Using CT texture analysis to differentiate between nasopharyngeal carcinoma and age-matched adenoid controls.

OBJECTIVES To investigate the use of texture analysis to quantitatively distinguish nasopharyngeal carcinoma (NPC) from normal adenoid on CT. METHODS In this IRB-approved, retrospective study, nasopharyngeal tissues in 13 patients with NPC and 13 control patients were manually contoured, segmented, and imported to an in-house developed texture analysis program, which extracted 41 texture features. Basic descriptive statistics were performed to evaluate for differences in texture parameters between NPC and controls. RESULTS Statistically significant differences between NPC and controls were seen in 32 of 41 texture features. These significant differences were present in 11 of 12 histogram features, 4 of 5 gray-level co-occurrence matrix features, 7 of 11 gray-level run length features, 4 of 4 gray-level gradient matrix features, and 6 of 9 Laws features. CONCLUSION Significant differences in many texture features were seen between NPC and normal adenoids. CT texture analysis may aid in differentiating NPC from normal adenoid tissue.