Using machine learning to uncover novel arrhythmic risk markers in a population of image-based virtual hearts

Elucidating new arrhythmic risk markers from ECGs remains difficult due to the lack of understanding in the relationship between underlying heart anatomy/electrical activity and the complex ECG signal. This project will advance knowledge in this front through the combination of biophysical simulations using a population of heart models and machine learning in order to gain insight into the relationship between heart electrical activity and the generated ECGs.
Master

Myocardial ischemia due to coronary artery occlusion promotes cardiac tissue remodeling that increases patient risk to lethal arrhythmias and sudden cardiac death (SCD). However, determining individual patient risk to SCD remains difficult and often involves invasive techniques. Cardiologists also rely on electrocardiograms (ECG) collected from the body surface to gain insight into the heart's electrical activity. However, correlating the complex signals with underlying anatomy remains incompletely understood. Further insight into this relationship could lead to identification of novel ECG risk markers.

Goal

The goal of this project is to identify novel ECG risk markers measured during sinus activity that could predict a patient’s susceptibility to future arrhythmias. To achieve this goal, the student will perform 3D biophysical simulations of sinus electrical activity in image-based models of patient hearts. The student will generate simulated ECGs from these simulations and then use signal analysis coupled with machine learning to discover novel markers that might be predictive of arrhythmic risk. The novel markers identified in this project will then be used in a retrospective patient data to determine if these markers have clinical utility.

Learning outcome

Upon completion of this project, the student will gain skills in heart physiology in health and disease, 3D cardiac electrophysiological simulations, image-based model construction, ECG annotation and analysis, signal processing, and machine learning.

Qualifications

Machine Learning
Programming (C or python or Matlab)

Supervisors