Project Grant R15LM013938

This $475,788 Project Grant from the Department of Health and Human Services National Institutes of Health National Institute of General Medical Sciences will support the development of automated clinical decision support tools for summarizing cardiac data. Specifically, the University of Central Oklahoma and collaborators at Baylor College of Medicine and Texas Children's Hospital will apply deep learning and subset selection techniques to classify and summarize electrocardiogram and vectorcardiogram waveforms recorded from pediatric cardiac patients.

Undergraduate students will evaluate variational autoencoder architectures to differentiate noise from arrhythmias and normal sinus rhythm in cardiac data. Additional models will be trained to recognize pathologies. These results will inform the subsequent summarization of patient records using techniques such as the Discrete Empirical Interpolation Method and related approaches. Students will also develop novel extensions of these methods for medical data classification and compare performance to existing literature. Ultimately, this project aims to provide clinicians with represented summaries of recent cardiac patient histories to support clinical decision making and predictive modeling, with the goal of improving pediatric cardiac outcomes.