Web-based modelling with applications to surgical training

Web-based surgical training has the potential to offer a cheap training platform, accessible from anywhere in the world, at any time. However the use of web-based VR for training faces a number of challenges including fast collision detection and real-time deformable modelling. In this thesis, these problems are addressed and solutions are given to meet these challenges.

In collaboration with the neurosurgeons at Leeds General Infirmary, the use of web-based VR to train neurosurgeons is investigated, using the percutaneous rhizotomy procedure as a case study.A number of surgical tasks are identified. The training requirements that can be met using a web-based approach, exploiting VRML and Java, are described. A novel solution for collision detection is presented. To provide an effective training platform, an assessment tool is developed and integrated in the training simulator to allow the progress of trainees to be monitored.

For the deformable modelling of soft objects, the extension of the existing 3D ChainMail algorithm, which is only able to model a volumetric object represented by a uniform rectilinear mesh, is achieved in two ways. First, the algorithm is modified so that it can be applicable to a surface represented as a non-uniform rectilinear mesh. The modified algorithm, called SurfaceChainMail in this thesis, is implemented in the simulation of the cutting of soft tissue. Secondly, the 3D ChainMail algorithm is further extended to arbitrary meshes, without sacrificing computation speed. This new algorithm is named as the Generalised ChainMail algorithm. An important aspect of this work is that both surfaces and volumes can be handled by the same basic approach: a surface represented as a triangular mesh in 3D space is deformed using the same algorithm as a volume represented as a 3D tetrahedral mesh.