Session: 10-04-01: Young Engineer Paper (YEP) Contest Fluids Engineering Division

Paper Number: 145903

145903 - A Benchtop Study of Fontan Circulation 

The Centers for Disease Control and Prevention (CDC) estimates that each year in the United States, approximately 1/3800 babies are born with only one functioning ventricle. The Fontan circulation is a result of the third stage procedure to correct single ventricle anatomy in children whose implementation has only led to a survival rate of ~50% into adulthood. One of the failure points recognized is increased inferior vena cava (IVC) pressure within the total cavopulmonary connection. The research team proposes to enhance flow into the pulmonary system by adding an injection jet shunt (IJS) to induce flow entrainment by drawing flow directly from the aortic arch, balanced by a fenestration to maintain a ratio of pulmonary flow (Qp) to systemic flow (Qs) of 1. Thus, oxygen tracking is critical to determining the optimal IJS-assisted Fontan model geometries for a given patient. In this study, we describe a dynamically controlled mock flow loop model (MFL) designed to quantitatively and qualitatively estimate the volume fraction of IJS flow (VF_IJS) absorbed by the fenestration in the IVC conduit using a high-speed fluorescence imaging technique called ultraviolet-induced fluorescence (UVIF). A patient generic surrogate model of the total cavopulmonary connection (TCPC) with average dimensions matching those of a 2–4-year-old patient is inserted in the MFL derived from a reduced lumped parameter model (LPM) representing the cardiovascular circulation which is used to generate the training data space for a deep-learning algorithm to estimate VF_IJS for various TCPC configurations. The LPM is comprised of four 2-element Windkessel compartments (compliance and resistance), approximating the upper and lower systemic circulations, and the right and left pulmonary circulations (RPA, LPA). The use of a surrogate model representative of a patient generalized TCPC has been proposed in this study to formulate the training dataset for the implementation of a deep learning algorithm to predict VF_IJS of any chosen patient. Surrogate modeling involves the use of high-temporal resolution UVIF imaging of the flow field within the FONTAN conduit which allows VF_IJS to be extracted for a set of experimental design parameters in the MFL. Using this technique to image at high temporal scales comes with the expense of low spatial resolution which is addressed by implementing and comparing two interpolation techniques to increase spatial resolution and preserve temporal resolution of the response surface: POD-RBF and genetic adversarial network (GAN). A response surface based on a radial basis function (RBF) interpolation network trained by a Proper Orthogonal Decomposition (POD) of the solution fields is used to extract coherent structures in the flow and is characterized by its efficiency and ability to generate response surfaces at a low cost. Conversely, we expect the GAN to generate a highly accurate response surface of FONTAN conduit VF_IJS but it requires more computational resources to achieve this result, making it desirable to use POD-RBF to formulate the VF_IJS training space. Using the two generated surrogate model data sets from POD-RBF and GAN, a deep-learning neural network will be trained and used to predict accurate approximations of oxygen content (VF_IJS) in the FONTAN conduit for various TCPC configurations where the effectiveness of POD-RBF interpolation and GAN will be compared. The experimental findings collected closely validate the CFD results of oxygen saturation in the IVC conduit across various TCPC configurations. 

Presenting Author: Anthony Damon Embry-Riddle Aeronautical University

Presenting Author Biography: Undergraduate student in the Aerospace Engineering department at Embry-Riddle Aeronautical University

Authors:

Anthony Damon Embry-Riddle Aeronautical University
Arka Das Embry-Riddle Aeronautical University
Ray Prather Embry-Riddle Aeronautical University
Alain Kassab University of Central Florida
Eduardo Divo Embry-Riddle Aeronautical University
William Decampli Arnold Palmer Children’s Hospital