To analyze the characteristics of chest CT and the pathological results of malignant small pulmonary nodules and establish the malignant prediction model for pulmonary nodules.

The clinical data of the patients with pulmonary nodules admitted to our hospital from 2014 to 2018 were retrospectively analyzed. The information including the patients′age, smoking history, pathological results, nodule diameter, spicule sign, vascular signs, sign of lobulation, vacuole sign, nodule nature and nodule site was collected and analyzed. The vascular signs were further classified as the blood vessels being located next or attached to the edge of the nodules and the blood vessels passing through the nodules. Single factor analysis of all the factors was carried out by SPSS software and the statistically significant factors were obtained (if P<0.05). And then the significant factors were analyzed by binomial Logistic Regression analysis and the model was established.

The predictive model was P=e^x/(1+ e^x), x=0.269+ (Age×0.051) + (Vessels located next to or attached to the edge of the lesion×0.722) + (Vessels passing through the nodules×4.196) + (Spicule sign 1.144)-(6.95×Calcification)-(3.77×Solid nodules)-(2.21×Ground glass nodules). In this study, the AUC was 0.958 (P<0.01), the predicted value was 0.789, the sensitivity was 87.1% and the specificity was 94.3%. Therefore, the accuracy of the predictive model for the malignant nodules was as high as expected.

The mathematical predictive model for malignant nodules established in this study has a high reliability. And the vascular signs have great values in the determination of the pulmonary nodules.