Detection of ectopic focus in the heart with machine learning on a large virtual population of electrophysiological simulations obtained from HPC based solvers Detection of ectopic focus in the heart with machine learning on a large virtual population of electrophysiological simulations obtained from HPC based solvers

During the last few years we have developed computational models to predict the site of origin (SOO) of ectopic foci in ventricular tachycardias placed in the outflow tracts of the heart ventricles. In collaboration with the Barcelona Supercomputer Centre (BSC) we built a pipeline to run hundreds of simulations to build large virtual populations that could help to train machine learning algorithms for faster SOO detection. The main goal of this project is to adapt the developed pipeline to work in the two ventricles of the heart, simulation pseudo-ECGs and ECGs (eventually after an automatic process segmenting the 4-chamber heart from CT images and orient them in a torso). Advanced machine learning techniques will be developed to compare real and simulated data, including transfer learning.

Supervisor: Oscar Camara