Comparison of radiomic features of the mandibular condyle between individuals with periodontitis and healthy controls

The aim of this retrospective study was to compare the radiomic features of the mandibular condyle between individuals with periodontitis and healthy controls using cone-beam computed tomography (CBCT) images and to evaluate the potential of radiomics for detecting the subclinical effects of periodontitis on the mandibular condyle. A total of 182 mandibular condyles from 115 individuals (96 healthy condyles and 86 periodontitis condyles) were included. The condyles were segmented using 3D Slicer software, and 107 radiomic features were initially extracted. After preprocessing, low-variance and highly correlated features were removed, leaving 41 radiomic features for analysis. Group comparisons were performed, and least absolute shrinkage and selection operator (LASSO) logistic regression was used for feature selection. Model performance was evaluated using receiver operating characteristic (ROC) analysis with patient-level grouped cross-validation. After correction for multiple comparisons, none of the radiomic features remained statistically significant, and the observed group differences were attenuated after adjustment for age. The radiomics-only model demonstrated limited discriminative performance [area under the curve (AUC) = 0.659]. The clinical model including age and condyle side achieved an AUC of 0.823, whereas the combined model incorporating radiomic features and clinical variables achieved an AUC of 0.818. These findings indicate that radiomic features of the mandibular condyle alone have limited ability to distinguish individuals with periodontitis from healthy controls and do not improve diagnostic performance beyond conventional clinical variables. The observed radiomic differences appear to be influenced predominantly by clinical factors, particularly age.