Background and purpose: Solitary pulmonary nodule (SPN) detected by computed tomography (CT) in breast cancer patients may be primary lung cancer (PLC) or metastatic pulmonary breast cancer (MBC)

 and definitive diagnosis of nodule is very important for personalized medical care of patients. In this study

we retrospectively analyzed high resolution CT (HRCT) features of MBC and PLC shown as SPN in breast cancer patients and evaluated their diagnostic values. Methods: From Oct. 2007 to Jun. 2014

data of 116 patients with 116 SPNs confirmed by pathology were assessed with respect to preoperative CT findings retrospectively. We reviewed CT signs of nodules to differentiate MBCs and PLCs by using multivariate logistic regression model and receiver operating characteristic (ROC) curve. Results: Of the 116 SPNs

77 (66.4%) were PLC and 39 were (33.6%) MBC. There were statistically significant differences in round or oval shape

sub-solid nodule

spiculation

lobulation

bubble lucency

air bronchogram

pleural tag

maximum diameter of nodule and portion of the solid volume component between MBCs and PLCs. Multivariate analysis showed that spiculation

presence of air bronchogram and portion of the solid volume component 24% were statistically significant independent predictors of PLCs

with the corresponding odds ratio (OR) being 0.219 (95%CI: 0.069-0.638)

0.061 (95%CI: 0.009-0.240) and 1.029 (95%CI: 1.010-1.052)

respectively. This multivariate regression model had a good performance in differentiating PLC from MBC

with an area under the ROC curve (AUC) of 0.903 (95%CI: 0.838-0.969)

an accuracy of 86.2%

a sensitivity of 88.3% and a specificity of 82.1%. Conclusion: In all SPNs of breast cancer patients

the probability of PLC is higher than that of MBC. Less solid component

absence of spiculation and air bronchogram are highly indicative of PLCs.