CDS&E: A Validated Hybrid Echo-CFD Framework for Patient-Specific Cardiac Assessment

Texas A&M Engineering Experiment Station, College Station TX

Investigators

Abstract

A computational model of the heart, built upon medical images, is invaluable for assessing cardiac function, managing therapy, optimizing biomedical devices for a specific patient, or better understanding the disease when data from a population are available. 2D echocardiography (echo) is the main imaging modality for a noninvasive evaluation of heart function due to its fast acquisition time, lower costs, portability, and wider availability compared to other imaging modalities. However, a computational model built upon 2D echo to benefit from its advantages, circumvent limitations of echo in anatomic depictions and yet, provide reliable and clinically useful applications, does not exist today. In this project, a computational heart model generated from 2D echo scans will be developed and validated to replicate cardiac flow and function and compute in vivo tissue and electrophysiological properties for specific patients. Considering that echo is the top imaging choice for evaluating heart disease, the top killer in the US, accounting for about 21% of deaths in 2020, a hybrid echo-CFD framework is anticipated to be most impactful compared to a framework coupled with other imaging modalities. Undergraduate students, in addition to graduate student, will be involved in the research (e.g., delineating the echo images) to broaden the impact. The long-term objective of this research is to develop a software package that can be utilized easily, based on echo images, to help basic science and medical researchers model the heart to diagnose heart disease, devise treatment strategies, optimize medical devices (e.g., valves, left-ventricular assist devices (LVAD), pacemakers, etc.) for specific patients, and contribute to better understanding the disease when data from a population become available. The goal of this project is to create a validated computational pipeline that takes standard echo scans as input, models cardiac flow and function, i.e., a hybrid echo-CFD framework, and computes in vivo mechanical and electrophysiological properties for a specific patient. To achieve this objective and capitalize on previous work, the walls of heart chambers and their valves will be identified in the 2D echo using deep learning. The 3D geometry will be reconstructed, and valves/atria geometries will be optimized by adopting an averaged-geometrical model. The resulting 3D geometry generating code will be coupled with an in-house CFD code based on a sharp-interface immersed boundary method to simulate large-deformation, fluid-structure interaction problems. The convergence of the Newton-Krylov solver of the CFD is accelerated by using an initial guess predicted from deep learning methods. The in vivo properties will be obtained by solving the inverse problem. Animal studies will be performed to obtain local reference flow, pressure measurements, and echo scans to validate the computational framework. The method will also be tested on retrospective clinical scans (human data). This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.