Abstract: Objective To observe and analyze differentiation of benign and malignant pulmonary nodules by computer-aided diagnosis(CAD) of digital chest radiography. Methods A total of 154 patients with singular nodule treated in our hospital between January 2016 and June 2017 were recruited as subjects for the retrospective analysis. The subjects include 80 patients with malignant nodule and 74 with benign nodule. All malignant pulmonary nodules were confirmed by pathology, while all benign pulmonary nodules were confirmed by CT scans and their reading by more than 2 radiologists. All chest X-ray films were obtained from the Digital Radiography (DR) system, which were read by 5 senior radiologists and 5 junior radiologists. The chest X-ray films with and without CAD were analyzed. The analysis was performed using the Receiver Operating Characteristic (ROC) curve for observers’ scores. Results The average area under the ROC curve (AUC) increased gradually from 0.713 (without CAD) to 0.792 (with CAD) (P<0.05) in junior radiologists group. But senior radiologists group showed no significant difference (P>0.05). The risk scores of AUC output from CAD was positively correlated with the malignant nodules. Therefore, the enhancement of AUC showed significant differences. Conclusion CAD of digital chest radiography can more effectively facilitate junior radiologists diagnosing and differentiating benign and malignant pulmonary nodules in patients with singular nodule. The patients with early accurate diagnosis thus can receive timely appropriate treatment. Therefore, CAD of digital chest radiography deserves wide promotion in clinical practice.

Key words: Digital chest radiography, Computer-aided diagnosis, Pulmonary nodule, Benign or malignant