In silico electrocardiogram from simplistic geometric and reaction diffusion model for detection of cardiac ventricular abnormalities through machine learning methods

UNCG Author/Contributor (non-UNCG co-authors, if there are any, appear on document)
Shane E. Loeffler (Creator)
The University of North Carolina at Greensboro (UNCG )
Web Site:
Joseph Starobin

Abstract: Millions of people around the globe die each year from ischemic heart disease. While previous work has focused on the detection of myocardial scarring after an ischemic event, there is little to no work that involves the detection of ventricular ischemia in the early stages of an ischemic event. Using simplified geometries and standard electrophysiological modeling, normal and ischemic conditions may be simulated within a left ventricle model. This work highlights the theoretical framework for determining the configuration of myocardial ischemia using analytical and deep learning methods. Using these methods, a two-dimensional model was used to find a theoretical threshold that can dictate when normal transmural myocardial ischemia is occurring. Using a three-dimensional model, 20,000 stochastic ischemic zones were simulated with associated ECGs to train a one-dimensional convolutional neural network to predict the configuration of the early stages of ischemia. In addition, a three-dimensional model was implemented to produce 10,000 stochastically growing ischemic configurations in which a one-dimensional convolutional and long-short term memory neural network was used to predict the future states of ischemia.

Additional Information

Language: English
Date: 2021
Artificial Intelligence, Cardiac, ECG, Ischemia, Machine Learning, Reaction Diffusion
Electrocardiography $x Computer simulation
Heart $x Electric properties $x Computer simulation
Ischemia $x Computer simulation

Email this document to