Computer-aided diagnosis: impact on nodule detection among community level radiologists. A multi-reader study

Early detection of lung nodules is an important clinical indication for obtaining routine CT studies of the thorax. To date, research has focused on the sensitivity of computer-aided diagnosis (CAD) compared with expert chest radiologists typically using data obtained from single detector CT scanners. The present study focuses on the use of CAD as a second reader supplementing four nonexpert "community level" radiologists using state-of-the-art multidetector high resolution data sets. Evaluations of 18 cases with a total of 87 nodules (average 4.8 per case) were subsequently validated by a panel of two expert dedicated chest radiologists. Only 21% of nodules were identified by all four readers; 17% were identified only by CAD. The mean sensitivity of readers before CAD was 49% while following CAD this improved to 72% (p<0.001). When analyzed by individual lobes, the percentage of these in which nodules could be identified increased from 36% prior to CAD to 44% following CAD (p<0.001). These data support the use of CAD as a second reader specifically for nonexpert radiologists in general clinical practice. (C) 2004 CARS and Elsevier B.V. All rights reserved.