A method, system, device and medium for determining coronary artery blood flow reserve fraction

Abstract

The present invention discloses a method, system, device, and medium for determining the fractional flow reserve (FFR) of a coronary artery. This method uses a fully connected network model to build a network model for determining the FFR distal to a stenosis. The model extracts the geometric features of the blood vessels from acquired IVUS or IVOCT image sequences and calculates the pressure loss across the stenosis. The geometric feature parameter set of the target vessel segment and the calculated pressure loss across the stenosis are input into the trained network model, which then outputs the FFR distal to the stenosis of the target vessel segment. The method utilizes deep learning technology, combining data-driven analysis with physical knowledge to integrate coronary artery physiology and morphology, thereby improving the accuracy of physiological assessments of coronary artery lesions.

Description

Technical Field

[0001] The present invention relates to the field of medical imaging technology, and in particular to a method, system, device and medium for determining coronary artery blood flow reserve fraction. Background Art

[0002] Coronary artery (abbreviated as coronary artery) Fractional Flow Reserve (FFR) is an internationally recognized physiological indicator for evaluating stable ischemic coronary artery stenosis. The flow reserve fraction is defined as the ratio of the maximum blood flow in the stenotic artery to the maximum blood flow in the same artery under normal conditions (i.e., when there is no stenosis). When the coronary resistance is minimal and relatively constant, the pressure in the lumen is linearly related to the blood flow. Therefore, the calculation of the flow reserve fraction can be simplified as the average pressure at the distal end of the coronary stenosis under maximum hyperemia (P d ) and the mean pressure at the aortic root or coronary artery ostium (P a The normal value of coronary flow reserve fraction is 1, and a flow reserve fraction less than or equal to 0.8 is usually defined as a stenotic lesion.

[0003] The current gold standard for clinical measurement of blood flow reserve fraction is to use a pressure guidewire to measure the pressure gradient across the stenotic lesion. This method has the following limitations: there is a risk of vascular damage; it significantly increases diagnostic costs and prolongs the interventional procedure time; it requires intravenous injection of vasodilators such as adenosine or adenosine triphosphate to induce congestion, and has high requirements for drug dosage and administration method. The patient's poor response and tolerance to the drug will affect the measurement accuracy. Patients with asthma, hypotension, or atrioventricular block are not suitable for the use of vasodilators.

[0004] Coronary imaging is an important adjunct to the clinical diagnosis of coronary artery stenosis, assessment of lesion severity, and treatment planning. In recent years, methods for calculating physiological indices based on coronary imaging have become a research hotspot in this field. Depending on the image data source used, the main solutions for estimating fractional flow reserve based on coronary imaging are categorized as noninvasive and invasive: using noninvasive computed tomography (CT) images and invasive quantitative coronary angiography (QCA) images to estimate fractional flow reserve. Because CT-based methods involve numerical simulation of hemodynamics, the overall computational time is very consuming, typically requiring dedicated computational fluid dynamics (CFD) software and supercomputers. Furthermore, these methods require high accuracy in 3D vascular reconstruction. When complex calcified lesions are present, it is difficult to accurately segment and reconstruct the 3D morphology of the diseased vessels from CT images. As an invasive imaging method, QCA has greater advantages than non-invasive CT in displaying the coronary lumen structure and lesion morphology. Therefore, the blood flow reserve fraction calculated based on QCA images is more accurate than the blood flow reserve fraction estimated based on CT images. The disadvantage is that QCA can only display the two-dimensional projection of the vascular lumen along the long axis, and there are problems such as projection overlap and perspective foreshortening. Since it is impossible to display the structure of the vascular lumen cross-section, it is generally assumed that the lumen cross-section is circular or elliptical when performing three-dimensional reconstruction of the blood vessels. In fact, when coronary artery stenosis occurs, the shape of the lumen is usually complex and diverse, and the stenosis is often eccentric and irregular. Therefore, the vascular lumen reconstructed based on this assumption cannot reflect the true anatomical morphology, which in turn affects the accuracy of the blood flow reserve fraction estimated on this basis.

[0005] Compared to QCA, intravascular ultrasound (IVUS) and intravascular optical coherence tomography (IVOCT) not only demonstrate lumen morphology but also reveal the cross-sectional structure of vessels, including plaques, vessel wall thickness, and plaque morphology. These imaging techniques are clinically valuable for diagnosing atherosclerotic lesions, guiding stent placement, and evaluating surgical outcomes. Estimating coronary artery physiology based on IVUS or IVOCT image sequences promises to enable accurate physiological assessment of vascular stenosis. Summary of the Invention

[0006] The purpose of the present invention is to provide a method, system, device and medium for determining coronary blood flow reserve fraction, which can improve the accuracy of physiological assessment of coronary artery lesions.

[0007] To achieve the above object, the present invention provides the following solutions:

[0008] A method for determining coronary artery blood flow reserve fraction, comprising:

[0009] Batch build vascular simulation models of the left anterior descending coronary artery with disease;

[0010] Performing steady-state hemodynamic simulation on each vascular simulation model to obtain a simulation value of the blood flow reserve fraction of each vascular simulation model;

[0011] Determining a set of geometric characteristic parameters of each blood vessel simulation model and a calculated value of pressure loss across a narrow region;

[0012] The geometric characteristic parameter set of each vascular simulation model and the calculated value of the pressure loss across the stenosis area are taken as input, and the simulated value of the blood flow reserve fraction of each vascular simulation model is taken as output to construct a data set;

[0013] A network model for determining the distal blood flow reserve fraction of stenosis was built based on a fully connected network model;

[0014] Using the data set to train the network model to obtain a trained network model;

[0015] Performing image segmentation on the acquired intravascular ultrasound image sequence or intravascular optical coherence tomography image sequence, and extracting the lumen and wall contours of the target vascular segment, thereby determining a set of geometric feature parameters of the target vascular segment and a calculated value of the pressure loss across the stenosis area;

[0016] The geometric characteristic parameter set of the target vascular segment and the calculated value of the blood flow reserve fraction are input into the trained network model, and the blood flow reserve fraction at the distal end of the stenosis of the target vascular segment is output.

[0017] Optionally, performing steady-state hemodynamic simulation on each vascular simulation model to obtain a simulation value of the blood flow reserve fraction of each vascular simulation model specifically includes:

[0018] The blood vessel simulation model is meshed using a hexahedral mesh, and the boundary layer of the blood vessel simulation model is set to three layers;

[0019] Set the fluid material and boundary conditions; the fluid material set includes setting blood as an incompressible Newtonian fluid with a density of 1050 kg / m 3 , the viscosity is 0.0035 Pa·s, the boundary conditions set include assuming that the blood vessel wall is an impermeable rigid wall, the blood vessel wall adopts a no-slip condition, the inlet boundary condition is set to a fixed congestion flow rate of 0.35 m / s, and the outlet boundary condition is set to 0 Pa;

[0020] Perform steady-state hemodynamic simulation on the vascular simulation model with set fluid materials and boundary conditions, and extract the simulated pressure drop ΔP of the stenosis segment CFD ;

[0021] Assume that the pressure at the blood vessel entrance is P a is 80 mmHg, using the formula Calculate the simulated numerical value FFR of the blood flow reserve fraction of the vascular simulation model CFD .

[0022] Optionally, determining the geometric characteristic parameter set of each blood vessel simulation model and the calculated value of the pressure loss across the stenosis area specifically includes:

[0023] A series of two-dimensional cross-sectional slices were obtained by sampling at the same interval along the lumen centerline of the vascular simulation model;

[0024] Extracting 10 geometric feature parameters from all two-dimensional cross-sectional slices to form a geometric feature parameter set for the vascular simulation model; the geometric feature parameter set includes the minimum lumen cross-sectional area, the proximal reference lumen cross-sectional area, the distal reference lumen cross-sectional area, the percentage of stenosis area, the lesion length, the entrance length, the exit length, the distance to the lesion, the lumen eccentricity at the minimum lumen area, and the average distance between the centers of the lumen and the media at the maximum stenosis;

[0025] Using the formula ΔP=FV+SV 2 , calculate the pressure loss ΔP across the stenosis area of ​​the vascular simulation model; where F and S are the local pressure loss coefficients caused by viscous friction and flow separation, respectively, and V is the congestion flow velocity.

[0026] Optionally, the calculation formula for the local pressure loss coefficient caused by viscous friction and flow separation is:

[0027]

[0028]

[0029] Where μ is the absolute blood viscosity, ρ is the blood density, L is the length of the stenosis, and A is the n is the reference lumen cross-sectional area, A s is the cross-sectional area of ​​the lumen of the stenosis segment.

[0030] Optionally, the network model for determining the distal blood flow reserve fraction of stenosis comprises: an input layer, a hidden layer, and an output layer;

[0031] The input layer has 11 variables, the hidden layer has 4 layers, and the output layer has 1;

[0032] The number of neurons in the four hidden layers is 256, 64, 16, and 4, respectively;

[0033] Each neuron in the input layer and hidden layer is connected to every neuron in the next layer, without using convolutional layers;

[0034] The linear rectification function is used as the activation function between the input and output of each layer of the input layer, hidden layer and output layer.

[0035] Optionally, using the data set to train the network model to obtain a trained network model specifically includes:

[0036] All samples in the dataset are randomly shuffled and divided into training set, validation set and test set in a ratio of 6:2:2;

[0037] Each layer in the network model is pre-trained as an autoencoder, and all weights are randomly initialized using the Xavier initialization method;

[0038] Import the training set, validation set, and test set through the pandas module;

[0039] Perform Min-max normalization on the sample data in the training set, validation set, and test set;

[0040] Based on the normalized training set, validation set, and test set, the mean square error loss function and stochastic gradient descent algorithm are used to adaptively optimize the learning parameters of the initialized network model and train the network model;

[0041] Output the trained network model.

[0042] A system for determining coronary artery blood flow reserve fraction, comprising:

[0043] A simulation model building module is used to batch build vascular simulation models of the left anterior descending coronary artery with disease;

[0044] A simulation module, used for performing steady-state hemodynamic simulation on each vascular simulation model to obtain a simulation value of the blood flow reserve fraction of each vascular simulation model;

[0045] a parameter determination module, used to determine a set of geometric characteristic parameters of each blood vessel simulation model and a calculated value of pressure loss across a narrow region;

[0046] a data set construction module, configured to take the geometric characteristic parameter set of each vascular simulation model and the calculated value of the pressure loss across the stenosis area as input, and output the simulated value of the blood flow reserve fraction of each vascular simulation model to construct a data set;

[0047] A model building module, used to build a network model for determining the distal blood flow reserve fraction of stenosis based on a fully connected network model;

[0048] A training module, configured to train the network model using the data set to obtain a trained network model;

[0049] an acquisition module, configured to perform image segmentation on the acquired intravascular ultrasound image sequence or intravascular optical coherence tomography image sequence, and extract the lumen and wall contours of the target vascular segment, thereby determining a set of geometric characteristic parameters of the target vascular segment and a calculated value of the pressure loss across the stenosis region;

[0050] The application module is used to input the geometric characteristic parameter set of the target blood vessel segment and the calculated value of the pressure loss across the stenosis area into the trained network model, and output the blood flow reserve fraction at the distal end of the stenosis of the target blood vessel segment.

[0051] An electronic device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the aforementioned method for determining coronary artery blood flow reserve fraction when executing the computer program.

[0052] A computer-readable storage medium stores a computer program, which, when executed, implements the aforementioned method for determining coronary artery blood flow reserve fraction.

[0053] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0054] The present invention discloses a method, system, device, and medium for determining the fractional flow reserve (FFR) of a coronary artery. This method uses a fully connected network model to build a network model for determining the FFR distal to a stenosis. The model extracts the geometric features of the blood vessels from acquired IVUS or IVOCT image sequences and calculates the pressure loss across the stenosis. The geometric feature parameter set of the target vessel segment and the calculated pressure loss across the stenosis are input into the trained network model, which then outputs the FFR distal to the stenosis of the target vessel segment. The method utilizes deep learning technology, combining data-driven analysis with physical knowledge to integrate coronary artery physiology and morphology, thereby improving the accuracy of physiological assessments of coronary artery lesions. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0056] Figure 1 A flowchart of a method for determining coronary artery blood flow reserve fraction provided by an embodiment of the present invention;

[0057] Figure 2 A schematic diagram of a method for determining coronary artery blood flow reserve fraction provided by an embodiment of the present invention;

[0058] Figure 3 A three-dimensional view of a simulated blood vessel model of the left anterior descending coronary artery provided in an embodiment of the present invention;

[0059] Figure 4 A cross-sectional view of a simulated blood vessel model of the left anterior descending coronary artery provided in an embodiment of the present invention;

[0060] Figure 5 A schematic diagram of the network model structure for determining the distal blood flow reserve fraction of stenosis provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0062] The purpose of the present invention is to provide a method, system, device and medium for determining coronary blood flow reserve fraction, which can improve the accuracy of physiological assessment of coronary artery lesions.

[0063] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0064] like Figure 1 As shown, an embodiment of the present invention provides a method for determining coronary artery blood flow reserve fraction, comprising:

[0065] Step 1: Batch build vascular simulation models of the diseased left anterior descending coronary artery.

[0066] According to the human coronary artery anatomical parameters and stenosis parameters (such as vessel segment length, entrance lumen diameter, stenosis area percentage, etc.), batch build the simulation model of the left anterior descending coronary artery with disease, for example Figure 3 and Figure 4 The anatomical parameters and stenosis parameter ranges of the left anterior descending artery are listed in Table 1.

[0067] Table 1 Anatomical parameters and stenosis parameter ranges of the left anterior descending artery

[0068]

[0069]

[0070] Step 2: Perform steady-state hemodynamic simulation on each vascular simulation model to obtain a simulation value of the blood flow reserve fraction of each vascular simulation model.

[0071] CFD technology was used to calculate the blood flow reserve fraction of the vascular simulation model. The simulated vascular model was imported into the Fluent module of the finite element analysis software ANSYS Workbench (ANSYS Inc., USA), and steady-state hemodynamic simulation was performed by solving the Navier-Stokes equation. The specific method is as follows: first, the three-dimensional vascular model was meshed using regular hexahedral meshing units. In order to improve the accuracy of the boundary layer calculation near the wall, the boundary layer was set to three layers; then, the fluid material and boundary conditions were set, and the blood was set to an incompressible Newtonian fluid with a density of 1050 kg / m 3 , the viscosity is 0.0035 Pa·s, the vascular wall is assumed to be an impermeable rigid wall, the vascular wall adopts a no-slip condition, the inlet boundary condition is set to a fixed congestion velocity of 0.35 m / s, and the outlet boundary condition is set to 0 Pa; finally, a fluid dynamics simulation is performed to extract the simulated pressure drop ΔP of the stenotic segment CFD Assuming that the pressure Pa at the blood vessel entrance is 80 mmHg, the hemodynamic simulation value of the blood flow reserve fraction is calculated according to the following formula, recorded as FFR CFD :

[0072]

[0073] Step 3: Determine the geometric characteristic parameter set of each blood vessel simulation model and the calculated value of the pressure loss across the stenosis area.

[0074] A series of two-dimensional cross-sectional slices were obtained by sampling at equal intervals along the lumen centerline of the simulated vascular model. Geometric characteristic parameters such as the lumen area and the lumen eccentricity at the maximum stenosis were extracted from each slice. Ten geometric features related to the hemodynamic characteristics of the measured vascular segment were then obtained, as listed in Table 2.

[0075] Table 2 Geometric characteristics

[0076]

[0077]

[0078] In Table 2, the average reference lumen area = (proximal reference lumen area + distal reference lumen area) / 2; the proximal reference point is the closest healthy vessel section along the coronary artery close to the heart at the stenosis; the distal reference point is the closest healthy vessel section along the coronary artery away from the heart at the stenosis.

[0079] The pressure loss across the narrow region, denoted as ΔP, is calculated using the following simplified fluid dynamics equation:

[0080] ΔP=FV+SV 2 (2)

[0081] Where V is a fixed hyperemic flow velocity of 0.35 m / s, F and S are the local pressure loss coefficients caused by viscous friction and flow separation, respectively, which are determined by the morphology of the stenotic segment and blood parameters:

[0082]

[0083]

[0084] Where μ is the absolute blood viscosity, ρ is the blood density, both of which are the average physiological values ​​of normal adults, L is the length of the stenosis segment, and A n is the reference lumen cross-sectional area, A s is the cross-sectional area of ​​the lumen of the stenotic segment.

[0085] Step 4: The geometric characteristic parameter set of each vascular simulation model and the calculated value of the pressure loss across the stenosis area are taken as input, and the simulated value of the blood flow reserve fraction of each vascular simulation model is taken as output to construct a data set.

[0086] Steps 1 to 4 use computer simulation methods to construct a deep learning dataset for calculating blood flow reserve fraction based on coronary intravascular image sequences.

[0087] Step 5: Build a network model based on the fully connected network model to determine the distal blood flow reserve fraction of stenosis.

[0088] like Figure 5 As shown in Figure 1, FFR-Net is built based on a fully connected network model, in which each neuron in each layer is connected to each neuron in the next layer, and no convolutional layer is used.

[0089] The hidden layers for deep learning were set up. Using the Dense layer in Keras, the input layer was defined as 11 variables (i.e., the 10 structural variables and ΔP in Table 2), the output layer was one (i.e., the fractional flow reserve at the distal end of the stenosis), and the hidden layer was set to four layers (containing 256, 64, 16, and 4 hidden nodes, respectively). The linear rectification function (ReLU) was used as the activation function between the input and output of each layer.

[0090] Step 6: Use the data set to train the network model to obtain a trained network model.

[0091] Step 6-1 Initialize the network parameters. Each layer is pre-trained as an autoencoder, and all weights are randomly initialized using the Xavier initialization method.

[0092] Step 6-2 imports the training set, validation set, and test set through the pandas module.

[0093] Step 6-3 performs Min-max normalization on the sample data, imports it into the constructed network model, and trains the network.

[0094] Step 6-4: Optimize network parameters. Based on the normalized training set, validation set, and test set, the network's learning parameters (such as learning rate, momentum, etc.) are adaptively optimized using the mean square error loss function and the stochastic gradient descent algorithm.

[0095] Step 6-5 outputs the trained network model.

[0096] Step 7: Perform image segmentation on the acquired intravascular ultrasound image sequence or intravascular optical coherence tomography image sequence, and extract the lumen and wall contours of the target vascular segment, thereby determining the geometric feature parameter set of the target vascular segment and the calculated value of the pressure loss across the stenosis area.

[0097] Step 8: The geometric characteristic parameter set of the target vascular segment and the calculated value of the pressure loss across the stenosis area are input into the trained network model, and the blood flow reserve fraction at the distal end of the stenosis of the target vascular segment is output.

[0098] like Figure 2 As shown in Figure 2, image segmentation is performed on clinically acquired IVUS or IVOCT image sequences to extract the lumen and wall contours. The 10 geometric features listed in Table 2 are then obtained through geometric measurement, and ΔP is calculated according to equations (2)-(4). The 10 geometric feature data and ΔP are input into the trained FFR-Net, which ultimately outputs the blood flow reserve fraction at the distal end of the stenosis of the target vessel segment.

[0099] The present invention first builds a fully connected network model, named FFR-Net; then, constructs a simulation dataset and uses it to train FFR-Net and optimize network parameters; finally, the geometric features of the blood vessels are extracted from clinically collected IVUS or IVOCT image sequences, and a simplified fluid dynamics equation is used to calculate the pressure loss across the stenotic area, recorded as ΔP. ​​The geometric features of the blood vessels and ΔP are input into the trained FFR-Net, and the network outputs the blood flow reserve fraction distal to the stenotic lesion.

[0100] The present invention adopts deep learning technology, combines data-driven with physical knowledge, integrates the physiological and morphological information of coronary arteries, improves the accuracy of physiological assessment of coronary artery lesions and reduces its cost.

[0101] An embodiment of the present invention further provides a system for determining coronary artery blood flow reserve fraction, comprising:

[0102] A simulation model building module is used to batch build vascular simulation models of the left anterior descending coronary artery with disease;

[0103] A simulation module, used for performing steady-state hemodynamic simulation on each vascular simulation model to obtain a simulation value of the blood flow reserve fraction of each vascular simulation model;

[0104] a parameter determination module, used to determine a set of geometric characteristic parameters of each blood vessel simulation model and a calculated value of pressure loss across a narrow region;

[0105] a data set construction module, configured to take the geometric characteristic parameter set of each vascular simulation model and the calculated value of the pressure loss across the stenosis area as input, and output the simulated value of the blood flow reserve fraction of each vascular simulation model to construct a data set;

[0106] A model building module, used to build a network model for determining the distal blood flow reserve fraction of stenosis based on a fully connected network model;

[0107] A training module, configured to train the network model using the data set to obtain a trained network model;

[0108] an acquisition module, configured to perform image segmentation on the acquired intravascular ultrasound image sequence or intravascular optical coherence tomography image sequence, and extract the lumen and wall contours of the target vascular segment, thereby determining a set of geometric characteristic parameters of the target vascular segment and a calculated value of the pressure loss across the stenosis region;

[0109] The application module is used to input the geometric characteristic parameter set of the target blood vessel segment and the calculated value of the pressure loss across the stenosis area into the trained network model, and output the blood flow reserve fraction at the distal end of the stenosis of the target blood vessel segment.

[0110] The system for determining the coronary blood flow reserve fraction provided in the embodiment of the present invention has similar working principles and beneficial effects to the method for determining the coronary blood flow reserve fraction described in the above embodiment, and therefore will not be described in detail here. For details, please refer to the introduction of the above method embodiment.

[0111] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the aforementioned method for determining coronary artery blood flow reserve fraction when executing the computer program.

[0112] In addition, when the computer program in the above-mentioned memory is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk.

[0113] Furthermore, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the aforementioned method for determining coronary artery blood flow reserve fraction when executed.

[0114] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0115] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A method for determining coronary artery blood flow reserve fraction, characterized in that: include: Batch build vascular simulation models of the left anterior descending coronary artery with disease; Perform steady-state hemodynamic simulation on each vascular simulation model to obtain a simulated value of the blood flow reserve fraction of each vascular simulation model; specifically, the following steps are involved: The blood vessel simulation model is meshed using a hexahedral mesh, and the boundary layer of the blood vessel simulation model is set to three layers; Set the fluid material and boundary conditions; the fluid material setting includes setting blood as an incompressible Newtonian fluid with a density of 1050 kg / m 3 , the viscosity is 0.0035 Pa·s; the boundary conditions set include assuming that the blood vessel wall is an impermeable rigid wall, the blood vessel wall adopts a no-slip condition, the inlet boundary condition is set to a fixed congestion flow rate of 0.35 m / s, and the outlet boundary condition is set to 0 Pa; Perform steady-state hemodynamic simulation on the vascular simulation model with set fluid materials and boundary conditions to extract the simulated pressure drop of the stenosis segment ; Assume that the pressure at the blood vessel entrance is is 80 mmHg, using the formula , calculate the simulation value of the blood flow reserve fraction of the vascular simulation model ; Determine the geometric characteristic parameter set of each blood vessel simulation model and the calculated value of the pressure loss across the stenosis area; specifically including: A series of two-dimensional cross-sectional slices were obtained by sampling at equal intervals along the lumen centerline of the vascular simulation model; Extracting 10 geometric feature parameters from all two-dimensional cross-sectional slices to form a geometric feature parameter set for the vascular simulation model; the geometric feature parameter set includes the minimum lumen cross-sectional area, the proximal reference lumen cross-sectional area, the distal reference lumen cross-sectional area, the percentage of stenosis area, the lesion length, the entrance length, the exit length, the distance to the lesion, the lumen eccentricity at the minimum lumen area, and the average distance between the centers of the lumen and the media at the maximum stenosis; Using the formula , calculate the pressure loss across the narrow area of ​​the blood vessel simulation model Where, and are the local pressure loss coefficients caused by viscous friction and flow separation, , ; is the absolute blood viscosity, is the blood density, is the length of the stenosis segment, is the reference lumen cross-sectional area, is the cross-sectional area of ​​the lumen of the stenosis segment; is the congestion flow rate; The geometric characteristic parameter set of each vascular simulation model and the calculated value of the pressure loss across the stenosis area are taken as input, and the simulated value of the blood flow reserve fraction of each vascular simulation model is taken as output to construct a data set; A network model for determining the distal blood flow reserve fraction of stenosis was built based on a fully connected network model; Using the data set to train the network model to obtain a trained network model; Performing image segmentation on the acquired intravascular ultrasound image sequence or intravascular optical coherence tomography image sequence, and extracting the lumen and wall contours of the target vascular segment, thereby determining a set of geometric feature parameters of the target vascular segment and a calculated value of the pressure loss across the stenosis area; The geometric characteristic parameter set of the target blood vessel segment and the calculated value of the pressure loss across the stenosis area are input into the trained network model, and the blood flow reserve fraction at the distal end of the stenosis of the target blood vessel segment is output.

2. The method for determining coronary artery blood flow reserve fraction according to claim 1, wherein: The network model for determining the distal blood flow reserve fraction of stenosis includes: an input layer, a hidden layer and an output layer; The input layer has 11 variables, the hidden layer has 4 layers, and the output layer has 1; The number of neurons in the four hidden layers is 256, 64, 16, and 4, respectively; Each neuron in the input layer and hidden layer is connected to every neuron in the next layer, without using convolutional layers; The linear rectification function is used as the activation function between the input and output of each layer of the input layer, hidden layer and output layer.

3. The method for determining coronary artery blood flow reserve fraction according to claim 2, wherein: Using the data set to train the network model to obtain a trained network model specifically includes: All samples in the dataset are randomly shuffled and divided into training set, validation set and test set in a ratio of 6:2:2; Each layer in the network model is pre-trained as an autoencoder, and all weights are randomly initialized using the Xavier initialization method; Import the training set, validation set, and test set through the pandas module; Perform Min-max normalization on the sample data in the training set, validation set, and test set; Based on the normalized training set, validation set, and test set, the mean square error loss function and stochastic gradient descent algorithm are used to adaptively optimize the learning parameters of the initialized network model and train the network model; Output the trained network model.

4. A system for determining coronary artery blood flow reserve fraction, characterized in that: include: A simulation model building module is used to batch build vascular simulation models of the left anterior descending coronary artery with disease; The simulation module is used to perform steady-state hemodynamic simulation on each vascular simulation model to obtain a simulation value of the blood flow reserve fraction of each vascular simulation model; specifically, it includes: The blood vessel simulation model is meshed using a hexahedral mesh, and the boundary layer of the blood vessel simulation model is set to three layers; Set the fluid material and boundary conditions; the fluid material setting includes setting blood as an incompressible Newtonian fluid with a density of 1050 kg / m 3 , the viscosity is 0.0035 Pa·s; the boundary conditions set include assuming that the blood vessel wall is an impermeable rigid wall, the blood vessel wall adopts a no-slip condition, the inlet boundary condition is set to a fixed congestion flow rate of 0.35 m / s, and the outlet boundary condition is set to 0 Pa; Perform steady-state hemodynamic simulation on the vascular simulation model with set fluid materials and boundary conditions to extract the simulated pressure drop of the stenosis segment ; Assume that the pressure at the blood vessel entrance is is 80 mmHg, using the formula , calculate the simulation value of the blood flow reserve fraction of the vascular simulation model ; The parameter determination module is used to determine the geometric characteristic parameter set of each blood vessel simulation model and the calculated value of the pressure loss across the stenosis area; specifically, it includes: A series of two-dimensional cross-sectional slices were obtained by sampling at equal intervals along the lumen centerline of the vascular simulation model; Extracting 10 geometric feature parameters from all two-dimensional cross-sectional slices to form a geometric feature parameter set for the vascular simulation model; the geometric feature parameter set includes the minimum lumen cross-sectional area, the proximal reference lumen cross-sectional area, the distal reference lumen cross-sectional area, the percentage of stenosis area, the lesion length, the entrance length, the exit length, the distance to the lesion, the lumen eccentricity at the minimum lumen area, and the average distance between the centers of the lumen and the media at the maximum stenosis; Using the formula , calculate the pressure loss across the narrow area of ​​the blood vessel simulation model Where, and are the local pressure loss coefficients caused by viscous friction and flow separation, , ; is the absolute blood viscosity, is the blood density, is the length of the stenosis segment, is the reference lumen cross-sectional area, is the cross-sectional area of ​​the lumen of the stenosis segment; is the congestion flow rate; a data set construction module, configured to take the geometric characteristic parameter set of each vascular simulation model and the calculated value of the pressure loss across the stenosis area as input, and output the simulated value of the blood flow reserve fraction of each vascular simulation model to construct a data set; A model building module, used to build a network model for determining the distal blood flow reserve fraction of stenosis based on a fully connected network model; A training module, configured to train the network model using the data set to obtain a trained network model; an acquisition module, configured to perform image segmentation on the acquired intravascular ultrasound image sequence or intravascular optical coherence tomography image sequence, and extract the lumen and wall contours of the target vascular segment, thereby determining a set of geometric characteristic parameters of the target vascular segment and a calculated value of the pressure loss across the stenosis region; The application module is used to input the geometric characteristic parameter set of the target blood vessel segment and the calculated value of the pressure loss across the stenosis area into the trained network model, and output the blood flow reserve fraction at the distal end of the stenosis of the target blood vessel segment.

5. An electronic device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method for determining the coronary blood flow reserve fraction according to any one of claims 1 to 3 when executing the computer program.

6. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed, the method for determining the coronary blood flow reserve fraction according to any one of claims 1 to 3 is implemented.

Images