A real-time reconstructed grid method for soft tissue cutting and haptic

The virtual surgery system is crucial for medical training and preoperative planning. To address the issues faced in existing research, such as distorted grids, cutting with volume loss, low computational efficiency, and surgical tool penetration, this paper proposes a parallel data structure optimized for GPU computation. This structure enables efficient cutting with minimal volume loss and provides force feedback effects without model penetration.

This paper applies the Dual Contouring algorithm within the deformable grid framework for real-time mesh reconstruction, which can be divided into surface reconstruction during deformation and cutting surface reconstruction. By precomputing all possible voxel grid connections and the calculation methods for the feature points of each mesh, feature point positions can be quickly determined within a single GPU thread, accelerating the reconstruction process. Additionally, the force feedback algorithm based on Signed Distance Field and the Finger-proxy model addresses the penetration issue when the surgical tool collides with soft tissue.

Simulation results show that, compared to traditional mesh cutting methods, the proposed approach retains the non-manifold surface features at the cut areas during real-time cutting, and the frame rate does not decrease as the number of cuts increases. The use of the Signed Distance Field for collision detection ensures that the tool has both collision volume and force feedback effects.

The proposed method can handle interactive cutting and collision operations in various surgical simulations and achieves an interactive frame rate that meets real-time requirements, contributing to more efficient and realistic virtual surgery simulations.