Leveraging Unlabeled Data for Machine Learning in the Electrocardiogram (James Brundage (UU HSC) )

Speaker: James Brundage (UU HSC)

Abstract: Supervised deep learning (DL) has become an increasingly common tool for advanced analysis of the electrocardiogram (ECG). These methods rely heavily on labeled datasets, in which there is a clinical annotation for each ECG. However, real world ECG datasets may not contain enough labeled recordings to facilitate robust feature extraction, preventing DL analysis for clinical problems with small datasets. Self-supervised learning (SSL) seeks to utilize cheaply labeled or unlabeled data to improve performance in a supervised learning task. This process consists of first training a model on a primary task with cheap data labels, followed by a second training process which attempts to learn the downstream task by initializing with weights learned from the first. While SSL has become a popular tool in many machine learning domains, it is only starting to be used in ECG based machine learning. Here, we demonstrate the progress we have made in applying SSL approaches to detect low left ventricular ejection fraction, a complex ECG detection task, using data extracted from the University of Utah.

Bio: James Brundage is a second year medical student (MSII) at the University of Utah. He completed a BS and MS in neuroscience at BYU with an emphasis in cellular neuro-electrophysiology. Since starting at the University of Utah, he has worked in the MacLeod lab, coordinating and carrying out a collaborative effort between the Scientific Computing and Imaging Institute (SCII), Nora Eccles Cardiovascular Research and Training Institute (CVRTI) and the cardiology team at University of Utah Hospital focused on analysis of the electrocardiogram (ECG) using machine learning. Following medical school, he hopes to pursue a career as a physician scientist with a focus on applied machine learning in the clinical setting.