A system for hemodynamic simulation and efficacy evaluation after heart valve replacement surgery

Abstract

This invention relates to the field of medical data processing technology and discloses a system for simulating hemodynamics and evaluating the efficacy of heart valve replacement surgery. The system includes: using a pre-defined standard three-dimensional geometric model of the artificial valve, replacing the valve region distorted by metal artifacts in computed tomography data to generate a static computational domain mesh; adding a time-varying momentum source term drag coefficient to the Navier-Stokes equations based on cardiac cycle temporal logic; and simulating the physical effects of valve leaflet opening and closing and solving the flow field under the static mesh by adjusting the fluid impedance properties of the spatial units. This invention establishes a virtual motion mechanism based on momentum source terms to replace complex dynamic mesh reconstruction, improving the numerical stability of simulation calculations and the timeliness of clinical processing while ensuring the fidelity of geometric boundaries.

Description

Technical Field

[0001] This invention relates to a system for simulating hemodynamics and evaluating the efficacy of heart valve replacement surgery, belonging to the field of medical data processing technology. Background Technology

[0002] Currently, after transcatheter aortic valve replacement (TAVR) for aortic stenosis, hemodynamic status is associated with patient prognosis. Assessing the flow field characteristics after artificial valve implantation is crucial for predicting leaflet thrombosis, paravalvular leakage, and valve durability. A three-dimensional computational model is constructed based on postoperative computed tomography (CT) data, and computational fluid dynamics (CFD) methods are used to solve the flow field, providing a general technical means to obtain key hemodynamic indicators such as wall shear force and transvalvular pressure gradient. However, simulation systems face a contradiction between the physical properties of the data source and the complexity of the computational model in clinical data processing. Artificial valve stents are mostly made of high-density nickel-titanium alloy, which generates beamhardening artifacts during CT imaging. These artifacts manifest as radial bright signal overflows, obscuring the true physical boundaries of the valve inner wall and leaflets. Conventional grayscale thresholding or gradient image segmentation algorithms cannot distinguish between artifact signals and real tissue, resulting in irregular sawtooth-shaped boundaries or false stenosis in the generated fluid computational domain. Geometric topological distortion alters the boundary layer flow field structure, leading to orders-of-magnitude deviations in the calculated wall shear force, failing to reflect the true blood flow state.

[0003] Existing simulation mechanisms suffer from path-dependent bias, emphasizing morphological teaching over pathological inversion. For example, Chinese invention patent CN115841776B discloses a VR and AR-based method and system for teaching hemodynamic simulation of congenital heart disease. While the solution utilizes the Unity3D engine to visualize and interact with the heart structure and blood flow direction, its core logic relies on 3DS Max to construct preset animation scripts. This falls under the category of phenomenological simulation within visual graphics, rather than physical solutions based on fluid dynamics control equations. Consequently, the processing method cannot respond to patient-specific boundary condition changes, cannot quantify the pressure gradient changes in complex flow fields caused by valve replacement, and struggles to support the quantitative assessment needs of postoperative complications in clinical practice. The dynamic control effect of valve opening and closing on the flow field is typically addressed using fluid-structure interaction methods in current technologies. This requires solving the fluid equations and the solid structure equations simultaneously, and performing dynamic mesh reconstruction or overlapping mesh interpolation at each time step to adapt to the large deformation motion of the valve leaflets. The computational load is enormous, with a single simulation taking tens of hours. When dealing with complex contact or large gradient flow fields, mesh distortion leads to computational divergence. Simplifying the valve leaflets to rigid bodies or using the immersion boundary method to reduce computational requirements still makes it difficult to complete high-precision simulations of the entire cardiac cycle within the clinical diagnosis and treatment time window. Existing simulation strategies often use steady velocity or general pressure waveforms as boundary conditions, without incorporating the patient's specific peripheral vascular impedance characteristics into the calculation system. As a result, the simulated pressure field cannot be verified and closed-loop by clinically measured blood pressure data.

[0004] Therefore, the technical problem to be solved by this invention is how to overcome the interference of metal artifacts, avoid the computational overhead of complex moving meshes, and rapidly reconstruct a high-fidelity personalized flow field hemodynamic simulation and efficacy evaluation system after heart valve replacement using clinical images and physiological measurement data. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: A system for hemodynamic simulation and efficacy evaluation after heart valve replacement surgery, comprising:

[0006] The geometric topology reconstruction module is used to acquire postoperative computed tomography (CT) scan data of the aortic root and the implanted prosthetic valve model code; extract the outer contour of the aortic root without metal artifacts from the CT scan data, and call the preset standard 3D geometric model of the prosthetic valve according to the model code; perform spatial rigid registration to map the standard 3D geometric model of the prosthetic valve to the interior of the outer contour of the aortic root; replace the original valve region boundary in the CT scan data that is distorted in grayscale due to metal artifacts with the surface boundary of the standard 3D geometric model of the prosthetic valve, and generate a static fluid computational domain mesh containing the standard geometry of the prosthetic valve;

[0007] The time-varying flow resistance field mapping module is used to mark the set of spatial units corresponding to the leaflet geometry in the standard three-dimensional geometric model of artificial valves in the static fluid computational domain grid; according to the preset cardiac cycle time sequence logic, a source term resistance coefficient that dynamically changes with time is added to the momentum conservation term of the Navier-Stokes equations; in the systolic phase, the source term resistance coefficient is assigned a low resistance threshold that allows fluid to pass through, and in the diastolic phase, the source term resistance coefficient is assigned a high resistance threshold that blocks fluid to pass through. By adjusting the fluid impedance properties of the spatial unit set, the physical effects of leaflet opening and closing are simulated under the topological constraints of the static grid.

[0008] The flow field solution module is used to solve the fluid control equations based on the static fluid computational domain grid and the dynamically changing source term drag coefficients, and output three-dimensional hemodynamic data.

[0009] Preferably, the system also includes a resistance parameter self-calibration module; the resistance parameter self-calibration module is used to acquire the measured transvalvular pressure difference data obtained by ultrasound Doppler detection of the patient; after the flow field calculation module outputs the preliminary three-dimensional hemodynamic data, it calculates the residual value between the simulated transvalvular pressure difference and the measured transvalvular pressure difference data; if the residual value exceeds the preset convergence range, a correction factor is generated based on the residual value, and the source term resistance coefficient of the systolic phase is iteratively updated using the correction factor and the flow field calculation module is triggered to recalculate until the residual value converges; based on the ratio of the finally converged source term resistance coefficient to the low resistance threshold, a quantitative index characterizing the degree of limitation of valve motion function is generated.

[0010] Preferably, the system further includes a perivalvular microleakage assessment module; the perivalvular microleakage assessment module is used to define a perivalvular annular voxel region between the outer surface of the standard three-dimensional geometric model of the artificial valve and the inner surface of the outer contour of the aortic root; extract the gray values ​​of each voxel unit in the perivalvular annular voxel region from the computed tomography data, and convert the gray values ​​into porous media permeability parameters based on a preset gray-scale permeability mapping function; wherein, gray values ​​higher than a preset calcification threshold are mapped to zero permeability, and gray values ​​lower than a preset tissue threshold are mapped to effective permeability, and the flow field calculation module is used to calculate the leakage flow rate in the perivalvular annular voxel region based on the porous media permeability parameters.

[0011] Preferably, the system further includes a physiological boundary reverse coupling module; the physiological boundary reverse coupling module is used to construct a lumped parameter model containing peripheral vascular impedance characteristics as the outlet boundary condition of the static fluid computational domain grid; it receives the patient's measured blood pressure data and cardiac output data, and uses the Levenberg-Marquardt algorithm to iteratively adjust the compliance parameter and peripheral resistance parameter of the lumped parameter model until the deviation between the pressure waveform output by the lumped parameter model and the measured blood pressure data is less than a preset threshold, and loads the time-varying pressure value output by the lumped parameter model after parameter convergence into the boundary equation of the flow field solution module in real time.

[0012] Preferably, when performing spatial rigid registration, the geometric topology reconstruction module identifies the high-density feature point cloud generated by the artificial valve metal stent in the computed tomography scan data; by minimizing the spatial Euclidean distance between the high-density feature point cloud and the stent node in the standard three-dimensional geometric model of the artificial valve, the pose coordinates of the standard three-dimensional geometric model of the artificial valve within the outer contour of the aortic root are determined.

[0013] Preferably, the source term drag coefficients added to the Navier-Stokes equations by the time-varying flow resistance field mapping module follow Darcy-Fochheimer's law, and the source term is defined as follows: The following relationship must be satisfied: ,in, The coefficient of viscosity resistance. The inertial drag coefficient, For fluid dynamic viscosity, For fluid density, For the fluid velocity vector, the low resistance threshold corresponds to and Values ​​approaching zero correspond to high resistance thresholds. and Values ​​greater than the preset cutoff value.

[0014] Preferably, after outputting three-dimensional hemodynamic data, the flow field calculation module extracts the wall shear force distribution data of the surface of the standard three-dimensional geometric model of the artificial valve; calculates the area of ​​the low shear force region where the wall shear force is lower than the preset thrombosis threshold, and calculates the percentage of the low shear force region area to the total surface area of ​​the valve leaflet, generating a thrombosis risk assessment report.

[0015] Preferably, the cardiac cycle timing logic based on the time-varying flow resistance field mapping module includes the start time of systole, the end time of systole, and the duration of diastole; the time-varying flow resistance field mapping module extracts R-wave feature points by analyzing the patient's electrocardiogram signal data to synchronously calibrate the time axis of the cardiac cycle timing logic.

[0016] Preferably, the system also includes a data visualization interaction module; the data visualization interaction module is used to map the three-dimensional hemodynamic data to the static fluid computational domain grid to generate a three-dimensional rendering scene containing streamline diagrams, pressure cloud diagrams and wall shear force distribution diagrams; in the three-dimensional rendering scene, the standard three-dimensional geometric model of the artificial valve is superimposed in a semi-transparent manner, and the spatial areas where the flow velocity exceeds the preset turbulence threshold are highlighted.

[0017] Preferably, the data visualization interaction module is also used to generate a virtual stent fit heat map; calculate the radial distance field between the outer surface of the standard three-dimensional geometric model of the artificial valve and the inner surface of the outer contour of the aortic root; map the radial distance field into a color gradient, and project the color gradient onto the stent surface of the standard three-dimensional geometric model of the artificial valve to intuitively display the geometric fit state between the stent and the blood vessel wall.

[0018] Compared with the prior art, the beneficial effects of the present invention are:

[0019] 1. In heart valve replacement surgery, the standard modal embedding and geometry construction module is used to solve the problem that metal artifacts prevent the reconstruction of the fluid computation domain in the heart valve implantation area. The system calls the pre-stored standard three-dimensional geometric model of the artificial valve and uses Boolean operations to replace the voxel regions in the computed tomography data that are distorted by metal hardening artifacts based on the spatial registration algorithm. The prior determinism of industrial standard data avoids the risk of jagged or broken boundaries produced by conventional image thresholding algorithms under strong artifact interference, ensuring the physical continuity and geometric fidelity of the fluid computation domain boundary. Under the condition of low-quality image data, a grid that meets the requirements of high Reynolds number flow field calculation is constructed to ensure the calculation benchmark of hemodynamic parameters sensitive to wall shear force.

[0020] 2. By utilizing the time-varying flow resistance field mapping module, a simulation mechanism for virtual leaflet motion based on momentum source term regulation is established. This overcomes the computational divergence risk and computational bottleneck of traditional fluid-structure interaction methods in dynamic mesh reconstruction. The computational domain mesh topology remains static throughout the process. Based on the temporal logic of the cardiac cycle, the viscous drag coefficient and inertial drag coefficient within the corresponding spatial unit of the leaflet are dynamically adjusted. The step change of scalar field parameters in the time dimension is used to equivalently reproduce the physical blocking and conduction effects of the physical leaflet on the flow field in the fluid control equation. The solution process of complex nonlinear structural motion is reduced to the process of assigning source terms in the fluid equation, eliminating negative volume mesh errors, reducing the computational cost per cardiac cycle, and meeting the timeliness requirements of clinical data processing.

[0021] 3. Utilizing the physiological boundary reverse coupling module, the simulation is reversed from general fluid dynamics simulation to patient-specific pathological state. A lumped parameter model is constructed as the downstream boundary of the three-dimensional flow field. An iterative optimization algorithm with the patient's measured blood pressure and transvalvular velocity data as the convergence target is introduced. The compliance and peripheral resistance parameters of the lumped parameter model are automatically calibrated. The output results of the closed-loop data processing logic simulation system are consistent with the clinically measured physiological indicators, eliminating boundary condition assumption bias and flow field distortion. This ensures that the calculated three-dimensional flow field data truly reflects the hemodynamic characteristics of the current cardiovascular load state of a specific patient. Attached Figure Description

[0022] Figure 1 This is a flowchart of the hemodynamic simulation data processing logic and algorithm of the present invention;

[0023] Figure 2 This is a comparison and verification diagram of the model output pressure waveform and measured data of the resistance self-calibration mechanism of the present invention.

[0024] Figure 3 This is a schematic diagram of the multi-terminal hardware deployment architecture and the functional distribution of the core computing engine of the present invention. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present invention, the technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0026] This invention provides a system for hemodynamic simulation and efficacy evaluation after heart valve replacement surgery. It comprises a geometric topology reconstruction module, a time-varying flow resistance field mapping module, a physiological boundary inverse coupling module, and a flow field solution module. Data flow begins with the input patient's computed tomography (CT) scan data and Doppler ultrasound data. Through three-dimensional matrix operations, parametric field mapping, and numerical solution of differential equations, it ultimately outputs three-dimensional hemodynamic quantitative indicators including pressure field, velocity field, and wall shear force distribution. Addressing the problem in image evaluation after transcatheter aortic valve replacement surgery where beam hardening artifacts from the metal stent cause the valve region's anatomical structure to appear as bright radial noise in CT images, making conventional gray-level gradient-based segmentation algorithms unable to capture the closed fluid domain boundary, the system employs a geometric topology reconstruction procedure based on standard modal embedding to read the patient's postoperative aortic root CT data sequence. The system extracts the contours of the aortic root and ascending aorta wall without artifact interference to construct a basic vascular fluid domain. Based on the implanted artificial valve model code, the system retrieves the corresponding standard 3D geometric model of the artificial valve from a preset database, which includes precise stent geometry and leaflet geometry. In the spatial registration stage, the system executes a rigid registration algorithm to identify the high-density feature point cloud corresponding to the artificial valve metal stent in the CT data, i.e., the set of voxels with a CT value greater than a preset threshold such as 2000 HU. Through iterative rotation and translation operations, the system minimizes the spatial Euclidean distance between the feature point cloud and the stent nodes in the standard 3D geometric model, thereby determining the precise pose of the standard model in the vascular fluid domain. Finally, the system performs 3D Boolean operations to replace the original valve region data in the CT data that is distorted due to artifacts with the surface boundary of the standard 3D geometric model, generating a static fluid computation domain mesh with physical realism.

[0027] To address the issues of long computation time and potential divergence caused by the need for dynamic mesh reconstruction in traditional fluid-structure interaction methods when simulating leaflet opening and closing, the system employs a time-varying flow resistance field mapping procedure. This maintains the computational domain mesh topology statically and simulates leaflet motion effects by dynamically adjusting the physical properties of spatial elements. Within the static fluid computational domain mesh, the system marks the set of spatial elements geometrically overlapping with the leaflets in the standard three-dimensional geometric model of the artificial valve and defines it as the control domain. The system constructs a fluid control model based on the Navier-Stokes equations and adds a source term to the momentum conservation equation. And it follows the Darcy-Forchheimer law, which is expressed mathematically as follows: ;in The coefficient of viscosity resistance. The inertial drag coefficient, For fluid dynamic viscosity, For fluid density, As a fluid velocity vector, the system incorporates a cardiac cycle timing controller, which receives the patient's electrocardiogram signal and extracts R-wave feature points to synchronize the time axis. During the simulation's time step progression, the controller updates the values ​​of each grid cell within the control domain in real time based on the current cardiac phase (systole or diastole). and During the contraction period, the system will... and A low-resistance threshold close to zero is assigned, allowing fluid to pass through the control domain without loss; during relaxation, the system will... and A high-resistance threshold greater than a preset cutoff value is assigned, and the fluid velocity is attenuated to zero using the momentum sink effect. This mechanism reduces the complex large structural deformation problem to a time-varying assignment problem of scalar field parameters. While ensuring the capture of dynamic characteristics of the flow field, it improves the stability and efficiency of numerical calculation. When the flow field solution module performs discretization, to avoid divergence in the solution of the Navier-Stokes equations due to a step change in the drag coefficient, a rectangular wave function is not directly used to switch between contraction and relaxation states. Instead, a hyperbolic tangent transition function is loaded through a time smoothing window with a time span of 2% to 5% of the cardiac cycle to control the viscous drag coefficient. With inertial drag coefficient A continuous monotonically nonlinear transformation is performed between low and high resistance thresholds to keep the time step condition number of the Jacobian matrix within the solution range. The solver continuously executes transient calculations for no less than four complete cardiac cycles. The first three cycles are numerical cleaning stages to eliminate errors in the initial flow field static assumption. Data from the fourth cycle is extracted as the output. The absolute value of the relative deviation of the intervalvular pressure difference at the same phase point in adjacent cycles is less than 1% as the periodic convergence termination criterion.

[0028] To address the issue that general boundary conditions cannot reflect the patient's specific cardiovascular load, the system executes a physiological boundary inverse coupling procedure. The system constructs a lumped parameter model (OD model) at the 3D flow field outlet, consisting of capacitance parameters representing vascular compliance and resistance parameters representing peripheral resistance. Before calculation, the system receives measured cuff blood pressure and cardiac output data from the patient as convergence targets. Using the Levenberg-Marquardt optimization algorithm, the system iteratively adjusts the capacitance and resistance parameters of the lumped parameter model in the frequency domain until the root mean square error between the pressure waveform output by the model and the measured blood pressure data is less than a preset convergence threshold, such as 5%. The converged lumped parameter model is then coupled to the 3D flow field calculation module. Its real-time output time-varying pressure value is loaded as a Neumann boundary condition at the 3D computational domain outlet. This step ensures that the simulated pressure field is numerically consistent with the patient's actual physiological measurements. To ensure the stability of the data exchange between the lumped parameter model and the 3D flow field, a relaxation factor update is performed at each time step of the data transfer interface. The flow field calculation module reads the pressure estimate for the next time step output by the lumped parameter model and calculates it according to the formula... Calculate the actual loaded mesh boundary pressure value and relaxation factor. The initial value is set to 0.2 to 0.5. If the rate of decrease of the nonlinear residual within the current time step is found to be lower than the preset linear convergence rate threshold, the automatic damping mechanism is triggered, and the relaxation factor is decreased in increments of 0.05 until the residual converges.

[0029] To address the challenge of capturing postoperative paravalvular leaks and tiny gaps through mesh generation, the system integrates a gray-gradient-based paravalvular microleakage assessment module. This module defines a fixed-thickness annular voxel region, or paravalvular assessment domain, between the outer surface of the standard 3D geometric model of the prosthetic valve and the inner surface of the aortic root. Instead of explicitly constructing the gap geometry, the system directly extracts the gray value (HU value) of each voxel unit in the original CT data within this region. The system then applies a preset nonlinear mapping function to convert the gray value into a porous media permeability parameter. For voxels with gray values ​​higher than a preset calcification threshold, such as 800 HU, and identified as calcification pressure points, their permeability is... The grayscale value is set to zero; for voxels with a grayscale value lower than a preset tissue threshold, such as 100 HU, the system determines them as potential leakage channels; in addition, to identify valvular dysfunction caused by subclinical leaflet thrombosis, the system includes a resistance parameter self-calibration module. After the initial simulation is completed, the system extracts the transvalvular pressure difference data calculated by the simulation and compares it with the transvalvular pressure difference data measured by the patient's Doppler ultrasound, calculates the pressure residual. If the measured pressure difference is higher than the simulated pressure difference and the residual exceeds the allowable range, the system determines that the leaflet has limited motion. At this time, the system activates the feedback loop, keeps the mesh geometry unchanged, and uses proportional-integral (PID) control logic to iteratively increase the inertial drag coefficient in the systolic control domain. The process continues until the simulated pressure difference approximates the measured pressure difference. The ratio of the final converged resistance coefficient to the standard low resistance threshold is output as the valve hardening index, serving as a technical indicator for quantitatively assessing the degree of damage to valve opening and closing function.

[0030] Example 1: In the follow-up assessment of patients after transcatheter aortic valve replacement, the patient had a 26mm self-expanding artificial valve implanted and severe irregular calcification at the aortic root. Postoperative computed tomography (CT) data produced strong radial beam hardening artifacts in the valve area due to the high-density attenuation of the metal stent, resulting in the complete loss of grayscale information of the leaflet and flow channel boundaries. Furthermore, clinical diagnosis and treatment require obtaining hemodynamic quantitative indicators, including transvalvular pressure gradient and leaflet surface shear force distribution, within 60 minutes. Faced with this engineering contradiction between data distortion and time constraints, the system performed geometric topology reconstruction. The system extracts the artifact-free outer contour of the aortic root, calls the pre-stored standard 3D geometric model of the artificial valve according to the artificial valve model code, and uses the high-density feature point cloud of the metal stent with the highest signal-to-noise ratio in the computed tomography data as a spatial anchor point. The system executes a rigid registration algorithm to minimize the spatial Euclidean distance between the point cloud and the stent node of the standard model, and embeds the standard model containing the leaflet geometry into the patient's aortic anatomical coordinate system. This step directly replaces the original image data distorted by artifacts with the deterministic topological structure of industrial standard data, and constructs a static fluid computational domain mesh with physical closure.

[0031] Based on this static grid, the system executes a time-varying flow resistance field mapping procedure. The system identifies grid spatial units that geometrically coincide with the valve leaflets of the standard model and defines them as control domains. According to the time axis synchronized with the patient's ECG R-wave signal, the system modulates the momentum source term of the Navier-Stokes equations. Viscous drag coefficient in and inertial drag coefficient Perform dynamic assignment, where For fluid dynamic viscosity, For fluid density, As the fluid velocity vector, during the contraction phase, the system will regulate the velocity vector within the control domain. and The value is set close to zero to simulate the conduction effect of leaflet opening; during the diastolic phase, the system will... and The system steps up to a high-resistance state greater than the preset cutoff value, using the momentum sink effect to simulate the blocking effect of valve closure. This mechanism replaces the high-dimensional fluid-structure interaction structure motion solution with the time-varying assignment of scalar field parameters, preserving the dynamic characteristics of the flow field while eliminating the risk of mesh distortion, thus meeting the timeliness requirements of rapid clinical assessment. During the flow field solution process, the system simultaneously executes the physiological boundary reverse coupling procedure, loads a lumped parameter model containing vascular compliance characteristics at the computational domain outlet, and receives the patient's measured cuff blood pressure and cardiac output data as physical constraint targets. The Levenberg-Marquardt optimization algorithm is used to iteratively adjust the impedance parameters of the lumped parameter model in the frequency domain until the residual between the pressure waveform output by the 0D model and the clinically measured value converges to the preset range. Finally, the system outputs three-dimensional hemodynamic data anchored to the patient's physiological state, accurately quantifies the local high-velocity jet region caused by irregular calcification, and identifies the subclinical thrombosis risk area caused by low wall shear force.

[0032] Example 2: This example verifies the computational accuracy and numerical stability of the hemodynamic simulation system under complex pathological scenarios. The experiment was conducted on a calibrated fluid dynamics workstation equipped with a dual-processor architecture (32 cores, 2.9 GHz base frequency, and 256 GB of memory). The operating environment was a customized fluid dynamics solver based on the Linux kernel. The experimental data came from an anonymous patient who experienced limited valve motion after transcatheter aortic valve replacement surgery. Data included postoperative enhanced CT scans and Doppler ultrasound measurements. The CT data was acquired from a dual-source CT scanner with a tube voltage of 120 kV and a reconstruction slice thickness of 0.6 mm, providing high-resolution information on the aortic root anatomy. However, due to metal artifacts, grayscale information in the valve leaflet area was lost. Ultrasound data... The data includes transvalvular pressure differential measured by continuous wave Doppler and flow velocity spectrum measured by pulse wave Doppler. The experimental design includes a comparative test between the sample group and the control group of this invention. The sample group of this invention adopts a complete technical solution of geometric topology reconstruction, time-varying flow resistance field mapping and resistance parameter self-calibration. The control group is designed as a partially missing solution, as follows: Control group A uses a conventional grayscale threshold segmentation algorithm to directly extract the fluid domain boundary in CT data without applying standard modal embedding; Control group B applies standard modal embedding but uses a fixed resistance coefficient and does not apply the resistance parameter self-calibration mechanism. All groups are run under the same computing hardware and initial boundary conditions. The inlet flow waveform is set according to the patient's cardiac output, and the outlet pressure waveform is provided by an uncalibrated general lumped parameter model.

[0033] After the experiment started, the system preprocessed the CT data. For the sample group of this invention, the geometric topology reconstruction module successfully identified and registered the artificial valve metal stent, replaced the artifact region with a standard geometric model, and generated a fluid computation domain mesh with smooth and closed boundaries. In contrast, the control group A was affected by metal artifacts, resulting in multiple breaks and non-physical jagged protrusions at the boundaries of the generated fluid domain, leading to a decrease in mesh quality. The proportion of poor-quality meshes with a Jacobian determinant of less than 0.2 exceeded 15%. In the flow field calculation stage, the sample group of this invention dynamically adjusted the drag coefficient in the control domain according to the time-varying flow resistance field mapping procedure and activated the drag parameter self-calibration module. Iterative correction of the systolic inertial drag coefficient was made using the measured transvalvular pressure difference residual. Key intermediate data are recorded as follows: In the initial calculation step, the average transvalvular pressure difference during systole calculated by the sample group of this invention was 8.5 mmHg, which deviated from the measured value of 22.0 mmHg. At this time, the drag parameter self-calibration module detected that the residual exceeded the limit and triggered the correction logic, adjusting the systolic inertial drag coefficient. From initial value Gradually improved, after 5 iterations, when Increase to At that time, the simulated differential pressure converged to 21.8 mmHg, and the residual decreased to 0.2 mmHg. The final output data comparison is shown in Table 1. The data in the table shows that the control group A failed to output effective results due to geometric boundary distortion and calculation divergence. Although the control group B could complete the calculation, it lacked resistance self-calibration, and the output transvalvular differential pressure was much lower than the measured value. It also failed to identify valve leaflet dysfunction. The sample group of this invention not only had a high degree of agreement between the output differential pressure data and the measured value with a relative error of less than 1%, but also accurately quantified the degree of valve hardening through the high resistance coefficient after convergence.

[0034] Table 1: Comparison of Simulation Results

[0035]

[0036] As can be seen from Table 1 and the above process, this invention solves the geometric modeling problem caused by artifacts by standard modal embedding, achieves efficient and stable dynamic simulation by time-varying flow resistance field mapping, and more importantly, improves the simulation from simple forward calculation to a reverse diagnostic tool that can invert pathological states through the drag parameter self-calibration mechanism.

[0037] Example 3: This example combines Figures 1 to 3 This document describes a system for simulating hemodynamics and evaluating the therapeutic effect after heart valve replacement surgery. Figure 1As shown, the entire data processing flow of the system begins with the input of computed tomography (CT) scan data and the standard geometric model of the artificial valve. The geometric topology reconstruction module calls the standard model with industrial-grade precision according to the model code, and replaces the artifact region through spatial rigid registration to generate a static fluid computational domain mesh. The time-varying flow resistance field mapping module adds momentum source terms and adjusts the viscous drag coefficient on the static mesh. With inertial drag coefficient The system simulates the physical effects of valve leaflet opening and closing. Simultaneously, the perivalvular microleakage assessment module maps permeability parameters based on grayscale data and defines porous media. The physiological boundary reverse coupling module combines clinical physiological monitoring data, namely measured blood pressure, cardiac output, and ECG R wave, to construct a lumped parameter model, i.e., an OD model, and uses the Leven-Beg-Marquardt algorithm for iterative optimization to provide outlet boundary conditions. The above parameters converge to the flow field solution module, which calculates three-dimensional hemodynamic data based on the Navier-Stokes equations using a static grid. The preliminary blood flow data is processed by the resistance parameter self-calibration module to calculate the simulation and measured transvalvular pressure differential residuals and iteratively update the systolic phase source term resistance coefficient to generate a correction factor until the convergence condition is met. The final convergence result is transmitted to the data visualization interaction and efficacy assessment module to generate streamline diagrams, pressure cloud maps, and wall shear force distributions, as well as generate thrombosis risk reports and fit heat maps.

[0038] like Figure 2 As shown, the horizontal axis represents cardiac cycle percentage, and the vertical axis represents pressure in mmHg. The solid line in the figure represents the measured blood pressure waveform, which serves as the convergence target of the system. The dashed line represents the pressure waveform output by the model before calibration, whose peak value and waveform characteristics deviate significantly from the measured value. The dotted line represents the pressure waveform output by the model after calibration, showing that after inverse coupling of physiological boundaries and self-calibration of resistance parameters, the simulated pressure curve matches the measured waveform in both systole and diastole. Figure 3 As shown, the architecture consists of a clinical data acquisition workstation, a prior knowledge base, a high-performance fluid simulation server, and a physician's diagnostic terminal. The clinical data acquisition workstation is responsible for uploading raw data, the prior knowledge base provides model calling support, and the core computing power carrier, namely the high-performance fluid simulation server, runs three engines in parallel: the geometry reconstruction engine is responsible for eliminating metal artifacts and embedding standard models, the flow resistance calculation engine performs time-varying resistance mapping, physiological boundary coupling, and perivalvular microleakage calculation, and the intelligent calibration engine is responsible for pressure residual monitoring, automatic parameter iteration, and hardening index generation. Finally, the completed evaluation results are pushed to the physician's diagnostic terminal in real time.

[0039] Example 4: This example uses a standardized algorithm to transparently complete the parameter identification process of the physiological boundary reverse coupling module. The input to this process is the patient's clinically measured systolic blood pressure. diastolic blood pressure and cardiac output measured by echocardiography The system is based on the formula Calculate the initial value of the total peripheral resistance and then apply it according to... (Proximal characteristic impedance) and The proportion of (far peripheral resistance) is allocated to the three-element Windkessel model based on empirical formulas. Set the initial value for overall compliance, where the time constant is... Initially set to twice the duration of the contraction period, initiate an iterative optimization loop: in each iteration, the current... , , Substitute the parameters into the 0D model to solve for the pressure waveform, and calculate the pulse pressure difference output by the model. The difference between measured pulse pressure and actual pulse pressure Error between ,like mmHg, then the compliance parameter is corrected using the gradient descent method. The correction step size is set as a linear function of the error. Until the error converges, this procedure transforms the abstract model parameter calibration into a closed-loop numerical calculation process based on measurable physiological indicators, ensuring the physiological consistency of boundary conditions.

[0040] Example 5: This example constructs a clear three-segment piecewise function model to eliminate the black box of permeability assignment. This model defines permeability based on the Hounsfield unit (HU) value of voxels in CT images. Voxels with a HU value greater than 800 are classified as calcifications or scaffold metals, based on zero permeability. To simulate rigid physical barriers; secondly, for voxels with a HU value less than 100, they are identified as blood or soft tissue spaces, and a maximum effective permeability is set. These values ​​were derived from hydrodynamic experiments calibrating blood dynamic viscosity and typical perivalvular gap dimensions; for voxels with HU values ​​between 100 and 800, a linear interpolation function was used. Permeability was calculated to simulate the flow resistance characteristics of partially calcified or mixed-density regions.

[0041] In addition, to verify the effectiveness of the above parameter calibration and mapping mechanism, this embodiment introduces a set of verification experiments involving gradient changes. Three groups of patients with different degrees of calcification, namely mild, moderate and severe, are selected, and simulations are performed using the above standardized procedures. The results show that as the calcification score increases from 200 to 800, the perivalvular leakage calculated based on the above mapping model shows a non-linear growth trend, and the leakage location obtained by simulation has more than 90% spatial overlap with the location of the color regurgitation jet shown by postoperative Doppler ultrasound.

[0042] Example 6: Regarding the construction process of the system's built-in standard geometric database and flow resistance parameter library, this example selects no fewer than fifty physical valve samples from different batches that meet the factory quality standards when constructing the standard three-dimensional geometric model library for artificial valves. A high-resolution scan is performed using an industrial-grade micro-computed tomography system under non-stress conditions. The voxel size is set to 10 micrometers to accurately capture the microscopic geometric features of the stent wire diameter and leaflet thickness. Statistical shape model analysis is performed on the point cloud data obtained from the scan, and the geometric mean position of each spatial node is calculated to generate a nominal CAD model that represents the general geometric characteristics of this valve model. A mesh independence verification test is then conducted on this nominal model. By progressively refining the mesh size until the relative error of the calculated key flow field variables is less than 1% of the required mesh convergence index, the optimal mesh density parameters for this valve model in the static computational domain mesh are determined and fixed into the database.

[0043] In addition, regarding the reference resistance coefficient used in the control domain of the time-varying flow resistance field mapping module, the system performs standardized in vitro fluid dynamics calibration procedures according to the ISO5840 heart valve prosthesis testing standard. A physical entity of a specific model of artificial valve is placed in the test section of the pulsating flow simulation circulation loop, and the volumetric flow rate through the valve is adjusted under steady-state flow conditions. The transvalvular pressure gradient across the valve was recorded simultaneously from 0 to 30 liters per minute. According to the theoretical form of the Darcy-Fochheimer Law The least squares nonlinear regression analysis was performed on the pressure difference-flow rate data points obtained from the experiment to calculate the viscous drag coefficient of this type of valve in the fully open state. With inertial drag coefficient The baseline values ​​are used as the basis for simulation under pulsating flow conditions. These baseline coefficients are then substituted into the fluid control equations for simulation, and the simulation results are compared with experimentally measured transient flow field data. If the root mean square error between the simulation and the experiment exceeds a preset tolerance limit, a Reynolds number-based approach is then introduced. Correction factor Dynamic compensation is applied to the drag coefficients, and the verified set of drag coefficients and their corresponding Reynolds number applicability ranges are ultimately encapsulated in the system's parameter library as a lookup table.

[0044] Example 7: This example describes a standardized pre-deployment calibration and debugging procedure to ensure the system's functional consistency and computational stability under different operating environments. After the system is initially installed on the server cluster of the target medical institution, a computing benchmark test is performed. By running a pre-set standardized fluid dynamics example, the floating-point operation capability and memory bandwidth of the hardware platform under different degrees of parallelism are measured. Based on the test results, the optimal parallel partitioning strategy and memory management scheme adapted to the hardware environment are automatically generated. Using a set of standardized phantom CT data containing different signal-to-noise ratios and resolutions, the edge extraction threshold and registration tolerance of the geometric topology reconstruction module are adaptively fine-tuned until the Hausdorff distance between the geometric model output by the system and the phantom's true value meets the preset acceptance criteria.

[0045] To address the uncertainty of physiological boundary condition parameters caused by individual patient differences, the system performs parameter initialization and Bayesian prior calibration based on a historical case database before formal operation. The system accesses the database of past TAVR surgery cases of medical institutions, extracts preoperative physiological parameters (including blood pressure, heart rate, and cardiac output) and postoperative hemodynamic measurement data of no less than one hundred patients, constructs a physiological characteristic distribution model of the patient population specific to the center, uses Bayesian inference algorithm to calculate the posterior probability distribution of each impedance parameter in the current patient ensemble parameter model, and uses the expected value of the posterior distribution as the initial guess value for iterative optimization.

[0046] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0047] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A post-cardiac valve replacement hemodynamics simulation and efficacy assessment processing system, characterized in that, The system comprises: a geometric topology reconstruction module, configured to acquire postoperative aortic root computed tomography data of a patient and a model code of an implanted artificial valve; extract an aortic root outer contour without metal artifacts in the computed tomography data, and call a preset standard three-dimensional geometric model of the artificial valve according to the model code; perform spatial rigid registration to map the standard three-dimensional geometric model of the artificial valve to the inside of the aortic root outer contour, replace the original valve area boundary in the computed tomography data due to gray scale distortion caused by metal artifacts with a surface boundary of the standard three-dimensional geometric model of the artificial valve, and generate a static fluid calculation domain grid containing a standard geometric of the artificial valve; a time-varying flow resistance field mapping module, configured to mark a set of spatial units corresponding to a valve leaflet geometry in the standard three-dimensional geometric model of the artificial valve in the static fluid calculation domain grid; according to a preset cardiac cycle timing logic, add a time-varying source term resistance coefficient to a momentum conservation term of the Navier-Stokes equation; assign the source term resistance coefficient to a low resistance threshold value allowing fluid to pass in the systolic phase, and assign the source term resistance coefficient to a high resistance threshold value blocking fluid from passing in the diastolic phase, and simulate the physical effect of the opening and closing of the valve leaflet under the constraint of the static grid topology by adjusting the fluid resistance properties of the set of spatial units; a flow field solving module, configured to solve a fluid control equation based on the static fluid calculation domain grid and the time-varying source term resistance coefficient, and output three-dimensional blood flow dynamics data.

2. The system according to claim 1, wherein, The system further comprises a resistance parameter self-calibration module; the resistance parameter self-calibration module is configured to acquire measured transvalvular pressure difference data detected by ultrasonic Doppler detection of the patient; after the flow field solving module outputs preliminary three-dimensional blood flow dynamics data, calculate a residual value of the simulated transvalvular pressure difference and the measured transvalvular pressure difference data; if the residual value exceeds a preset convergence range, generate a correction factor based on the residual value, iteratively update the source term resistance coefficient in the systolic phase using the correction factor, and trigger the flow field solving module to recalculate until the residual value converges; based on a ratio of the finally converged source term resistance coefficient to the low resistance threshold value, generate a quantitative index representing the degree of restriction of the valve motion function.

3. The system according to claim 1, wherein, The system further comprises a paravalvular microleakage evaluation module, which is configured to define a paravalvular annular voxel region between an outer surface of the standard three-dimensional geometric model of the artificial valve and an inner surface of the aortic root outer contour; extract gray scale values of each voxel unit in the paravalvular annular voxel region in the computed tomography data, and convert the gray scale values into porous medium permeability parameters based on a preset gray scale permeability mapping function; wherein the gray scale values higher than a preset calcification threshold value are mapped to zero permeability, and the gray scale values lower than a preset tissue threshold value are mapped to effective permeability; the flow field solving module is configured to calculate leakage flow in the paravalvular annular voxel region according to the porous medium permeability parameters.

4. The system according to claim 1, wherein, The system further comprises a physiological boundary reverse coupling module, which is configured to construct a lumped parameter model containing peripheral vascular impedance characteristics as an outlet boundary condition of a static fluid calculation domain grid; receive patient measured blood pressure data and cardiac output data, and iteratively adjust the compliance parameter and peripheral resistance parameter of the lumped parameter model using the Levenberg-Marquardt algorithm until the deviation between the pressure waveform output by the lumped parameter model and the measured blood pressure data is less than a preset threshold; and load the time-varying pressure value output by the lumped parameter model after parameter convergence into the boundary equation of the flow field solving module in real time.

5. The system for post-cardiac valve replacement hemodynamics simulation and therapy evaluation processing according to claim 1, wherein, The geometric topology reconstruction module identifies a high-density feature point cloud generated by the metal stent of the artificial valve in the computed tomography data when performing spatial rigid registration; determines the pose coordinates of the standard three-dimensional geometric model of the artificial valve within the outer contour of the aortic root by minimizing the spatial Euclidean distance between the high-density feature point cloud and the stent nodes in the standard three-dimensional geometric model of the artificial valve.

6. The system for post-cardiac valve replacement hemodynamics simulation and therapy evaluation processing according to claim 1, wherein, The time-varying flow resistance field mapping module adds a source term in the Navier-Stokes equation, the resistance coefficient of which follows Darcy-Forchheimer law, which defines the source term satisfies the following relationship: wherein, is the viscous resistance coefficient, is the inertial resistance coefficient, is the dynamic viscosity of the fluid, is the fluid density, is the fluid velocity vector; the low resistance threshold corresponds to and a value that tends to zero, the high resistance threshold corresponds to and a value greater than a preset cutoff value.

7. The system for post-cardiac valve replacement hemodynamics simulation and therapy evaluation processing according to claim 1, characterized in that, The flow field solving module extracts the wall shear stress distribution data on the surface of the standard three-dimensional geometric model of the artificial valve after outputting the three-dimensional blood flow dynamics data; calculates the area of the low shear region where the wall shear stress is lower than a preset thrombus formation threshold, and calculates the percentage of the low shear region area in the total surface area of the valve leaflet, to generate a thrombus risk assessment report.

8. The system for post-cardiac valve replacement hemodynamics simulation and therapy evaluation processing according to claim 1, characterized in that, The time-varying flow resistance field mapping module includes the start time of the systole phase, the end time of the systole phase, and the duration of the diastole phase; the time-varying flow resistance field mapping module extracts the R-wave feature points by analyzing the patient's electrocardiogram signal data to synchronize and calibrate the time axis of the cardiac cycle timing logic.

9. The system for post-cardiac valve replacement hemodynamics simulation and therapy evaluation processing according to claim 1, characterized in that, The system further comprises a data visualization interaction module; the data visualization interaction module is configured to map the three-dimensional blood flow dynamics data to the static fluid calculation domain grid to generate a three-dimensional rendering scene containing flow line diagrams, pressure cloud diagrams, and wall shear stress distribution diagrams; in the three-dimensional rendering scene, the standard three-dimensional geometric model of the artificial valve is displayed in a semi-transparent manner, and the spatial region where the flow velocity exceeds a preset turbulent flow threshold is highlighted.

10. The system for post-cardiac valve replacement hemodynamics simulation and therapy evaluation processing according to claim 9, wherein, The data visualization interaction module is further configured to generate a virtual stent fitment heat map; calculate the radial distance field between the outer surface of the standard three-dimensional geometric model of the artificial valve and the inner surface of the outer contour of the aortic root; map the radial distance field to a color gradient, and project the color gradient to the stent surface of the standard three-dimensional geometric model of the artificial valve to display the geometric fit state of the stent and the blood vessel wall.

Images