Optimizing the performance of ventricular assist devices

Abstract

A method is described for determining the effect of initiating or operating a ventricular assist therapy, such as a ventricular assist device 1. One or more parameters of a patient's heart 3 and one or more parameters of the ventricular assist therapy are determined. A mechanical and fluid dynamic model of at least one ventricle in which the ventricular assist therapy is or will be used is then used to calculate a predicted quantity of interest of the patient's heart 3 by initiating or operating the ventricular assist therapy. This may be performed iteratively until a predetermined configuration objective is achieved.

Description

[Technical field]

[0001] A ventricular assist device (VAD) is an implantable artificial pump intended to partially or completely replace the function of the heart. VADs are most frequently designed to assist the left ventricle (LVAD), but can be used to assist the right ventricle (RVAD) or both ventricles (BiVAD). [Background technology]

[0002] VADs provide life-saving therapy for patients with advanced heart failure, and are used as a bridge to future heart transplantation or as a targeted therapy (with the intention that dependency on the VAD will be permanent). Despite improved survival, multiple side effects are associated with these treatments. The main complications remain bleeding or thrombosis, occurring in approximately 20% of cases [4,6].

[0003] One important issue is that changes in the left ventricular (LV) flow pattern resulting from the use of a VAD can induce LV stagnant areas. Subsequent generations of LVADs have introduced an "artificial pulse" pattern, a modulation of pump speed, that attempts to break up stagnant areas. However, this is not a complete solution to the problem, and several other variables in the implantation of a VAD have been shown to affect the risk of thrombosis, particularly the alignment and placement of the cannula [10,11]. However, without a trial-and-error approach, it is difficult to determine the success or failure of such an approach or to identify the best conditions for placement or operation of a VAD for an individual.

[0004] There is also evidence of inflammation associated with VAD use [2], which leads to an increased risk of thrombosis. This, combined with the risk of endothelial lesions and the described abnormal flow patterns [5], are the three components of Virchow's triad [3] for thrombus formation. Local flow conditions also influence the type of thrombus generated. High velocities and high shear stresses lead to platelet activation [8], which aggregates fibrin and thus forms so-called white clots. In contrast, stagnant, slow recirculating flows and low shear stresses lead to aggregation of all blood components, including red blood cells [9], as well as leukocytes infiltrated with fibrin [7], resulting in so-called red clots.

[0005] In this context, VADs are generally composed of centrifugal or axial pumps that remove blood from the ventricles by suction. The highest velocities are generated around the rotor. The lowest velocities are generated in the far field of the suction domain (in this case the LV itself). Thus, white thrombi are generated near the rotor and red thrombi are generated far from the suction domain (LV). The problem of white thrombi formation is generally addressed by appropriate pump design, for example by the use of hydrodynamic or magnetic bearings.

[0006] After the flow is diverted from the ventricle to the inside of the cannula pump, the rotor, and the remainder of the flow path to the outlet graft for connection with the aorta, the flow becomes a purely engineering problem that is likely influenced by the patient's anatomy or condition, with negligible effect.The main practical difficulty that still needs to be effectively addressed is the problem of red thrombus formation in the LV itself.

[0007] In particular, reliable data on the efficacy of VAD procedures are difficult to obtain. Clinical trials and animal studies of VAD are an accurate representation of reality, but they include the following: High cost. Low (often statistically irrelevant) number of samples. Little or no control over patient parameters. ·Ethical issues. · Little or no access to important measurable information (e.g. fluid dynamics). Bench testing of VADs has been found to have several advantages and disadvantages compared to clinical trials (or numerical modeling), including the following: · Requirement for expensive low-use equipment. ·Insufficient control over variables. · Poor measurements. User error.

[0008] In contrast, computer simulations are less expensive than clinical, animal and bench studies, do not require specific equipment, allow a high level of control over variables, and allow quantities of interest to be obtained at a high level of detail. Summary of the Invention

[0009] In a first aspect, the present disclosure provides a method for determining the effect of initiating or operating a ventricular assist therapy, the method comprising the steps of determining one or more parameters of a patient's heart, determining one or more parameters of the ventricular assist therapy, and calculating a predicted quantity of interest in the patient's heart by initiating or operating the ventricular assist therapy using a mechanical and fluid dynamic model of at least one ventricle in which the ventricular assist therapy is or will be used.

[0010] The method may involve varying one or more of the parameters, and then continuing the calculation steps for the varied parameters until a determined set of parameters and associated predicted quantities of interest are obtained to meet a predetermined objective. The objective may be to identify an operating space over a range of variable values. The objective may be to determine parameters of a ventricular assist therapy that result in a predicted quantity of interest that meets predetermined criteria (such as performance, safety and efficacy indicators). The ventricular assist therapy may be the implantation and operation of a ventricular assist device, such as a left ventricular assist device.

[0011] In principle, parameters from a patient's heart can be obtained for a particular patient, but an approach of particular interest is to determine these parameters for an uncertain population. The patient's heart parameters may relate to the subject's anatomy or cardiac activity (e.g., cardiac morphology and other subject-specific geometry, ventricular volumes such as end-diastolic volume, ejection fraction and heart rate) or the subject's general condition (e.g., arterial mean pressure, arterial resistance or arterial capacitance). Ventricular assist therapy parameters for a VAD may include the subject's cannula implantation specifications (implantation location, insertion depth, cannula angle, cannula shape, cannula geometry) or VAD pump operating conditions (pump speed waveform, pump synchronization with intrinsic heart rate, pump performance function (HQ function)).

[0012] The quantities of interest may be obtained directly from calculations (e.g., ventricular velocity and pressure field) or may be derived (such as residence time, velocity magnitude, kinetic energy, strain time and pulsatility index). The predetermined condition may, for example, relate to avoiding regions of high risk of thrombus formation, which are characterized by one or more of high residence time, low kinetic energy, high strain time and low pulsatility index.

[0013] The calculations may be performed over two meshes: a solid mechanics mesh determined from the original heart shape, and a fluid mechanics mesh created by extruding inlets and outlets to generate flow from the solid. The calculations may include the use of the Navier-Stokes equations.

[0014] Thus, a framework of tools is provided that aims to enable accurate computational modelling of VAD treatment procedures, which framework enables, inter alia, the following: - Creating patient populations with different stages of heart failure - Apply user-defined pump conditioning protocols Measurement of Quantities of Interest (QoI) related to fluid stagnation Measurement of QoI related to pressure and stress on valves and ventricular walls Iterating the model to achieve one or more of the following: - Optimize rate modulation protocols for uncertain populations -Find the best speed modulation for your patient -Find the best cannula shape, angle and position for your patient These tools can be used to enable the VAD to be most effectively implanted and configured, for example in the case of an LVAD, by proper cannulation of the left ventricle to provide a fluid channel to the LVAD pump, and by pre-setting the VAD operation (especially velocity modulation) appropriate for the patient.

[0015] Disclosure It should be noted that although LVADs are considered in detail herein, the techniques described above may be applied to VADs more generally (e.g., to RVADs and BiVADs as well), and thus, although the term VAD is used generally below, it is actually an LVAD that any issues specifically related to LVADs are discussed. In context, it is useful to consider the standard process for configuration of a VAD. After VAD placement, optimization of VAD speed is routinely performed in post-implant patients using ramp studies. Transthoracic echocardiographic measurements of cardiac geometry and function are performed while slowly increasing the VAD speed over a wide range. For LVADs, the final pump speed is selected by balancing overall cardiac output, efficiency of left ventricular (LV) unloading, and maintenance of flow pulsatility. Several variables are assessed from standard echo views, including LV end-diastolic dimension, LV end-systolic diameter, frequency of aortic valve (AoV) opening, degree of valvular regurgitation, right ventricular (RV) systolic pressure, blood pressure, and heart rate (HR) at each speed setting. Additionally, VAD pump output, pulsatility index, and flow rate are recorded. For example, for the Thoratec HeartMate II, the ramp speed protocol starts at a speed of 8k [rpm] and increases by 400 [rpm] every 2 minutes until a speed of 12k [rpm] is reached. As LVAD speed increases, LV volume decreases, as does the frequency of AoV opening and flow pulsatility. Excessive LV unloading at higher LVAD speeds increases demand on the right heart, causing tricuspid regurgitation and may also result in aspiration events that disrupt flow to the LVAD inflow cannula.

[0016] Clinical practice for LVAD speed selection first ensures that hemodynamics are compatible with longevity, e.g., mean arterial pressure above 65 mmHg and a minimum cardiac index of 2.2 [L / min / m2] of body surface area (BSA). To optimize LV unloading, the ventricular septum position should not bend toward either the left or the right. If these conditions are met, the LVAD speed is selected to achieve intermittent AoV opening while maintaining mild mitral regurgitation or aortic regurgitation. The development of de novo AoV dysfunction in LVAD patients is associated with a lack of AoV opening. Interestingly, AoV dysfunction occurred in the majority (66%) of LVAD patients whose AoV remained closed during support, but rarely in LVAD patients whose AoV was regularly open (8%).

[0017] Patient-specific LVAD rate calibration is important to ensure adequate cardiovascular support and minimize the frequency of adverse events associated with long-term support. However, ramp echo studies are not routinely performed after the first month after implantation due to cost and inconvenience. Thus, computational tools that predict cardiac output and aortic valve opening for subject characteristics can reduce the requirement for ramp testing and contribute to rate adjustments required over time while supporting a rate adjustment paradigm that contributes to recovery. The computational tool can be used with the highest risk devices to ensure that LVAD rate selection is within the valid range from placement, minimizing the likelihood that significant corrections will be required as a result of ramp studies, reducing risk to the patient.

[0018] Specific embodiments of the present disclosure will now be described, by way of example, with reference to the accompanying drawings, in which: [Brief description of the drawings]

[0019] [Figure 1] An overview of a model for use in embodiments of the present disclosure is provided showing the role of parameters and quantities of interest. [Diagram 2] 1 illustrates an exemplary computing architecture for implementing embodiments of the present disclosure. [Diagram 3] FIG. 11 is a scatter plot showing a comparison between numerical and experimental results according to an implementation of the present disclosure. [Figure 4A] Box plots and scatter plots for a population of nine patients for one pump speed adjustment protocol are shown, showing the results for aortic flow. [Figure 4B] FIG. 4A shows box plots and scatter plots for a population of nine patients for different pump speed adjustment protocols, and illustrates the results for kinetic energy. [Figure 4C] 4A and 4B show box plots and scatter plots for a population of nine patients for different pump speed adjustment protocols, showing the results for the pulsatility index. [Diagram 5]1 illustrates generally the systems involved in a real-world implementation of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0020] FIG. 5 generally illustrates a system involved in a real-world implementation of the present disclosure. A ventricular assist device (VAD) 1 is shown implanted in a human body 2 to assist the operation of the human heart 3. Here, the VAD 1 is a left ventricular assist device (LVAD) comprising a pump 10 in a pumping chamber 11 and a connecting tube 12 so that blood is pumped from the left ventricle 3a to the aorta 3b. The VAD 1 is controlled by a control unit 4 and powered by a battery 5 (here shown as powering the control unit 4 with a power connection to the VAD 1, although other configurations are of course possible). The control unit 4 is programmed via a computer 6, and in the implementation of the present disclosure, the computer programming is supported by modeling using a high performance computer 7 and a remote computer 8. As explained below, this modeling can determine the operating parameters of the VAD 1, but can also be used to determine the features involved in the implantation of the VAD, such as the cannulation position for implantation.

[0021] Hereinafter, the physical basis and computational implementation of a computational tool for modeling left ventricular flow after LVAD implementation that can be used for effective LVAD placement and initial configuration is described, but more generally related to VADs, with a specific application to LVADs being presented here. We also provide below the outline steps required to demonstrate that a computational model is fit for purpose. The V&V40[1] standard provides a framework for evaluating the relevance and suitability of completed Verification, Validation, and Uncertainty Quantification (VVUQ) activities that establish the reliability of a computational model. As a first step, the standard requires identifying: (1) the question of interest, which is the question for which the tool finds an answer; (2) the context of use (CoU), which is the specific role and scope of the computational model; (3) quantities of interest (QoI), which are the measurements related to the CoU used to predict; (4) model impact, which is the contribution of the model in making a decision; and (5) model risk and decision consequences, which is the likelihood that an erroneous model result will result in patient harm.

[0022] Usage (CoU) The cardiac-LVAD computational model can be used to aid in the preclinical design and development of LVADs by identifying operating conditions that may result in stagnant ventricular flow. It can therefore be used to ensure that baseline operating criteria are selected to be safe for the general population, and can also be used to provide prior probability adjustments based on patient-specific data, if used in the model. This is illustrated here by code and numerical validation by calculating Uncertainty Quantification (UQ) with mixed areal-epistemic inputs, using convergence rates observed in fabricated solution and mesh convergence studies; validation against bench experiments with 12 operating conditions.

[0023] A heart-LVAD computational model is then used to extract quantities of interest (QoIs) that highlight regions associated with flow stagnation and potentially associated with thrombus formation (relevant properties include but are not limited to stagnation time, kinetic energy and strain).Unlike animal studies, the proposed model provides insight into the working conditions in severe heart failure.

[0024] Input Parameters It is contemplated that the techniques described herein may include additional input parameters that reflect a particular subject's individualized and / or proposed treatment technique. Relevant parameters include parameters that characterize the patient's geometry, the implantation of the pump, and the operation of the pump. Such inputs may improve the accuracy of the generated model, and thus the reliability of measurements that may be derived from the model, and any recommended treatment actions.

[0025] Exemplary input parameters that may be used include the following, which may be used independently or in any combination: These parameters may be identified from a particular subject, may be determined by a user, and / or may be generated from population data. For example, known population averages may be used, which may be a general healthy population or populations selected for various criteria such as disease state, age, sex, etc. A. Parameters Related to Subject Anatomy and / or Cardiac Activity A.1. Subject-specific geometry (e.g., cardiac morphology and trabecular meshwork formation) A.2. Ventricular volumes, such as end-diastolic volume (EDV) A.3. Ejection fraction (EF) A.4. Heart rate (HR) B. Subject's General Condition: B.1. Mean arterial pressure B.2. Arterial resistance B.3. Arterial Capacitance C. Subject Cannula Implantation Specifications C.1. Cannula Position Implantation C.2. Cannulation Depth C.3.Cannula angle C.4. Cannula Shape C.5. Cannula Shape (e.g., Cannula Diameter) D. Pump operating conditions D.1. Time-varying pump speed waveforms (including constant speed) D.2. Time-varying pump speed synchronization point with intrinsic heart rate. D.3. Characterization of pump performance function with speed (also called HQ function, where H is the pressure head and Q is the flow rate).

[0026] The use of some of these variables is further explained below in the context of specific implementations. It should be noted that the above list is non-limiting. It should be noted that the parameters in sections A and B are characteristics of the subject and therefore do not themselves form part of the output from the modeling process. In contrast, the parameters in sections C and D are features that can be altered or controlled in practice as part of the infusion process (section C) or by pump selection or control (section D). The relationships between the parameters, models, and quantities of interest are shown in FIG. 1.

[0027] Output: Quantity of Interest (QoI) The models described herein generate QoIs that can be measured directly from the models or provided by further processing of these measurements. The QoIs are:

[0028] A. Raw QoIs: In the model shown below, these are the QoIs extracted immediately after the solution of the Navier-Stokes equations. Modifications of the model to include other structural elements or physical considerations (such as electrophysiology) may involve the solution of other equations and lead to additional or different QoIs. These require additional post-processing steps (items B, C, D...) to extract the relevant information for industrial and clinical use. A.1. Ventricular velocity field A.2. Ventricular pressure area B. Integrated or Differentiated QoI (eg, Volumetric QoI): Operations performed using velocity and pressure fields that are volumetrically calculated and augment the obtained information. B.1. Residence Time: The time spent by a fluid particle within a domain. B.2. Velocity Magnitude: The root mean square of the pointwise flow velocity. B.3. Kinetic Energy: The point-mean-square of the flow velocity multiplied by the fluid density. B.4. Strain Time: A measure of flow strain obtained from the first invariant of the symmetric strain tensor. B.5. Pulsatility Index (maximum flow velocity variation in a local area, where a higher pulsatility index may mean a greater acceleration / deceleration of blood.)

[0029] C. Curvilinear QoI: Volume integral of the integrated QoI or differentiated QoI (B) that allows reduction of the 3D results to time-varying curves. They represent the behavior in the Region of Interest (ROI). For example, the ROIs of these curvilinear QoIs can be the entire LV, the cannula pocket and the aortic root region, but also any other region that may be a potential thrombus formation region, including single localizations. C.1. QoI Averaging: Average of QoI in ROI per time step. C.2. Max / Min QoI: The maximum and minimum in the domain for each time step.

[0030] D. Scalar QoI: A singular representation of the curve QoI. It is obtained by some kind of time integration of the curve QoI during a particular time window. D.1. Time-averaged scalar QoI: Average of the curved QoI between several time boundaries (e.g., 1.23 [s] to 3.24 [s] - time window may be the systole and may be the rise in phase of the LVAD velocity modulation). D.2. Max / Min Scalar QoIs: The minimum and maximum values ​​of the curve.

[0031] Curve QoIs and scalar QoIs can be used as a way to reduce the complexity of the results and increase the ability to detect differences compared to the video output. Each volume QoI has an associated curve QoI. For example, a kinetic energy output has a field, curve and scalar representation.

[0032] Correspondence with physical results and selection of optimal method From the above QoI, it is possible to identify characteristics of the output model that can provide information about the likely outcome of a given VAD procedure, such as areas of stagnation. Areas of stagnation, and therefore areas at high risk of thrombus formation, are areas with one or more of the following: A) High residence time B) Low kinetic energy C) High strain time D) Low pulsatility index

[0033] When comparing two devices or proposed approaches, the "best" device or approach is one that demonstrates the opposite, i.e. A) Decrease in residence time B) higher kinetic energy C) Low distortion time D) High pulsatility It is.

[0034] Reliability of the numerical model VVUQ does not mandate the use of simulation results as design guidance or regulatory evidence, but it does ensure the reliability of the results. For example, ASMEV&V40[1] requires experimental data to compare simulation results and ensure their reliability.

[0035] Easily accessible and reproducible bench experiments are a particularly effective way to assess the reliability of a model. Such bench experiments can include at least an idealized ventricular model made of flexible material, including the connecting tubing for the VAD. Such bench experiments should be able to obtain measurements of flow and pressure at the inlet, outlet, and cavity, as well as particle image velocimetry (PIV) of the cavity. These devices are known as pulse duplicators, and the reader can find commercially available options at the following link: https: / / vivitrolabs.com / product / pulse-duplicator / .

[0036] Model operation principle An exemplary analysis is provided below. A Sequence for one run: 1. Create a baseline patient or population by selecting ventricular end diastolic volume (EDV), ejection fraction (EF), and heart rate (HR). If the prediction is for a single patient, these parameters are selected as single values. If the prediction is for a population, the parameters are characterized as probability distributions, and then the distributions are sampled to obtain a sample of patients from the population to calculate.

[0037] 2. The patient's heartbeat is calculated by a fluid-structure interaction (FSI) simulation that deforms the ventricular mesh to fit the patient-specific parameters. This approach can be accurate yet computationally relatively inexpensive. This technique can also be used to modify other geometric features, such as the insertion or angulation of the cannula, or any other input parameters mentioned above. New geometries can be used to change the position and / or shape of the cannula, or the original geometry can be deformed using displacement boundaries to obtain the desired geometric configuration.

[0038] FSI in this embodiment involves the construction of two meshes. The solid mechanics mesh is created directly from the original geometry. The CFD mesh is created by closing the solid domain and extruding the inlets and outlets to ensure flow development. The mesh is spatially discretized using linear Teslahedra, pyramids, and pentagons. A first order trapezoidal rule is used for time discretization. 3. Describe the processing that will be applied, including: Time-variable pump speed (i.e., speed modulation). This is the magnitude of the baseline speed, and the number, magnitude, direction (up or down), and frequency that the speed changes. b. When used for rate modulation, it provides a synchronization point with the intrinsic heart rate (e.g., the cyclic upstroke occurs during isovolumic contractions). c. HQ curves representing the pump performance at operating conditions, including at least the HQ curves for the extreme cases of operating conditions. Intermediate HQ curves are obtained automatically via a linear interpolator.

[0039] 4. Select the speed modulation and pump characteristics (i.e. the treatment to apply). This is done by the pump's HQ performance curve (HQ is a graphical expression of how much flow (Q) can be produced due to an increase in head / pressure (H) applied to the fluid by the pump). There is a single HQ curve for each pump speed, and the HQ curves are then linearly interpolated to find the unknown speeds.

[0040] 5. Run the simulation. The simulation engine is based on the Finite Element Method (FEM) and includes the incompressible Newton-Navier-Stokes equations using an arbitrary Lagrangian-Euler formulation that allows to deform the mesh in the fluid problem. JPEG2024540389000002.jpg17149 and JPEG2024540389000003.jpg17149Here, μ is the dynamic viscosity of the fluid, ρ is the density, v i is the velocity, p is the mechanical pressure, f i is the body force term, TIFF2024540389000004.tif10149 is the domain velocity. As the fluid domain deforms due to imposed boundary displacements, the deformations of the interior nodes are calculated via a diffusion equation. The resolved domain includes the intraventricular fluid and the inflow / outflow pathways.

[0041] The mitral valve inlet is at a constant pressure P LA The aortic model has an RC parallel impedance R p A resistor R in series with s , and C p The deforming ventricular wall, as well as the rest of the fluid domain, are imposed with a domain deformation rate. This is JPEG2024540389000005.jpg10149, where: JPEG2024540389000006.jpg12149 is the velocity at the deformed boundary.

[0042] A pressure dependent flow boundary condition is provided at the LVAD outflow, as further described below. The remaining boundaries are JPEG2024540389000007.jpg12149, where the initial fluid velocity is The image is JPEG2024540389000008.jpg11144.

[0043] The mitral and aortic valves are modeled via a pressure-driven porous layer in the valve region. This porous medium has the shape Add an isotropic force to the right hand side of the momentum equation with JPEG2024540389000009.jpg11144, in this case: JPEG2024540389000010.jpg11144, where P is the material porosity and I ij is the identity matrix. This strategy provides a robust numerical scheme for potential mal-state stages of the trapped fluid. To ensure a smooth transition with potentially abrupt changes in the transvalve pressure, the porosity is driven through a hyperbolic tangent as follows: JPEG2024540389000011.jpg16144, where P max is the maximum possible porosity, s is the slope of the curve, and Δp V is the transvalve pressure drop, TIFF2024540389000012.tif9144 is the reference pressure gradient. In reality, Δp V >> In the case of TIFF2024540389000013.tif9144, the valve is closed and Δp V << If TIFF2024540389000014.tif9144, the valve is open. To avoid spurious valve openings and / or closures due to transient peaks in the transvalvular pressure gradient, the measurements are filtered using a median filter.

[0044] The LVAD boundary conditions use a pressure-flow transfer function given by the characteristic performance curves of the pump at a determined speed. These pressure-flow curves (also called HQ curves) provide the relationship between the pressure difference between the pump inlet and outlet and the flow rate the pump can deliver at that speed. If the pump speed changes during the simulation as a result of a time-varying speed, the corresponding HQ curve for that speed is used. Δp VAD =P Ao -P LV is the pressure difference between the pump outlet and inlet, and Q VAD If is the flow rate through the pump, the pressure-flow relationship can be approximated as: JPEG2024540389000015.jpg15151 each Δp VAD There is a single Q VAD and vice versa, this relationship can be used as a boundary condition so that the calculated pressure difference Δp VAD and thus the flow rate is constrained to satisfy this equation.

[0045] As mentioned above, the aortic outlet is coupled with a Windkessel (second-order R-RC model) that models the lumped systemic arteries. The backflow at the outlet and inlet is stabilized through a high viscosity layer.

[0046] 6. Extract Quantities of Interest (QoIs): QoIs are calculated on the fly along with volume averages, means, maxima, minima, and standard deviations for multiple Regions of Interest (ROIs) and time windows. These are listed above and can be defined as follows: a. Raw and derived QoI: i. Velocity field - The velocity at each point in the domain is the solution of the Navier-Stokes equations and is the quantity that determines the ventricular fluid motion. It is determined by the kinetic energy (K), residence time T R , strain time T D and is also used to calculate the pulsatility index.

[0047] ii. Pressure field: The pressure obtained after forcing continuity in the Navier-Stokes equations. iii. Residence Time field - This is the time spent in the domain of interest by a fluid particle. Regions with high TR are potentially red clot formation danger areas, but do not necessarily imply a clot risk. For example, a vortex is a region with an equally high residence time, but does not pose a clot formation risk. Thus, T R is the kinetic energy (K) and strain time (T D ) should be considered together with other quantities such as

[0048] iv. Strain Time Field - This is a measure derived from the strain tensor with units of time that represents the time it takes for a fluid to shear. Laminar flow has a high T due to little shear. D However, highly turbulent flows have small T because vortices generate high shear with viscous energy dissipation. D has.

[0049] v. Kinetic Energy Field - This is TIFF2024540389000016.tif15151ρv2 where ρ is the fluid density and v is the magnitude of the velocity. This is a measure of fluid energy. Regions of high kinetic energy may also have high residence times but do not carry with them a thrombus risk. However, since K is not a Galilean invariant, there may be regions of high K with little or no distortion (e.g. a bolus of fluid moving from the atrium to the ventricle) and therefore this may need to be taken into account in the vorticity measure.

[0050] vi. Pulsatility index - this is a beat-by-beat measurement of volumetrics. For each beat, at each point in the domain, the maximum v max , min v min , and the average v avg The velocity is calculated. Then, for each point in the domain, the pulsatility index is calculated as follows: PI=(v max -v min ) / v avg was obtained beat by beat. Since PI is a measure of fluid velocity fluctuations, for steady flows PI=0[-] and for highly transient flows PI>>0[-].

[0051] 7. These raw and processed QoIs are integrated in volume and / or time to obtain corresponding curves and scalar QoIs that predict the stagnation regions induced by a particular patient-device combination.

[0052] It should be noted that the above approach is exemplary and may be modified in several ways. One factor that can be addressed in many ways is the motion of the ventricular (endocardial) wall. To accomplish this, a position vs. time series can be imposed on each node / element of the endocardial VAD problem surface mesh using a suitable interpolation algorithm. This series may be expressed as a function or provided as a table. This can be done by: · As a unidirectional scheme derived from patient images showing wall motion; · As a unidirectional method derived from electromechanical simulation - ventricular mechanical motion is driven by electrophysiology and a coupled electromechanical simulation is used to obtain wall motion; · As a unidirectional scheme derived from solid mechanics simulations – in this case the ventricular mechanical motion is driven by the pericardial pressure; As an interactive method from the electromechanical simulation - a decoupled electromechanical fluid dynamics problem is solved and the resulting endocardial wall motion is used to impose a motion of the VAD problem surface mesh onto the endocardium; and As a bidirectional scheme with a fully coupled electromechanical-hydrodynamic problem solved in the same domain as the VAD problem, There are several ways it can be derived. B) A single simulation defined by the automation tool (A) is executed in the HPC environment as a cluster and treated as a building block of a future larger simulation process, which effectively treats this building block as a black box. In the exemplary approach shown here (see Fig. 2), an external system (DARE, DAkota SeRvEr) is responsible for automating the black box execution from one or more input parameters and returning potential results. More generally, the automation tool automates the job configuration, execution, and result retrieval by reading an input sample of patient / device combinations (even if it is a single patient or a population) and creating a job input file that is later submitted to the HPC environment for execution. The process starts with a template set of files containing flags that are replaced by the automation tool with the final parameters. Once the inputs are created, the files are exchanged with the HPC environment and the job is submitted to a queue. The HPC queue manager is responsible for assigning a priority to each task. After submitting the job, the automation tool waits a variable amount of time (depending on the job size and length) before calling for successful job termination. Once a job is completed, the automation tool executes a user-defined post-processing script (e.g., the script could contain a set of instructions to extract a single column from the output file) to obtain the final scalars. These scalars may be the final desired results or may be part of an optimization loop. Multiple instances of the automation tool can be launched in parallel, allowing for full parallel efficiency. Robustness is provided by the implementation of SSH (Secure Shell) exception handling to avoid termination due to potentially faulty connections.

[0053] The operation of the automation tool is as follows: · Create remote (cluster) and local directory trees. Create a symbolic link to the template file. Modify the necessary template files. Start execution. Wait for completion and / or detect errors · Run post-processing scripts. Get results.

[0054] C) Sampling Tools An exemplary sampling tool is Dakota (Sandia), but any other suitable sampling tool can be used to select new parameters for a run that the automation tool uses to configure and run the simulation. Select parameters from range / distribution, Optimize the inputs, if necessary, to get the desired output, Collect the results, Create a surrogate model of performance using Stochastic Collocation (SC), Polynomial Chaos Expansion (PCE), or any other Reduced Order Model (ROM) technique (case-dependent); It is necessary.

[0055] D) Create scatter plots and box plots showing trends and statistical associations of data retrieval inputs / outputs. In this way, the QoI can be used to answer specific questions of interest. The computational techniques used here are explained in more detail. Numerical code validation is performed according to section 2 of

[12] for the 2D Poiseuille and 3D Wormsley flow problems in cylindrical pipes. These problems have nontrivial analytical solutions that are used as the truth values. In both cases, the discretization error is monitored as the grid is systematically refined by halving as in

[13] . The ratio between the mesh partitions is r i,j =r i / r j Then, r 1,2 =r 2,3 = r = 2.0, which is considerably larger than the recommended minimum value of 1.3

[12] . The velocity field is the quantity of interest (QoI) to be validated, since it is also a raw variable obtained from a numerical model.

[0056] It must be established that the computational mesh converges properly. For example, the root mean square error (RMSE) between solutions i and j is the L2 norm Defined by TIFF2024540389000017.tif15151. JPEG2024540389000018.jpg15151All cases are calculated, and the calculated error ε 1、2 and ε 2、3 Once calculated, the observed convergence order can be calculated as follows

[14] : JPEG2024540389000019.jpg32151In this case, r i,j =r j,k =r=2.0, so q i,k (p i,k )=0:0, so the previous system of equations reduces to: JPEG2024540389000020.jpg19138For the three computed velocity fields u1,u2,u3;u1, where u1 is the coarsest, for the three levels of refinement, the order of convergence of the numerical scheme is given by: Using the observed p-values, the Grid Convergence Index (GCI) can be calculated as

[14] : JPEG2024540389000022.jpg19138This uncertainty estimate is relative to the true mathematical value f T Interval f±U within the range 95% with a 95% probability.

[0057] In implementation, a VVUQ plan can be used to illustrate the application of a particular model (as well as provide a basis for determining whether the model can be used effectively). An example for an LVAD can be:

[0058] Questions of interest: For apically implanted LVADs, which pump speed was selected to achieve the following outcomes: (a) complete aortic valve opening (Q Ao >5[cm 3 / s); and (b) cardiac output compatible with life ( TIFF2024540389000023.tif13138[cm 3 / s]).

[0059] Context of Use (CoU): As shown above, the heart-LVAD computational model can be used to aid in the preclinical design and development of the LVAD by characterizing the flow in the aortic root, the LVAD and the LV for a given pump speed.

[0060] Quantities of Interest (QoI): The QoIs used during validation are the maximum and average flow through the outlet boundary (LVAD flow Q VAD and aortic root flow Q Ao ).

[0061] The model provides information about stagnation areas within the domain. Here, a population of nine subjects with different EDV and EF was treated with three velocity-modulated protocols (fixed, L1, L2). The model does not show statistical differences in aortic flow and kinetic energy, but there is a clear increase in the pulsatility index (PI). This indicates that L1 and L2 may be less susceptible to thromboembolism than the fixed treatment. This is a single example of the output of the tool. This is shown in Figure 3 and Figure 4A-C.

[0062] Figure 3 shows a scatter plot comparing the numerical results (blue) with the experimental results (orange). It shows a good correlation for most of the QoIs. Figures 4A-C show boxplots and scatter plots of a population of nine patients divided into three categories (fixed, L1, L2). There is no statistical difference between them for aortic flow (Figure 4A) or kinetic energy between the three categories (Figure 4B), but there is a statistically significant increase in the pulsatility index (Figure 4C).

[0063] Exemplary Applications of the Embodiments of the Present Disclosure The present disclosure, in its embodiments, describes a tool that can predict important operational metrics (such as performance, safety and efficacy metrics) for VAD therapy (LVAD therapy in the illustrated model) from a given set of inputs describing either a single patient or a patient population. A set of exemplary, non-limiting applications are described below.

[0064] 1. Optimizing LVAD Therapy for Uncertain Populations The parameterized uncertain population can be described by a probability distribution. Modeling can be performed over the space defined by this parameterized population and a set of treatment parameters. This allows treatment parameters (e.g., safety, efficacy and performance measures) to be optimized across the population. This can be used, for example, to create a preferred baseline for the application of a treatment.

[0065] 2. Surgical guidance for LVAD implantation location and configuration It is possible to model optimized parameters of the LVAD implantation process for predefined patient parameters. These may include injection locations and motion configurations. This process can be performed during the motion planning stage to help establish a motion plan. In embodiments, such guidance can also be performed intraoperatively, which may require different computational strategies (e.g., using reduced sequence models to accelerate computation time).

[0066] 3. Virtual lamp survey (as in-box information) The tool can be used to create tabulated data provided "in the box" with the VAD device, which can be used to determine the configuration of the device in light of the patient's condition. For example, one important parameter after LVAD implantation is the maximum pump speed that can be set while still allowing aortic valve opening (required to avoid long-term regurgitation) (thus maximizing cardiac output). Such tabulated information can provide initial values ​​for pump speed depending on the patient's condition, which can subsequently reduce the amount of testing and optimization time required.

[0067] 4. Computational biomarker discovery and development The tool can be used to discover and subsequently develop computational biomarkers and indices for developing parameterization of VADs and LVADs. The goal of this development is typically to develop and optimize the operation of the device with the aim of maximizing the likelihood of a positive outcome. Differential markers can include statistical and risk indices based on any combination of quantities, functions of quantities, and results (both directly and indirectly calculated) from the tool, with the practical utility provided by such biomarkers (in providing the most practically effective description of the system) being particularly considered in their development. Here, time-based and transient factors may prove to be important.

[0068] 5. Accurate predictive patient-specific LVAD performance assessment to assist clinicians This may be time-based or lifetime, and different approaches may be used in each case. a. Constant-time Sensitivity analysis can be performed (often using) for patient-specific physiological information input. VAD parameters can then be optimized to provide the best scenario for a given patient at a precise time point, for example, to minimize the risk of thrombus formation (such as dwell time) and minimize the structural impact of the VAD on the patient (e.g., avoiding suction of the heart wall or forced valve opening, limiting stress and strain imposed on the ventricle or elsewhere). For the treatment to be performed, this can include a scenario analysis considering plausible placement, appropriate cannula shape, implant depth, pump speed and pressure-related profile. Physiological sensitivity can also be considered here. b. Longitudinal (lifetime) prediction The above "constant-time" approach can be followed while considering the HF deterioration pathway. Currently, this is an empirical process; it is known that HF ​​becomes more likely over time and VAD operating conditions need to be updated, but there is no systematic process to achieve this. This tool can be used to establish a patient alert table that links potential future patient conditions to associated optimized operating parameters, with deterioration pathways based on observed medical events. Associated biomedical markers can be used to create effective patient / device management guidelines that are tailored to the patient and adapted to support clinical decision making.

[0069] References [1] American Society of Mechanical Engineers. Assessing the Reliability of Computational Modeling Through Verification and Validation: Applications to Medical Devices-V 40-2018. Asme V&V 40-2018, page 60, 2018. [2] Carlo R Bartoli, David Zhang, Jooeun Kang, Samson Hennessy-Strahs, David Restle, Jessica Howard, Gretchen Redline, Christian Bermudez, Pavan Atluri, and Michael A Acker. Clinical and in vitro evidence that subclinical hemolysis contributes to thrombosis. Annals of Thoracic Surgery, 105(3):807{814,2018. [3] Gordon DO Lowe. Revisiting Virchow's triad: abnormal flow. Pathophysiology of Hemostasis and Thrombosis, 33(5-6):455-457, 2003. [4] Joseph G Rogers, Francis D Pagani, Antone J Tatooles, Geetha Bhat, Mark S Slaughter, Emma J Birks, Steven W Boyce, Samer S Najjar, Valluvan Jeevanandam, Allen S Anderson et al. Intrapericardial left ventricular assist devices for severe heart failure. New England Journal of Medicine,376(5):451{460,2017. [5] Lorenzo Rossini, Oscar O Braun, Michela Brambatti, Yolanda Benito, Adam Mizeracki, Marissa Miramontes, Cathleen Nguyen, Pablo Martinez-Legazpi, Shone Almeida, Megan Kraushaar et al. Intraventricular flow patterns in patients treated with left ventricular assist devices. ASAIO Journal, 2020. [6] MS Slaughter. Heartmate II Investigators: Advanced heart failure treated with continuous left ventricular assist devices. N Engl J Med, 361:2241{2251, 2009. [7] Kiat T Tan and Gregory YH Lip. White blood cells: the importance of proper thrombus treatment. Archives of Internal Medicine, 163(20):2534-2535, 2003. [8] David Varga-Szabo, Irina Pleines, and Bernhard Nieswandt. Mechanisms of cell adhesion in platelets. Atherosclerosis, Thrombosis, and Vascular Biology, 28(3):403-412, 2008. [9] Rui Zhao, Joie N Marhefka, Fangjun Shu, Samuel J Hund, Marina V Kameneva, and James F Antaki. Microflow visualization of red blood cell-enhanced platelet concentration during rapid expansion. Annals of Biomedical Engineering, 36(7):1130, 2008.

[10] Neidlin et al., Understanding Left Ventricular Assist Device Inflow Cannula Placement and the Risk of Ventricular Thrombosis. BioMed Eng OnLine (2021) 20:47 https: / / doi.org / 10.1186 / s12938-021-00884-6

[11] Ong et al., Numerical Study of the Effect of Cannula Placement on Thrombosis, Theoretical Biology and Medical Modeling 2013,10:35

[12] American Society of Mechanical Engineers, 2009. “Standard for Verification and Validation in Computational Fluid Dynamics and Heat Transfer: ASME V&V 20.” American Society of Mechanical Engineers (ASME).

[13] Houzeaux, G., de la Cruz, R., Owen, H., and Vazquez, M., 2013. “Parallel Uniform Mesh Multiplication Applied to Navier-Stokes Solvers.” Computers and Fluids, 80(1), pp. 142-151.

[14] Roache, P.J., 1998. "Verification and Validation in Computational Science and Engineering." Vol. 895. Hermosa Albuquerque, NM.

[15] Roache, P.J., 2002. "Code Verification with Manufacturing Solutions". J. Fluids Engineering, 124(1), pp.4-10.

[16] Pedley, TJ, and Luo, X., 1995. Fluid mechanics of large vessels. Shaanxi People´s Press. Effect of rabbit cycle timing on ventricular washout in an in vitro flow visualization setting. THANANYA KHIENWAD, ASAIO 2021 Effect of LVAD implantation site on ventricular blood stagnation, Prisco. ASAIO 2017 Left ventricular assist device cannula inflow angle and risk of thrombosis. Chivukula. Circulatory heart failure. A clinical method for mapping and quantifying blood congestion in the left ventricle. Rossini. Journal of Biomechanics. 2015.

Claims

1. 1. A method of configuring a ventricular assist device, comprising: determining one or more parameters of the patient's heart; determining one or more parameters of the ventricular assist device; using a mechanical and fluid dynamic model of at least one ventricle in which the ventricular assist device is or will be used, to calculate a predicted quantity of interest for the patient's heart by operating the ventricular assist device using the determined parameters; Varying the one or more parameters of the patient's heart and / or the one or more parameters of the ventricular assist device and re-calculating the predicted quantities of interest, repeating this step until a set of parameters and associated predicted quantities of interest determined to meet predetermined configuration objectives is developed; configuring the ventricular assist device in accordance with the predetermined configuration objectives; A method comprising:

2. The method of claim 1 , wherein the configuration objectives include one or more of the predicted quantities of interest that meet predetermined criteria.

3. The method of claim 2 , wherein the predetermined criteria is one or more of the following indicators: performance, safety, and efficacy.

4. The method of claim 1 , wherein the parameter of a patient's heart relates to the heart of any patient in an uncertain population of patients.

5. The method of claim 1 , wherein the parameters of the patient's heart include one or more anatomical parameters related to the patient's anatomy or cardiac activity.

6. The method of claim 5 , wherein the one or more anatomical parameters include one or more of cardiac morphology, ventricular volume, ejection fraction, and heart rate.

7. The method of claim 1 , wherein the parameters of the patient's heart include one or more systemic parameters related to the patient's system state.

8. The method of claim 7 , wherein the one or more systemic parameters include one or more of arterial main pressure, arterial resistance, and arterial capacitance.

9. The method of claim 1 , wherein the one or more parameters of the ventricular assist device include a pump operating state parameter.

10. 10. The method of claim 9, wherein the one or more pump operating state parameters include one or more of a pump speed waveform, a pump synchronization with an intrinsic heart rate, and a pump performance function.

11. The method of claim 1 , wherein the quantities of interest include one or more quantities derived from the calculation.

12. The method of claim 11 , wherein the quantities of interest include one or more of dwell time, velocity magnitude, kinetic energy, strain time, and pulsatility index.

13. The method of claim 12 , wherein the predetermined condition relates to the likelihood of thrombus formation.

14. 2. The method of claim 1, wherein the step of calculating the predicted quantity of interest includes modeling over a first solid mechanics mesh determined from the original cardiac geometry and a second fluid mechanics mesh created by extruding inlets and outlets to generate flow.

15. The method of claim 14 , wherein the step of calculating a predictor comprises solving the Navier-Stokes equations.

16. The method of claim 1, wherein configuring the ventricular assist device includes determining a range of velocity modulation for the ventricular assist device.

Images