We evaluated a computer-aided diagnosis (CAD) system with automatic detection of pulmonary nodules for lung cancer screening with computed tomography (CT). Five hundred and eighteen participants were examined with low-dose helical CT during a lung cancer screening by three respiratory physicians according to the General Rule edited by the Japan Lung Cancer Society. Four cases were detected by CAD and pathologically diagnosed as lung cancer. We compared the detection capability of the physician and CAD in 301 participants. Three physicians determined 75/301 (24.9%) participants as "e" (suspicious of lung cancer) in consensus without CAD, while 3 participants were added to "e" with CAD. Three physicians did not independently judge as "e" in 14 (18.7%), 16 (21.3%) and 16 (21.3%) out of 75 participants. CAD could not identify 17 (22.7%) nodules of 75 participants, and all 17 were less than 6 mm in diameter. The CAD system offers a useful second opinion when physicians examine patients at lung cancer CT screenings.