Physics informed learning for high fidelity medical simulators

Details

Medical robots development and validation are time consuming, arduous processes that heavily rely on building fixed hardware architectures and embedded software to be tested in controlled, structured synthetic environments. This approach generally leads to their spurious adoption in clinical practices. With the advances of machine learning (ML), AI and virtual reality platforms, high fidelity simulation of human anatomy models and medical instruments has become ubiquitous for surgical training and pre-planning.

We need high fidelity test environments for fast prototyping of medical devices and validation of their physical design and control policies to bring optimised solutions in a sustainable and robust manner.

The synergy between ML models from relevant data and physics based knowledge of the system offers a unique opportunity to enhance the understanding, prediction, and optimization of body dynamics and their physical interactions with the environment.

Low fidelity ML models, without the knowledge of governing physics and uncertainties in data, rely only on the quality and quantity of machine/environment interaction data. The intricate machine body dynamics and complex interactions of medical devices with the environment are not sufficiently represented in physics-based models which creates a significant disparity between the simulated and real world.

• With physics informed machine learning, and data collected from high fidelity experimental set ups, these complex interactions could be modelled with higher accuracy and transferred into the virtual world for testing and optimisation.

This project will focus on one of common surgical and/or endoscopic procedures.

We will collect relevant data with help of clinical specialist in the area and project collaborator. The project will be carried out in medical robotics group in Sheffield Robotics at University of Sheffield. The group is led by Prof. Dogramadzi (Sheffield Robotics | Sheffield Robotics | The University of Sheffield).

• Eligibility: First degree and Masters in one of engineering and computing fields

• Standard departmental requirements: First Class

• Experience in physical modelling and machine learning, interest in medical robotics.