Simulating Electrical Waves in the Heart using Physics Informed Neural Networks
Physics Informed Neural Networks (PINNs) are an exciting new technology that enable the seamless combination of measured data with theoretical principles expressed as partial differential equations. Being based on Neural Networks, PINNs can be evaluated very quickly, thereby providing substantial computational speed-ups over classical numerical methods. This has implications for the simulation of electrical waves in the human heart, and especially digital twins.
Classical cardiac electrical simulation methods require several hours of compute time on high performance clusters to simulate a typical human heartbeat. This is adequate for planning medical interventions in advance, such as catheter ablation and drug selection, as the simulations can be run before any treatment decision is made. However, digital twin scenarios are more complex, as they involve continuous data transfer between a patient and their digital heart representation. This will require substantial computational speed-ups that can be provided by PINNs.
In this talk I will present the basics of cardiac electrophysiology (EP) simulation, as well as recent work on EP-PINNS by myself and other researchers. Overall, the field of EP-PINNs is in its infancy and rapidly evolving, with good possibilities to do exciting work within the context of a Master’s thesis.
As usual, pizza and refreshments will be served in advance.