Joel Garland Fletcher MD, Presenter: Nothing to Disclose

Fargol Booya MD, Abstract Co-Author: Nothing to Disclose

Ronald M. Summers MD, PhD, Abstract Co-Author: Nothing to Disclose

David Roy MD, Abstract Co-Author: Nothing to Disclose

Lutz Guendel, Abstract Co-Author: Nothing to Disclose

Bernhard Schmidt PhD, Abstract Co-Author: Nothing to Disclose

Jeff Lynn Fidler MD, Abstract Co-Author: Nothing to Disclose

C.Daniel Johnson MD, Abstract Co-Author: Nothing to Disclose

Cynthia H. McCollough PhD, Abstract Co-Author: Nothing to Disclose

Nicole Gregor, Abstract Co-Author: Nothing to Disclose

et al, Abstract Co-Author: Nothing to Disclose

To compare two current CAD systems (Siemens Polyp-enhanced Viewing [PEV], and Summers NIH) to determine their performance in a cohort of patients undergoing CTC and endoscopy.

A library of 65 clinical CTC datasets with complete colonoscopy correlation (31 true positive, 34 false positive by prospective radiologist assessment) was constructed. An unblinded radiologist not participating in the CAD analyses documented the exact location of every endoscopically verified polyp. Datasets did not include stool tagging agents, and had 2.5 mm/1.25 thickness/RI. The datasets were evaluated by two scientific teams with proprietary CAD systems (Siemens PEV, prototype of SyngoColon; and Summers NIH), blinded to all endoscopic results. CAD detections were compared to the previously specified reference standard. With respect to counting false positive detections per patient, Summers CAD automatically removes rectal tube detections and predicts ileocecal valve detections, so we manually subtracted any Siemens false positive detections on the rectal tube or ileocecal valve for comparative purposes. Patterns and potential reasons for false negative examinations were assessed.

Siemens CAD had a sensitivity of 73% (16/22) for detecting lesions greater than or equal to 1 cm, while the Summers CAD tool had a sensitivity of 95% (21/22) for similarly sized lesions. Siemens CAD had 0.6 false positive detections/patient, while NIH CAD had 4.6 false positive detections/patient. Of the 6 polyps >= 1 cm missed by Siemens CAD, 4 were flat polyps, a morphology for which their system is not optimized.

1. Similar to ROC differences between radiologists that interpret exams with high sensitivity or high specificity, one CAD system (Summers CAD) performed with higher sensitivity (95% vs. 73%), while the other (Siemens CAD) performed with higher specificity (0.6 vs. 4.6 false positive detections/patient). 2. Current CAD systems require radiologist visual assessment of the ileocecal valve, low rectum and rectal tube. 3. Prospective trials utilizing CAD systems as a primary means of CTC interpretation appear warranted.

R.M.S.: Employees of National Institute of Health Patents pending/awarded in subject area of this abstract.L.G.,B.S.: Employees of Siemens Medical SolutionsJ.G.F.,C.H.M.,N.G.: Grant recipients-Siemens Medical SolutionsC.J.: Software License - GE Healthcare

Fletcher, J, Booya, F, Summers, R, Roy, D, Guendel, L, Schmidt, B, Fidler, J, Johnson, C, McCollough, C, Gregor, N, et al, , Comparative Performance of Two Available CAD Systems in CT Colonography (CTC).  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4417912.html