To explore the CT imaging features of thymic epithelial tumors, and evaluate the reproducibility of preoperative staging diagnosis.

The clinical data of 30 patients with thymic epithelial tumors confirmed by surgery and pathology in Shanghai Chest Hospital affiliated to Shanghai Jiao Tong University between April 2010 and July 2010 were retrospectively collected. Based on the CT imaging diagnosis criteria proposed by International Thymic Malignancy Interest Group, the comparative analysis of description of tumor characteristics by three observers with CT was done, and the reproducibility of assessment of extent of tumor invasion before operation and consistency with postoperative pathological diagnosis were evaluated. The senior surgeon was served as primary observer, and one thoracic surgeon with 5 years working experience and one radiologist with 10 years working experience were served as control observers. Kappa value of 0 to 0.40 indicated a poor diagnosis consistency, Kappa value of 0.41 to 0.75 a fair diagnosis consistency, and Kappa value of 0.76 to 1.00 a good diagnosis consistency.

For imaging features of thymic epithelial tumors, such as the diagnosis of tumor location, density and contour, three observers agreed well(Kappa value, 0.634-0.821). For judgment of tumor contour, heterogeneous attenuation and calcification, the primary observer mostly consisted with the other two observers(Kappa value, 0.4-0.626), while the diagnosis of the other thoracic surgeon and radiologist showed poor consistency(Kappa value, 0.216-0.313). For judgment of tumor invasion to mediastinal fat or lung, the diagnosis consistency of three observers was poor(Kappa value, 0.104-0.585), while the diagnosis consistency for judgment of tumor invasion to mediastinal pleural, pericardial or mediastinal great vessels and phrenic nerve was fair(Kappa value, 0.38-0.839). The diagnosis of invasion to mediastinal fat and pleura by three observers was poorly consistent with postoperative pathological results(Kappa value, 0.203-0.493). The diagnosis of tumor invasion to lung by primary observer was poorly consistent with postoperative pathological diagnosis(Kappa value, 0.135), while the other two observers had good consistency(Kappa value, 0.366-0.712). The judgment of tumor invasion to pericardial, mediasinal great vessels and phrenic nerve by three observers were all mostly consistent with postoperative pathological diagnosis(Kappa value, 0.366-0.87).

CT could be used to describe the imaging features of thymic epithelial tumors. CT has better reproducibility in assessment of extent of tumor invasion before operation, and has better consistency with postoperative pathological diagnosis, which may enjoy a high value in preoperative tumor staging.