Predictive Analytics in Endodontics: Machine Learning Approaches for Treatment Success and Failure Prediction

Endodontic treatment success or failure prediction continues to be an important challenge as a result of the intricate relationship between clinical, radiographic, and procedural factors. Current prediction models either make use of patient data in structured form with the help of conventional machine learning (ML) algorithms or analyze radiographic images through deep learning, both of which are limited by accuracy and generality. This research suggests a hybrid model that incorporates Random Forest (RF) and Convolutional Neural Networks (CNN) to improve prediction performance by combining structured clinical data with radiographic analysis. The study utilized a 2,000-case dataset including patient demographics, clinical symptoms, radiographic signs, and procedural information. Standalone models were outperformed by the hybrid RF + CNN model, with 85.6% accuracy, 87% precision, 83% recall, and an F1-score of 85%, against 80.5% accuracy for RF and 83.2% accuracy for CNN. Root canal morphology, preoperative infection status, and obturation quality were major predictive factors found. These results prove that the hybrid method greatly enhances predictive performance, providing endodontists with a more accurate tool for clinical decision-making, risk evaluation, and treatment optimization.