Methods for Echocardiographic Biomechanical Measurements

Brett A Meyers, Purdue University

Abstract

Heart Failure, a life-threatening condition in which the heart does not pump enough oxygenated blood to meet the needs of the body, affects more than 6 million children and adults in America alone. The failure which occurs is characterized as diastolic dysfunction when the ventricles cannot fill sufficiently, or as systolic dysfunction when the ventricles are too weak to pump. Heart failure remains difficult to detect, with one in three patients going undiagnosed during initial hospital visits with acute symptoms, as the heart remodels to compensate for insufficient pumping. Currently the most widely used non-invasive imaging method for assessing heart function is echocardiography. While echocardiography technology has advanced over the past 50 years, many of the image-based measurements performed are unable take advantage of these new capabilities. This work presents the development of several new methods to extract clinically relevant biomechanics measurements from echocardiography and demonstrates how these quantities can provide insight into understanding as well as detecting abnormal heart function. A new framework for echocardiographic particle image velocimetry (echoPIV) for the first time enables the technique to be reliably used on patient data collected during routine examination. A novel method for reconstructing the underlying two-dimensional, two-component blood velocity vector field from Color Doppler echocardiograms provides a fully non-invasive modality for visualizing flow in the heart and enables the quantification of the flow physics robustly. Global deformation of the heart chambers is quantified using a novel framework based on the seldom used logarithm-scaled Fourier correlation kernel. These methods are each presented with a test dataset. These tools are utilized to demonstrate the functional changes of cardiac biomechanics that occur in utero versus ex uterofor healthy children and those with hypoplastic left heart syndrome (HLHS). Quantities characterizing the deformation, shape, and hemodynamics of the left and right ventricle show how the heart adapts to self-sustained breathing in normal individuals and the challenges that arise from this in HLHS patients. Finally, methods of machine learning are employed to train a model to predict patients at high risk of heart failure based on fundamental cardiac function parameters. This tool shows a 30% improvement in correctly determining patients at high risk versus current guidelines and recommendations used widely by clinicians.

Degree

Ph.D.

Advisors

Vlachos, Purdue University.

Subject Area

Biomechanics|Electrical engineering|Fluid mechanics|Mathematics|Mechanics|Medical imaging|Medicine|Physiology|Public health

COinS