Computer-aided diagnosis for the detection of the pulmonary nodules on digital chest radiography in lung cancer screening / 中华放射学杂志

ABSTRACT

Objective To evaluate the value of computer-aided detection (CAD) system for pulmonary nodule detection using digital chest radiography in lung cancer screening. Methods One hundred consecutive digital chest radiographs from 6280 outpatients for lung cancer screening were independently reviewed by a thoracic radiologist and a computer-aided pulmonary nodule detection system.The radiographs were also reviewed by two experienced thoracic radiologists and the true nodules confirmed by two radiologists with reference to the CT images were marked and stored as a gold standard in the CAD system. The sensitivity and false positive of the radiologist and the CAD system for the detection of nodules on digital chest radiographs were compared. Results Ninety-five and 304 nodules were identified by radiologist and the CAD system, respectively. Of 134 nodules marked as true nodules by experienced radiologists, 82 (61.2%) and 105 (78. 4% ) nodules were identified by the radiologist and the CAD,respectively. The radiologist missed 35 true nodules which were only detected by CAD. The CAD system missed 10 true nodules which were only detected by radiologist. One hundred and twelve (83.6%) nodules were identified by radiologist with the CAD system. One hundred and ninety-nine nodules identified by CAD were false-positive with a rate of 2. 0 ( 199/100 ) per case. Conclusion Combining review of digital radiographs by radiologist with CAD system can improve the detection of pulmonary nodules in lung cancer screening.