Chin A. Yi MD, PhD, Presenter: Nothing to Disclose

David Andrew Olsen MS, Abstract Co-Author: Nothing to Disclose

Justus E. Roos MD, Abstract Co-Author: Nothing to Disclose

David Seungwon Paik PhD, Abstract Co-Author: Spouse is consultant, 23andMe, Inc

Sandy Napel PhD, Abstract Co-Author: Medical Advisory Board, Fovia, Inc Medical Advisory Board, Vital Images, Inc Stockholder, Hologic, Inc Stockholder, General Electric Company

Geoffrey D. Rubin MD, Abstract Co-Author: Speaker, Bracco Group

To understand the reasons why TP nodule candidates by CAD are rejected by radiologists so that we may ultimately improve radiologist performance using CAD.

Twenty adult outpatients were examined with chest MDCT (1.25-mm section thickness and 0.6-mm interval) for pulmonary nodules. Three radiologists independently performed a free search of the scans. Subsequently, additional CAD detections were presented to the radiologists as a list of candidates ordered by CAD score. Radiologists assigned confidence levels of 0-5 to each detection. We considered confidence levels of 0-2 and 3-5 as “rejected” and “accepted” lesions, respectively. The reference standard was established by two additional thoracic radiologists in consensus who performed free search and blindly evaluated all reader and CAD detections. All noncalcified pulmonary nodules ≥ 3-mm in diameter in the reference standard were evaluated for size, location, relationship to bronchovascular bundle, contour, margin, and solid vs. non-solid. The characteristics of rejected nodules were compared to those of accepted nodules using Mann-Whitney test and Fisher’s exact test.

Of 190 ≥ 3-mm noncalcified nodules in the reference standard, 127 were missed by at least one radiologist on free search. 89 of these 127 were detected by CAD. Following radiologist review of CAD detections, 49/89 were accepted by all three radiologists, but 40/89 were rejected by at least one radiologist. When compared with the accepted nodules, the rejected nodules were significantly smaller (4.7 ± 1.7 vs. 3.8 ± 0.8 mm; p<0.01), had indistinct margins (0/49 vs. 11/40; p<0.01), and were non-solid (0/49 vs. 8/40; p<0.01).

Pulmonary nodules that are detected by CAD, but subsequently rejected by radiologists, are significantly smaller, more likely to have indistinct margins, and non-solid.

CAD sensitivity is limited by a tendency for radiologists to reject true positive CAD detections. Knowledge of the character of rejected nodules may help radiologists make better use of CAD for improving sensitivity.

Yi, C, Olsen, D, Roos, J, Paik, D, Napel, S, Rubin, G, Characteristics of True Positive (TP) Lung Nodules Detected by Computer-aided Detection (CAD) But Subsequently Rejected by Radiologists on Chest MDCT Scans.  Radiological Society of North America 2007 Scientific Assembly and Annual Meeting, November 25 - November 30, 2007 ,Chicago IL. http://archive.rsna.org/2007/5011869.html