Modelling Soft Tissue Deformations using Principal Component Analysis and Haptic Devices

This paper describes a new technique to incorporate accurate mechanical deformations of soft tissue into interactive training simulators designed to teach medical procedures. The technique is based on pre-computing a representative set of deformations accurately, then encoding these using principal component analysis such that they can be reconstructed quickly using a small number of shape parameters. Learning algorithms provide the link between haptic devices and the shape parameters. The investigation has been carried out with simulated data representing deformations to an elastic sphere, and it has been shown that real time performance with realistic deformations can be achieved.