Kouichi Nagata, Presenter: Nothing to Disclose

Hiroyuki Yoshida PHD, Abstract Co-Author: Patent holder, Hologic, Inc Patent holder, Hologic, Inc (R2 Technology, Inc) Patent holder, MEDIAN Technologies

Michael Ethan Zalis MD, Abstract Co-Author: Research grant, General Electric Company

Philippe Lefere MD, Abstract Co-Author: Consultant, E-Z-EM, Inc Researcher, Vital Images, Inc Medical Advisory Board, Vital Images, Inc

Stefaan Gryspeerdt MD, Abstract Co-Author: Researcher, Vital Images, Inc Medical Advisory Board, Vital Images, Inc Researcher, E-Z-EM, Inc Medical Advisory Board, E-Z-EM, Inc

Janne Johannes Nappi PhD, Abstract Co-Author: Nothing to Disclose

To compare CAD performance with unaided human reading in laxative and non-laxative CT colonography (CTC).

Database DB1 included 47 non-laxative multicenter CTC cases with fecal tagging by barium (n=33) or non-ionic iodine (n=14). Database DB2 included 47 consecutive laxative fecal-tagging CTC cases from the Walter Reed Army Medical Center database. The cases were analyzed by our fully automated CAD scheme that had been trained to detect polyps ≥6 mm by use of an anthropomorphic fecal-tagging phantom and 146 laxative CTC cases including 62 polyps ≥6 mm. The CAD detection sensitivity was compared with that of unaided human reading at <5 FP CAD detections per scan on average.

DB1 had 5 (7) adenomas ≥10 mm (6-9 mm) in 7 patients, and DB2 had 4 (3) adenomas ≥10 mm (6-9 mm) in 7 patients. Unaided human CTC reading sensitivity was 80% (86%) in DB1 and 100% (67%) in DB2 for adenomas/carcinomas ≥10 mm (6-9 mm). The corresponding CAD sensitivity was 100% (50%) with 2.3 (4.7) FPs per scan in DB1 and 100% (100%) with 0.1 (4.0) FPs per scan in DB2. CAD detected 1 (1) adenomas ≥10 mm (6-9 mm) missed by unaided CTC reading in two suboptimally tagged DB1 cases, and a 7-mm adenoma missed by unaided reading in DB2. Most FP CAD detections in DB1 were caused by folds and small pieces of untagged feces, whereas in DB2 most FP CAD detections were caused by folds.

CAD can detect polyps missed by unaided human readers in both laxative and non-laxative CTC. CAD can detect adenomas ≥10 mm at high sensitivity, but presence of untagged stool can reduce detection sensitivity for smaller polyps.

A computer-aided detection system with high performance in both laxative and non-laxative CT colonography would have wide utility in a clinical setting.

Nagata, K, Yoshida, H, Zalis, M, Lefere, P, Gryspeerdt, S, Nappi, J, Performance Comparison of Computer-aided Detection in Laxative and Non-laxative CT Colonography.  Radiological Society of North America 2007 Scientific Assembly and Annual Meeting, November 25 - November 30, 2007 ,Chicago IL. http://archive.rsna.org/2007/5011971.html