Modeling Realistic Tool-Tissue Interactions with Haptic Feedback: A Learning-based Method

摘要

Surgical simulators present a safe, practical, and ethical method for surgical training. In order to enhance realism and provide the user with an immersive training experience, simulators should have the capability to provide haptic feedback to the user. High-fidelity surgical simulators also require accurate modeling of the interaction between surgical instruments and organs. Linear elasticity-based models are commonly used to simulate tool-tissue interaction due to computational considerations, although real soft tissues exhibit nonlinear viscoelastic behavior. In this paper, we use a learning algorithm to train a linear 2D mass-spring-damper system that behaves similarly to a high-fidelity nonlinear finite element (FE) model. The spring parameters are trained off-line using data from an FE simulation of brain tissue deformation using simultaneous perturbation stochastic approximation, a model-free optimization algorithm. The model is implemented in a real-time soft tissue simulator with haptic interaction provided through the PHANTOM Omni haptic device. Our model's response is significantly closer to the desired response of the FE model than that of a linear heuristic model.