Abstract:

Objective To explore the diagnostic value of the radiomic model based on panoramic radiographs in differentiating single or multiple root canals in mandibular first premolars. Methods A retrospective analysis was conducted on 68 patients with CBCT and panoramic radiographs from the affiliated Stomatological Hospital of Nanjing Medical University between July 2022 and November 2023. One hundred and thirty one mandibular first premolars (105 single-root and 26 multiple-root canals) were included in this study. Using stratified random sampling, the mandibular first premolars were divided into a training cohort (104 teeth) and a test cohort (27 teeth) at a ratio of 8∶2. CBCT was used to confirm the root canal configuration, and radiomic features were extracted from panoramic radiographs. After feature screening, radiomic models were constructed using extratrees, LightGBM, logistic regression(LR), multi-layer perceptron(MLP), random forest, support vector machine(SVM), and XGBoost. The optimal radiomic model was selected based on receiver operating characteristic(ROC) curve, accuracy, sensitivity, specificity, positive predictive value(PPV), negative predictive value(NPV), and decision curve analysis(DCA). Results A total of 107 radiomicfeatures were extracted from panoramic radiographs. After feature filtering, 4 radiomic features were ultimately selected using the Lasso model. The XGBoost model showed higher predictive efficiency and stability, with an AUC of 0.962 in the training cohort and 0.770 in the test cohort. DCA curves also indicated that the XGBoost model was optimal. Conclusion The radiomic model based on panoramic radiographs demonstrates good performance in differentiating single or multiple root canals in mandibular first premolars.

Key words: radiomics, mandibular first premolar, root canal classification, panoramic radiographs