Catherine Beigelman-Aubry MD, Presenter: Nothing to Disclose

Marie-Josee Dallaire MD, Abstract Co-Author: Nothing to Disclose

Catherine Hill MBChB, FRCR, Abstract Co-Author: Nothing to Disclose

Anne Laure Brun MD, Abstract Co-Author: Nothing to Disclose

Delphine Leclerc, Abstract Co-Author: Nothing to Disclose

Philippe Andre Grenier MD, Abstract Co-Author: Nothing to Disclose

The purpose of this study was to evaluate a CAD software for the detection of non-solid pulmonary nodules which have a higher prevalence of malignancy than solid nodules.

Twelve CT examinations from 10 patients were selected for having at least one non-solid nodule. There were 7 adenocarcinomas (2 bronchioloalveolar carcinoma and 5 mixed adenocarcinomas). Acquisitions were performed using a 16 detector-row CT scanner (Lightspeed, GE Healthcare, Milwaukee, WI) and a 40 detector-row CT scanner (Brilliance 40, Philips Medical Systems, Best, Netherlands), reconstructed with a lung kernel and a slice thickness / interval of 1.25 mm / 0.6 mm (n=7), and of 0.8 mm / 0.4 mm (n=5) respectively. The ground truth was established by 2 experienced thoracic radiologists in consensus and consists in 13 solid nodules and 265 non-solid nodules. The examinations were read by 2 radiologists and analyzed by the CAD system (MEDIAN Technologies, France). As the CAD software was especially designed for lesions over 3 mm, the performances were evaluated for the sub-category of non-solid nodules greater than 3 mm (n=95). This category was composed of 63 pure groung glass opacities (GGO) and 31 mixed (part solid) nodules.

The sensitivities for non-solid nodules of any size and greater than 3 mm were respectively 26% and 52% for radiologist 1; 28% and 45% for radiologist 2; 38% and 61% for the CAD software. The overall predictive positive value of the CAD was 95%. When adding the findings found by the CAD and by each radiologist, the sensitivities were 52% and 82% for radiologist 1; 55% and 79% for radiologist 2.

The use of a CAD-system as a second observer can significantly improve the detection of non-solid nodules.

Non-solid pulmonary nodules are easely missed and they have a higher prevalence of malignancy than solid nodules. CAD-system as a second observer can significantly improve their detection.

Beigelman-Aubry, C, Dallaire, M, Hill, C, Brun, A, Leclerc, D, Grenier, P, Evaluation of a Computer-aided Detection (CAD) System for Non-solid Pulmonary Nodules.  Radiological Society of North America 2007 Scientific Assembly and Annual Meeting, November 25 - November 30, 2007 ,Chicago IL. http://archive.rsna.org/2007/5012619.html