Complex physics and biochemistry, the wide range of temporal and spatial scales to be captured, and the significant inter- and intraindividual variations make the description of the blood flow in the human aorta a difficult task. Although computational tools are mature, said aspects challenge aortic biomechanical models, and the approach must always be tailored to the question at hand. Besides being highly pulsatile, the curvature and tortuosity of the aortic geometry strongly impacts the flow dynamics with aortic pathologies may even lead to turbulent flow. In the following chapter, the characteristics of blood and different modeling approaches for rheology, species transport and thrombus/stenosis development will be addressed.

Dissolution-permeation (D/P) experiments are widely used during preclinical development due to producing results with better predictability than traditional monophasic experiments. However, it is difficult to compare absorption across in vitro setups given the propensity to only report apparent permeability. We therefore developed an approach to predict the concentration boundary layer for any D/P device by using computational fluid dynamics (CFD). The Navier-Stokes and continuity equation in 2D were solved numerically in MATLAB and by finite element methods in COMSOL v6.1 to predict the momentum (ηf′) and concentration ηg boundary layer for a flow over a flat plate, i.e. the classical Blasius boundary layer flow. A MATLAB algorithm was developed to calculate the edge of either boundary layer. The methodology to determine the concentration boundary layer based on Blasius's analysis provided an accurate estimate for both ηf′ and ηg, resulting in, ηf′/ηg, at high Schmidt numbers (Sc ∼ 1000) within 14 % of the Blasius solution and 6.6 % of the accepted Schmidt number correlation (Sc1/3=ηf′/ηg). The methodology based on the Blasius analysis of the concentration boundary layer using velocity and concentration profiles computed using CFD presented herein will enable characterization/analysis of complex D/P apparatuses used in preclinical development, where an analytical solution may not be available.

Background: : Increasingly, computational fluid dynamics (CFD) is helping explore the impact of variables like: cannula design/size/position/flow rate and patient physiology on venovenous (VV) extracorporeal membrane oxygenation (ECMO). Here we use a CFD model to determine what role cardiac output (CO) plays and to analyse return cannula dynamics. Methods: : Using a patient-averaged model of the right atrium and venae cava, we virtually inserted a 19Fr return cannula and a 25Fr drainage cannula. Running large eddy simulations, we assessed cardiac output at: 3.5–6.5 L/min and ECMO flow rate at: 2–6 L/min. We analysed recirculation fraction (Rf), time-averaged wall shear stress (TAWSS), pressure, velocity, and turbulent kinetic energy (TKE) and extracorporeal flow fraction (EFF = ECMO flow rate/CO). Results: : Increased ECMO flow rate and decreased CO (high EFF) led to increased Rf (R = 0.98, log fit). Negative pressures developed in the venae cavae at low CO and high ECMO flow (high CR). Mean return cannula TAWSS was >10 Pa for all ECMO flow rates, with majority of the flow exiting the tip (94.0–95.8 %). Conclusions: : Our results underpin the strong impact of CO on VV ECMO. A simple metric like EFF, once supported by clinical data, might help predict Rf for a patient at a given ECMO flow rate. The return cannula imparts high shear stresses on the blood, largely a result of the internal diameter.

We investigated variations in haemodynamics in response to simulated microgravity across a semi-subject-specific three-dimensional (3D) continuous arterial network connecting the heart to the eye using computational fluid dynamics (CFD) simulations. Using this model we simulated pulsatile blood flow in an upright Earth gravity case and a simulated microgravity case. Under simulated microgravity, regional time-averaged wall shear stress (TAWSS) increased and oscillatory shear index (OSI) decreased in upper body arteries, whilst the opposite was observed in the lower body. Between cases, uniform changes in TAWSS and OSI were found in the retina across diameters. This work demonstrates that 3D CFD simulations can be performed across continuously connected networks of small and large arteries. Simulated results exhibited similarities to low dimensional spaceflight simulations and measured data—specifically that blood flow and shear stress decrease towards the lower limbs and increase towards the cerebrovasculature and eyes in response to simulated microgravity, relative to an upright position in Earth gravity.

Venovenous extracorporeal membrane oxygenation is used for respiratory support in the most severe cases of acute respiratory distress syndrome. Blood is drained from the large veins, oxygenated in an artificial lung, and returned to the right atrium (RA). In this study, we have used large eddy simulations to simulate a single-stage “lighthouse” drainage cannula in a patient-averaged model of the large veins and RA, including the return cannula. We compared the results with previous experimental and numerical studies of these cannulas in idealized tube geometries. According to the simulations, wall proximity at the drainage holes and the presence of the return cannula greatly increased drainage through the tip (33% at 5 L/min). We then simulated a multi-stage device in the same patient-averaged model, showing similar recirculation performance across the range of extracorporeal membrane oxygenation (ECMO) flow rates compared to the lighthouse cannula. Mean and maximum time-averaged wall shear stress were slightly higher for the lighthouse design. At high ECMO flow rates, the multi-stage device developed a negative caval pressure, which may be a cause of drainage obstruction in a clinical environment. Finally, through calculation of the energy spectra and vorticity field, we observed ring-like vortices inside the cannula originating from the side holes, most prominent in the proximal position. Our work highlights the important differences between a patient-derived and simplified venous model, with the latter tending to underestimate tip drainage. We also draw attention to the different dynamics of single-stage and multistage drainage cannulas, which may guide clinical use.

Venovenous extracorporeal membrane oxygenation (ECMO) can be performed with two single lumen cannulas (SLCs) or one dual-lumen cannula (DLC) where low recirculation fraction (Rf) is a key performance criterion. DLCs are widely believed to have lower Rf , though these have not been directly compared. Similarly, correct positioning is considered critical although its impact is unclear. We aimed to compare two common bi-caval DLC designs and quantify R f in several positions. Two different commercially available DLCs were sectioned, measured, reconstructed, scaled to 27Fr and simulated in our previously published patient-averaged computational model of the right atrium (RA) and venae cavae at 2–6 L/min. One DLC was then used to simulate ± 30° and ± 60° rotation and ± 4 cm insertion depth. Both designs had low Rf (< 7%) and similar SVC/IVC drainage fractions and pressure drops. Both cannula reinfusion ports created a high-velocity jet and high shear stresses in the cannula (> 413 Pa) and RA (> 52 Pa) even at low flow rates. Caval pressures were abnormally high (16.2–23.9 mmHg) at low flow rates. Rotation did not significantly impact Rf . Short insertion depth increased Rf (> 31%) for all flow rates whilst long insertion only increased Rf at 6 L/min (24%). Our results show that DLCs have lower Rf compared to SLCs at moderate-high flow rates (> 4 L/min), but high shear stresses. Obstruction from DLCs increases caval pressures at low flow rates, a potential reason for increased intracranial hemorrhages. Cannula rotation does not impact Rf though correct insertion depth is critical.

Venovenous extracorporeal membrane oxygenation is a treatment for acute respiratory distress syndrome. Femoro-atrial cannulation means blood is drained from the inferior vena cava and returned to the superior vena cava; the opposite is termed atrio-femoral. Clinical data comparing these two methods is scarce and conflicting. Using computational fluid dynamics, we aim to compare atrio-femoral and femoro-atrial cannulation to assess the impact on recirculation fraction, under ideal conditions and several clinical scenarios. Using a patient-averaged model of the venae cavae and right atrium, commercially-available cannulae were positioned in each configuration. Additionally, occlusion of the femoro-atrial drainage cannula side-holes with/without reduced inferior vena cava inflow (0-75%) and retraction of the atrio-femoral drainage cannula were modelled. Large-eddy simulations were run for 2-6L/min circuit flow, obtaining time-averaged flow data. The model showed good agreement with clinical atrio-femoral recirculation data. Under ideal conditions, atrio-femoral yielded 13.5% higher recirculation than femoro-atrial across all circuit flow rates. Atrio-femoral right atrium flow patterns resembled normal physiology with a single large vortex. Femoro-atrial cannulation resulted in multiple vortices and increased turbulent kinetic energy at > 3L/min circuit flow. Occluding femoro-atrial drainage cannula side-holes and reducing inferior vena cava inflow increased mean recirculation by 11% and 32%, respectively. Retracting the atrio-femoral drainage cannula did not affect recirculation. These results suggest that, depending on drainage issues, either atrio-femoral or femoro-atrial cannulation may be preferrable. Rather than cannula tip proximity, the supply of available venous blood at the drainage site appears to be the strongest factor affecting recirculation.

The right atrium (RA) combines flows from the inferior (IVC) and superior vena cava (SVC). Here RA mixing is simulated using computational fluid dynamics, comparing four modeling approaches. A patient-averaged model (11 M cells) was created from four volunteers. We compared: (1) unsteady k–ω Reynolds-averaged Navier–Stokes (URANS) (2) implicit large eddy simulation with second-order upwind convection scheme (iLES-SOU) (3) iLES with bounded-central difference convection scheme (iLES-BCD) and (4) LES with wall-adapting local eddy-viscosity (LES-WALE). A constant inlet flow rate of 6 L/min was applied with both IVC/SVC contributions ranging from 30–70%. A higher density mesh (37 M cells) was also simulated for models 2 and 4 (equal IVC/SVC flow) to assess the accuracy of models 1–4. Results from the 11 M cell LES-WALE model showed good agreement with the 37 M cell meshes. All four 11 M cell models captured the same large-scale flow structures. There were local differences in velocity, time-averaged wall shear stress, and IVC/SVC mixing when compared to LES-WALE, particularly at high SVC flow. Energy spectra and velocity animations from the LES-WALE model suggest the presence of transitional flow. For the general flow structures, all four methods provide similar results, though local quantities can vary greatly. On coarse meshes, the convection scheme and subgrid-scale (SGS) model have a significant impact on results. For RA flows, URANS should be avoided and iLES models are sensitive to convection scheme unless used on a highly resolved grid.

The right atrium (RA) combines the superior vena cava (SVC) and inferior vena cava (IVC) flows. Treatments like extracorporeal membrane oxygenation (ECMO) and hemodialysis by catheter alter IVC/SVC flows. Here we assess how altered IVC/SVC flow contributions impact RA flow. Four healthy volunteers were imaged with computerized tomography (CT), reconstructed and combined into a patient-averaged model. Large eddy simulations (LESs) were performed for a range of IVC/SVC flow contributions (30%-70% each, increments of 5%) and common flow metrics were recorded. Model sensitivity to reconstruction domain extent, constant/pulsatile inlets, and hematocrit was also assessed. Consistent with literature, a single vortex occupied the central RA across all flowrates with a smaller counter-rotating vortex, not previously reported, in the auricle. Vena cava flow was highly helical. RA turbulent kinetic energy (TKE; P = 0.027) and time-averaged wall shear stress (WSS; P < 0.001) increased with SVC flow. WSS was lower in the auricle (2 Pa, P < 0.001). WSS in the vena cava was equal at IVC/SVC = 65/35%. The model was highly sensitive to the reconstruction domain with cropped geometries lacking helicity in the venae cavae, altering the RA flow. The RA flow was not significantly affected by constant inlets or hematocrit. The commonly reported vortex in in the central RA is confirmed; however, a new, smaller vortex was also recorded in the auricle. When IVC flow dominates, as is normal, TKE in the RA is reduced and WSS in the venae cavae equalize. Significant helicity exists in the vena cava, as a result of distal geometry and this geometry appears crucial to accurately simulating RA flow. NEW & NOTEWORTHY Right atrium turbulent kinetic energy increases as the proportion of flow entering from the superior vena cava is increased. Although the commonly reported large right atrium vortex was confirmed across all flow scenarios, a new smaller vortex is observed in the right auricle. The caval veins exhibit highly helical flow and this appears to be the result of distal venous morphology.