CT texture analysis for differentiating odontogenic cysts and tumors: A pilot study

Objectives: To evaluate the diagnostic utility of computed tomography texture features in differentiating odontogenic cysts and tumors. Methods: This cross-sectional pilot study included 30 patients with odontogenic cysts and tumors. Lesions were manually segmented using MRIcron software. Histogram, gray-level co-occurrence matrix, and run-length matrix features were extracted. Independent t-tests, logistic regression modeling, and receiver operating characteristic curve analysis were performed for statistically significant features. Results: Histogram mean, median, and mode values were significantly higher in cystic lesions compared to tumors (p < 0.05). Combined logistic regression modeling demonstrated a diagnostic accuracy of 66.7% with an area under the curve of 0.826, sensitivity of 0.875, and specificity of 0.643. Individual features demonstrated limited discriminatory ability when analyzed independently. Conclusions: CT texture analysis demonstrated promising diagnostic potential when combined feature models were applied. These findings support the role of texture analysis as a non-invasive adjunct tool for lesion characterization, warranting validation in larger cohorts.