Design of an Iterative Unified Enclosure-Guided Machine Learning Framework for Real-Time Dental Defect Diagnosis, Treatment Planning, and Procedural Intelligence Capabilities

Resumen

Current endodontics lacks intraoperative imaging-based defect evaluation and procedural guidance. Fluid artifacts, unstable views, and limited canal topological understanding make static radiography and narrow-field video operator-dependent for diagnostic accuracy, safety margins, and treatment planning. A unified, enclosure-aware computational workflow validates clinical intelligence from raw endodontic video to circumvent these limitations. We offer five analytical methods. For physics-consistent 4D reconstruction, E-STeDIC includes integrated spatio-temporal calibration to improve geometric quality and suppress motion and smear distortions. This calibrated volume feeds RCT-GDP, a topology-driven defect profiling tool that creates a Root Canal Topology Graph including anatomy, defect types, and uncertainty estimates. This representation allows SAFE-TRaC to prescribe treatment paths and enforce safety limits utilizing counterfactual risk estimation using causal modeling and clinical principles. An incremental, tooth-type-aware estimator that adjusts procedure trajectories using intraoperative signals, IT-TAPE successfully forecasts step durations, complexity, and deviations live for the process. MuSRS V's final validation and optimization layer uses a digital-twin generator and multi-scenario robustness analysis to quantify generalization, safety, and performance in different operating conditions. These technologies create a tightly connected system that enhances next-generation intelligent endodontic system fault identification, procedural risks, decision reliability, and reproducibility.