Artificial intelligence assisted identification of therapy history from periapical films for dental root canal

Abstract In this work, we aim to develop and validate an AI-assisted method for identifying the history of root canal therapy by using periapical films. First, we propose a pre-processing method to extract the regions of interest (ROI) containing the root canals. Then, in order to improve the generalization ability, data augmentation is adopted to expand the dataset. Finally, three machine learning methods, including SIFT-SVM, CNN, and transfer learning are used. All the models are validated based on the receiving operating characteristic (ROC) curve analysis. The accuracies for the three machine learning methods are above 95%. The AUC, sensitivity, and specificity of the best model are also presented and analyzed.