Method for training prediction model for predicting blood vessel deformation after stent implantation

Abstract

The invention discloses a method for training a prediction model for predicting blood vessel deformation after stent implantation. Comprising the following steps: respectively performing three-dimensional reconstruction processing on acquired pre-operation brain image data before stent implantation and post-operation brain image data after stent implantation to obtain a pre-operation blood vessel model and a post-operation blood vessel model; mapping the pose information of the stent in the postoperative craniocerebral image data to the preoperative blood vessel model to generate a stent model; constructing a target vascular graph structure based on the preoperative vascular model and the stent model, and constructing a target stent graph structure based on the stent model; a first node on the target vascular graph structure carries first physical coding information; a second node on the target support graph structure carries second physical coding information; and inputting the target blood vessel graph structure, the target stent graph structure and the postoperative blood vessel model into a graph neural network model for training until a set training ending condition is reached, thereby obtaining a prediction model. The rapid and accurate prediction of the blood vessel deformation after the stent is implanted is realized.

Description

Technical Field

[0001] The present disclosure generally relates to the field of artificial intelligence and, more specifically, to a method for training a prediction model for predicting vascular deformation after stent implantation. Background Art

[0002] With the widespread use of vascular interventional therapy for vascular diseases (e.g., intracranial aneurysms, intracranial vascular stenosis, etc.), stent implantation in blood vessels has become a common clinical treatment method. After implantation, the stent will cause local expansion and reconstruction of the blood vessel, thereby causing deformation of the blood vessel wall. In actual stent implantation surgery, before implanting the stent in the blood vessel, it is necessary to predict the deformation of the blood vessel after the stent is implanted in the blood vessel, which can assist doctors in selecting the best stent implantation plan during surgery.

[0003] Currently, finite element methods (FEM) or other numerical simulation methods are commonly used to predict vascular deformation. Although these methods have high prediction accuracy, they are computationally intensive and time-consuming, making them difficult to meet the requirements of real-time surgical planning and intraoperative monitoring.

[0004] In view of this, there is an urgent need to provide a solution for training a prediction model for predicting vascular deformation after stent implantation, so as to use the trained prediction model to quickly and accurately predict vascular deformation after stent implantation. Summary of the Invention

[0005] In order to at least solve one or more of the technical problems mentioned above, the present disclosure proposes, in multiple aspects, a solution for training a prediction model for predicting vascular deformation after stent implantation.

[0006] In a first aspect, the present disclosure provides a method for training a prediction model for predicting vascular deformation after stent implantation, the method comprising: obtaining a pair of cranial brain image training data; the pair of cranial brain image training data comprises preoperative cranial brain image data before stent implantation and postoperative cranial brain image data after stent implantation; performing three-dimensional reconstruction processing on the preoperative cranial brain image data and the postoperative cranial brain image data respectively to obtain a preoperative vascular model and a postoperative vascular model; mapping the posture information of the stent in the postoperative cranial brain image data to the preoperative vascular model to generate a stent model; constructing a target vascular graph structure based on the preoperative vascular model and the stent model, and constructing a target stent graph structure based on the stent model; wherein the first node on the target vascular graph structure is a sampling point on the vascular wall in the preoperative vascular model, and the first node carries first physical coding information; the second node on the target stent graph structure is a sampling point on the stent in the stent model, and the second node carries second physical coding information; the target vascular graph structure, the target stent graph structure and the postoperative vascular model are input into a constructed graph neural network model for training until a set training end condition is reached to obtain the prediction model.

[0007] In some embodiments, constructing a target vascular graph structure based on the preoperative vascular model and the stent model includes: calibrating the contact area between the stent and the blood vessel on the preoperative vascular model; sampling the vertices on the blood vessel wall on the preoperative vascular model to obtain a first node; physically encoding the first node in the contact area on the preoperative vascular model based on the node information and stent information of the first node to obtain first physical coding information of the first node; and physically encoding the first node in the non-contact area on the preoperative vascular model based on the node information of the first node to obtain first physical coding information of the first node to form the target vascular graph structure.

[0008] In some embodiments, the graph neural network model includes: an encoder, an attention module and a decoder; the attention module includes a multi-head attention unit; the target vascular graph structure, the target stent graph structure and the postoperative vascular model are input into the constructed graph neural network model for training, including: inputting the target vascular graph structure and the target stent graph structure into the encoder for feature extraction to obtain vascular features and stent features respectively; inputting the vascular features and the stent features into the attention module for feature fusion to obtain fused features; inputting the fused features into the decoder for decoding to obtain a deformed vascular model; calculating a loss value based on the postoperative vascular model and the deformed vascular model, and adjusting the model parameters of the graph neural network model based on the loss value.

[0009] In some embodiments, the loss value includes: a first loss value and a second loss value; the first loss value is calculated by the mean square error loss function, and the second loss value is calculated by Figure 1 The consistency regularization loss function is calculated.

[0010] In some embodiments, the graph neural network model includes: a TAGConv architecture.

[0011] In some embodiments, after obtaining the prediction model, the method further includes: obtaining a target vascular model to be used for vascular deformation prediction and a candidate stent model of at least one candidate stent; for each candidate stent model, constructing a candidate vascular map structure based on the target vascular model and the candidate stent model, and constructing a candidate stent map structure based on the candidate stent model; for each candidate stent map structure, inputting the candidate stent map structure and the candidate vascular map structure corresponding to the candidate stent map structure into the prediction model for prediction to obtain a candidate vascular prediction model; and determining a target stent from at least one candidate stent based on each candidate vascular prediction model.

[0012] In some embodiments, the method further includes: obtaining an actual blood vessel model after the target stent is implanted into the blood vessel; and optimizing the prediction model based on the actual blood vessel model and a candidate blood vessel prediction model corresponding to the target stent.

[0013] In a second aspect, the present disclosure provides a device for training a prediction model for predicting vascular deformation after stent implantation, the device comprising: a cranial brain image training data pair acquisition module for acquiring cranial brain image training data pairs; the cranial brain image training data pairs comprising preoperative cranial brain image data before stent implantation and postoperative cranial brain image data after stent implantation; a three-dimensional reconstruction module for performing three-dimensional reconstruction processing on the preoperative cranial brain image data and the postoperative cranial brain image data, respectively, to obtain a preoperative vascular model and a postoperative vascular model; a stent model generation module for mapping the stent posture information in the postoperative cranial brain image data to the preoperative vascular model to generate a stent model; a graph structure construction model A block is used to construct a target vascular graph structure based on the preoperative vascular model and the stent model, and to construct a target stent graph structure based on the stent model; wherein, the first node on the target vascular graph structure is a sampling point on the vascular wall in the preoperative vascular model, and the first node carries first physical coding information; the second node on the target stent graph structure is a sampling point on the stent in the stent model, and the second node carries second physical coding information; a training module is used to input the target vascular graph structure, the target stent graph structure and the postoperative vascular model into the constructed graph neural network model for training until the set training end condition is reached to obtain the prediction model.

[0014] In a third aspect, the present disclosure provides an electronic device comprising: a processor configured to execute program instructions; and a memory configured to store the program instructions, wherein when the program instructions are loaded and executed by the processor, the processor executes the steps of the method for training a prediction model for predicting vascular deformation after stent implantation as described in the first aspect or any optional embodiment of the first aspect.

[0015] In a fourth aspect, the present disclosure provides a computer-readable storage medium storing program instructions, characterized in that when the program instructions are loaded and executed by a processor, the processor executes the steps of the method for training a prediction model for predicting vascular deformation after stent implantation as described in the first aspect or any optional embodiment of the first aspect.

[0016] Through the method provided above for training a prediction model for predicting vascular deformation after stent implantation, the disclosed embodiment obtains a preoperative vascular model and a postoperative vascular model by performing three-dimensional reconstruction processing on the preoperative cranial imaging data and the postoperative cranial imaging data respectively, and then maps the posture information of the stent in the postoperative cranial imaging data to the preoperative vascular model to generate a stent model, and constructs a target vascular map structure and a target stent map structure to train a prediction model, so that the prediction model can predict vascular deformation. Compared with the use of finite element method (FEM) or other numerical simulation methods for deformation prediction, it can perform prediction more quickly and accurately, thereby realizing real-time prediction of complex vascular deformation. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and other objects, features and advantages of the exemplary embodiments of the present disclosure will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present disclosure are shown in an illustrative and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0018] Figure 1 An exemplary flow chart showing a method for training a prediction model for predicting blood vessel deformation after stent implantation according to some embodiments of the present disclosure is shown;

[0019] Figure 2 A schematic diagram of a model architecture of some embodiments of the present disclosure is shown;

[0020] Figure 3 An exemplary structural block diagram of an apparatus for training a prediction model for predicting blood vessel deformation after stent implantation according to some embodiments of the present disclosure is shown;

[0021] Figure 4 An exemplary structural block diagram of an electronic device according to some embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of this disclosure in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this disclosure, not all of them. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this disclosure.

[0023] It should be understood that the terms “include” and “comprising” used in the specification and claims of the present disclosure indicate the presence of described features, integers, steps, operations, elements and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0024] It should also be understood that the terminology used in this disclosure is for the purpose of describing specific embodiments only and is not intended to limit the disclosure. As used in this disclosure and the claims, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly indicates otherwise. It should be further understood that the term "and / or" as used in this disclosure and the claims refers to any and all possible combinations of one or more of the associated listed items, including and including these combinations.

[0025] As used in this specification and claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0026] The specific embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0027] Example application scenarios

[0028] In medicine, especially in the treatment of vascular diseases, stent implantation is a common surgical procedure. However, predicting vascular deformation after stent implantation is crucial for surgical planning and postoperative evaluation. Traditional prediction methods often use the finite element method (FEM) or other numerical simulation methods, but these methods are computationally intensive and time-consuming, making them difficult to meet the requirements of real-time surgical planning and intraoperative monitoring.

[0029] In view of this, the disclosed embodiments provide a method for training a prediction model for predicting blood vessel deformation after stent implantation, so as to use the trained prediction model to quickly and accurately predict blood vessel deformation after stent implantation.

[0030] Figure 1 An exemplary flow chart of a method 100 for training a prediction model for predicting vascular deformation after stent implantation according to some embodiments of the present disclosure is shown. It will be appreciated that the method 100 for training a prediction model for predicting vascular deformation after stent implantation can be performed by any suitable device with data processing capabilities, including but not limited to a terminal device, a processor, and a server.

[0031] like Figure 1 As shown, the method 100 provided in the present disclosure for training a prediction model for predicting vascular deformation after stent implantation includes: step S110: obtaining a pair of cranial brain image training data; step S120: performing three-dimensional reconstruction processing on the preoperative cranial brain image data and the postoperative cranial brain image data respectively to obtain a preoperative vascular model and a postoperative vascular model; step S130: mapping the posture information of the stent in the postoperative cranial brain image data to the preoperative vascular model to generate a stent model; step S140: constructing a target vascular graph structure based on the preoperative vascular model and the stent model, and constructing a target stent graph structure based on the stent model; wherein the first node on the target vascular graph structure is a sampling point on the vascular wall in the preoperative vascular model, and the first node carries first physical coding information; the second node on the target stent graph structure is a sampling point on the stent in the stent model, and the second node carries second physical coding information; step S150: inputting the target vascular graph structure, the target stent graph structure and the postoperative vascular model into the constructed graph neural network model for training until the set training end condition is reached to obtain a prediction model.

[0032] For example, in the disclosed embodiments, the cranial imaging training data pair in step S110 refers to a set of cranial imaging data obtained from clinical practice for training the prediction model. Specifically, the cranial imaging training data pair includes preoperative cranial imaging data before stent implantation and postoperative cranial imaging data after stent implantation.

[0033] In the embodiment disclosed herein, the above-mentioned preoperative brain imaging data and postoperative brain imaging data are image data of the internal structure of the brain obtained through medical imaging technology, which can be, for example, any one of CT angiography data (CTA), magnetic resonance angiography data (MRA) or digital subtraction angiography data (DSA).

[0034] Based on this, the above-mentioned preoperative brain imaging data and postoperative brain imaging data are collected by professional medical imaging equipment, here, professional imaging equipment such as computed tomography (CT) equipment, magnetic resonance imaging (MRI) equipment, ultrasound imaging (Ultrasound) equipment, positron emission tomography (PET) equipment, etc.

[0035] In specific implementation, before performing a stent implantation surgery on a cerebral vascular stent, a professional imaging device (CT device, MRI device or DSA device) is used to scan the patient's brain and collect preoperative brain imaging data, which can reflect the original shape, position and size of the blood vessels before stent implantation, etc.; after completing the cerebral vascular stent implantation surgery, the above imaging device is also used to scan the patient's brain and collect postoperative brain imaging data, which can reflect the changes in blood vessels and other tissues after stent implantation, including blood vessel expansion, changes in blood vessel curvature, etc. In this embodiment, for each stent implantation surgery, the preoperative brain imaging data and the postoperative brain imaging data can be stored so that they can be directly called up when needed.

[0036] It should be noted that the above-mentioned preoperative and postoperative cranial imaging data can only be used for model training with the patient's consent or after necessary privacy processing.

[0037] For example, in the disclosed embodiments, in the aforementioned step S120, the three-dimensional reconstruction process refers to a method of converting two-dimensional medical imaging data into a three-dimensional model, which can intuitively display the spatial structure of blood vessels, including the branching and tortuosity of the blood vessels. In the disclosed embodiments, during the three-dimensional reconstruction of preoperative cranial imaging data and postoperative cranial imaging data, specialized medical image processing software can be used to convert the two-dimensional preoperative cranial imaging data and postoperative cranial imaging data into three-dimensional preoperative vascular models and postoperative vascular models through a series of algorithms (e.g., surface reconstruction algorithms, voxel reconstruction algorithms, etc.).

[0038] Specifically, using preoperative cranial imaging data as an example, we first extract vascular contours from the preoperative cranial imaging data using an image segmentation algorithm. These contours are then reconstructed using 3D reconstruction techniques to obtain a preoperative vascular model. The process for 3D reconstruction of the postoperative cranial imaging data to obtain the postoperative vascular model is similar to the process described above and will not be repeated here.

[0039] For example, in the disclosed embodiment, the stent posture information in step S130 may include information such as the stent's position, orientation, angle, shape, size, and curvature, which may characterize the stent's specific placement within the blood vessel. In the disclosed embodiment, the stent posture information may be extracted from postoperative cranial imaging data using methods such as image recognition and image registration.

[0040] Mapping the posture information of the stent to the preoperative vascular model refers to spatially aligning the extracted posture information of the stent with the preoperative vascular model. Specifically, a specified algorithm (for example, a rigid registration algorithm, a non-rigid registration algorithm, etc.) can be used to map the posture information of the stent to the preoperative vascular model, and a virtual three-dimensional stent model is constructed in the preoperative vascular model based on the mapped stent posture information. Of course, the embodiment disclosed herein can also simulate the target stent model in other ways, such as finite element (FEA) simulation, which is not specifically limited in the embodiment disclosed herein.

[0041] Illustratively, in the disclosed embodiment, the target vascular graph structure in the above step S140 is a graph structure constructed based on the preoperative vascular model, and the first node on the target vascular graph structure is a sampling point on the vascular wall in the preoperative vascular model, and the first node carries first physical coding information for describing the physical characteristics and other related information at the node. In the disclosed embodiment, the first physical coding information may include the geometric information of the node and the mechanical information of the node. Of course, the first physical coding information may vary depending on the area where the node is located (for example, the contact area or non-contact area between the stent and the blood vessel). For example, when the first node is a node in the contact area between the stent and the blood vessel, the first physical coding information may also include stent information. The disclosed embodiment enriches the information of the node by encoding the stent information into the node of the graph structure when constructing the graph structure.

[0042] In the disclosed embodiments, the geometric information of the nodes may include the coordinates of the nodes, the connection relationship of the nodes, the normal vector of the nodes, the Gaussian curvature of the nodes, etc.; the mechanical information of the nodes may include the mass of the nodes, the stiffness of the nodes, and the damping coefficient of the nodes, etc.; the stent information may include the stent model (e.g., drug-eluting type, bare metal type, etc.), the stent size (e.g., diameter, length), etc. The disclosed embodiments do not specifically limit the geometric information of the nodes, the mechanical information of the nodes, and the stent information, and they can be determined according to actual conditions.

[0043] In the disclosed embodiments, the target support structure is a graph structure constructed based on the support model. The second nodes on the graph structure are sampling points on the support in the support model. The second nodes carry second physical coding information. Here, the second physical coding information may include geometric information and mechanical information of the nodes. The geometric information and mechanical information of the nodes can be found in the description of the above embodiments and will not be repeated here.

[0044] As for how to construct the target vessel map structure based on the preoperative vessel model and the stent model, and how to construct the target stent map structure based on the stent model, the following embodiments provide examples for description, which will not be elaborated here.

[0045] For example, in the embodiment of the present disclosure, the graph neural network model in the above step S150 is a neural network model specifically used to process graph structure data, which can learn the features of nodes and edges in the graph structure and the relationship between them. Specifically, it can adopt a lightweight and efficient GNN architecture, such as convolution mechanisms such as TAGConv and GCN. As a specific example of the present disclosure, Figure 2 As shown, the graph neural network model may include an encoder and a decoder, and both the encoder and the decoder may include multiple layers (e.g., 3 layers) of graph convolution layers. In the disclosed embodiment, in addition to the encoder and the decoder, the graph neural network model may also introduce a cross-region multi-head attention mechanism (e.g., Figure 2 The cross-attention module in

[15] enables the model to capture the complex interaction effects between the stent and the vascular region.

[0046] In the disclosed embodiment, the above-mentioned training end condition can be set in many ways, for example, reaching a set number of iterations (for example, 500 times), or for example, loss value convergence, etc. The disclosed embodiment does not specifically limit the above-mentioned training end condition.

[0047] In the disclosed embodiment, the target vascular map structure, the target stent map structure and the postoperative vascular model are input into the constructed graph neural network model for training. Specifically, feature extraction, feature fusion and feature decoding are performed on the target vascular map structure and the target stent map structure based on the graph neural network model to obtain the deformed vascular model, and the postoperative vascular model is used as the target data to calculate the loss value. The model parameters of the graph neural network model are adjusted based on the calculated loss value until the above-mentioned set training end conditions are met, and the prediction model can be obtained.

[0048] The disclosed embodiment performs three-dimensional reconstruction on preoperative cranial imaging data and postoperative cranial imaging data respectively to obtain a preoperative vascular model and a postoperative vascular model, and then maps the posture information of the stent in the postoperative cranial imaging data to the preoperative vascular model to generate a stent model, and constructs a target vascular map structure and a target stent map structure to train a prediction model so that the prediction model can predict vascular deformation. Compared with deformation prediction using the finite element method (FEM) or other numerical simulation methods, prediction can be performed more quickly and accurately, thereby realizing real-time prediction of complex vascular deformation.

[0049] The following describes a specific implementation method for constructing a target vascular graph structure based on the preoperative vascular model and the stent model in step S140. Specifically, the following steps are performed: The contact area between the stent and the vessel is calibrated on the preoperative vascular model; vertices on the vessel wall of the preoperative vascular model are sampled to obtain first nodes; For each first node in the contact area of ​​the preoperative vascular model, physical encoding is performed based on the node information of the first node and the stent information to obtain first physical encoding information of the first node; and, for each first node in the non-contact area of ​​the preoperative vascular model, physical encoding is performed based on the node information of the first node to obtain first physical encoding information of the first node, thereby forming the target vascular graph structure.

[0050] For example, in the disclosed embodiment, after the stent model is constructed in the preoperative vascular model, the contact area between the stent and the blood vessel can be directly calibrated on the preoperative vascular model. In the disclosed embodiment, when specifically calibrating the contact area, the result of manual calibration can be accepted, or the spatial distance between each vertex on the preoperative vascular model and the vascular model can be calculated. If the spatial distance is less than a set threshold (for example, 0.1 mm), the vertex is considered to belong to the contact area. The disclosed embodiment does not specifically limit the method for determining the contact area.

[0051] In the embodiment of the present disclosure, when calibrating the contact area, it can be calibrated by numbers, for example, marking the vertices of the contact area as "1"; it can also be calibrated by color, for example, marking the vertices of the contact area as "red"; the embodiment of the present disclosure does not specifically limit the specific calibration method.

[0052] In the disclosed embodiments, when constructing the target vascular graph structure, physically encoding every vertex on the preoperative vascular model would be computationally intensive. Therefore, in the disclosed embodiments, prior to constructing the target vascular graph structure, vertices on the preoperative vascular model are sampled to obtain first nodes (i.e., sampling points). The specific sampling method may be, for example, uniform sampling or non-uniform sampling based on the geometric characteristics of the vascular wall (e.g., curvature, thickness, etc.). The disclosed embodiments do not specifically limit the sampling method.

[0053] After obtaining the first node, physical encoding is performed on each first node to form a target vascular map structure. Specifically, if the first node is a vertex in the contact area, physical encoding is performed based on the node information of the first node (i.e., the geometric information and mechanical information of the node) and the stent information (i.e., the stent model, stent size, etc.) to obtain first physical encoding information for the first node. If the first node is a vertex in the non-contact area, physical encoding is performed based on the node information of the first node to obtain first physical encoding information for the first node.

[0054] In the disclosed embodiment, in order to obtain a physical feature code with higher discrimination ability, when physically encoding each first node, the adaptive fusion of the above-mentioned node information, bracket information, etc. can be achieved through a multi-layer perception mechanism (MLP) or a self-attention mechanism.

[0055] The following describes a specific implementation method for constructing a target stent graph structure based on the stent model in step S140. Specifically, vertices on the stent model are sampled to obtain second nodes; and for each second node, physical encoding is performed based on the node information of the second node to obtain second physical encoding information of the second node, thereby forming the target stent graph structure.

[0056] For example, in the disclosed embodiment, when constructing the target support graph structure, if each vertex on the support model is physically encoded, the amount of calculation is very large. Therefore, in the disclosed embodiment, before constructing the target support graph structure, the vertices on the support model are sampled to obtain the second node (i.e., the sampling point). The specific sampling method is the same as the sampling method for the first node described above and will not be repeated here.

[0057] In the embodiment of the present disclosure, after obtaining the second node, physical encoding is performed on each second node to form a target support graph structure. Specifically, for each second node, physical encoding is performed based on the node information of the second node (i.e., node coordinates, node connection relationship, node normal vector, node Gaussian curvature, support material, support shape, etc.) to obtain the second physical coding information of the second node. Similarly, in the embodiment of the present disclosure, in order to obtain a physical feature coding with higher discrimination ability, when physically encoding each second node, the adaptive fusion of the above-mentioned node information, support information, etc. can be achieved through a multi-layer perception mechanism (MLP) or a self-attention mechanism.

[0058] The following describes an example of a training method for a prediction model. Specifically, the graph neural network model includes an encoder, an attention module, and a decoder; the attention module includes a multi-head attention unit; the target vascular graph structure, the target stent graph structure, and the postoperative vascular model are input into the constructed graph neural network model for training, including: inputting the target vascular graph structure and the target stent graph structure into the encoder for feature extraction to obtain vascular features and stent features; inputting the vascular features and stent features into the attention module for feature fusion to obtain fused features; inputting the fused features into the decoder for decoding to obtain a deformed vascular model; calculating a loss value based on the postoperative vascular model and the deformed vascular model, and adjusting the model parameters of the graph neural network model based on the loss value.

[0059] Exemplarily, the encoder is used to extract features from the input graph structure data. In the disclosed embodiment, the encoder encodes the node features in the target vessel graph structure and the target stent graph structure to extract the vessel features and stent features. The encoder is usually composed of a multi-layer neural network, each layer transforms and extracts the input features, and finally obtains a high-dimensional feature vector, for example, Figure 2 In

[15] , the encoder consists of 3 layers of graph convolutional layers.

[0060] For example, the function of the attention module is to fuse the blood vessel features and stent features output by the encoder. In the disclosed embodiment, it includes a multi-head attention unit that can perform weighted fusion of features from different angles, so that the model can capture the complex interaction effects between the stent and the blood vessel, thereby fusing to obtain a more comprehensive fusion feature. For example, Figure 2 In

[15] , the attention module consists of 3 multi-head attention units.

[0061] For example, the decoder is used to decode the fused features and reconstruct the deformed graphics to obtain a deformed blood vessel model. Specifically, the decoder predicts the deformation of the blood vessel after stent implantation based on the fused features, including the displacement of the blood vessel wall, the change in the size of the blood vessel cavity, etc. Figure 2 In

[15] , the decoder consists of 3 graph convolutional layers.

[0062] Exemplarily, after obtaining the deformed vascular model, a loss value is calculated based on the postoperative vascular model and the deformed vascular model, and the loss value can reflect the difference between the predicted result and the actual result. In the embodiment disclosed herein, the loss value is calculated by a loss function. The loss value in the embodiment disclosed herein may include a first loss value and a second loss value; wherein the first loss value is calculated by a mean squared error (MSE) loss function, which can be used to measure the error between the predicted deformation and the actual deformation; the second loss value is calculated by Figure 1The graph consistency regularization (GCR) loss function is used to calculate the GCR loss, which can ensure the continuity and smoothness of vascular deformation by constraining the local gradient of the prediction results.

[0063] In the disclosed embodiment, the above loss value may be the sum of the first loss value and the second loss value, or may be a weighted average of the first loss value and the second loss value. The disclosed embodiment does not specifically limit the above loss value. After obtaining the loss value, the model parameters of the graph neural network model (e.g., the model weights and biases, etc.) are adjusted based on the loss value.

[0064] As an optional embodiment of the present disclosure, after obtaining the prediction model, prediction can be performed during surgery based on the trained prediction model, specifically including: obtaining a target vascular model for vascular deformation prediction and a candidate stent model of at least one candidate stent; for each candidate stent model, constructing a candidate vascular map structure based on the target vascular model and the candidate stent model, and constructing a candidate stent map structure based on the candidate stent model; for each candidate stent map structure, inputting the candidate stent map structure and the candidate vascular map structure corresponding to the candidate stent map structure into the prediction model for prediction to obtain a candidate vascular prediction model; and determining a target stent from at least one candidate stent based on each candidate vascular prediction model.

[0065] For example, in the disclosed embodiments, the target vascular model for which vascular deformation prediction is to be performed refers to a specific vascular model for which vascular deformation after stent implantation needs to be predicted, which can be obtained by three-dimensional reconstruction of the cranial imaging data of a specific patient. Candidate stents refer to a series of stents of different types, sizes, structures, etc., which are candidates for possible implantation into the target blood vessel. Candidate stent models refer to three-dimensional models of candidate stents. In the disclosed embodiments, the above-mentioned candidate stent models can be obtained from a pre-constructed stent model database or other sources.

[0066] In the disclosed embodiment, the method for constructing the candidate vessel map structure and the candidate stent map structure is the same as the method for constructing the target vessel map structure and the target stent map structure described above, and will not be repeated here.

[0067] In the disclosed embodiments, for each candidate stent map structure, the candidate stent map structure and the candidate vessel map structure corresponding to the candidate stent map structure are input into a prediction model for prediction, resulting in a candidate vessel prediction model. A target stent is then determined from at least one candidate stent based on each candidate vessel prediction model. Here, the target stent refers to the stent that is most suitable for implantation in the target vessel. During the selection process, the target stent can be determined from the at least one candidate stent based on certain evaluation indicators (e.g., rationality of vessel deformation, compatibility of the stent with the vessel, etc.).

[0068] As an optional embodiment of the present disclosure, the above-mentioned method 100 for training a prediction model for predicting vascular deformation after stent implantation also includes: obtaining an actual vascular model after the target stent is implanted into the blood vessel; and optimizing the prediction model based on the actual vascular model and the candidate vascular prediction model corresponding to the target stent.

[0069] For example, in the disclosed embodiments, the actual vascular model refers to a model that accurately reflects the vascular morphology, obtained through 3D reconstruction and acquired using medical imaging equipment after the target stent is implanted in the vessel. After obtaining the actual vascular model, a comparison analysis is performed between the actual vascular model and the candidate vascular prediction model corresponding to the target stent. Based on the differences between the two, an appropriate optimization algorithm is used to adjust the prediction model parameters, further improving the accuracy and reliability of the prediction model.

[0070] The trained prediction model in the disclosed embodiment can be deployed on a surgical planning platform to achieve preoperative simulation and real-time monitoring during surgery, assisting doctors in selecting the best implantation plan based on different stent parameters.

[0071] The disclosed embodiment organically integrates physical modeling with data-driven deep learning methods, and introduces additional stent parameters on the basis of the original vascular map structure and physical parameters, and encodes the stent model (such as drug-eluting type, bare metal type, etc.), size (diameter, length) and placement position (relative to the geometric center of the blood vessel and local characteristic area), etc., which not only takes into account the accuracy of traditional physical simulation methods, but also has the advantages of high efficiency and strong generalization ability of deep learning models, thereby realizing real-time prediction of complex vascular deformation. This technical solution can be widely used in clinical interventional treatment, surgical planning and risk assessment.

[0072] Figure 3 An exemplary structural block diagram of an apparatus 300 for training a prediction model for predicting blood vessel deformation after stent implantation according to some embodiments of the present disclosure is shown.

[0073] like Figure 3As shown, the present disclosure provides an apparatus 300 for training a prediction model for predicting vascular deformation after stent implantation, comprising: a cranial image training data pair acquisition module 310 for acquiring cranial image training data pairs; the cranial image training data pairs include preoperative cranial image data before stent implantation and postoperative cranial image data after stent implantation; a three-dimensional reconstruction module 320 for performing three-dimensional reconstruction processing on the preoperative cranial image data and the postoperative cranial image data, respectively, to obtain a preoperative vascular model and a postoperative vascular model; a stent model generation module 330 for mapping the posture information of the stent in the postoperative cranial image data to the preoperative vascular model, to generate a stent model. type; a graph structure construction module 340, used to construct a target vascular graph structure based on the preoperative vascular model and the stent model, and to construct a target stent graph structure based on the stent model; wherein the first node on the target vascular graph structure is a sampling point on the vascular wall in the preoperative vascular model, and the first node carries first physical coding information; the second node on the target stent graph structure is a sampling point on the stent in the stent model, and the second node carries second physical coding information; a training module 350, used to input the target vascular graph structure, the target stent graph structure and the postoperative vascular model into the constructed graph neural network model for training until the set training end condition is reached to obtain a prediction model.

[0074] As an optional embodiment of the present disclosure, the above-mentioned three-dimensional reconstruction module 320 is specifically used to: calibrate the contact area between the stent and the blood vessel on the preoperative vascular model; sample the vertices on the vascular wall on the preoperative vascular model to obtain a first node; for the first node in the contact area on the preoperative vascular model, physically encode the first node based on the node information of the first node and the stent information to obtain the first physical coding information of the first node; and, for the first node in the non-contact area on the preoperative vascular model, physically encode the first node based on the node information of the first node to obtain the first physical coding information of the first node to form a target vascular map structure.

[0075] As an optional embodiment of the present disclosure, the graph neural network model includes: an encoder, an attention module and a decoder; the attention module includes a multi-head attention unit; the above-mentioned training module 350 is specifically used to: input the target vascular graph structure and the target stent graph structure into the encoder for feature extraction, respectively, to obtain vascular features and stent features; input the vascular features and stent features into the attention module for feature fusion, to obtain fusion features; input the fusion features into the decoder for decoding, to obtain a deformed vascular model; calculate the loss value based on the postoperative vascular model and the deformed vascular model, and adjust the model parameters of the graph neural network model based on the loss value.

[0076] As an optional embodiment of the present disclosure, the above-mentioned loss value includes: a first loss value and a second loss value; the first loss value is calculated by the mean square error loss function, and the second loss value is calculated by Figure 1 The consistency regularization loss function is calculated.

[0077] As an optional embodiment of the present disclosure, the graph neural network model includes: TAGConv architecture.

[0078] As an optional embodiment of the present disclosure, the device 300 for training a prediction model for predicting vascular deformation after stent implantation also includes: a model acquisition module, used to obtain a target vascular model to be used for vascular deformation prediction and a candidate stent model of at least one candidate stent; a candidate graph structure construction module, used to construct a candidate vascular graph structure based on the target vascular model and the candidate stent model for each candidate stent model, and to construct a candidate stent graph structure based on the candidate stent model; a prediction module, used to input the candidate stent graph structure and the candidate vascular graph structure corresponding to the candidate stent graph structure into the prediction model for prediction, so as to obtain a candidate vascular prediction model; and a target stent determination module, used to determine a target stent from at least one candidate stent based on each candidate vascular prediction model.

[0079] As an optional embodiment of the present disclosure, the device 300 for training a prediction model for predicting blood vessel deformation after stent implantation also includes: an actual blood vessel model acquisition module, used to obtain the actual blood vessel model after the target stent is implanted into the blood vessel; and an optimization module, used to optimize the prediction model based on the actual blood vessel model and the candidate blood vessel prediction model corresponding to the target stent.

[0080] Correspondingly, the present disclosure also provides Figure 3 The hardware structure diagram of the device shown is as follows Figure 4 As shown, the electronic device 400 may be a device for implementing the above method 100. Figure 4 As shown, the electronic device 400 includes a processor 410 and a memory 420. The memory 420 is configured to store program instructions; the processor 410 is configured to load and execute the program instructions stored in the memory 420 to implement the corresponding embodiment of the method for training a prediction model for predicting vascular deformation after stent implantation as shown above.

[0081] As an embodiment, the memory 420 can be any electronic, magnetic, optical, or other physical storage device that can contain or store information, such as program instructions, data, and the like. For example, the memory 420 can be a volatile memory, a non-volatile memory, or a similar storage medium. Specifically, the memory 420 can be a RAM (Random Access Memory), a flash memory, a storage drive (such as a hard disk drive), a solid-state drive, any type of storage disk (such as a CD, DVD, etc.), or a similar storage medium, or a combination thereof.

[0082] Although a plurality of embodiments of the present disclosure have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art may conceive of many modifications, changes, and alternatives without departing from the ideas and spirit of the present disclosure. It should be understood that in practicing the present disclosure, various alternatives to the embodiments of the present disclosure described herein may be adopted. The appended claims are intended to define the scope of protection of the present disclosure and therefore cover equivalents or alternatives within the scope of these claims.

Claims

1. A method for training a prediction model for predicting vascular deformation after stent implantation, characterized in that: The method comprises: Acquire a pair of cranial brain image training data; the pair of cranial brain image training data includes preoperative cranial brain image data before stent implantation and postoperative cranial brain image data after stent implantation; performing three-dimensional reconstruction processing on the preoperative cranial brain image data and the postoperative cranial brain image data respectively to obtain a preoperative vascular model and a postoperative vascular model; Mapping the position information of the stent in the postoperative cranial imaging data to the preoperative vascular model to generate a stent model; Constructing a target vessel graph structure based on the preoperative vessel model and the stent model, and constructing a target stent graph structure based on the stent model; wherein a first node on the target vessel graph structure is a sampling point on the vessel wall in the preoperative vessel model, and the first node carries first physical coding information; a second node on the target stent graph structure is a sampling point on the stent in the stent model, and the second node carries second physical coding information; The target blood vessel graph structure, the target stent graph structure and the postoperative blood vessel model are input into the constructed graph neural network model for training until the set training end condition is reached to obtain the prediction model.

2. The method according to claim 1, characterized in that The constructing of a target vascular map structure based on the preoperative vascular model and the stent model includes: Marking the contact area between the stent and the blood vessel on the preoperative blood vessel model; Sampling vertices on the blood vessel wall of the preoperative blood vessel model to obtain a first node; For the first node in the contact area on the preoperative vascular model, physical encoding is performed based on the node information and stent information of the first node to obtain the first physical encoding information of the first node; and for the first node in the non-contact area on the preoperative vascular model, physical encoding is performed based on the node information of the first node to obtain the first physical encoding information of the first node to form the target vascular graph structure.

3. The method according to claim 1, characterized in that The graph neural network model includes: an encoder, an attention module, and a decoder; the attention module includes a multi-head attention unit; the target blood vessel graph structure, the target stent graph structure, and the postoperative blood vessel model are input into the constructed graph neural network model for training, including: Inputting the target blood vessel graph structure and the target stent graph structure into the encoder for feature extraction to obtain blood vessel features and stent features respectively; Inputting the blood vessel features and the stent features into the attention module for feature fusion to obtain fused features; Inputting the fused features into the decoder for decoding to obtain a deformed blood vessel model; A loss value is calculated based on the postoperative vascular model and the deformed vascular model, and model parameters of the graph neural network model are adjusted based on the loss value.

4. The method according to claim 3, characterized in that The loss value includes: a first loss value and a second loss value; the first loss value is calculated by a mean square error loss function, and the second loss value is calculated by a graph consistency regularization term loss function.

5. The method according to claim 1, wherein The graph neural network model includes: TAGConv architecture.

6. The method according to claim 1, wherein After obtaining the prediction model, the method further includes: Acquire a target blood vessel model for blood vessel deformation prediction and a candidate stent model of at least one candidate stent; For each candidate stent model, constructing a candidate vessel graph structure based on the target vessel model and the candidate stent model, and constructing a candidate stent graph structure based on the candidate stent model; For each candidate stent map structure, input the candidate stent map structure and the candidate blood vessel map structure corresponding to the candidate stent map structure into the prediction model for prediction, thereby obtaining a candidate blood vessel prediction model; A target stent is determined from at least one candidate stent based on the prediction models of the candidate blood vessels.

7. The method according to claim 6, characterized in that The method further comprises: Acquiring an actual blood vessel model after the target stent is implanted into the blood vessel; The prediction model is optimized based on the actual blood vessel model and the candidate blood vessel prediction model corresponding to the target stent.

8. A device for training a prediction model for predicting vascular deformation after stent implantation, characterized in that: The device comprises: A cranial brain image training data pair acquisition module is used to acquire cranial brain image training data pairs; the cranial brain image training data pairs include preoperative cranial brain image data before stent implantation and postoperative cranial brain image data after stent implantation; a three-dimensional reconstruction module, configured to perform three-dimensional reconstruction processing on the preoperative cranial brain image data and the postoperative cranial brain image data, respectively, to obtain a preoperative vascular model and a postoperative vascular model; a stent model generation module, configured to map the stent posture information in the postoperative cranial brain imaging data to the preoperative vascular model to generate a stent model; a graph structure construction module, configured to construct a target vessel graph structure based on the preoperative vessel model and the stent model, and to construct a target stent graph structure based on the stent model; wherein a first node on the target vessel graph structure is a sampling point on the vessel wall in the preoperative vessel model, and the first node carries first physical coding information; and a second node on the target stent graph structure is a sampling point on the stent in the stent model, and the second node carries second physical coding information; The training module is used to input the target blood vessel graph structure, the target stent graph structure and the postoperative blood vessel model into the constructed graph neural network model for training until the set training end condition is reached to obtain the prediction model.

9. An electronic device, characterized in that: include: a processor configured to execute program instructions; as well as A memory configured to store the program instructions, which, when loaded and executed by the processor, causes the processor to perform the steps of the method for training a prediction model for predicting vascular deformation after stent implantation as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing program instructions, characterized in that: When the program instructions are loaded and executed by the processor, the processor is caused to perform the steps of the method for training a prediction model for predicting blood vessel deformation after stent implantation as described in any one of claims 1 to 7.

Images