Histogram analysis based on unenhanced CT for identifying thymoma and lymphoma among prevascular mediastinal incidentalomas

Abstract

Objective To determine whether histogram analysis based on unenhanced CT can play a role in the differential diagnosis of thymoma and lymphoma from thymic hyperplasia and cyst (mean CT attenuation > 10 HU). Materials and methods This retrospective study included consecutive asymptomatic participants who have prevascular mediastinal lesions incidentally detected by unenhanced CT between December 2013 and August 2020, and with definitive diagnosis by pathology or additional radiologic work-ups. A total of thirteen histogram parameters on enhanced CT were calculated for each lesion, then were compared between tumor (thymoma + lymphoma) and non-tumor (hyperplasia + cyst). Receiver operating characteristic analysis was conducted to investigate the performance of histogram parameter for identifying tumor. Results The study population included 192 patients (106 men and 86 women) with a mean age of 50.5 years at the time of CT examination. Of them, 94 patients have tumor (87 thymomas and 7 lymphoma) and 98 have non-tumor (48 thymic hyperplasia and 50 cysts). Nine of the thirteen histogram parameters revealed significant difference between the two groups, including median, minimum, range, 10th percentile, 90th percentile, kurtosis, skewness, uniformity and entropy. No significant difference was observed in the mean CT attenuation between groups. Higher median was found to be independent predictors for distinguishing tumor from non-tumor, and can achieve an area under the curve (AUC) of 0.785 (95% confidence interval [95% IC], 0.720–0.841). Conclusions Histogram analysis based on unenhanced CT may be able to provide some help in the differential diagnosis of incidental lesions in prevascular mediastinal. Grand support This study was sponsored by Natural Science Foundation of Shanghai (No. 21ZR1459700).