Zsolt Tarjan MD,PhD, Presenter: Nothing to Disclose

Jozsef Koloszar, Abstract Co-Author: Nothing to Disclose

Balazs Forgacs, Abstract Co-Author: Nothing to Disclose

Gabor Csongor Kovacs, Abstract Co-Author: Nothing to Disclose

Mary Hart MD, Abstract Co-Author: Nothing to Disclose

Hans-Juergen Daus MD,PhD, Abstract Co-Author: Nothing to Disclose

To assess the performance of a CAD system and readers with different experience with/without CAD support in evaluating CT colonography.

The data set of 92 colonoscopy correlated CT colonography cases were used to test an own developed 3D morphology based CAD system integrated to an academic virtual colonoscopy application (ColVis). The data set consisted of 36 single and 56 multidetector CT examinations. In 53 CT colonography data sets all together 129 colonoscopically proven polyps >5mm were included and 39 CT colonographies with no polyps. Two radiologists (R1: experienced, R2: beginner in CT colonography) blinded to the colonoscopy results evaluated the sets in random order first with 2D based and than with 3D based reading. The evaluation times were measured in each case. Adding the CAD markings (transparent red signs on polyp candidates) to the 2D and virtual endoscopic views, the evaluation was done again. The diagnostic yield of the CAD system and the readers with and without the use of CAD markings were calculated.

Per polyp sensitivity of R1 was 89%/83% (>10mm/>5mm polyps). Sensitivity of R2 was 81%/67% (>10mm/>5mm polyps). The same sensitivities of the CAD alone were 96%/94%. In an average of 8.2 false positive markings per patient were made by the CAD. After CAD marked the polyp candidates R1 and also R2 read the cases with 96%/95% sensitivity (>10mm/>5mm polyps). Mean evaluation time with 2D based reading was 14/32 minutes, and 11/23 minutes with 3D based reading (R1/R2). Mean evaluation time after the CAD marked the polyp candidates was shorter for both observers in the 2D based reading method (3/5 minutes (R1/R2)) but remained similar using the 3D based method (7/20 minutes) (p < 0.05).

Morphologic CAD can detect colonic polyps with similar sensitivity to well trained readers. In small polyps the CAD seems to be even more sensitive. Less trained readers benefit using CAD and may have similar results to senior readers. CAD markings decrease evaluation time for experienced readers too, especially in 2D based reading.

Tarjan, Z, Koloszar, J, Forgacs, B, Kovacs, G, Hart, M, Daus, H, Effect of CAD in CT Colonography: Diagnostic Accuracies and Evaluation Times Using an Academic Software.  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4418464.html