Using Machine Learning to Analyze and Classify Echocardiogram Results

Abstract

Heart arrhythmias can be difficult to diagnosis simply from external observation, and many fail to present themselves through concerning symptoms until damage has been done to the heart and the rest of the body. One of the most common tools used to identify these problems is an electrocardiogram (ECG) test, which records impulses from the patient’s heart (heartbeats) and can reflect the state of the individual’s cardiovascular system. ECG test outputs are labored over by cardiologists, which can be time consuming and subject to misinterpretation. An efficient and accurate analysis of an ECG test is critical, as early detection—particularly with more serious arrhythmias—is extremely influential in treatment success. This research explores the potential of using machine learning algorithms to read and analyze echocardiograms based on several input factors. Using a computer algorithm can reduce the amount of time cardiologists spend analyzing and understanding the output of an ECG, while also potentially improving accuracy. This poster details the machine learning algorithms used in the diagnosis, as well as their individual performances.