Machine learning approaches for interactive surgical simulation
Supervisor: Dr Zeike Taylor

The potential impact of surgical simulators on medical training is comparable to that of flight simulators on pilot training. However, central to the effectiveness of any simulator as a training tool is its realism. Accurate and robust physical models of the organs and other soft tissues are therefore required to give realistic visual and haptic feedback to the user. And all of this must be achieved in real-time.

Normally, established techniques like finite element methods are used to solve the underlying physical models of organ behaviour and physiological response. While significant improvements have been made in this area in recent years, their computational speed remains the bottleneck in the overall process. In this project you will develop a new class of simulation tools based on machine learning (ML) techniques. These will be used first to learn the behaviour of regular FE-based models, and will then to replace the latter in an interactive simulation environment. Such approaches are expected to be several orders of magnitude faster than FE techniques, allowing use of far more elaborate and realistic models than is currently possible. While the focus here is on surgical simulation, naturally such technologies will have applications wherever simulation-based engineering is employed.

For more information contact: Dr Zeike Taylor