OBJECTIVE: This study aimed to develop and validate a clinico-radiomic nomogram, integrating Cone-Beam Computed Tomography (CBCT) radiomic features with clinical characteristics, to predict the difficulty of mandibular third molar (MTM) extraction. METHODS: A retrospective cohort of 600 patients undergoing MTM extraction was divided into training and validation sets. Radiomic features were extracted from preoperative CBCT images. The Least Absolute Shrinkage and Selection Operator (LASSO) method was used to select key features and build a radiomic score (Rad-score). Multivariable logistic regression identified independent clinical predictors. A final nomogram was established by combining the Rad-score and clinical factors. The model's performance was assessed for discrimination (AUC), calibration (Hosmer-Lemeshow test), and clinical utility (Decision Curve Analysis - DCA). RESULTS: Eleven robust features were selected to construct a radiomic signature. The Rad-score was significantly higher in the high-difficulty group (P < .001). Five independent predictors were identified: Age, BMI, Pell & Gregory classification, Root Curvature, and the Rad-score. The combined clinico-radiomic nomogram demonstrated superior predictive performance in both training (AUC = 0.892) and validation (AUC = 0.865) cohorts, significantly outperforming models based solely on clinical or radiomic factors alone. The model showed excellent calibration between predicted and observed probabilities and demonstrated substantial clinical net benefit via DCA. CONCLUSION: The novel CBCT-based clinico-radiomic nomogram provides a noninvasive, accurate, and visual tool for the preoperative stratification of MTM extraction difficulty. This facilitates more informed surgical planning and personalised patient counselling.
Xu et al. (Thu,) studied this question.