A method for completing blood flow parameters after TAVR based on CT-MRI fusion

Abstract

This invention discloses a method for completing postoperative blood flow parameters based on CT-MRI fusion, belonging to the interdisciplinary field of medical image processing and computational fluid dynamics. This invention employs CT valve registration technology to reconstruct the invisible three-dimensional structure of the artificial valve leaflet using valve frame information from CT images. This reconstruction is coupled with a dynamic heart model to construct a complete valve mechanics and functional simulation platform, solving the imaging blind zone problem. Furthermore, by combining 4D Flow MRI to acquire hemodynamic parameters at the inlet and outlet of the postoperative region, the invention comprehensively reconstructs the blood flow conditions upstream and downstream of the artificial valve region during systole and diastole at each stage after TAVR, greatly improving the spatiotemporal resolution and physical realism of boundary conditions in fluid-structure interaction simulation. This yields a more realistic passive response of the valve leaflet under blood flow, providing a reliable basis for the quantitative and mechanistic risk analysis of postoperative complications such as thrombosis, paravalvular leak, and leaflet fatigue.

Description

Technical Field

[0001] This invention relates to a method for completing blood flow parameters after TAVR based on CT-MRI fusion, belonging to the interdisciplinary field of medical image processing and computational fluid dynamics. Background Technology

[0002] TAVR (Transcatheter Aortic Valve Replacement) is a minimally invasive procedure that involves implanting an artificial valve to treat aortic stenosis. Postoperatively, computed tomography (CT) and ultrasound imaging are usually required to assess the patient's morphological and functional status to guide clinical decisions such as re-intervention for valve failure or anticoagulation therapy.

[0003] For clinical hemodynamic assessment (flow velocity, pressure gradient, and risk of complications), current methods mainly rely on echocardiography (ultrasound) and four-dimensional phase-contrast magnetic resonance imaging (4D Flow MRI). Echocardiography can quickly provide information on cardiac systolic / diastolic function, valvular blood flow velocity, and some pressure gradient information (such as aortic valve orifice velocity and mitral valve blood flow spectrum). However, the accuracy of its results is highly dependent on the operator's experience and skills, and the blood flow information obtained by echocardiography is based on two-dimensional cross-sections, making it difficult to fully reconstruct the complex structure of the three-dimensional blood flow field. Although 4D Flow MRI can reconstruct the three-dimensional flow field, in the post-TAVR scenario, the metal stent of the artificial valve can cause magnetic susceptibility artifacts, leading to signal loss in local areas (such as the aortic root).

[0004] In recent years, some scholars have proposed using CT or MRI to reconstruct anatomical models and then employing computational fluid dynamics (CFD) and fluid-structure interaction (FSI) methods to predict hemodynamic parameters, as illustrated in the article "Votta E, Le TB, Stevanella M, et al. Towardpatient-specific simulations of cardiac valves: state-of-the-art and future directions[J]. Journal of biomechanics, 2013, 46(2):217-228." However, it is difficult to accurately identify the geometry of the prosthetic valve (especially the leaflet structure) in CT images after TAVR, causing the CFD / FSI method to fail. Therefore, existing technologies cannot complete the hemodynamic parameters of the prosthetic valve region after TAVR, and consequently cannot provide a reliable basis for the quantitative mechanism of postoperative complications such as thrombosis and leaflet failure. Summary of the Invention

[0005] To achieve complete hemodynamic parameters in the prosthetic valve region after TAVR surgery and to quantify the mechanisms of postoperative complications, this invention provides a method for completing hemodynamic parameters in the prosthetic valve region after TAVR surgery, capable of restoring flow field information that was previously invisible in 4D Flow MRI images after TAVR. Specifically, it includes:

[0006] Step 1: Based on the CT images after TAVR, extract the complete geometric model of the aortic valve root after surgery; the complete geometric model of the aortic valve root after surgery includes the implanted artificial valve and the aortic valve root.

[0007] Step 2: Based on the 4D Flow MRI images after TAVR surgery, extract the blood flow boundaries upstream and downstream of the prosthetic valve area;

[0008] Step 3: Calculate the hemodynamic parameters of the artificial valve region using the complete geometric model of the aortic valve root reconstructed in Step 1 and the blood flow boundary obtained in Step 2.

[0009] Optionally, step 1 includes:

[0010] Step 1.1: Obtain CT images after TAVR surgery;

[0011] Step 1.2: Extract the geometric model of the aortic root based on CT image data after TAVR surgery;

[0012] Step 1.3: Obtain the geometric model of the artificial valve in the patient's body based on the CT image data after TAVR surgery;

[0013] Step 1.4: Combine the geometric model of the aortic root with the geometric model corresponding to the artificial valve in the patient's body to obtain the complete geometric model of the aortic valve root after surgery.

[0014] Optionally, in step 1.2, an image segmentation and 3D reconstruction algorithm is used to extract the geometric model of the aortic root.

[0015] Optionally, image segmentation algorithms include thresholding, region growing, and deep learning algorithms.

[0016] Optionally, step 1.3 includes:

[0017] Step 1.3.1: Establish a three-dimensional geometric model of the artificial valve in its natural state and generate the corresponding discretized model;

[0018] Step 1.3.2: Obtain the displacement of the valve frame from the initial state to the deformed state, where the initial state corresponds to the natural state and the deformed state corresponds to the state within the patient's body.

[0019] Step 1.3.3: Obtain the deformation of the leaflet and skirt under the deformed valve frame through the non-uniform shape function field reconstruction algorithm, and obtain the three-dimensional discrete model of the deformed valve, and then obtain the geometric model of the deformed valve, that is, the geometric model corresponding to the artificial valve in the patient's body.

[0020] Optionally, step 1.3.1 includes:

[0021] The first step is to construct three-dimensional geometric models of the valve frame, leaflets, and skirt respectively, based on the physical structure of the artificial valve, using parametric modeling or measurement.

[0022] The second step involves extracting a typical symmetrical feature unit from the valve frame, artificial leaflet, and skirt, taking into account the periodicity and symmetry of the artificial valve structure.

[0023] The third step involves discretizing the extracted symmetrical feature units and using functions such as 3D mirroring, symmetry, and rotation to generate a complete 3D discrete model of the petal frame, petal leaf, and skirt based on the discrete region.

[0024] Optionally, step 1.3.2 includes:

[0025] The first step is to establish the deformable flap after surgery based on postoperative CT images through image segmentation or three-dimensional reconstruction.

[0026] The second step is to deform the initial valve frame to fit the voxel model of the deformed valve frame.

[0027] The third step is to calculate the spatial coordinate difference between the corresponding nodes of the initial petal frame and the deformed petal frame, and obtain the displacement of each node of the petal frame.

[0028] Optionally, step 1.3.3 includes:

[0029] The first step is to define the material parameters for the artificial petals and skirt;

[0030] The second step is to construct the connection positions and methods of the valve frame, artificial leaflets, and skirt according to the connection method of the actual artificial valve;

[0031] The third step is to define the contact type and contact attributes;

[0032] The fourth step, based on the principle of minimum potential energy, is to derive the stiffness matrix and equivalent nodal force vector of each discrete region in the three-dimensional discrete model of the artificial valve; assemble all discrete regions according to nodal degrees of freedom to form the global stiffness matrix [K] and the global equivalent nodal load vector {F}; and solve the linear / nonlinear system equations [K]{u} using the direct solution method / iterative solution method. e}[N]={F}, obtain the updated global node displacement vector {U}; [N] represents the shape function matrix, u e This represents the displacement of each discrete region;

[0033] Fifth step, based on the updated discrete region displacement vector {U}, use equation P deformed =P initial +{U}, to obtain an accurate reconstruction of the three-dimensional discrete model of the deformed valve in vivo;

[0034] The sixth step is to obtain the geometric model of the deformable valve based on the three-dimensional discrete model of the deformable valve, which is the geometric model corresponding to the artificial valve in the patient's body.

[0035] Optionally, step 2 includes:

[0036] Step 2.1: Obtain 4D flow MRI images after TAVR surgery;

[0037] Step 2.2: Extract the geometric model of the aortic root based on the 4D Flow MRI image data after TAVR;

[0038] Step 2.3: Determine the upstream and downstream boundary plane positions of the artificial valve region in a single phase of 4D Flow MRI images;

[0039] Step 2.4: Use image registration to obtain the upstream and downstream boundary plane positions of the artificial valve region in multiple phases of 4D Flow MRI images;

[0040] Step 2.5: Obtain blood flow information at the boundary plane.

[0041] Optionally, step 3 includes:

[0042] Step 3.1: Use image registration to fuse post-TAVR CT and 4D Flow MRI images;

[0043] Step 3.2, select the artificial valve material;

[0044] Step 3.3, set the boundaries;

[0045] Step 3.4, Fluid-structure Interaction Calculation;

[0046] Step 3.5: Output the hemodynamic parameters of the artificial valve area, including the energy loss index (ELI) value, the pressure and velocity distribution on which the calculation was based, the energy loss distribution visualization, and the comparative analysis results with normal valves.

[0047] Optionally, step 3.4 includes:

[0048] Step 3.4.1, Fluid Calculation: Boolean operations are performed on the vascular region of the aortic root model obtained in Step 2.2 and the artificial valve model region reconstructed in Step 1. The artificial valve model is removed from the aortic root vascular region, and the remaining region is used as the fluid computation domain. The computation domain is discretized using the finite volume method. By integrating the three-dimensional incompressible Navier-Stokes equations within each finite volume element, the partial differential equations are transformed into a system of algebraic equations. For each control volume, Gauss's theorem is used to transform the terms into flux forms on the volume surface. Based on the boundary conditions set in Step 3.3, an iterative algorithm is used to solve the transformed algebraic equations to obtain hemodynamic parameters and wall force data.

[0049] Step 3.4.2, solid calculation: The artificial valve model is divided into a finite number of elements using the finite element method. The displacement field inside the element is approximated by the interpolation function. The hemodynamic load obtained in step 3.4.1 is applied to the surface of the artificial valve. Combining the connection relationship and constitutive model, the momentum balance equation is transformed into a set of algebraic equations of nodal displacements using the finite element method. The stress, strain and deformation distribution of the valve structure are obtained through numerical solution.

[0050] Step 3.4.3, Information Interaction: The force information of the artificial valve in the fluid calculation and the displacement information of the artificial valve in the solid calculation are exchanged to ensure that the fluid calculation and the solid calculation satisfy the solution conditions.

[0051] Optionally, step 3.5 includes:

[0052] Step 3.5.1: Output the three-dimensional transient velocity field data calculated in the fluid domain as text format data to obtain post-processable flow field information; the flow field information includes three-dimensional velocity field, three-dimensional pressure field, and wall shear stress;

[0053] Step 3.5.2: Based on the fluid calculation results, the energy loss index is obtained by calculating the energy change of blood flow before and after passing through the artificial valve, including the ELI value, the pressure and velocity distribution on which the calculation is based, the visualization of the energy loss distribution, and the comparative analysis with normal valves.

[0054] The beneficial effects of this invention are:

[0055] 1. By using high temporal resolution 4D Flow MRI to cover the complete cardiac cycle, the system collects hemodynamic parameters of the postoperative region's inlet and outlet, comprehensively reconstructing the blood flow conditions upstream and downstream of the artificial valve region at each stage of systole and diastole after TAVR, greatly improving the spatiotemporal resolution and physical realism of boundary conditions in fluid-structure interaction simulation.

[0056] 2. CT valve frame registration technology was employed to reconstruct the invisible three-dimensional structure of the artificial valve leaflet using valve frame information from CT images. This structure was then coupled with a dynamic heart model to construct a complete valve mechanics and function simulation platform. This not only solved the imaging blind spot problem but also provided a realistic and accurate structural basis for postoperative valve mechanics analysis and risk assessment.

[0057] 3. By integrating the boundary conditions obtained from MRI with the registered and reconstructed artificial valve leaflet model through fluid-structure interaction, the true passive response of the valve leaflet under blood flow is simulated. It can simultaneously output the local mechanical parameters (such as stress and deformation) and local hemodynamic state of the valve leaflet, providing a reliable basis for the quantitative and mechanistic risk analysis of postoperative complications such as thrombosis, paravalvular leakage, and valve leaflet fatigue, and significantly improving the scientific nature and pertinence of risk prediction. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a schematic diagram of a technique for completing hemodynamic parameters after TAVR provided by the present invention;

[0060] Figure 2 This is a schematic diagram of the geometric model of the aortic root extracted from CT images after TAVR surgery;

[0061] Figure 3 This is a schematic diagram of the geometric reconstruction results of the artificial valve in CT images after TAVR surgery;

[0062] Figure 4 This is a schematic diagram of the complete geometric model of the aortic root extracted from CT images after TAVR surgery;

[0063] Figure 5 This is a schematic diagram of the upstream and downstream boundary planes and information in 4D Flow MRI images after TAVR surgery;

[0064] Figure 6 This is a schematic diagram of image fusion of CT images and 4D Flow MRI after TAVR surgery;

[0065] Figure 7 This is a schematic diagram of the reconstructed flow field in the artificial valve area after TAVR surgery based on CT-MRI fusion;

[0066] Figure 8 This is a schematic diagram of a CFD method for reconstructing the flow field at the aortic root based on MRI boundaries;

[0067] Figure 9 This is a schematic diagram comparing the results of the reconstructed flow field (CFD) and the 4D flow MRI flow field at the aortic root.

[0068] Figure 10 This is a schematic diagram of the FSI method for reconstructing the flow field at the aortic root based on MRI boundaries;

[0069] Figure 11 This is a schematic diagram comparing the results of the reconstructed flow field (FSI) at the aortic root with the flow field of 4D Flow MRI. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0071] The basic knowledge involved in the solution of this invention is introduced as follows:

[0072] Hemodynamics is the science that studies the mechanical laws governing blood flow in the cardiovascular system. It mainly explores the relationship between blood flow, blood flow resistance, and blood pressure. After TAVR, hemodynamic parameters can be used to determine whether the implanted artificial valve is functioning properly and to predict the patient's long-term prognosis. These parameters mainly include transvalvular pressure gradient, valvular blood flow velocity, perivalvular regurgitation / regurgitation fraction, valvular arterial impedance, cardiac output, and stroke volume.

[0073] The aortic valve is an important valve in the heart, located between the left ventricle and the aorta. It plays a crucial role in the heart's blood circulation, ensuring that blood flows unidirectionally from the left ventricle to the aorta, and then to all organs and tissues throughout the body. The aortic valve consists of three semilunar leaflets, commonly known as the left coronary valve, right coronary valve, and non-coronary valve. These leaflets open when the left ventricle contracts, allowing blood to flow into the aorta; and close when the left ventricle relaxes, preventing blood from flowing back into the left ventricle.

[0074] Aortic stenosis (AS) is a heart valve disease characterized by narrowing of the blood flow passage between the left ventricle and the aorta, leading to increased resistance to heart pumping. It can be treated by replacing the aorta with an artificial valve.

[0075] The artificial valve used in TAVR surgery is typically made of metal tubing (nickel-titanium or cobalt-chromium alloy) and animal (such as pig or cow) heart valve tissue. An artificial valve usually consists of a valve stent, leaflets, and a skirt.

[0076] A constitutive model is a mathematical expression describing the stress-strain relationship of a material under external forces, and it forms the foundation for studying and simulating the mechanical behavior of materials. Through constitutive models, the mechanical responses of different materials under various deformation conditions such as tension, compression, and shear can be accurately reflected. The rational selection and establishment of constitutive models are of great significance for engineering structural design, numerical simulation, and the development of new materials. Especially in the study of nonlinear, large-deformation materials such as soft tissues and rubber, a suitable constitutive model can effectively predict their complex mechanical properties, providing theoretical basis and technical support for related applications.

[0077] 4D Flow MRI is a magnetic resonance imaging technique capable of acquiring blood flow velocity information in three dimensions throughout the entire cardiac cycle. It encodes the blood flow velocity vector field by combining the velocity encoding gradient with the conventional imaging gradient, thereby obtaining the velocity magnitude and direction information of each voxel in three-dimensional space at different phases of the cardiac cycle.

[0078] Example 1

[0079] This embodiment provides a method for completing blood flow parameters after TAVR based on CT-MRI fusion, used to obtain hemodynamic parameters of patients after TAVR, such as... Figure 1 As shown, the method includes:

[0080] Step 1: Extract the complete geometric model of the aortic valve root based on the CT images after TAVR.

[0081] The complete geometric model of the aortic valve root after surgery is a geometric model of the aortic valve root including the implanted artificial valve.

[0082] Step 1.1: Obtain CT images after TAVR surgery;

[0083] Post-TAVR CT images are CT images of a patient after TAVR surgery. These images are obtained by taking CT images of the aorta (usually the thoracic aorta) post-surgery, including the entire heart, aortic sinus, ascending aorta, aortic arch, and descending aorta. Figure 2 As shown, the geometric model of the aortic root and the valve frame of the implanted artificial valve can be seen.

[0084] Step 1.2: Extract the geometric model of the aortic root based on CT image data after TAVR surgery;

[0085] For CT images after TAVR, image segmentation and 3D reconstruction algorithms can be used to extract the geometric model of the aortic root, including: left ventricle, aortic sinus, native valve leaflets, calcification (if present), coronary arteries, artificial valve stents, and ascending aorta.

[0086] Image segmentation algorithms that can be used include threshold segmentation, region growing, and deep learning.

[0087] In this embodiment, a deep learning method is used to obtain the segmentation result of the aortic root geometric model, and surface rendering based on the segmentation result is used for three-dimensional reconstruction to obtain the aortic root geometric model.

[0088] Step 1.3: Obtain the geometric model of the artificial valve in the patient's body based on the CT image data after TAVR surgery;

[0089] Considering the poor visibility of artificial valve leaflets in CT images, making them difficult to directly use for geometric model reconstruction, this invention proposes a method for constructing an artificial valve geometric model. First, a three-dimensional geometric model of the artificial valve in its natural state is established. Second, the displacement of the valve frame from its initial state to its deformed state is obtained, where the initial state corresponds to the natural state and the deformed state corresponds to the state within the patient's body. Finally, the deformation of the leaflets and skirt under the deformed valve frame is obtained using a non-uniform shape function field reconstruction algorithm, resulting in the geometric model of the deformed valve, i.e., the geometric model of the artificial valve when it is within the patient's body.

[0090] The three-dimensional geometric model of an artificial valve in its natural state refers to the three-dimensional geometric model of the artificial valve when it is not subject to any external constraints. In contrast, when the artificial valve is inside the patient's body, it is subject to the forces of the surrounding human tissues, that is, it is subject to external constraints. Therefore, compared with the natural state, the artificial valve has a certain deformation when it is inside the patient's body. So the state of the artificial valve when it is inside the patient's body is called the deformed state.

[0091] In practical applications, obtaining the geometric model of the artificial valve in the patient's body includes:

[0092] Step 1.3.1: Establish a three-dimensional geometric model of the artificial valve in its natural state and generate a corresponding three-dimensional discretized model;

[0093] The first step is to construct three-dimensional geometric models of the valve frame, leaflets, and skirt respectively, based on the actual structure of the artificial valve, using parametric modeling or measurement.

[0094] The second step involves extracting a typical symmetrical feature unit from the valve frame, artificial leaflet, and skirt, respectively, considering the periodicity and symmetry of the artificial valve structure. These units are typically 1 / 12 of the valve frame structure, 1 / 3 of the leaflet structure, and 1 / 24 of the skirt structure.

[0095] The third step involves discretizing the extracted symmetrical feature units. The discretized size of the petiole frame is 0.2mm to 0.4mm, and the discretized size of the artificial leaflets and skirt is 0.4mm to 0.6mm. Based on this, using functions such as 3D mirroring, symmetry, and rotation, a complete 3D discrete model of the petiole frame, leaflets, and skirt is generated based on the discrete regions.

[0096] Step 1.3.2: Obtain the displacement of the petiole frame from the initial state to the deformed state;

[0097] The initial state of the valve frame refers to the state corresponding to the three-dimensional discretized model of the valve frame created in step 1.3.1. The valve frame in this state is called the initial valve frame. The deformed state of the valve frame refers to the state of the valve frame in the patient's body after TAVR. The valve frame in this state is called the deformed valve frame, which is obtained from postoperative CT images.

[0098] The first step is to establish the deformable flap after surgery based on postoperative CT images through image segmentation or three-dimensional reconstruction.

[0099] Image segmentation algorithms that can be used include thresholding, region growing, and deep learning. 3D reconstruction algorithms that can be used include surface rendering based on segmentation results, direct volume rendering, and point cloud reconstruction.

[0100] In this embodiment, a deep learning method (such as 3D convolutional neural networks, U-Net variants, etc.) is used to obtain a voxel model of the deformed valve arch in postoperative CT. A voxel model is a three-dimensional data model composed of countless stacked voxels, which together describe the shape, internal structure, and properties of the valve arch. The data generated by CT scans is voxel-level data (such as 3D volume data in DICOM format). The deep learning model (such as a segmentation network) can directly output binary voxel labels (such as valve arch regions), and then generate a continuous voxel model through post-processing.

[0101] The second step involves deforming the initial valve frame to fit the voxel model of the deformed valve frame.

[0102] The initial valve frame is the three-dimensional discretized model of the valve frame created in step 1.3.1. The deformable valve frame is the valve frame in the patient's body after TAVR surgery. The deformable valve frame is obtained by deforming the initial valve frame through a deformation method.

[0103] The deformation methods that can be used include point set-based registration algorithms, deformation field-based registration methods, and model-based parametric deformation methods.

[0104] The third step is to calculate the spatial coordinate difference between the corresponding nodes of the initial petal frame and the deformed petal frame, and obtain the displacement of each node of the petal frame.

[0105] The initial and deformed lobes have the same geometric connections and topology. The spatial coordinates of each node on the initial lobe... Spatial coordinates of each node on the deformable flap. Calculate the difference, the displacement of each node of the petal frame.

[0106] Step 1.3.3: Obtain the deformation of the leaflet and skirt under the deformed valve frame through the non-uniform shape function field reconstruction algorithm, and obtain the three-dimensional discrete model of the deformed valve, and then obtain the geometric model of the deformed valve, that is, the geometric model corresponding to the artificial valve in the patient's body.

[0107] The first step is to define the material parameters for the artificial petals and skirt.

[0108] Based on a review of relevant literature, the collected material data were used to define constitutive models and material parameters for the artificial petals and skirts, whereby the constitutive model defines the stress-strain relationship of the material.

[0109] In this embodiment, the artificial leaflet adopts the Ogden model (an isotropic hyperelastic model based on principal stretch ratio, a constitutive model), with shear modulus term 1 being 0.35 MPa, exponent term 1 being 15.2, shear modulus term 2 being -0.12 MPa, exponent term 2 being -10.8, and Poisson's ratio being 0.5.

[0110] The second step involves constructing the connection positions and methods for the valve frame, artificial leaflets, and skirt based on the connection methods of the actual artificial valve. Connection methods include tie constraints, shared nodes, and multi-point constraints (MPC).

[0111] In this embodiment, the valve frame and skirt, the artificial leaflet and skirt, and the artificial leaflet and skirt are all bound together.

[0112] The third step is to define the contact type and contact attributes. Contact attributes refer to the contact types and attributes added between the petiole frame, artificial leaflets, and skirt. Contact types include self-contact, surface-to-surface contact, etc., and contact attributes include hard contact, penalty function contact, etc.

[0113] In this embodiment, the artificial leaflet and the skirt adopt surface self-contact, and the contact attribute is hard contact; the artificial leaflet and the petiole frame, and the skirt and the petiole frame, all adopt surface-to-surface contact, and the contact attribute is hard contact.

[0114] The fourth step, based on the principle of minimum potential energy, is to derive the stiffness matrix and equivalent nodal force vector for each discrete region in the three-dimensional discrete model of the artificial valve. All discrete regions are assembled according to nodal degrees of freedom to form the global stiffness matrix [K] and the global equivalent nodal load vector {F}; the linear / nonlinear system equations [K]{u} are solved using direct solution methods / iterative solution methods (such as the preprocessed conjugate gradient method). e}[N]={F}, obtain the updated global node displacement vector {U}. [N] represents the shape function matrix, u e This represents the displacement of each discrete region.

[0115] Fifth step, based on the updated discrete region displacement vector {U}, use equation P deformed =P initial +{U} yields an accurate reconstruction of the three-dimensional discrete model of the deformed valve in vivo.

[0116] The sixth step is to obtain the geometric model of the deformable valve based on the three-dimensional discrete model of the deformable valve, which is the geometric model corresponding to the artificial valve in the patient's body.

[0117] The geometric model obtained from the discrete model can be achieved using methods such as voxelization, reverse engineering-based 3D reconstruction, or tensor completion-based 3D imaging. These methods are described below:

[0118] ① Based on the voxelization method: the three-dimensional discrete model is converted into a voxel model, the voxel model is imported into finite element software (such as FEBio, Pycalculix), mesh generation is performed, and the mesh corresponding to the gaps is deleted, thereby constructing a three-dimensional geometric model.

[0119] ② 3D reconstruction method based on reverse engineering: Extract point cloud data from a 3D discrete model, generate a continuous geometric surface through interpolation and fitting algorithms, construct the geometric and attribute information of the model surface using the topological relationship of discretized triangle planes, set material attributes and connection attributes, and realize a realistic 3D model using graphics libraries such as OpenGL.

[0120] ③ Tensor completion-based 3D imaging method: The target in the 3D discrete model is decomposed into a finite number of scattering points to establish a 3D radar signal model. Tensor completion technology in the embedded space is used to solve the problem of missing data, restore the complete 3D tensor, and perform 3D imaging through the restored high-order tensor to obtain a complete 3D geometric model.

[0121] This application does not limit the method for obtaining geometric models based on discrete models.

[0122] Figure 3 This is a schematic diagram of the geometric reconstruction results of the artificial valve in CT images after TAVR surgery. It can be seen that it includes the valve frame, leaflets, and skirt of the artificial valve.

[0123] Step 1.4: Assemble the complete geometric model of the aortic valve root after surgery;

[0124] Based on the aortic root geometric model obtained in step 1.2 and the artificial valve geometric model obtained in step 1.3, the complete geometric model of the aortic valve root after surgery is extracted as follows: Figure 4 The diagram shows a geometric model of the left ventricle, aortic sinus, native leaflet, calcification (if present), coronary arteries, ascending aorta, and artificial valve (valve frame, skirt, artificial leaflet).

[0125] Step 2: Based on the 4D Flow MRI images after TAVR surgery, extract the blood flow boundaries upstream and downstream of the prosthetic valve area.

[0126] The upstream region of the prosthetic valve region typically refers to the left ventricular outflow tract (LVOT), the area in front of the prosthetic valve after blood is pumped out of the left ventricle. Measuring the blood flow velocity and pressure upstream is fundamental for calculating core parameters such as transvalvular pressure gradient and effective valve orifice area. The downstream region of the prosthetic valve region refers to the aorta, where blood enters the aorta after passing through the prosthetic valve. Assessing downstream blood flow, especially the blood flow in the aorta distal to the valve, is crucial for determining the presence of valvular stenosis, perivalvular regurgitation, and assessing aortic compliance.

[0127] Step 2.1: Obtain 4D flow MRI images after TAVR surgery;

[0128] Post-TAVR 4D Flow MRI images refer to 4D Flow MRI images of the patient after TAVR surgery. These images are obtained by performing 4D Flow MRI on the patient's aorta (usually the thoracic aorta) after the surgery, including the whole heart, ascending aorta, aortic arch, and descending aorta.

[0129] In this embodiment, the slice thickness of the 4D Flow MRI image is 0.75 mm.

[0130] Step 2.2: Extract the geometric model of the aortic root based on the 4D Flow MRI image data after TAVR;

[0131] For 4D Flow MRI images after TAVR, image segmentation and 3D reconstruction algorithms can be used to extract the geometric model of the aortic root, including the left ventricle, ascending aorta, and prosthetic valve region.

[0132] Image segmentation algorithms that can be used include threshold segmentation, region growing, and deep learning.

[0133] Since the prosthetic valve stent exhibits significant signal loss in 4D Flow MRI images, this embodiment directly uses threshold segmentation to obtain the segmentation results of the prosthetic valve region. For regions that do not show signal loss in 4D Flow MRI images, this embodiment uses a combination of threshold segmentation and region growing to obtain the segmentation results of the left ventricle and ascending aorta.

[0134] Algorithms that can be used for 3D reconstruction include surface rendering based on segmentation results, direct volume rendering, and point cloud reconstruction.

[0135] In this embodiment, a three-dimensional reconstruction is performed using surface rendering based on the segmentation results of the left ventricle, ascending aorta, and artificial valve regions to obtain the geometric model of the aortic root.

[0136] Step 2.3: Determine the upstream and downstream boundary plane positions of the artificial valve region in a single phase of 4D Flow MRI images;

[0137] Because significant signal loss is observed in the prosthetic valve area, the upstream and downstream boundary planes should be located at the left ventricle and ascending aorta, respectively. Figure 5 As shown. Based on this and combined with prior anatomical knowledge, this embodiment sets the upstream and downstream boundary surfaces according to the following principles:

[0138] (1) Upstream boundary: Located 2-5 mm above the upper edge of the artificial valve area, perpendicular to the center line of the ascending aorta, and avoiding the coronary artery opening;

[0139] (2) Downstream boundary surface: located 2-5 mm below the lower edge of the artificial valve, perpendicular to the center line of the left ventricular outflow tract.

[0140] And generate a uniformly distributed set of spatial points on the determined boundary surface.

[0141] Step 2.4: Use image registration to obtain the upstream and downstream boundary plane positions of the artificial valve region in multiple phases of 4D Flow MRI images;

[0142] 4D Flow MRI images contain velocity information throughout the entire cardiac cycle. The heart and aorta as a whole will move, rotate and deform in different phases. Therefore, it is necessary to determine the upstream and downstream boundary surfaces at the same physical location in different phases.

[0143] Possible image registration methods include feature-point-based registration, surface-based registration, intensity-based registration, deep learning and reinforcement learning methods, multimodal image fusion methods, and template-based registration methods. The principles of each image registration method are introduced below:

[0144] ① Feature-point-based registration method: Registration is achieved by identifying and matching feature points in the image (such as the aortic sinus at the aortic root, the ascending aorta, etc.).

[0145] ② Surface-based registration methods: Registration is performed by matching surface features between two images, and is typically used for registration between 3D models. In 4D Flow MRI images, surface registration methods can be used to align aortic models (ascending aorta, aortic arch, and descending aorta) at different phases.

[0146] ③ Intensity-based registration method: Registration is achieved by comparing the grayscale or intensity information of the images, and it is usually used for the registration of images of different modalities.

[0147] ④ Deep learning and reinforcement learning methods: Utilize deep learning and reinforcement learning algorithms to automatically learn the registration relationship between images without manually labeling feature points.

[0148] ⑤ Multimodal image fusion method: Combining the advantages of multiple imaging modalities (such as CT, MRI, echocardiography, etc.), image fusion technology can be used to achieve a more comprehensive display of anatomical information. For details, please refer to "Xie Mingxing. (2022). Application of medical imaging artificial intelligence in interventional diagnosis and treatment of heart valve disease. Journal of Clinical Cardiovascular Diseases".

[0149] In this embodiment, a surface-based registration method is used to obtain the upstream and downstream boundary plane positions of the patient, such as... Figure 3 As shown.

[0150] Step 2.5: Obtain blood flow information at the boundary plane;

[0151] Specifically, for the spatial points on the boundary plane mentioned in step 2.3, the velocity-time curve for each spatial point over the entire cardiac cycle is extracted. The specific extraction process is as follows:

[0152] (1) Each spatial point on the boundary plane is mapped to the voxel data of each phase of the 4D Flow MRI image through the registration method in step 2.4, so as to obtain the voxel neighborhood of the point in each phase and the distance from the voxel center point.

[0153] (2) In each phase, the velocity voxel values ​​in the neighborhood of a spatial point are weighted and averaged to obtain the velocity vector of that point in that phase. Common weighting methods include voxel distance weighting and voxel similarity weighting.

[0154] ① Voxel distance weighting: The closer a voxel is to a spatial point, the greater its weight. Gaussian function, trilinear interpolation function, etc. can be used as weighting functions.

[0155] ② Voxel similarity weighting: The closer the gray value is to the gray value of the center of the neighborhood of a spatial point, the greater the weight of the voxel. The inverse of the gray value, the local correlation coefficient, etc. can be used as similarity measures.

[0156] This embodiment uses a voxel distance weighting method.

[0157] (3) Decompose the velocity vector obtained above according to the local coordinate system direction of the boundary plane to obtain axial (plane normal) and radial (plane tangential) velocity components. The former reflects the main flow change across the valve, and the latter reflects the secondary flow within the boundary plane.

[0158] (4) The obtained velocity component sequence is smoothed appropriately to eliminate the influence of noise. Commonly used methods include moving average, polynomial fitting, etc.

[0159] This embodiment uses the moving average method.

[0160] Step 3: Calculate the hemodynamic parameters of the artificial valve region using the complete geometric model of the aortic valve root reconstructed in Step 1 and the blood flow boundary obtained in Step 2.

[0161] Step 3.1: Use image registration to fuse post-TAVR CT and 4D Flow MRI images;

[0162] Anatomical landmarks such as the aortic root were selected, and spatial alignment was performed using a rigid + flexible registration algorithm, such as... Figure 6 As shown in the figure, the final registration transformation matrix is ​​applied to the valve model to achieve spatial consistency between MRI and CT image data, providing a unified coordinate basis for subsequent hemodynamic analysis.

[0163] Step 3.2, setting the artificial valve material;

[0164] Based on a review of relevant literature, the constitutive models and material parameters of the artificial petals and skirts were defined using the collected material data.

[0165] In this embodiment, the artificial leaflet adopts the Ogden model (an isotropic hyperelastic model based on principal stretch ratio), with shear modulus term 1 being 0.35 MPa, exponent term 1 being 15.2, shear modulus term 1 being -0.12 MPa, exponent term 1 being -10.8, and Poisson's ratio being 0.5.

[0166] Step 3.3, Boundary setting;

[0167] The upstream and downstream boundary planes of the artificial valve region selected in step 2.3 are used as the inlet and outlet boundaries for fluid simulation, respectively. The blood flow information at the boundary plane in step 2.5 is applied to the inlet / outlet boundary. The aortic root vessel wall obtained in step 2.2 is set as a no-slip wall boundary condition. The artificial valve surface reconstructed in step 1 is set as a no-slip wall boundary condition.

[0168] Step 3.4, Fluid-structure Interaction Calculation;

[0169] Step 3.4.1, fluid calculation.

[0170] Boolean operations are performed on the vascular region in the aortic root model obtained in step 2.2 and the artificial valve model region reconstructed in step 1 to remove the artificial valve model from the aortic root vascular region, and the remaining region is used as the fluid computation domain.

[0171] This example uses the finite volume method to discretize the computational domain. By integrating the three-dimensional incompressible Navier-Stokes equations within each finite volume element, the partial differential equations are transformed into a system of algebraic equations. For each control volume, Gauss's theorem is used to transform the terms into flux forms on the volume surface.

[0172] Based on the boundary conditions set in step 3.3, an iterative algorithm is used to solve the transformed algebraic equations to obtain hemodynamic parameters and wall force data.

[0173] Step 3.4.2, Solid Calculation.

[0174] In this example, the area where the artificial valve is bound to the skirt is fully constrained; the artificial valve leaflet uses surface-to-surface contact, and the contact attribute is hard contact.

[0175] The artificial valve region reconstructed in step 1 is discretized using the finite element method to perform solid mechanics calculations.

[0176] This embodiment uses the finite element method, which divides the artificial valve model into a finite number of elements and uses an interpolation function to approximate the displacement field inside the element.

[0177] The relationship between stress and strain is determined by the constitutive model. In this embodiment, the artificial leaflet material adopts the Ogden hyperelastic constitutive model, and its strain energy density function is:

[0178]

[0179] Where, μ i The shear modulus / material parameter of the i-th component determines the material's response strength to shear deformation; α i λ1, λ2, and λ3 are the power-law parameters of the i-th component, controlling the nonlinear sensitivity of strain energy to the stretch ratio; λ1, λ2, and λ3 are the principal stretch ratios, i.e., the ratios of tensile deformation of the material in the three principal directions.

[0180] The hemodynamic load obtained in step 3.4.1 is applied to the surface of the artificial valve. Combining the connection relationship and constitutive model, the momentum balance equation is transformed into a set of algebraic equations of nodal displacements using the finite element method. The stress, strain and deformation distribution of the valve structure are obtained through numerical solution.

[0181] Step 3.4.3, Information Interaction.

[0182] Through a computer interface, the force information of the artificial valve in fluid calculations and the displacement information of the artificial valve in solid calculations are exchanged, so that the fluid calculations and solid calculations can satisfy each other's solution conditions.

[0183] Step 3.5: Output the hemodynamic parameters of the prosthetic valve area;

[0184] Step 3.5.1: Output the three-dimensional transient velocity field data calculated in the fluid domain as text format data to obtain post-processable flow field information; the flow field information includes three-dimensional velocity field, three-dimensional pressure field, and wall shear stress.

[0185] Figure 7 This is a schematic diagram of the reconstructed flow field in the prosthetic valve region after TAVR surgery based on CT-MRI fusion. It shows the flow information at different time points. In the stent region, the velocity amplitude corresponds to the flow rate, indicating that the flow field in the prosthetic valve region after TAVR surgery can be reconstructed relatively completely and accurately based on CT-MRI fusion.

[0186] Step 3.5.2: Based on the fluid calculation results, the energy loss index (ELI) is obtained by calculating the energy change of blood flow before and after passing through the artificial valve. This includes the ELI value, the pressure and velocity distributions used in the calculation, a visualization of the energy loss distribution, and a comparative analysis with a normal valve.

[0187] The energy loss index is a hemodynamic indicator used to quantify the energy loss of blood flow through valves or stenotic sites. By comparing the pressure and velocity of the fluid before and after the valve, it quantitatively expresses the mechanical energy lost by blood flow through the valve, reflecting the valve's resistance characteristics to blood flow. The calculation formula based on Bernoulli's equation is as follows:

[0188]

[0189] Where p1 and p2 are the pressures before and after the artificial valve, respectively, ρ is the blood density, and v1 and v2 are the blood flow velocities before and after the artificial valve, respectively.

[0190] Step 4: Result Verification

[0191] To verify the accuracy and reliability of the CT-MRI fusion-based post-TAVR blood flow parameter completion method described in this invention, this step performs multi-dimensional result verification on the calculated hemodynamic parameters of the artificial valve region. Specifically, ① the hemodynamic parameters of a normal aortic root at a certain moment are reconstructed using the method provided in this application and compared with the actual hemodynamic parameters of the aortic root; ② the hemodynamic parameters of a normal aortic root at different moments are reconstructed using the method provided in this application and compared with the actual flow field information.

[0192] ① Comparison of hemodynamic calculation results;

[0193] Based on the method for obtaining the upstream and downstream blood flow boundaries of 4D Flow MRI in step 2, the blood flow information of the upstream and downstream boundary planes of 4D Flow MRI at a certain moment is extracted. Then, combined with the calculation method of hemodynamic parameters in the artificial valve region in step 3, the preoperative hemodynamic parameters of the aortic root are reconstructed, such as... Figure 8 As shown, the parameter information is compared with the 4D Flow MRI flow field data at that moment, and the comparison diagram is shown below. Figure 9 As shown, the aortic reconstructed flow field obtained by the method of this application is basically consistent with the 4D Flow MRI flow field, which verifies the accuracy and reliability of the blood flow parameter completion method of this invention.

[0194] ② Comparison of fluid-structure interaction calculation results;

[0195] like Figure 10 As shown, boundary condition information for fluid-structure interaction (FSI) calculations is obtained based on upstream and downstream blood flow boundary information from 4D Flow MRI at different time periods. This boundary condition information is applied to the FSI calculation domain, with the valve considered as the FSI interface, and FSI calculations are performed. The FSI at the same time point is compared with the corresponding 4D Flow MRI flow field information; a comparison diagram is shown below. Figure 11 As shown, the fluid-structure interaction at the same time point is basically consistent with the preoperative 4D Flow MRI flow field information, which verifies the accuracy and reliability of the blood flow parameter completion method of the present invention.

[0196] Some steps in the embodiments of the present invention can be implemented using software, and the corresponding software program can be stored in a readable storage medium, such as an optical disc or a hard disk.

[0197] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for completing blood flow parameters after TAVR based on CT-MRI fusion, characterized in that, The method includes: Step 1: Based on the CT images after TAVR, extract the complete geometric model of the aortic valve root after surgery; the complete geometric model of the aortic valve root after surgery includes the implanted artificial valve and the aortic valve root. Step 2: Based on the 4D Flow MRI images after TAVR surgery, extract the blood flow boundaries upstream and downstream of the prosthetic valve area; Step 3: Calculate the hemodynamic parameters of the artificial valve region using the complete geometric model of the aortic valve root reconstructed in Step 1 and the blood flow boundary obtained in Step 2; Step 1 includes: Step 1.1: Obtain CT images after TAVR surgery; Step 1.2: Extract the geometric model of the aortic root based on CT image data after TAVR surgery; Step 1.3: Obtain the geometric model of the artificial valve in the patient's body based on the CT image data after TAVR surgery; Step 1.4: Combine the geometric model of the aortic root with the geometric model corresponding to the artificial valve in the patient's body to obtain the complete geometric model of the aortic valve root after surgery. Step 1.3 includes: Step 1.3.1: Establish a three-dimensional geometric model of the artificial valve in its natural state and generate the corresponding discretized model; Step 1.3.2: Obtain the displacement of the valve frame from the initial state to the deformed state, where the initial state corresponds to the natural state and the deformed state corresponds to the state within the patient's body; Step 1.3.3: Obtain the deformation of the leaflet and skirt under the deformed valve frame through the non-uniform shape function field reconstruction algorithm, and obtain the three-dimensional discrete model of the deformed valve, and then obtain the geometric model of the deformed valve, that is, the geometric model corresponding to the artificial valve in the patient's body. Step 1.3.3 includes: The first step is to define the material parameters for the artificial petals and skirt; The second step is to construct the connection positions and methods of the valve frame, artificial leaflets, and skirt according to the connection method of the actual artificial valve; The third step is to define the contact type and contact attributes; The fourth step, based on the principle of minimum potential energy, is to derive the stiffness matrix and equivalent nodal force vector of each discrete region in the three-dimensional discrete model of the artificial valve; assemble all discrete regions according to nodal degrees of freedom to form the global stiffness matrix [K] and the global equivalent nodal load vector {F}; and solve the linear / nonlinear system equations [K]{u} using the direct solution method / iterative solution method. e }[N]={F}, obtain the updated global node displacement vector {U}; [N] represents the shape function matrix, u e This represents the displacement of each discrete region; Fifth step, based on the updated discrete region displacement vector {U}, use the equation This allows for the accurate reconstruction of a three-dimensional discrete model of the deformed valve in vivo. The sixth step is to obtain the geometric model of the deformable valve based on the three-dimensional discrete model of the deformable valve, which is the geometric model corresponding to the artificial valve in the patient's body.

2. The method according to claim 1, characterized in that, Step 2 includes: Step 2.1: Obtain 4D flow MRI images after TAVR surgery; Step 2.2: Extract the geometric model of the aortic root based on the 4D Flow MRI image data after TAVR; Step 2.3: Determine the upstream and downstream boundary plane positions of the artificial valve region in a single phase of 4D Flow MRI images; Step 2.4: Use image registration to obtain the upstream and downstream boundary plane positions of the artificial valve region in multiple phases of 4D Flow MRI images; Step 2.5: Obtain blood flow information at the boundary plane.

3. The method according to claim 2, characterized in that, Step 3 includes: Step 3.1: Use image registration to fuse post-TAVR CT and 4D Flow MRI images; Step 3.2, select the artificial valve material; Step 3.3, set the boundaries; Step 3.4, Fluid-structure Interaction Calculation; Step 3.5: Output the hemodynamic parameters of the artificial valve area, including the energy loss index (ELI) value, the pressure and velocity distribution on which the calculation was based, the energy loss distribution visualization, and the comparative analysis results with normal valves.

4. The method according to claim 3, characterized in that, Step 3.4 includes: Step 3.4.1, Fluid Calculation: Boolean operations are performed on the vascular region of the aortic root model obtained in Step 2.2 and the artificial valve model region reconstructed in Step 1. The artificial valve model is removed from the aortic root vascular region, and the remaining region is used as the fluid computation domain. The computation domain is discretized using the finite volume method. By integrating the three-dimensional incompressible Navier-Stokes equations within each finite volume element, the partial differential equations are transformed into a system of algebraic equations. For each control volume, Gauss's theorem is used to transform the terms into flux forms on the volume surface. Based on the boundary conditions set in Step 3.3, an iterative algorithm is used to solve the transformed algebraic equations to obtain hemodynamic parameters and wall force data. Step 3.4.2, solid calculation: The artificial valve model is divided into a finite number of elements using the finite element method. The displacement field inside the element is approximated by the interpolation function. The hemodynamic load obtained in step 3.4.1 is applied to the surface of the artificial valve. Combining the connection relationship and constitutive model, the momentum balance equation is transformed into a set of algebraic equations of nodal displacements using the finite element method. The stress, strain and deformation distribution of the valve structure are obtained through numerical solution. Step 3.4.3, Information Interaction: The force information of the artificial valve in the fluid calculation and the displacement information of the artificial valve in the solid calculation are exchanged to ensure that the fluid calculation and the solid calculation satisfy the solution conditions.

5. The method according to claim 4, characterized in that, Step 3.5 includes: Step 3.5.1: Output the three-dimensional transient velocity field data calculated in the fluid domain as text format data to obtain post-processable flow field information; the flow field information includes three-dimensional velocity field, three-dimensional pressure field, and wall shear stress; Step 3.5.2: Based on the fluid calculation results, the energy loss index is obtained by calculating the energy change of blood flow before and after passing through the artificial valve, including the ELI value, the pressure and velocity distribution on which the calculation is based, the visualization of the energy loss distribution, and the comparative analysis with normal valves.

Images