Predicting histology of tracheobronchial neoplasms: a CT based differentiation model

Abstract Background: Tracheobronchial (TB) tumors follow same pathological classification as lung neoplasms, however some entities are known to favor airways. Distinction of pathological types is necessary for suggesting appropriate management strategy. Purpose: To evaluate utility of multidetector CT (MDCT) in differentiation of primary TB tumors; and assess validity of a scoring system based on imaging biomarkers to differentiate tumor types. Methods: MDCT features of 45 patients were analysed for location, shape, calcification, attenuation, parenchymal changes, bronchoceles, extraluminal extension, lymphadenopathy,metastases. The two largest groups were compared with each other and remaining entities using Chi square tests. Six point scoring system combining the differentiating features was devised and receiver operating characteristic (ROC) curve analysis performed. Results: The most frequent type was neuroendocrine tumors (NET) (51.1%), followed by salivary gland tumors (SGT) (20%); including adenoid cystic carcinoma (ACC) (13.3%) and mucoepidermoid carcinoma (MEC) (6.7%). Comparing NETs with other entities as a whole, and independently with SGTs, significant difference was found among location (p=0.05 and 0.001 respectively), shape (p On ROC curve analysis, areas under curve for NET, SGT and ACC were 0.913, 0.872 and 0.962 respectively. Suggested cut-off values were >3.5 for carcinoid (sensitivity 70%, specificity 91%), Conclusion: Use of a scoring system enables maximum diagnostic accuracy in MDCT differentiation of TB tumors.