness, 1.5-mm reconstruction intervals, and routine (non-reduced) mAs. Only polyps 1 cm and larger were evaluated using the prototype CAD system, he said.

According to the results, the CAD software detected 27 of the 34 polyps or masses confirmed by colonoscopy, with an average of 2.2 false-positives per patient. Overall per-patient sensitivity was 88%, with supine contrast-enhanced CT performing slightly better than the non-contrast prone exam.

Summers pointed out that these findings are preliminary, but concluded that the CAD scheme yielded a high sensitivity and a low false-positive rate for the detection of polyps and masses 1 cm or larger -- and that sensitivity that is improved with intravenous contrast administration.

By Jerry Ingram contributing writer
May 18, 2004

Related Reading

VC researchers push for study quality, consistency, March 17, 2004

Experience sharpens role of IV contrast-enhanced VC, January 19, 2004

Mass appeal: VC CAD doesn't stop at polyps, October 10, 2003

Copyright © 2004

Page 1 of 655
Next Page