The Effect of Flow Field on the Initial Formation of Thrombus in Ventricular Assist Devices

Thrombosis is the main reason for the failure of ventricular assist devices (VADs). It has been acknowledged that the platelet activation induced by the nonphysiological blood flow leads to the increased thrombotic risk. However, due to the complicated influence of the VADs’ flow field and the difficulty in real‐time in situ observation, the mechanisms and process of thrombus formation in VADs remain unclear. In this work, the process of thrombus formation in VADs in vitro experiments is observed. The thrombus is found to form on the middle of the inlet guide vanes first and it is mainly caused by the immediate activation of platelets induced by the high shear rate of the flow field around the vanes and also affected by the rotation of the impeller. Then, subsequent thrombus is found in the tail of guide vanes and around the axle journal, where the blood flow is stagnated and the platelets are activated by the accumulated bioagonists. These findings clarify that the thrombus formed on the inlet guide vanes and the axle journal are dominated by two different mechanisms. This work provides unique insights into the initial formation of the thrombus on VADs and helps to reduce the thrombotic risk.


Introduction
Heart failure caused by left ventricular systolic dysfunction has become a global epidemic.Ventricular assist devices (VADs) have emerged as effective alternative treatments alongside heart transplantation, significantly improving patients' survival and quality of life. [1]However, with an incidence of 4% to 9% events per patient-year, thrombus is still a severe complication of VADs, which often induces the malfunction and inefficiency of the VADs and causes device-related thrombosis in the body. [2]Despite extensive research on VAD thrombosis, the initial formation of thrombus on VADs remains unclear due to its limited visibility.

DOI: 10.1002/admi.202300683
[5][6] The biological thrombotic processes, including coagulation and platelet activation and aggregation, are well known to associate with hydrodynamics. [7,8]The nonphysiological blood flow makes the interplay of biological pathways complex. [9]High shear stress affects mechanosensitive proteins and cells, even induces cell damage. [10,11]Low velocity strengthens transport and diffusion of proteins, increasing the residence time of prothrombotic components and inducing cell adhesion. [7,12]The process of formation of thrombus usually includes an initial stage and a rapidly developing stage. [13]Once the thrombus initially forms on the surface, it usually turns to become the adhesion point and causes the flow deterioration, which leads to the rapid increase of the thrombus in the subsequent stage.After the initial stage, the thrombus grew rapidly.[16] The researchers have identified severe thrombosis in several types of VADs following long-term tests in both in vivo and in vitro experiments.[15][16] Unfortunately, the exact process of the initial formation of the thrombus in VADs and the related hydrodynamic-biological coupling mechanism are still unclear due to complicated interaction between the blood and the VADs, and the difficulty of real-time observation inside these metallic devices.
Computational simulations have been utilized to study the process of the formation of thrombus in VADs.The effects of fluid dynamics, [23][24][25][26][27][28][29][30][31][32] cells, [33][34][35][36][37][38] and bioagonists [39][40][41][42][43][44] have been taken into account individually or jointly in these models.Some of them drew the conclusion that the thrombus generated first in the tail of the front stator's guide vanes. [30]The others proposed that the thrombus should initiate at the rear of front stator vanes and gradually grew both downstream into the bearing region and upstream along the vanes. [44]The descriptions of the initial thrombus formation process were very different in these computational simulation methods.In a word, overcoming the difficulties of direct observation and proposing a verifiable, accurate computational simulation method were main challenges of deeply understanding the effect of the flow field in VADs on the thrombus formation in initial stage.
In this work, we observed the initial formation of thrombus in a VAD with a transparent shell within an in vitro mock loop.The spatial location and the time sequence of the thrombus in the initial stage is captured with the fluorescent monitoring, and the flow field in the VAD is measured with the particle image velocimetry (PIV).The mechanism of the initial forming of thrombus on VAD is revealed and the effect of the flow field is clarified.According to the relationship between the spatial-temporal features of the thrombus inception and the flow field, a model of the initial formation of thrombus in VADs is established, which takes both the delayed shear-induced activation and the accumulative bioagonist induced activation into account.With this model, the initial formation of thrombus in VADs is well explained.

Materials
Blood for the in vitro experiment was collected from healthy swine using a venipuncture and stored in a 200 mL blood bag containing sodium citrate to prevent coagulation.The whole blood was centrifuged at 1500 RPM for 5-10 min, and the resulting supernatant was then centrifuged at 3000 RPM for 10 min to obtain the blood plasma.Alexa-594 labelled human fibrinogen (final concentration 12.5 μg mL −1 , Thermo Fisher) was added to both the whole blood and blood plasma to visualize the fibrin deposition.The dyed fibrinogen has red fluorescence under the green light excitation.During the experiment, the labelled blood was rapidly mixed with a recalcification buffer, prepared by dissolving 75 × 10 −3 m CaCl 2 and 35 × 10 −3 m MgCl 2 in physiological saline, to restore blood coagulation.The solution was injected at a rate of 0.5 mL min −1 for 10 min.

In Vitro Experiment in the Mock Loop
The design of the mock loop followed the ASTM.F1841 design standard and complied with the requirements for hemolysis and coagulation tests in a standard extracorporeal circulation experiment.The experiment platform consisted of four parts: 1) circulation loop module, including a circulation loop, flow sensors, pressure sensors, an injection port and a reservoir, 2) optical observation module, including a PIV laser device, a CCD camera, an excitation source of fluorescence, a digital single lens reflex camera and optical filters, 3) experimental object module with a heart pump, and 4) temperature control module, including temperature sensors and heaters, as shown in Figure 1A.The in vitro experiment platform was designed to monitor the essential technical parameters, including flow rate, pressure, rotational speed, voltage, current, and power in real time.This data helped estimate the thrombus formation process in VAD.The flow rate was measured by the flow rate sensor in the circulation loop.The flow rate sensor was FD-XS20 (KEYENCE (CHINA) CO., LTD.) and the measurement range was 0 to 20 L min −1 .The absolute error was 6.3 mL min −1 .The pressure of the left ventricular assist device (LVAD)'s inlet and outlet were monitored by two pressure sensors in the circulation loop.The pressure sensors were AP-10S (KEYENCE (China) CO., LTD.) and the measurement range was 0 to 100 kPa.The absolute error was 0.5 kPa.The rotational speed, voltage, current, and power of the LVAD were monitored and calculated by the controller of the LVAD.The LVAD used was an axial flow pump (Fuwai Hospital, Chinese Academy of Medical Sciences).The flow rate of the LVAD was maintained between 2 and 5 L min −1 and the pressure difference of the LVAD was maintained at 100 ± 5 mmHg.The flow rate used in the experiment was 2 ± 0.2 L min −1 , which was the same as the mean flow rate of the VAD in the in vivo experiment. [45]More details about the LVAD could be found in Zheng et al. [45,46] The optical observation module was designed for observing the flow field and thrombus formation.Particle image velocimetry (PIV) was used to study the flow field of VAD.The PIV laser device (Beamtech Optronics Co., Ltd.) provided a green planar pulsed laser that excited the fluorescence of particle in fluid during the experiments.The direction of observation was perpendicular to the plane of the PIV laser, as shown in Figure S1A  Fluorescence was used to observe thrombus formation in situ.The excitation source of fluorescence is Model SFA Fluorescence Adapter for Keyence VHX (NIHGTSEA, Electron Microscopy Science), which excited the red fluorescence of the thrombus in the experiments.The direction of observation was almost parallel to the direction of the excitation source, as shown in Figure S1B (Supporting Information).The area of the labeled thrombus could be clearly observed by the digital single lens reflex camera.The fibrin was larger and easier to observe than platelet.The activation and adhesion of platelets induced the generation of fibrin.Alexa-594 labelled human fibrinogen (final concentration 12.5 μg mL −1 ) was added into the whole blood and blood plasma to visualize the deposition of the fibrin and mark the thrombus.
Before experiment, normal saline solution was used to rinse and wet all the blood-contacting surfaces by circulating the solution for approximately 10 to 20 min.After rinsing the circulation loop, the normal saline was drained and filled it with 200 ± 50 mL of plasma which was labeled by Alexa-594 labeled human fibrinogen in the PVC loop and the reservoir.Air in the reservoir and the circulation loop should be eliminated.The pressure sensors were placed at the inlet and outlet of the VAD and measured the pressure in real time.The flow rate sensor was placed at the outlet side of the VAD and measured the flow rate in real time.The key technical parameters of in vitro experiment platform, including flow rate, pressure, rotational speed, voltage, current and power, were set to mimic the physiological conditions for which the VAD was designed.The rotational speed of VAD was 9000 ± 500 RPM.The flow rate was 2 ± 0.2 L min −1 .The mean pressure difference was 100 ± 5 mmHg.The motor power was 4.32 ± 0.5 W. The parameter fluctuation was 10 percent.The temperature was 37 ± 1 °C.The volume of circulation loop was 200 ± 20 mL.The low volume of circulation loop could enhance the VAD-related hemolysis and thrombosis. [47]

Mathematical Model
A simulation model was developed for thrombus formation by considering platelet activation and adhesion, while taking into account the relevant chemical and biological species.The model consisted of two parts: the hydrodynamics of blood and the biological reaction in thrombus formation.The hydrodynamics of blood included equations of motion that determine the pressure and velocity fields.A set of coupled convection-diffusion-reaction (CDR) equations governed the transport and inter-conversion of chemical and biological species.The biological reactions in thrombus formation were presented as source terms in CDR equations.The model was implemented by several modules of COMSOL Multiphysics 5.6, including Chemistry module, Turbulent flow module, Transport of Diluted Species module, Surface Reactions module and Multiphysics module.
k- turbulent flow model was applied to account for the hydrodynamics of blood.The blood was considered as a Newtonian fluid. [48,49]The viscosity of whole blood can be viewed as constant for shear rates higher than 100 s −1 .The shear rate in the specific regions of the vane and bearing in our work exceeded 100 s −1 , ranging from about 200 to 4000 s −1 .The density was set to  = 1.06 g cm −3 , and viscosity was set to μ = 0.00466 Pa s.Since the model did not consider the deformation of computational domain shape, a cylindrical computational domain was built, deleted the part of the VAD, and divided the remaining area into grids.The diameter of a cylindrical computational domain was 12.5 mm and the length was 18 mm.The computational domain was meshed with COMSOL Multiphysics 5.6 Meshing tool using unstructured tetrahedral elements.The computation model's geometries, mesh and computational condition were shown in Figure 2A.No-slip boundary conditions were prescribed at the VAD's walls.The boundary condition of inlet was static pressure and the boundary condition of outlet was velocity field with the effect of the impeller rotation.
The biological reaction in thrombus formation was modeled using a set of coupled CDR equations.The model of thrombus formation included 9 species: 5 states of platelets (RP: resting platelets in the flow field; AP: activated platelets in the flow field; RP d : deposited (trapped) resting platelets on the surface, AP d : deposited activated platelets on the surface, and AP s : Extruded and stabilized platelets on the surface), adenosine diphosphate (ADP), thromboxane A2 (TxA2), prothrombin (PT), thrombin (TB).
The model simulated the following mechanisms of thrombus formation, which occurred in blood through interactions among the nine chemical and biological species: 1) Platelet activation: RP could be converted to AP by contacted with critical levels of bioagonists, shear rate, and surface of VAD.It was assumed that this transformation causes platelets to adhere to the surface of the biomaterial and release ADP.Besides, it was assumed that there will be an activation delay time (about 6 ms) for platelets to be activated after reaching the activation conditions, rather than immediately changing states from RP to AP. 2) Generation of coagulation-related agonists: The release of ADP was caused by the conversion of AP to RP. Thromboxane A2 (TxA2) was synthesized from AP. Thrombin (TB) was synthesized from prothrombin (PT) on the phospholipid membrane of platelets.3) Platelet adhesion: Both RP and AP in free flow could deposit on the surfaces of biomaterial, such as channel walls, VAD, so they were designated RP d and AP d respectively.The process was modeled mathematically by the adhesion reaction rate of RP d and AP d on the boundary.There were differences in adhesion reaction rate of different kinds of biomaterials.4) Thrombus dissolution or erosion: Shear stress exerted by the fluid could clean deposited platelets and thus corresponded to a phenomenon called surface "cleaning".Mathematically, this was modeled by the thrombus clearance rate associated with shear stress.The adhesion of free platelets and the clearance of adhesive platelets were in dynamic equilibrium.5) Thrombus stabilization: Newly deposited active platelets (AP d ) were transformed at a constant rate into stable clots that cannot be removed by hydrodynamic (diffusion or shear) clearance.Additional details and descriptions of the models were provided in Wu et al. [40,42,44] Based on these assumptions, the spatiotemporal kinetics of the bio-agonists in the flow field, as included in the model, were governed by a set of CDR equations in the following form: where [C i ] denoted the concentration of species i, D i denoted the diffusivity of species i, R i represented the reaction source term for species i and ⃗ u represented the fluid velocity vector (assume that biochemical species has the same velocity as fluid velocity).The source term R i was used to implement interactions, such as platelet activation, adhesion, and bioagonists generation.Deposited platelets (RP d , AP d , and AP s ) do not have convection and diffusion terms associated with them, and are each governed by a concentration rate equation: Platelet-surface adhesion and all the other reactions are modeled by surface-flux boundary conditions.Here, similar to the reaction terms in the internal domain, a negative and a positive flux corresponds to consumption and generation, respectively.Additional details of the physical meaning of all the terms were provided in a prior publication. [40,42,44]

Statistical Analysis
All experimental data were expressed as mean values ± SD and are analyzed using SPSS 19.0 (SPSS).The sample size (n) for each statistical analysis is 3.

In Vitro Experiment of Thrombus Formation in VADs
We designed an experimental platform of blood circulation to simulate the physiological environment in the body (including pressure, temperature, flow rate, etc.) to capture the spatial location and the time sequence of the thrombus in the initial stage, as shown in Figure 1A.The VAD consisted primarily of a front stator (including 3 guide vanes and a front bearing), an impeller and a rear stator, as depicted in Figure 1B.The front stator and front bearing of VADs had a high possibility of initial thrombus formation, [17][18][19][20][21][22] so we chose the area around front stator (red frame in Figure 1B) as the observation area.The blood in circulation loop was driven by the VAD and circulated for 10 h.To counteract the anticoagulant effect of sodium citrate added during blood collection and recover the coagulation capacity of the blood, a solution (75 × 10 −3 m CaCl 2 and 35 × 10 −3 m MgCl 2 in physiological saline) was injected at a rate of 0.5 mL min −1 for 10 min via injection port.The solution was injected while blood was flowing in the loop and mixed by coflowing with the fluid in the section of straight tubing upstream of the VAD.The streams were well mixed because the volume of solution injected was extremely small compared to the flow rate.In order to observe the thrombus in real time during the experiments, we used not only whole blood but also blood plasma to form thrombus in VADs in the in vitro experiment.The whole blood was collected from healthy swine using a venipuncture and stored in blood bag (200 mL) containing sodium citrate to prevent coagulation.The whole blood was centrifuged at 1500 RPM for 5-10 min and the supernatant was taken to centrifuged at 3000 RPM for 10 min to get the supernatant as blood plasma.The blood plasma was almost transparent and rich in platelets, fibrin, and coagulation factors.The thrombus formed by plateletrich plasma were consisted of platelets and fibrin.The fibrin was larger and easier to observe than platelets, so we added Alexa-594 labelled human fibrinogen (final concentration 12.5 μg mL −1 ) into the whole blood and blood plasma to visualize the deposition of the fibrin and mark the thrombus.The process of thrombus formation and the flow field in VAD were observed by fluorescent observation and particle image velocimetry (PIV) of optical observation module.The observation area was shown in Figure 1C1, including the front stator and the axle journal.
The streamlines of flow field around the guide vane of the front stator in the observation area were shown in Figure 1C2.We observed that the flow beneath the guide vane was basically parallel to the flow direction, while vortices existed above the guide vane.The directions of flow changed around the axle journal, influenced by the rotation of the impeller.The distribution of velocity around the guide vane were shown in Figure S2A (Supporting Information).We found that the velocity beneath the guide vane was higher than the velocity above the guide vane.The velocity decreased obviously in the front of guide vane and a low velocity region appeared in the tail of guide vane.The distribution of shear rate around the guide vane were shown in Figure S2B (Supporting Information).We found that there were high shear rate zones near the front and tail of the guide vane.The shear rate beneath the guide vane was also higher than the shear rate above the guide vane.We considered that the asymmetrical flow field around the guide vane was caused by the rotation of impeller.The impeller rotation induced the difference of streamlines, velocity and shear rate in the region facing to the rotation flow (beneath the guide vane in Figure 1C2) and the region sheltered from the rotation flow by the guide vane (above the guide vane in Figure 1C2).The average velocity in the observation area was 1.4 ± 0.4 m s −1 .The average shear rate in the observation area was 2200 ± 300 s −1 .However, due to the block of the PIV laser by the guide vanes and the precision of the PIV observation, more details of the flow field were not revealed.
As shown in Figure 1D1-D4, the process of thrombus formation is presented.At T = 2 h, we found that there was red fluorescence of thrombus above the middle of guide vanes.At T = 6 h, the red fluorescence of thrombus above the middle of guide vanes continued to grow downstream.The red fluorescence of thrombus also appeared under the middle of guide vanes and in the rear of guide.At T = 10 h, the thrombus area in the middle part and the rear of the guide vanes continued to increase.The thrombus area in the rear of the guide vanes grew faster.We also found that the thrombus formed on the guide vanes of front stator before the guide vanes of rear stator.Only a few of thrombi appeared on the edge of the rear guide vanes at T = 10 h, as shown in Figure S3B (Supporting Information).So the formation of thrombus on the guide vanes of rear stator was not analyzed in detail.It was known that high shear rate was considered as the main reason of platelet activation and thrombus formation.However, the data in Figure S2B (Supporting Information) showed low shear rates in the middle of the guide vanes, where the thrombus appeared first in the experiment.Therefore, we assumed that thrombus first appeared in the middle of guide was induced by high shear rate in the front of the guide vane.And then the thrombus grew in the rear of the guide vanes was induced by low velocity.These assumptions were verified by the following computational results and the mechanisms were revealed.
Based on the observed flow field, we discovered that the impeller rotation induced the variations in velocity in the region facing to the rotation flow (the region beneath the guide vane) and the region shielded from the rotational flow by the guide vane (the region above the guide vane).This was the asymmetrical flow field around the guide vanes caused by the rotation of impeller.According to the results of thrombus formation, we found that thrombus first appeared above the middle of guide vanes and then appeared under the middle of guide vanes.This was the temporal asymmetry of thrombus formation.We also found that the thrombus above the guide vanes was larger than the thrombus beneath the guide vanes.This was the spatial asymmetry of thrombus formation.We considered that the temporal and spatial asymmetry of thrombus formation were caused by the asymmetric flow field around the guide vanes.

In Silico Analysis of Thrombus Formation in VADs
According to the experimental results, we found that the thrombus formation on VADs was associated with both the high shear rate and low velocity in initial stage.We proposed a novel transient model, including the temporal and spatial effects of thrombus formation to illustrate the mechanism of thrombus formation on VADs in initial stage and the relation between flow field and thrombus formation.The model considered 2 types of platelet activation: shear-induced platelet activation and bioagonists-induced platelet activation.We used the increase in the concentration of activated platelets to depict the thrombus formation.The shear platelet activation rate K spa was calculated based on shear rate, while the bioagonists platelet activation rate K apa was determined by concentrations of thrombosis related bioagonists.We also considered the impact of platelet activation delay time and flow asymmetry to elucidate the spatial distribution and asymmetry of thrombus.The effect of activation delay on the thrombus's location was particularly notable in high-velocity flows, necessitating its inclusion in the VAD thrombus model.

Computational Model of Thrombus Formation
Given the mechanical sensitivities of platelets and nonphysiological shear rate caused by VAD, platelets were activated by high shear rate. [50]In this work, we proposed shear platelet activation rate K spa .K spa was given by the following formulas: where  was shear stress (units: dyne cm −2 , 1 dyne cm −2 = 0.1 Pa), which was calculated by shear rate du/dr (units: s −1 ) and constant viscosity μ (units: Pa s) according to the following formula:  = 10μdu/dr.t ct,spa was the calculated activation time, t act was the physical limit time of platelet activation (t act = 0.1 s).Because the platelet activation rate existed physical limitation, the shear activation rate K spa did not continue to increase when the shear stress exceeded a threshold value (the calculated activation time was bigger than t act ).
On the other hand, the platelets could not only be activated by mechanical factor (Shear stress), but also be activated by chemical factor (bioagonists). [51] In this work, we proposed bioagonists pathway of platelet activation, which correspond to platelet bioagonists activation rate K apa and took 4 bioagonists into account, namely adenosine diphosphate (ADP), thromboxane A2 (TxA2), prothrombin (PT), thrombin (TB).The release of ADP was caused by the conversion of AP to RP. TxA2 was synthesized from AP. TB was synthesized from PT on the phospholipid membrane of platelets.High concentration of ADP, TxA2 and TB could lead to activation of platelets.K apa was given by the following formulas: where (x, t) was concentration coefficient.t ct was the characteristic time, which could be used for adjusting the activation rate, and here we chose t ct = 1 s.
[TB] thr was 0.10 × 10 6 U m −3 and w j of TB was 30.We summarized the abbreviation and source term forms of all species of the model in Table 1 and listed the units and initial (inlet) condition of all species in Table S1 (Supporting Information).Additional details of our thrombus formation model, including the explanation of the values and physical meanings of these terms, the values and origins of all coefficients, were provided in prior publications. [40,42,44]esides, Yazdani et al. found that platelet needed a period of time to transform from resting state to activated state and the activation delay time was about 6 ms after reaching the activation condition. [36]These results had been verified by experiments and computational model.Therefore, we proposed that the platelets activation needed a delay time (6 ms) before the transformation from RP to AP, rather than immediately changing states.A corresponding computing method was established, which took both the delayed shear induced activation and the accumulative bioagonist induced activation into account.This model could simulate more accurate location of thrombus in high velocity flow than other VAD's thrombus formation model.
First, we divided the delay time into several episodes (i) and calculated a set of new coordinates through iteration: where x 0 , y 0, and z 0 denoted the original coordinates, x i , y i and z i denoted the new coordinates before delay time.u i , v i and w i denoted the components of the flow velocity in the x, y, and z directions at corresponding coordinates (x i , y i and z i ).
Secondly, we read the data of the shear rate and the bioagonists' concentration at the new coordinates (x i , y i and z i ) and calculated the value of K spa and K apa .These value of K spa and K apa represented the platelet activation rates at the original coordinates (x 0 , y 0 and z 0 ).
The flow field around the front stator was simulated and the streamlines of the area around front stator were shown in Figure 2B.The flow was steady in the area around the front stator and upstream of the impeller.We observed a significant decrease in velocity in front of the guide vane due to the guide vane's obstruction.We also found that a low velocity region appeared in the tail of guide vane caused by the flow separation.The vortices formed around the axle journal.The velocity and directions of flow changed rapidly around the axle journal, influenced by the rotation of the impeller.In order to describe the flow field and the distribution of surface concentration on the guide vanes, we defined a line on the middle of the guide vane and the axle journal as the Z′-axis and the flow direction was the positive direction, as shown in Figure 2C.

First-Stage Shear Induced Thrombus Following High Shear Rate Region
High shear rate was considered as the main reason of platelet activation and thrombus formation.The presence of high shear rate zones near both the front and tail of the guide vane was observed, as shown in Figure 2D.The high shear rate regions in the front of the guide vane resulted from the rapid change in flow velocity at the front of the guide vane, caused by the obstruction effect of the guide vane.The high shear rate zones near the tail of the guide vane were due to the difference of flow velocity between the tail of the guide vane and other regions around the guide vane caused by the flow separation of the guide vane.Therefore, we discussed these 2 high shear rate zones separately.
Regardless of the platelet activation delay time, the distribution of the shear activation rate K spa was similar to the distribution of high shear rate, as described in Equations ( 1) and (2).But this was not consistent with the initial location of thrombus observed in vitro experiment.The platelet activation delay time (6 ms) resulted in a disparity in the position between the high shear region and high K spa region along the direction of flow.The high shear rate region in the front of the guide vane induced the high K spa region in the front and middle of the guide vane.The high shear rate region in the tail of the guide vane did not lead to the thrombus formation on guide vane because the platelet activation delay time resulted in the activated platelets moving away from the guide vane, as shown in Figure 2E.These activated platelets might adhere to other downstream areas of VAD, such as the impeller and the rear stator.Due to the high speed of the impeller rotation, the thrombus formation on the impeller occurred later than the thrombus formation on the front stator.Furthermore, we also observed that the asymmetrical flow caused by impeller's rotation led to the asymmetry of thrombus formation on VAD's guide vanes in initial stage.Figure 2F showed the simulation results of the asymmetrical thrombus formation at T = 6 min and the length of red lines represents the surface concentration of APs on the guide vane.By adjusting the direction of the outlet flow, we simulated the thrombus formation under asymmetrical flow.The asymmetrical flow conditions were shown as Figure S4A (Supporting Information).The asymmetrical flow field was shown as Figure S4B,C (Supporting Information).The asymmetrical concentration distribution of AP was shown as Figure S4D (Supporting Information).Because of the asymmetrical flow, the flow velocity on both sides of the guide vane was different.The flow above the guide vane existed vortices, as shown in Figure 1C2.We hypothesized that the activated platelets were more likely to deposit in low velocity region, instead of high velocity region.The vortices increased the possibility that activated platelets interacted with the surface above the guide vanes and induced more platelets adhesion.Therefore, the adhesion of platelets was asymmetrical, which was the same as the experiment result.

Second-Stage Bioagonists Induced Thrombus in Stagnation Region
As demonstrated by the results, the shear activation model explained why thrombus formation occurred in the middle of the guide vane rather than at the rear of the guide vane.Therefore, we used the bioagonists activation mechanism, which mimicked the biochemical processes of platelet activation to illustrate the reason of these thrombus formations.These two platelets activation mechanisms existed simultaneously in the model with the different corresponding conditions and time of activation.The shear activation mechanism was triggered immediately after platelets entered the high-shear region.The bioagonists activation mechanism did not start until the concentration of bioagonists that accumulated from the blood and the release of activated platelets reached the threshold of platelets activation.
There were low velocity regions in the tail of the guide vane and the axle journal of the VAD, as shown in Figure 3A1.The low velocity region in the tail of the guide vane was due to the obstruction effect of the guide vane and the flow separation occurred near the tail of the guide vane, as shown in Figure 3A2.The low velocity region around the axle journal was also due to the flow separation caused by the difference of diameters between the front stator and the axle journal, as shown in Figure 3A3.
The mechanism of the thrombus formation in the rear of guide vanes, including the thrombus formation in the tail of the guide vane and around the axle journal, was bioagonists activation.Unlike the activation of platelets by high shear rate, concentrations of biological agonists need to accumulate over time to reach the threshold of platelets activation.Low flow velocity led to accumulation of biological agonists.Once the concentration of bioagonists had accumulated and (x, t) exceeded 1, it initiated the activation of platelets at high concentrations of bioagonists.The distribution of (x, t) and the contribution of TB, ADP and TxA2 to (x, t) along the Z'-axis at different time were shown in Figure 3B1-B3.We found that TxA2 made the dominant contribution to bio-agonists activation.Therefore, it was found that there were high concentration regions in the tail of the guide vane (as shown in Figure 3C1-C5) and the axle journal (as shown in Figure 3D1-D5).The results indicated that the concentration coefficient (x, t) required at least around 6 minutes to reach the threshold and initiate the bio-agonists activation.This also explained why thrombus formation took more time in the rear of the guide vanes during the in vitro experiment.The velocity of flow on the middle of guide vane was higher than it in the rear of guide vane and the bioagonists were hard to accumulate in high velocity regions.Besides, bioagonists activation released more bioagonists and further promoted the accumulation of biological agonists.So, thrombus formed in the rear of the guide vanes after a period of time in vitro experiment.
Additionally, there were variations in specific timing of thrombus formation between the simulation and the experimental results, with the simulation showing earlier thrombus formation compared to the experiments.These differences might be caused by the insufficient observation accuracy of in vitro experiment The observation accuracy was not good enough to gain the image of thrombus at the initial time of thrombus formation.We could only catch the fluorescence image of thrombus when the concentrations of deposited platelets and fibrin were high enough but the concentrations were extremely higher than the concentrations when the bio-agonists activation started in simulation results.

Composition Analysis of Thrombus Formation in VADs
We analyzed the composition of plasma thrombus and whole blood thrombus on the guide vane and the axle journal by martius scarlet blue (MSB) staining, hematoxylin-eosin (H&E) staining, platelet immunofluorescence staining, platelet immunehistochemical staining and scanning electron microscope (SEM).The whole blood thrombus was formed under the same conditions as the plasma thrombus.As shown in Figure 4A1,B1, the staining results of MSB revealed that the components of the whole blood thrombus contained abundant fibrin, while the thrombus around the axle journal contained a significant number of red blood cells (indicated by the yellow area with a red arrow).The SEM images also indicated that the thrombus on the surface of guide vane contained few red blood cells, while the thrombus around the axle journal contained a lot of red blood cells, as shown in Figure 4A2,B2.The staining results of H&E and the SEM images revealed that the components of plasma thrombus contained rich fibrin, as shown in Figure S5A1,B1,A3,B3 (Supporting Information).The fibrin in the plasma thrombus around the axle journal was denser than that on the surface of the guide vane.Additionally, gaps and holes were observed in the thrombus on the surface of the guide vane.According to the results of platelet immunofluorescence staining, the thrombus also contained a large number of platelets, indicating that the thrombus formation was caused by platelet activation and adhesion.These results confirmed the reliability and validity of the platelet simulation model.Furthermore, we found that the concentration of platelets in guide vane's thrombus was higher than the concentration in the axle journal's thrombus, as shown in Figure S5A2,B2 (Supporting Information).
These differences of composition were similar with the differences in the composition of thrombus in arteries and veins.Artery thrombus was composed primarily of fibrin and platelets.The fibrin was found mainly in the form of bundles.Venous thrombus was composed primarily of RBCs and fibrin fibers.The RBC content was significantly higher in venous thrombus than in artery thrombus.The platelet content was significantly lower in venous thrombus than in artery thrombus. [52]Therefore, we speculated that the distinction of composition might be caused by the difference of flow velocity and shear rate around the guide vanes and the axle journal.The shear rate on the guide vanes was higher than that around the axle journal and the flow velocity around the axle journal was lower than that on the guide vanes.The high shear rate region was more likely to form the thrombus rich in fibrin and platelets.The low flow velocity region was more likely to form the thrombus rich in RBCs and fibrin.These findings clarified that the different flow field of VAD influenced not only the location and sequence of thrombus formation, but also the composition of thrombus through two corresponding mechanisms.

Discussion
In this work, we found that the middle and the rear of the guide vanes were prone to thrombus growth in vitro experiment.Thrombus formation initially occurred on the middle of the guide vanes before extending to the rear of the guide vanes.We also found that the formation of thrombus existed the temporal and spatial asymmetry, which caused by the asymmetric flow field around the guide vanes.
In order to illustrate the mechanism of thrombus formation on VADs in initial stage, we established a computational model of thrombus formation.In this model, platelet activation had two pathways: shear activation and bioagonists activation, which corresponded to platelets shear activation rate K spa and bioagonists activation rate K apa respectively.There were 2 regions with high concentration of APs: the middle of the guide vanes and the rear of the guide vanes, indicating that activated platelets deposited preferentially at these regions and formed thrombus.This was basically the same as in vitro experiments.Platelets were activated in the high shear rate regions in the front of the guide vane and continued to flow with the blood due to the activation delay time.After the activation delay time, they changed to the activation state and adhered to the surface of the guide vanes, as shown in Figure S6A,B (Supporting Information).Bioagonists accumulated from the blood and the release of activated platelets occurred in the rear of the guide vanes.After the concentration of bio-agonists achieved the threshold value of activation, platelets were activated by the high concentration of biological agonists, as shown in Figure S6A,C (Supporting Information).A limitation of our model was that it did not simulate the changes of the flow field induced by thrombus formation, as thrombus formation was represented by surface concentration of APs.Our model was mainly based on the original flow field of the VAD, which was valid and effective for the initial stage of thrombus formation.Wu et al used a similar thrombus formation model and set that the permeability of the thrombus was inversely related to the volume fraction of the deposited platelets.The change of thrombus' permeability influenced the flow field in the vicinity of the surface of the VAD.This method might be a possible solution and improvement. [44]e found 2 representative phases in initial stage of thrombus formation, as shown in Figure 5: 1) Platelets were first activated by shear rate and the thrombus began to grow on the middle of the guide vanes, as shown in Figure 1D2 and Figure S7B (Supporting Information).Because shear induced thrombus formation started in high shear rate regions immediately.2) Platelets were still activated by shear rate and the bio-agonists from the blood and the release of activated platelets were transported to the low velocity regions in the rear of guide vanes.Platelets began to be activated by high concentration of bioagonists, as shown in Figure 1D3 and Figure S7C (Supporting Information).And then, bioagonists activation made a dominant contribution to the thrombus formation and platelets mainly deposited at the axle journal of VADs, as shown in Figure 1D4 and Figure S7D (Supporting Information).After the initial stage of thrombus formation, the thrombus grew rapidly and covered the entire front stator quickly, as shown in Figure S8A (Supporting Information).Rapidly growing thrombus blocked the flow channel of VADs and interfered the function of VADs in several minutes, as shown in Figure S8B (Supporting Information).The thrombus formed by platelet-rich plasma was shown in Figure S8C (Supporting Information).The thrombus formed by whole blood was shown in Figure S8D (Supporting Information).The thrombus of VADs eventually formed in vitro experiments were consistent with the thrombus eventually formed in animal experiments with the same type of heart pumps. [45]he location, formation sequence and composition of thrombus, along with the two corresponding mechanisms, hold significant reference value for the design of new VADs.In order to avoid shear activation, we could design the geometry and flow of the VAD to decrease the shear rate before the platelets suffered high shear rate and were activated.In order to avoid bio-agonists activation, specific scouring structure could be added in the VAD to accelerate the flow of low velocity region and strengthen the convection of bioagonists.Concerning the delay time of platelet activation, we could predict the locations easy to be adhered by activated platelets more precisely and coat them with antiadhesion coating.Concerning the different composition of thrombus, we could choose corresponding kinds of anticoagulant coating designed for artery thrombus and venous thrombus to achieve the best antithrombotic effect.Selecting the appropriate VAD design and anticoagulation strategy is crucial for effectively preventing VAD thrombosis in the long term.Besides, this model could be used not only for the analysis of thrombus formation in axial flow heart pumps, but also for other types of heart pumps, such as centrifugal heart pumps and pulsating heart pumps.The location, formation sequence of thrombus changed according to the flow field of different pumps.Therefore, it was also helpful for other types of VADs to reduce the thrombotic risk.

Conclusion
In this work, the initial formation of thrombus in a VAD with transparent shell was observed in an in vitro mock loop.The spatial location and the time sequence of the thrombus in the initial stage is captured in real time with the fluorescent monitoring, and the flow field in the VAD is measured with the particle image velocimetry (PIV).We identified a unique process of thrombus formation, wherein the thrombus first developed in the middle of the guide vanes, followed by growth in the tail of the guide vanes and around the VAD's axle journal.The mechanism of the initial forming of thrombus on VAD is revealed and the effect of the flow field is clarified.According to the relationship between the spatial-temporal features of the thrombus inception and the flow field, a model of the initial formation of thrombus in VADs is established, which takes both the delayed shear induced activation and the accumulative bioagonist induced activation into account.The thrombus on the middle of the guide vanes were mainly caused by shear activation and the thrombus in the tail of the guide vanes and the axle journal were mainly caused by bioagonists activation.Furthermore, we also found variations in the composition of thrombus in different locations, influenced by the flow field.The thrombus on the middle of the guide were rich in fibrin and platelets.The thrombus around the axle journal were rich in RBCs and fibrin.With this model, the initial formation of thrombus in VADs is well explained.This study revealed how the flow field in VADs affected the location, formation sequence and composition of thrombus.These unique insights of the initial thrombus formation on the VADs were valuable for enhancing the hemodynamic and anti-thrombotic design of VADs.
(Supporting Information).The charge coupled device (CCD) camera used was PowerView Plus 11MP (TSI INCORPORATED) and the image acquisition and analysis system software was Insight3G.The system could control the laser device and CCD camera by TSI LaserPulse synchronizer (TSI INCORPORATED).The CCD camera had a resolution of 4K × 2.6K pixels and a maximum frame rate of 4.8 frames per second.The PIV Frame Mode was set to straddle.The laser pulse delay was 405 μs.The time between 2 laser pulse was 200 μs and the PIV exposure time was 450 μs.The frames captured by CCD were imported and processed by several modules in the DYNAMIC STUDIO software, including Adaptive PIV, Vector Masking, UV Scatter Plot Validation, Average Filter, Vector Statistics and Numeric Export.The streamlines of flow field were plotted by TECPLOT software.The distributions of velocity and shear rate were plotted by Origin software.

Figure 1 .
Figure 1.The flow field around the front stator and the process of thrombus formation on guide vanes.A) The experimental platform of blood circulation.B) The geometry model of the axial VAD (The red box area represented the observation area.).C1) Observation area:The front stator and the axle journal with fluorescent particles of PIV.C2) The streamline of flow field by PIV.(Scale Bar: 3 mm) D) In vitro experiment of VAD's thrombus formation with Stained plasma.D1-D4) Fluorescent image of the thrombus formation process (T = 0 h, 2 h, 6 h,10 h).The red fluorescent areas represented thrombus.(Scale Bar: 2 mm).

Figure 2 .
Figure 2. High shear rate region and shear induced thrombus formation.A) The geometries, mesh and computational condition of the flow field around the front stator of VAD:Flow rate (2 L min −1 ) and rotation rate (8000 RPM).B) The streamline of the flow field around the front stator of VAD.C) A red line was defined on the middle of th e guide vane and the axle journal as the Z'-axis, and the flow direction is the positive direction.The front view of front stator showed that the Z'-axis included the little part of front stator, the upper surface of the guide vane and the axle journal.The top view of guide vane showed that the Z'-axis was in the middle of upper surface of the guide vane.D) The distribution of shear rate along the Z'-axis (2 L min −1 , 8000 RPM).Illustration: The distribution of shear rate in the sectional view of guide vanes at x = 4.5 mm.E) The distribution of K spa along the Z'-axis.Illustration: The distribution of K spa in the sectional view of guide vanes at x = 4.5 mm.F) The asymmetrical flow caused by the impeller rotation resulted in the asymmetry of APs' concentration on VAD's guide vanes.The length of red lines represents the surface concentration of APs on the guide vane (T = 6 min).

Figure 4 .
Figure 4. Composition of whole blood thrombus.A) Composition of whole blood thrombus on the surface of the guide vane: A1) MSB staining (Scale Bar: 100 μm).A2) SEM images (Scale Bar: 20 μm).B) Composition of whole blood thrombus around the axle journal of the VAD: B1) MSB staining (scale bar: 100 μm).B2) SEM images (scale bar: 20 μm).The color of RBCs in A2 and B2 is processed by Photoshop to clearly indicate their locations.

Figure 5 .
Figure 5. Schematic of thrombus formation on the VAD.There were 2 pathways of platelet activation: (1) high shear rate (orange region), (2) high concentration of bio-agonists (blue region).There were high shear rate areas in the front and tail end of the guide vane, and there were obvious low flow velocity areas in the end of the guide vane.High shear rates directly caused shear activation.Low flow velocity caused the accumulation of biological agonists and induced bio-agonists activation.

Table 1 .
The abbreviation and source term forms of all species in the model.