Design and implementation of a medical education module on in-silico simulations using high-performance computing

Abstract

Outflow tract ventricular arrhythmias (OTVAs) are diagnosed by interpreting suggestive features from 12-lead electrocardiograms (ECGs) and they are commonly treated by radiofrequency ablation (RFA). High success rates are obtained with RFA provided that the site of origin (SOO) of an arrhythmia is correctly inferred. An approach to predict the SOO of OTVAs using patient-specific cardiac electrophysiological simulations was previously proposed by Dr. Doste. However, the pipeline was too complicated for nonexpert users to work on sophisticated high-performance computing (HPC). In this work, the pipeline was adapted to be user-friendly in order to train future generations of medical practitioners within the field of computational models with the use of HPC. Clinicians acquiring technical knowledge will help to upgrade the use of technological abilities in clinical practice and to enhance the potential of assisting clinical-decision making. As part of the CompBioMed2 European project, an elective subject was proposed and is expected to be taught in the academic year 2021-2022. The educational module is based on the use of electrophysiological simulations with HPC to generate simulated pseudo- ECGs that reproduce OTVAs from computational meshes reconstructed out of heart computed tomography (CT) images. As part of the study, the variability of the simulated pseudo-ECGs was analyzed and the most relevant ECG features were correlated with clinical literature. Finally, simulated pseudo-ECGs were used to train a machine learning algorithm to predict the SOO of the reproduced OTVAs, and quantitative validation was performed using a correlation metric.