Interactive endodontics: Modeling, simulation and experimental validation

Endodontic or root canal treatment is one of the most important dental procedures employed in modern dentistry. Although there have been many results in surgical simulation in the past two decades, little work has been done on endodontics. The objective of this thesis is to explore critical issues for simulating root canal procedures effectively and efficiently, and to develop an interactive virtual environment featured with both visual and haptic feedback to facilitate endodontic training, which opens new roads to the research of root canal simulation and endodontic training. Beyond endodontics, the results of this work can be extended and applied to other kinds of medical simulation and biomechanical modeling, especially the simulations featured with interactions between soft tissue and solid tool objects.;This work covers both theories and methodologies related to the interactive endodontic simulation and validation, including dynamic modeling, visual and haptic display, model validation and statistical learning. We focus on the simulation of the most critical step in the entire endodontic procedure---shaping root canal with endodontic files. There are four major contributions of this thesis. First, we propose a dynamic model to simulate endodontic shaping operations, which is a smoothed particle based dynamic model derived for the pulpal tissue coupled with a finite element model for the endodontic files. This approach effectively characterizes the special properties and constraints associated with both the pulpal tissue and the endodontic files. Second, we implement the derived dynamic model and build up the virtual environment for endodontic simulation with both graphic and haptic interfaces. Third, we design a novel Robotic Endodontic Measurement System to acquire real haptic data of interactions between root canal and endodontic files during endodontic shaping, and propose an experimental validation method to evaluate the performance of derived dynamic model. Finally, we propose a support vector regression model to accurately characterize the input-output haptic mapping for endodontic shaping operation. The optimized parameters of this model can be learned from robotic endodontic measurements with RBF kernel.