*This webpage will be updated and the materials for the tutorial will be uploaded to the GitHub repository in due course.
This tutorial aims to provide an introduction to physics-informed neural networks (PINNs) using examples from medical imaging. It is expected that by the end of the session attendees will be able to use PINNs to solve their own problems related to medical imaging and beyond. PINNs are a type of deep learning framework that explicitly incorporates physical equations into the learning process as a regulariser, such that the network approximates the data whilst conforming to the equations and associated boundary and initial conditions. This enables them to learn with a relatively small amount of data compared to conventional neural networks and provides trust that the NNs’ inferences obey the known physics laws.
The tutorial will:
In addition, there will be a hands-on code-based session that covers examples from cardiovascular medicine and neuroscience. In detail:
Marta Varela (National Heart and Lung Institute, Imperial College London), Ching-En Chiu (Department of Electrical and Electronic Engineering, Imperial College London), Christoforos Galazis (National Heart and Lung Institute, Imperial College London), Yu Hon On (National Heart and Lung Institute, Imperial College London), Danilo Mandic (Department of Electrical and Electronic Engineering, Imperial College London), Phil Livermore (School of Earth and Environment, University of Leeds), Zack Xuereb Conti (Data-Centric Engineering / TRIC:DT, The Alan Turing Institute, London)