The effect of arterial cannula tip position on differential hypoxemia during venoarterial extracorporeal membrane oxygenation

Interaction between native ventricular output and venoarterial extracorporeal membrane oxygenation (VA ECMO) flow may hinder oxygenated blood flow to the aortic arch branches, resulting in differential hypoxemia. Typically, the arterial cannula tip is placed in the iliac artery or abdominal aorta. However, the hemodynamics of a more proximal arterial cannula tip have not been studied before. This study investigated the effect of arterial cannula tip position on VA ECMO blood flow to the upper extremities using computational fluid dynamics simulations. Four arterial cannula tip positions (P1. common iliac, P2. abdominal aorta, P3. descending aorta and P4. aortic arch) were compared with different degrees of cardiac dysfunction and VA ECMO support (50%, 80% and 90% support). P4 was able to supply oxygenated blood to the arch vessels at all support levels, while P1 to P3 only supplied the arch vessels during the highest level (90%) of VA ECMO support. Even during the highest level of support, P1 to P3 could only provide oxygenated VA-ECMO flow at 0.11 L/min to the brachiocephalic artery, compared with 0.5 L/min at P4. This study suggests that cerebral perfusion of VA ECMO flow can be increased by advancing the arterial cannula tip towards the aortic arch.


3
Background Venoarterial extracorporeal membrane oxygenation (VA ECMO) is form of mechanical circulatory support for patients with refractory cardiac or cardio-pulmonary failure. A typical VA ECMO circuit consists of a drainage cannula that extracts venous blood, a centrifugal pump, an oxygenator and an arterial cannula that passes replenished blood back into the body. In peripheral VA ECMO, the arterial cannula is positioned in the arterial system and while multiple different arterial access sites can be used, the femoral artery is the most common having been demonstrated to have faster and easier access, with a lower risk of bleeding complications [1][2][3]. When cannulating via the femoral artery, the cannula tip typically lies within the abdominal aorta or common iliac artery, with flow from the cannula traveling in a retrograde direction, against the forward flow produced by the heart. The junction of flow from the cannula and antegrade flow from the left ventricle forms a watershed region or mixing zone. The location of this region can impact VA ECMO perfusion of parts of the upper and lower body. Patients with impaired pulmonary function coupled with increasing cardiac output can lead to a state whereby proximal aortic branch vessels are perfused with blood of lower oxygen saturation. In contrast, distal to the mixing zone, tissues and organs are perfused by oxygenated VA ECMO flow. This phenomenon is known as differential hypoxemia. It can result in cerebral and coronary hypoxia, and is associated with increased rates of adverse clinical outcomes, including neurological complications and reduced survival [4][5][6].
In a single center study that involved 720 ECMO patients, Rupprecht et al. reported an incidence rate of 8.8% for complications arising due to upper body hypoxemia [5]. This was the second most frequent complication they observed and often required intervention to improve upper body oxygen saturation levels. Differential hypoxemia is associated with poor perfusion of oxygenated blood to the brain. For example, Pozzebon et al. reported cerebral desaturation (defined as < 60% oxygen saturation for > 5% ECMO duration) in 43 (74%) ECMO patients using near-infrared spectroscopy applied to the patients' foreheads [7]. Moreover, 18 (42%) of these patients went on to develop acute cerebral complications such as stroke and brain death. However, patients with no cerebral desaturation experienced no acute cerebral complications.
Clinical interventions for preventing differential hypoxia include placing the cannula tip in the subclavian artery via axillary or subclavian cannulation [8], placing the cannula tip in the brachiocephalic artery via a mini-upper sternotomy (central sport model) [9] or direct carotid cannulation [10]. However, these techniques are technically challenging to perform and require open surgical cannulation. In addition, they require meticulous surgical training and care to perform in the first place, transport of the patient to an operating theatre, and cannot be used alongside simultaneous chest compressions [8,9]. Some studies have also suggested advancing the drainage cannula into the superior vena cava as a possible strategy [11][12][13]. This ensures drainage of poorly oxygenated blood from the upper body, thereby preventing its circulation to the arterial system. Another commonly proposed method of treating differential hypoxemia after femoral cannulation is the conversion of the circuit to a veno-arteriovenous configuration [5,6]. The addition of an extra return cannula through the jugular vein introduces additional risk of bleeding, risk of recirculation between the return and drainage cannula and the division of return flow between the venous and arterial system requires clamps to adjust for the different pressure regimes, thereby increasing the risk of thrombosis [6,14]. Therefore, it is clear that current clinical practices in peripheral VA ECMO are not systematically structured to prevent differential hypoxemia, thereby prompting the need for alternative intervention methods.
In recent years, computational fluid dynamics (CFD) have been increasingly used as research tools to obtain insights in how to optimize mechanical circulatory support strategies. CFD enables researchers to investigate the effect of different cannula positions on pressure and flow fields, including mixing zones, with spatial and temporal resolution unattainable by clinical methods. Previous studies have used CFD to investigate the effect of increasing cannula flow rate on mitigating differential hypoxemia [15][16][17][18][19]. Investigators concluded high ECMO flow rates (> 4.5 L/min) were required to mitigate differential hypoxemia, and even then, small quantities of cardiac output were able to shift the mixing zone distally, thereby reducing flow rates of oxygenated blood to the upper body. Advancing the cannula tip to a more proximally located position has been suggested as a possible method to mitigate differential hypoxemia, but the hemodynamics of such a solution has not been studied previously [20].
The aim of this study was to evaluate how altering cannula tip position impacts mixing zone location and flow field characteristics which might be conducive to cerebral hypoxia. For this purpose, a CFD simulation was constructed using patient specific geometry and physiologically representative boundary conditions. Cannula tip positions were varied in simulations of different degrees of cardiac dysfunction and VA ECMO support. We hypothesized that VA ECMO perfusion of the aortic arch branch vessels could be improved by positioning the cannula tip closer to the aortic arch.

Geometry
A 3D model of the aorta and its major branches was extracted from a computed tomography (CT) scan of the chest, abdomen and pelvis of a male patient undergoing VA ECMO. The patient had VA ECMO initiated in an extracorporeal cardiopulmonary resuscitation (ECPR) setting due to pulseless electrical activity arrest secondary to an STelevation myocardial infarction. The patient was 44 years of age, 175 cm in height and weighed 80 kg. Image segmentation was performed using MIMICS 21.0 and 3-Matic 21.0 image processing software (Materialise NV., Leuven, Belgium) where contrast thresholds were manipulated to obtain the blood volume of the aorta. The aortic branches extracted for this study included the brachiocephalic artery (BCA), left common carotid artery (LCCA), left subclavian artery (LSCA), coeliac axis/trunk, superior and inferior mesenteric arteries (SMA and IMA, respectively), left and right renal arteries and left and right common iliac arteries (LCI and RCI, respectively). The model was then imported into SpaceClaim (SpaceClaim Corporation, MA, USA) to perform 3D design modelling. A 19 Fr Maquet HLS (Getinge, Rastatt, Germany) arterial cannula tip was placed in the right common iliac artery (P1). Three other cannula tip positions, each progressively more proximal, were created to represent advancement of the arterial cannula towards the aortic arch: position 2 (P2) in the abdominal aorta; position 3 (P3) in the distal portion of the descending aorta; and position 4 (P4) in the distal section of the aortic arch (Fig. 1). The meshes used in this study contained between 0.6 and 1 million poly-hexcore elements and was meshed using Fluent 20R2 (ANSYS, Canonsburg, Pennsylvania, USA). Further details regarding the meshing process and mesh sensitivity analysis for this study can be found in the supplementary

Boundary conditions and model set-up
The two inlets for this CFD model consisted of the cannula tip and the aortic valve. The cannula produced a constant flow whilst a pulsatile flow waveform was prescribed at the aortic valve inlet using patient derived measurements from the literature [21]. Total flow (aortic valve flow + VA ECMO flow) entering the system was kept constant at 5 L/ min throughout all simulations as per previous studies [15,16]. Native output and VA ECMO flow were then varied accordingly to reflect various degrees of cardiac dysfunction and VA ECMO support. The following scenarios were simulated: 50% support (2.5 L/min native flow and 2.5 L/ min from VA ECMO), 80% support (1 L/min native flow and 4 L/min from VA ECMO), and 90% support (0.5 L/min native flow and 4.5 L/min from VA ECMO). A baseline healthy scenario (5 L/min native flow and 0 L/min from VA ECMO) without a cannula in-situ was also simulated to serve as a control for vessel perfusion flow rates. The healthy transaortic valve flow waveform was scaled to produce the aforementioned degrees of failure ( Fig. 2) [21].
Boundary conditions at the outlets of each arterial branch were created using a 3-element Windkessel model to incorporate the effects of the distal vasculature and the dynamic nature of the systemic circulation [22]. The 3-element Windkessel functions lump components of the systemic circulation into peripheral and distal resistances whilst also incorporating compliance effects. Further details regarding these boundary conditions are outlined in the supplementary material (Supplementary File 2).
All vessel walls were assumed to be rigid and a no-slip boundary condition was imposed. The non-Newtonian behavior of blood was modelled using the Carreau model to maintain physiological accuracy, accounting for pulsatility and shear thinning properties that may arise during higher ECMO support levels [23,24]: where is the local viscosity, ∞ = 0.00345 kg/(m.s), 0 = 0.056 kg/(m.s), is the local shear rate, = 3.313 s, and n = 0.3568 [23].
The implicit formulation of the Volume of Fluid multiphase model was chosen to distinguish between blood from the left ventricle (LV) and the VA ECMO circuit. Turbulence was modelled using the k-ω Shear Stress Transport model (Re > 4000 for all simulations) and the Pressure-Implicit with Splitting of Operators algorithm was adopted for the pressure-velocity coupling method. The time step was set as 0.001 s and 15 cardiac cycles were simulated to ensure convergence was achieved and only the last five cycles were used for all data processing purposes. Additionally, the solution at each time step was considered converged when the scaled residuals value decreased below 10 -4 . All simulations were performed on a high-performance cluster (Multi-modal Australian ScienceS Imaging and Visualisation Environment) at Monash University, Melbourne, Australia.

Results
Results from each simulation were obtained quantitatively and qualitatively. The flow rates of blood received from VA ECMO perfusion at all aortic branches are presented in Fig. 3. These results were averaged across the last five cardiac cycles. Aortic branch vessel flow rates for a simulated healthy adult (without VA ECMO) is also shown in Fig. 3. Mixing zone locations for cannula tip positions according to increasing level of VA ECMO support are shown in Fig. 4. This data reflects mixing zone locations at end-diastole.

50% VA ECMO support
During this scenario, P4 was the only position which perfused all aortic arch branch vessels with blood from VA ECMO. Total flow to these vessels were comparable in magnitude to healthy conditions as it contained blood from both the LV (deoxygenated) and blood from VA ECMO (oxygenated). However, the flow rates of blood from VA ECMO to the BCA, LCCA and LSCA were 0.29, 0.06 and  0.32 L/min, respectively (Fig. 3). Compared to a healthy adult, these flow rates reflect 54%, 55% and 76% of normal flow rates usually seen in the BCA, LCCA and LSCA, respectively. However, flow rates of oxygenated blood to all other vessels decreased during P4. For example, during P1 and the healthy simulated case, the LCI received 0.55 L/min of oxygenated blood. This value decreased to 0.22 L/min during P4. These results were supported by their associated mixing zones (Fig. 4). At P1, the mixing zone was located in the abdominal aorta and a clear separation was observed between blood from VA ECMO and the LV. However, as the Fig. 4 VA ECMO blood distribution represented as volume fractions for each support level and cannula position at end-diastole. In this figure, the red color (volume fraction of 1) refers to oxygenated blood provided by the circuit, whereas blue refers to blood from the LV, and hence, an absence of oxygenated blood (volume fraction of 0). P1, 2, 3 and 4 Position 1, 2, 3 and 4, BCA brachiocephalic artery, LCCA left common carotid artery, LSCA left subclavian artery, L-Renal left renal, R-Renal right renal, SMA superior mesenteric artery, IMA inferior mesenteric artery, RCI right common iliac, LCI left common iliac cannula was advanced, blood from the LV was able to reach distal portions of the aorta, thereby disrupting the distinct boundary observed in P1. This is accompanied with gradual advancements of blood from VA ECMO during P2 and P3 to the descending aorta. P4 allowed blood from VA ECMO to be distributed throughout the aorta with high concentrations located at the aortic arch and lower concentrations distal to the arch. The interaction between cannula and aortic flow during P4 resulted in chaotic flow characteristics within the arch of the aorta; thereby preventing the usual development of aortic flow around the bend of the aortic arch.

80% VA ECMO support
With an increased cannula flow rate, P4 was able to provide increased flow from the ECMO circuit to the aortic arch branches (Fig. 3). The BCA, LCCA and LSCA received 0.43, 0.07 and 0.46 L/min of blood from the VA ECMO circuit, respectively. Compared to a healthy adult, these flow rates are 81%, 67% and 109% of normal flow rates usually seen in the BCA, LCCA and LSCA, respectively. The arch vessels did not receive blood from VA ECMO during P1, P2 and P3. Instead, increased perfusion of blood from VA ECMO was seen in all branches below the arch.
Visualizing the distribution of blood from VA ECMO shows a higher concentration of oxygenated blood distributed throughout the aorta during P4 (Fig. 4). Due to the cannula tip position during P4, the outer bend of the aortic arch only consisted of blood from VA ECMO. Conversely, aortic flow traveled around the inner bend of the aortic arch. At P3, the mixing zone was located at the distal end of the aortic arch. P2 showed slight advancement of the mixing zone within the descending aorta compared to P1. Unlike during 50% support, all branches below the arch were completely perfused with blood from the VA ECMO circuit between P1 and P3. This was due to the lower inertia of aortic flow being unable to disrupt the cannula flow in the distal regions of the aorta.

90% VA ECMO support
During this scenario, P4 showed even greater perfusion of blood from VA ECMO to the arch branches. Compared to a healthy adult, flow to the aortic arch branches from the VA ECMO circuit were 94%, 110% and 109% of normal flow rates usually seen in the BCA, LCCA and LSCA, respectively. The BCA received much greater perfusion of oxygenated blood when compared to other positions: 0.50 L/min at P4, compared to 0.07, 0.09 and 0.11 L/min at P1, P2 and P3 respectively (Fig. 3). Interestingly, total flow (combination of native blood and blood from VA ECMO) to the arch vessels decreased between P3 and P4 (by a total of 0.04 L/ min). Due to continuity, this decrease was associated with an increase in total flow to all branches below the arch. For example, total flow to the BCA decreased by 0.03 L/min and total flow to the coeliac axis increased by 0.01 L/min. Unlike in 80% support, the BCA, LCCA and LSCA received blood from VA ECMO at P1, P2 and P3 during 90% support. Total upper body perfusion of VA ECMO blood increased to 0.59, 0.64 and 0.68 L/min during P1, P2 and P3, respectively. While a majority of this flow was supplied to the LSCA which does not supply the cerebral circulation, upper body oxygenation levels are expected to improve due to this increase in VA ECMO support.
Due to low cardiac output, the location of the mixing zones for P1, P2 and P3 was in the aortic arch, adjacent to the BCA, with minimal variation for all three positions (Fig. 4). Accordingly, blood from the ECMO circuit perfused the LCCA and LSCA for P1, P2, and P3, comparable to native perfusion in a healthy state. Similar to 80% support, P1, P2 and P3 resulted in complete perfusion of VA ECMO blood to all branches below the arch along with the presence of a distinct boundary between blood from VA ECMO and the LV. Furthermore, due to combined effect of a proximal mixing zone location and a distally position cannula tip, P1, P2 and P3 were unable to provide flow with sufficient inertia to disrupt the aortic arch mixing zone, thereby resulting in similar flow features and mixing zone results. For P4, however, homogenous mixing of VA ECMO blood was seen throughout the aorta. Much less aortic flow was able to advance past the aortic arch compared to 50% and 80% support and flow from the cannula was able to advance around the outer bend of the arch, reaching the aortic root.

Afterload assessment
The effect of cannula advancement on cardiac afterload is presented in Table 1. Slight increases in systolic and diastolic pressure were observed as the cannula was advanced from P1 to P3. Mean aortic pressures (MAP) also slightly increased with a maximum increase of 6 mmHg during 90% VA ECMO support between P1 and P3. Interestingly, MAP decreased between P3 and P4 during all support cases. As expected, as VA ECMO support increased, a narrowing of pulse pressure was seen, with a decrease in systolic pressure and increase in diastolic pressure.

Discussion
In a simulation setting, we demonstrated that more proximal arterial cannula tip positioning within the aorta improves VA ECMO blood perfusion to the aortic arch branches. This finding was present even during low levels of support when the cannula tip was placed in P4. We also found much higher perfusion of blood from VA ECMO to the brachiocephalic artery (BCA) during P4 at 90% support compared to all other positions, thereby demonstrating the capacity to reduce the incidence of differential hypoxemia. Advancement of the arterial cannula greatly impacted flow characteristics within the aorta. Mixing zones are usually associated with a distinct boundary that shifts proximally as VA ECMO support is increased [15,16,19]. Regions distal to the mixing zone are completely perfused with blood from VA ECMO. Conversely, regions of the aorta proximal to the mixing zone are completely perfused with blood from the LV. Our study demonstrated such results during all P1 simulations. However, the mixing zone was disrupted when the cannula was advanced. In particular, all P4 simulations demonstrated mixing to occur longitudinally throughout the length of the aorta. Furthermore, the interaction between cannula and aortic flow resulted in unique and chaotic flow structures that prevented physiological development of native flow structures such as Dean vortices distal to the aortic arch [25].
Perfusion of blood from VA ECMO to the brachiocephalic artery (BCA) has not been observed in previous simulation studies unless maximum cannula flow rates were used [15,16,19]. These studies showed that high cannula flow rates (4 L/min and above) were required to establish a mixing zone in the aortic arch whilst a cannula flow rate of 5 L/ min was required to adequately perfuse the BCA with blood from the VA ECMO circuit. In our study, however, the BCA received oxygenated blood even during lower levels (50%) of support when placed in P4. Additionally, during 90% VA ECMO support and P4, blood from the cannula was homogenously distributed throughout the entirety of the aorta.
Despite not modelling the coronary arteries in this model, blood from VA ECMO appeared to reach the aortic valve during P4 at 90% support. Therefore, perfusion of the coronary vessels was likely achieved during this scenario. However, this observation should be confirmed in a more comprehensive coronary flow study. In other CFD and in-vitro based studies, the potential for improved coronary perfusion with blood from the ECMO circuit was not observed, even at high ECMO flow rates (> 4.5 L/min) [15-17, 19, 26, 27]. For example, Hoeper et al. demonstrated that clinical VA ECMO support with an arterial cannula tip placed within the common iliac artery and a flow rate of 4.5 L/min (similar to P1 during our 90% support case) resulted in a mixing zone in the aortic arch, identified using CT [27]. These results, which agree with those obtained in our simulation study, concluded that despite maximal increases in cannula flow rates (and for typically positioned cannulae), the ascending aorta and coronary arteries do not receive oxygenated blood from the VA ECMO circuit thereby potentially resulting in cardiac hypoxia and inadequate conditions for cardiac recovery. An advanced arterial cannula tip has been attempted clinically as described by Rodriguez and Maharajh who implemented this technique in two pediatric patients [20]. In their study, a 19 Fr drainage cannula was used in an offlabel manner in place of an arterial cannula because there are no commercially available arterial cannulae capable of proximal positioning. The cannula tip was advanced until the tip lay distal to the left subclavian artery (similar to P4 in our study) and a flow rate of 83 ml/kg/min was used with a premembrane pressure of 240 mmHg. Both patients showed no signs of differential hypoxemia during ECMO support and showed improved hemodynamics and saturations.
An important implication of our results is the use of P4 in cases of ECPR where maximizing cerebral perfusion is a priority [28]. Despite the high cannula blood flow rate during 90% support (4.5 L/min), in P1 only 12% of the flow to the BCA consisted of blood from VA ECMO. This is concerning as P1 reflects the most common arterial tip position used in current clinical practice [3]. Instead, our results demonstrated much higher cerebral perfusion of oxygenated blood to the BCA is possible at P4 with the added benefit of possible coronary perfusion at 90% support. This is particularly important in cardiogenic shock after acute myocardial infarction and in the setting of ischemic heart disease. Therefore, cannula tip position may be an important factor in preventing not only neurological injury in patients treated with ECPR but also cardiac injury.
While placement of the cannula in P4 resulted in increased oxygenated VA ECMO perfusion of the aortic arch vessels in our study, there was a decrease in the proportion of VA ECMO blood reaching all other vessels. In particular, VA ECMO perfusion of the two common iliac vessels decreased by more than half with P4 compared to P1 during 50% VA ECMO support. This is supported by numerical results from Bongert et al. who showed that advancing the cannula tip from the femoral artery to the abdominal aorta resulted in lower perfusion of blood from VA ECMO to the lower limbs [29].
We also found advancing the cannula caused a minor increase in afterload compared to other cannula positions. Only a maximum increase of 6 mmHg was observed in P3 1 3 compared to P1. However, a slight decrease in MAP between P3 and P4 was found for all levels of support. This can be attributed to an increase in total flow to the lower branches (from LV and VA ECMO) and a decrease in total flow to the arch branches in P3.
An increase in arterial cannula insertion length may cause concern due to the increase in cannula resistance and its potential for inducing hemolysis. However, the maximum pressure-drop observed between the cannula tip and cannula inlet surfaces was 160 mmHg (P4 at 90% support). This pressure-drop can be achieved by clinically used VA ECMO pumps by increasing pump speed accordingly as usually done when using arterial cannulas of smaller diameter [30]. For example, Stephens et al. demonstrated that a 15 Fr arterial cannula was able to provide targeted full ECMO support with a pressure drop of 282 mmHg across the cannula [31]. Due to the Hagen-Poiseuille flow relationship, increased resistance due to a decrease in cannula diameter is much more substantial than due to an increase in length. Given that smaller arterial cannulae are being readily used with low risk of complications [32,33], we expect similar outcomes with a long arterial cannula.
A long arterial cannula does increase the blood-cannula surface interface within the aorta. This may increase the risk of thrombus formation on the outer surface of the cannula as sometimes observed on drainage cannulae. However, incidence of such "thrombus sheaths" forming on drainage cannulae during VA ECMO can be attributed to the larger cannula size and the vastly disparate flow dynamics in the venous system [34]. This includes flow stasis and endothelial damage arising from vessel collapse. Even still, the incidence associated with such thrombus formation is low when considering VA ECMO [34]. An advanced arterial cannula tip may even mitigate thrombus formation that is observed at the aortic root. This is seen when poor LV function allows for a region of stagnant blood to clot in the ascending aorta. In our study, flow from the cannula placed in P4 showed elimination of this stagnation region when native cardiac output was low (90% support), thereby potentially reducing risk of thrombus formation in the proximal regions.
Various studies have previously investigated mixing zone location during VA ECMO using a CFD model [15][16][17][18][19]. However, these studies primarily involved variation of VA ECMO flow rates with a constant cannula position throughout. Cannulas were placed more distal to P1 from our study but tip placement was still within a common iliac artery. Comparison of results between our study (during P1) and Stevens et al. show good agreement at 50% support [16]. However, their exclusion of major vessels such as the SMA and IMA result in a much higher mixing zone location at 80% support. Nezami et al. used an idealized geometry which contained all the major vessels used in our study, and showed good agreement with respect to mixing zone locations and vessel perfusion et al. support levels when our cannula was placed in P1 [15]. In both aforementioned papers, however, cannula length was excluded from their models and was addressed as a limitation.

Limitations
We assumed the aortic walls to be rigid in our simulations. Incorporating wall deformability would require much higher computational power than was feasible for the number of simulations conducted. However, Nezami et al. included wall deformability and their mixing zone results showed negligible difference to simulations conducted with rigid vessel walls [17]. Secondly, the resistance and compliance parameters used in the Windkessel model reflect healthy patient conditions. Thus, any differences in these parameters associated with the heart failure state or vasopressor drugs were not simulated. Additionally, cardiac chamber pressure volume relations and autonomic nervous system autoregulatory mechanisms such as the baroreflex were not modelled in this study. Thus, cardiac and vascular changes in response to VA ECMO implementation were not simulated. Lastly, the results produced in this study have not been validated using in-vivo or in-vitro data. Therefore, all results should be interpreted with caution until experimental or clinical validation is performed.

Conclusion
In a simulation study, advancing the arterial cannula tip further into the aorta provides increased perfusion of oxygenated blood from the VA ECMO circuit to the aortic arch vessels at all levels of VA ECMO support. If translated to the clinical setting, this approach may reduce the incidence of differential and/or cerebral hypoxemia. In comparison, standard arterial cannula tip position was predicted to result in differential hypoxemia at both 50% and 80% support with only a moderate improvement at 90% support. These findings can inform clinicians in their choice of cannula length and position, and can form the basis of new cannula design.
MS, AV and MK aided in the interpretation of the data and provided vital contributions to the drafting of the manuscript. AV and MK also provided assistance with CFD modelling. JR and RB provided valuable clinical input and contributed towards the final draft of the manuscript contents. SG designed the study, aided in the interpretation of data and contributed towards the final draft of the manuscript.
Funding This work was supported by Monash University. Shaun D Gregory is the recipient of a Fellowship (102062) from the National Heart Foundation of Australia.

Conflict of interest
The authors declare that they have no competing interests.
Ethical approval and consent to participate Anonymized patient imaging and demographic data was used in this article in accordance with ethics approval from The Alfred Ethics Committee. Therefore, informed consent was not required.

Consent for publication
This manuscript has been approved for publication for this journal by all authors. Informed consent was not required from the patient as their imaging and demographic data was used in accordance with ethics approval from The Alfred Ethics Committee.