Session T01: Biofluids: Cardiac Flows II

All-In-One Left Ventricular Vector Flow, Pressure, & Clotting Risk Mapping by Multi-Physics-Informed Neural Network

Color-Doppler echo is widely used to assess left ventricular (LV) flow. However, it is limited to the velocity component parallel to the ultrasound beam. Vector flow mapping (VFM) infers the cross-beam velocity but it is limited by relying only on mass conservation and offering little tolerance to gappy data. Moreover, VFM does not compute pressure or clotting risk; those involve secondary analyses complicating clinical translation.

 

We present MPI-VFM, a multi-physics-informed vector flow mapping method that applies deep learning to partial or complete Color-Doppler data. Its underlying models are continuity, Navier-Stokes, de-aliasing, and transport equations to infer clotting risk.  We analyze MPI-VFM on CFD-generated ground-truth data vs. imaging parameters like spatial and temporal resolution, probe angle, aliasing, and Doppler sector size. We apply MPI-VFM to clinical Color-Doppler sequences and compare head-to-head with 4D flow MRI.   

Assessment of right ventricle intracardiac flow in congenital heart disease: Comparison of enhanced echocardiography and 4D flow MRI measurements

Abnormal flow in the human heart play a crucial role in the pathology of congenital heart defects. Non-invasive flow imaging techniques can aid CHD management, especially 4D flow MRI and color Doppler echocardiography (echo) flow reconstruction. Although both methods have been independently validated against computational models, a comprehensive assessment of flow measurements and hydrodynamics between these methods has yet to be shown. In this presentation, we address the limitations and strategies associated with comparing flow measurements from 4D flow MRI and echo reconstructions. Using a cohort of repaired Tetralogy of Fallot, dilated right ventricles from atrial shunts, and control subjects, we investigate the agreement between these modalities by analyzing velocity magnitude, vorticity, flow energy loss, kinetic energy, and relative pressure fields. Our findings reveal a high correlation (median R ≥ 0.87) and low random differences, demonstrating the potential for obtaining complementary measurements between imaging modalities. These results establish color Doppler echo flow reconstruction as a complementary measurement tool to 4D Flow MRI to reliably resolve underlying intracardiac flow patterns.   

Bayesian Intraventricular Vector Flow Mapping: Influence of imaging parameters & algorithmic choices on output uncertainty

Color-Doppler echocardiography remains the workhorse of clinical left ventricular (LV) flow evaluation because of its harmlessness, low cost, portability, and quick acquisition. Offline analysis of Color-Doppler data by vector flow mapping (VFM) allows for reconstructing 2D flow fields, quantifying LV vortex dynamics, and delineating stagnant regions. However, sensitivity to image noise and lack of uncertainty quantification (UQ) limits the clinical translation of VFM­­­.

 

We present a Bayesian VFM (B-VFM) algorithm combining physics-informed priors (mass conservation, endocardial boundary conditions) and input data uncertainty (color-Doppler, endocardial position) to infer LV velocity fields and propagate imaging noise forward.  Maximum a-posteriori estimation locally weighs input noise with priors to automatically handle Doppler artifacts and LV wall segmentation errors. Using synthetic ground-truth data and an ultrasound simulator, we quantify B-VFM's performance vs. imaging parameters and algorithmic choices. Of note, we find that the usual polar-coordinate implementation of VFM augments uncertainty in the LV apex, and offer strategies to avoid this issue by preconditioning the discretized divergence operator.   

Fluid-structure interaction and optimal bending stiffness of bioprosthetic aortic valves made from Pulmonary Visceral Pleura (PVP)

Transcatheter Aortic Valve replacement (TAVR) is a procedure used for the deployment of bioprosthetic valves for patients with valvular heart disease. Although widely popular in its usage, these prosthetic valves – which are usually made with bovine pericardium (BP) tissue – are prone to calcification and structural failure. Recently, a newly identified biomaterial, Pulmonary Visceral Pleura (PVP), has shown promise in resilience, elasticity, and resistance to calcification due to its high elastin composition. In the present work, we perform a series of 3D fluid-structure interaction (FSI) simulations of the PVP aortic valve with varying leaflet thicknesses and thus bending stiffnesses.  From these simulations, we compare the performance of the PVP valve with that of the conventional BP valve, and we study the effect of the normalized bending stiffness on the valve opening/closing dynamics as well as on the flow behavior. The previously identified optimal range of bending stiffness will also be examined using this new valve material.   

Reduced Order Modeling (ROM) of vortex propagation in abdominal aneurysm

Abdominal aortic aneurysm (AAA) is a local dilation of the abdominal aorta. If left untreated, the aneurysm can rupture leading to life-threatening conditions. It has been shown that there exists a vortex ring formation during systole in the AAA, which is confirmed by in-vitro experiments, numerical simulations, and in-vivo measurements. However, the complex interaction between this vortex ring and the endothelial wall poses significant challenges to study it under in-vivo condition. In this study, we perform numerical simulations of blood flow within an idealized AAA to understand the vortex formation and breakdown processes. Our simulation shows the formation of a vortex ring propagating along the length of the aorta during the systole phase of the cardiac cycle. The blood flow velocity profiles are used to supply as inputs for our ROMs. We use both linear (Dynamic Mode Decomposition - DMD) and non-linear (an autoencoder based deep neural network - ADNN) to extract the prominent flow patterns from the simulations. The ROM results show a good correlation between the vortex propagation with the geometrical changes during aneurysm progression.   

A computational fluid dynamics study of the impact of temporal synchronization of the pulsatility mode on platelet activation in left ventricular assist device patients

Left ventricular assist devices are centrifugal pumps implanted in patients with advanced heart failure. Despite design improvements that reduced in-pump thrombosis, the risk of stroke remains high. A speed-modulation was introduced to promote in-pump washout that ramps the pump up and down in rpm for a fraction of a second, every two seconds. However, this LVAD modulation is not synchronized with the cardiac cycle. We investigate the effect of LVAD pulsatility temporal synchronization with the ventricle contraction in promoting intraventricular washout. Four timings of the LVAD pulsatility are investigated, via computational fluid mechanics. Lagrangian particle tracking captures the effect of pulsatility on platelet shear stress history and residence time, which are hemodynamics quantities associated with platelet activation. A virtual angiogram is performed by injecting a passive scalar to understand washout from an Eulerian perspective. Both Lagrangian and Eulerian metrics indicate that synchronizing the speed ramp up with peak systole promotes optimum intraventricular washout. The comparison of the simulated flow fields to those in an in-silico particle image velocimetry experiment in a patient left ventricle phantom implanted with a HeartMate3 commercial LVAD, show very good agreement, validating the simulations and confirming this preliminary finding.   

Time-spectral methods for cardiovascular simulations

Traditionally, cardiovascular flows are simulated by means of time integration. That requires the use of thousands of time steps to accurately resolve the flow in time and allow for cycle-to-cycle convergence. In this talk, I propose a much cheaper alternative that eliminates the need for time integration and simulation of many cardiac cycles for convergence. The idea is to simulate cardiac flows in frequency rather than time domain. The resulting time-spectral method significantly reduces the cost of cardiorespiratory simulations as it enables accurate discretization of the solution using a handful of frequencies. The number of required frequencies, owing to the predominantly periodic and smooth nature of cardiac flows, is typically less than 10. As a result, we are able to reduce the cost of these calculations by several orders of magnitude. Additionally, time-spectral formulations are parallelizable in the frequency domain, further reducing the time-to-solution. In this talk, I present a few cases that demonstrate the capability of the proposed time-spectral formulation, showing how it reduces solution turnover time from hours to minutes.