A biological tissue model structure optimization model, method, device and medium

Abstract

The present invention discloses a biological tissue model structure optimization model, method, device, and medium. The optimization model is constructed by: obtaining a learning sample set, the learning sample set including multiple learning samples, each learning sample including the spatial coordinates of all points extracted from a biological tissue model sample; pre-constructing a base model, the base model including a deformation network and a template network, the deformation network configured to transform the learning samples into a preset three-dimensional space to obtain a predicted biological tissue model, and the template network configured to determine the SDF value of the predicted biological tissue model; inputting the learning sample set into the base model, and optimizing the base model using a preset total loss function to obtain a biological tissue model structure optimization model; the total loss function is related to the SDF loss and / or geometric gradient loss of the predicted biological tissue model. The present invention can predict and repair defects and deformations in the biological tissue model structure, improving the accuracy of the model structure.

Description

Technical Field

[0001] The present invention relates to the fields of medicine and computer simulation technology, and in particular to a biological tissue model structure optimization model, method, device and medium. Background Art

[0002] Interventional surgery has been widely used in the treatment of various diseases due to its minimally invasive nature, and is particularly important in the treatment of vascular diseases (such as aneurysms, arteriosclerosis, cancer-related vascular reconstruction, etc.). Accurate vascular structure reconstruction is the key to ensuring surgical safety and improving treatment efficacy. However, existing vascular structure optimization technologies still face many challenges, such as segmentation algorithms leading to reduced vessel diameter, reduced image resolution, and accuracy issues in discrete voxel processing. In clinical applications, these problems may lead to large errors and discontinuities in the obtained vascular model, which in turn affects the accuracy of interventional treatment and increases the operational burden on doctors.

[0003] Existing methods typically model vascular structures by processing image stacks layer by layer, so the shape is inevitably affected by the resolution of discrete voxels. Furthermore, as the diameter of the vessel decreases, the clarity of the CT image decreases significantly. These factors result in a rougher vascular structure and susceptibility to interference from spike-like noise. Although Laplacian smoothing can reduce some of the noise, it cannot guarantee high fidelity of the vascular surface while eliminating the interference from spike-like noise.

[0004] With the development of convolutional neural networks (CNNs), automated vascular structure optimization methods have been widely used to reduce manual workload. These methods can use preoperative data to infer intraoperative data from the same patient, thereby reducing the burden on repeated physicians. However, CNN-based methods generally still model vascular structures in discrete voxel space and therefore still face the above-mentioned problems. In addition, these methods are prone to breakage problems during the inference process, resulting in discontinuity of the vascular tree, which in turn interferes with the preoperative path planning algorithm and may even lead interventional physicians to draw incorrect conclusions.

[0005] The disclosure of the above background technology content is only used to assist in understanding the inventive concept and technical solution of the present invention. It does not necessarily belong to the prior art of the present application, nor does it necessarily provide technical guidance. In the absence of clear evidence that the above content has been disclosed before the filing date of the present application, the above background technology should not be used to evaluate the novelty and creativity of the present application. Summary of the Invention

[0006] The purpose of the present invention is to provide a biological tissue model structure optimization model, method, device and medium. By introducing the joint modeling of implicit templates and spatial transformation functions, it is possible to more accurately capture the complex geometric characteristics of biological tissue structures, predict and repair defects and deformations in biological tissue model structures, and thus improve the accuracy of surface reconstruction of biological tissue model structures.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0008] A biological tissue model structure optimization model is constructed in the following way:

[0009] Acquire a learning sample set, the learning sample set including a plurality of learning samples, each learning sample including spatial coordinates of all points extracted from a biological tissue model sample; the points on the biological tissue model sample are generally points on the surface of the biological tissue model, also called vertices or surface vertices, and each biological tissue model includes a plurality of points / vertices;

[0010] Pre-constructing a neural network basic model, the basic model including a deformation network and a template network, the deformation network being configured to transform the learning sample into a preset three-dimensional space to obtain a predicted biological tissue model, and the template network being configured to determine an SDF value, i.e., a signed distance function value, of the predicted biological tissue model;

[0011] The learning sample set is input into the basic model, and the basic model is optimized and trained using a preset total loss function to obtain a biological tissue model structure optimization model; the total loss function is related to the SDF loss and / or geometric gradient loss of the predicted biological tissue model.

[0012] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the total loss function is expressed as:

[0013] L sdf =k1L sdf-val +k2L sdf-geo

[0014] Among them, L sdf Represents the total loss function, L sdf-wal represents the SDF loss of the predicted biological tissue model, L sdf-geo It represents the geometric gradient loss of the predicted biological tissue model, k1 is the first coefficient, k1>0, k2 is the second coefficient, k2>0.

[0015] Furthermore, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, the SDF loss L of the predicted biological tissue model is calculated by the following formula: sdf-wal :

[0016]

[0017] Among them, Ω i represents the entire space of the i-th predicted biological tissue model, S i represents the surface of the i-th predicted biological tissue model, p represents the spatial coordinates of the point, α i represents the characteristic vector of the i-th biological tissue model sample, β represents the characteristic vector of a class of biological tissue model samples, s' represents the true SDF value, F(p; α i ,β) represents the SDF value predicted for p, φ(F(p;α i ,β))=exp(-δ·||F(p;α i ,β)||) is a penalty for the surface point prediction of the non-biological tissue model, aiming to make its SDF value close to zero, where δ>>1, ω s Indicates that L sdf-wal The weight of the SDF error value in ,ω φ Indicates that L sdf-wal The weight of the spatial distribution of SDF in .

[0018] Furthermore, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, the geometric gradient loss L of the predicted biological tissue model is calculated by the following formula: sdf-geo :

[0019]

[0020] Among them, Ω i represents the entire space of the i-th predicted biological tissue model, S i represents the surface of the i-th predicted biological tissue model, p represents the spatial coordinate of the point, n' represents the normal vector of the plane where the surface p of the biological tissue model sample is located, S cos Represents cosine similarity, which is used to measure the consistency between the predicted gradient and the true normal vector, represents the gradient direction of the surface point p of the biological tissue model sample, Indicates that the gradient stride is consistent with the modulus of the unit normal vector, ω n The coefficient representing the cosine similarity, ω Eik A coefficient representing the magnitude.

[0021] Furthermore, according to any one of the technical solutions or a combination of multiple technical solutions mentioned above, k1=1, k2=1.

[0022] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the basic model may be optimized and trained in the following manner:

[0023] A latent code is configured for the learning sample. The deformable network is configured to transform the learning sample into a preset three-dimensional space based on the latent code to obtain a predicted biological tissue model. The template network is configured to determine the SDF value of the predicted biological tissue model based on the latent code, and optimize the latent code during each round of training. The latent code is related to the shape characteristics of the learning sample. Specifically, the latent code is a high-dimensional vector representing the shape characteristics of the learning sample. At the beginning of training, the latent code is a randomly initialized high-dimensional vector; the latent code is optimized during each round of training. After the training is completed, the latent code is a high-dimensional vector that accurately represents the learning sample. By introducing the latent code, the trained deformable network and template network can accurately learn the overall structural characteristics of the learning sample. In practical applications, the overall structural characteristics of the 3D biological tissue model with defects to be optimized can be determined more efficiently, quickly, and accurately.

[0024] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the learning sample is transformed into a preset three-dimensional space using the deformation network and the SDF value of the predicted biological tissue model is obtained using the template network based on the following formula:

[0025] F(p∈R 3 ,c)=T(D(p,c))

[0026] Where F represents the signed distance function, the output of which is the SDF value, p represents the spatial coordinates of the point on the biological tissue model sample, c represents the potential code of the learning sample, R 3 represents the three-dimensional space of the learning sample, the function D is used to map the spatial coordinates of point p to a preset three-dimensional space to obtain the corresponding three-dimensional coordinates, and T is an implicit function used to return the SDF value of a point in the preset three-dimensional space.

[0027] Furthermore, based on any one of the technical solutions or a combination of multiple technical solutions described above, the deformation network includes multiple layers of fully connected layers and nonlinear activation functions, whose input is the point coordinates and latent codes in three-dimensional space, and its output is a three-dimensional displacement vector, and the latent code is related to the shape characteristics of the biological tissue model.

[0028] Furthermore, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, the biological tissue model structure optimization model satisfies: taking part of the learning sample as input, the output obtained includes not only the input but also the spatial coordinates of points that are not included in the input but included in the learning sample; and / or,

[0029] The preset three-dimensional space is a template space, which is a space representing an implicit template of a biological tissue model structure. Based on the template space, the deformation network and the template network perform unified modeling of different shapes.

[0030] Furthermore, based on any one of the technical solutions or a combination of multiple technical solutions described above, the biological tissue model includes one or more of a blood vessel model, a bone model, and an organ model.

[0031] According to another aspect of the present invention, a method for optimizing the structure of a biological tissue model is provided, comprising the following steps:

[0032] Determining input data, wherein the input data includes spatial coordinates of all vertices on the 3D biological tissue model to be optimized and an initial latent code, wherein the initial latent code is determined by random initialization;

[0033] Inputting the input data into a pre-built biological tissue model structure optimization model to obtain an optimized 3D biological tissue model;

[0034] The biological tissue model structure optimization model is constructed in the following way:

[0035] Acquire a learning sample set, wherein the learning sample set includes a plurality of learning samples, and each learning sample includes the spatial coordinates of all points extracted from a biological tissue model sample;

[0036] Pre-constructing a neural network basic model, the basic model including a deformation network and a template network, the deformation network being configured to transform the learning sample into a preset three-dimensional space to obtain a predicted biological tissue model, and the template network being configured to determine an SDF value of the predicted biological tissue model;

[0037] The learning sample set is input into the basic model, and the basic model is optimized and trained using a preset total loss function to obtain a biological tissue model structure optimization model; the total loss function is related to the SDF loss and / or geometric gradient loss of the predicted biological tissue model.

[0038] Furthermore, based on any one of the above technical solutions or a combination of multiple technical solutions, the input data is optimized using the biological tissue model structure optimization model to obtain an optimized 3D biological tissue model, including:

[0039] Based on the input data, transforming the 3D biological tissue model into a preset three-dimensional space using the trained deformable network to obtain a current 3D biological tissue optimization model, and determining an SDF value of the current 3D biological tissue optimization model using the trained template network;

[0040] determining whether the preset total loss function converges according to the 3D biological tissue optimization model and the SDF value, and if the total loss function converges, determining that the current 3D biological tissue optimization model is an optimized 3D biological tissue model;

[0041] If the total loss function does not converge, optimizing the neural network parameters and the current potential code, and performing a next round of optimization of the 3D biological tissue model based on the optimized neural network parameters and the current potential code until the total loss function converges;

[0042] The current potential code corresponding to the first round of optimization is obtained by random initialization, and the current potential code corresponding to each subsequent round of optimization is determined according to the optimization result of the previous round.

[0043] Specifically, the latent code of the 3D biological tissue model to be optimized, that is, the model to be repaired, is obtained, and the maximum a posteriori probability estimation method in statistics is adopted, that is, N different random SDF space samplings are performed on the model to be repaired, which can be used as the observation value. The latent code that satisfies the minimum average error of the N observations or the highest similarity is the latent code of the repair model.

[0044] According to another aspect of the present invention, the present invention provides a biological tissue model structure optimization device, in which a biological tissue model structure optimization model as described in any of the above embodiments is provided, and the biological tissue model structure optimization device is configured to optimize and reconstruct the input 3D biological tissue model to be optimized.

[0045] According to another aspect of the present invention, the present invention provides a computer-readable storage medium for storing program instructions, wherein the program instructions are configured to be called by a processor to execute the steps of the method described in any one of the above technical solutions or a combination of multiple technical solutions.

[0046] The beneficial effects brought about by the technical solution provided by the present invention are as follows:

[0047] a. By introducing a joint modeling approach of implicit templates and spatial transformation functions, this method can more accurately capture the complex geometric features of biological tissue structures. Traditional DeepSDF methods often struggle to accurately reflect subtle changes in shape when processing complex shapes. However, this method links geometric changes in shape with differences in the general template SDF. Based on a deep implicit neural field and geometric gradient constraints, fractured regions in the original data are jointly modeled using the implicit template and spatial transformation function during the reconstruction process. Through optimization of the geometric gradient constraints, the SDF values ​​at the fractures can be accurately predicted and completed.

[0048] b. The present invention performs conditional modeling of implicit templates and spatial transformation functions through latent coding. The shape features extracted from the input data using latent coding can represent the geometric differences of different biological tissue model instances. During the training process, the parameters and latent coding of the implicit template and spatial transformation function are jointly optimized to jointly capture the global and local geometric features of the shape, ensuring the consistency between the implicit template and the spatial transformation function, avoiding the problem of global and local feature mismatch that may occur in traditional methods, and effectively improving the accuracy of surface reconstruction. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0050] Figure 1 A flowchart of constructing a vascular structure optimization model provided by an exemplary embodiment of the present invention;

[0051] Figure 2 A schematic diagram of the architecture of a vascular structure optimization model provided by an exemplary embodiment of the present invention;

[0052] Figure 3 A diagram of a blood vessel model before repair provided by an exemplary embodiment of the present invention;

[0053] Figure 4 A diagram of a repaired blood vessel model provided by an exemplary embodiment of the present invention;

[0054] Figure 5 A flow chart of a blood vessel structure optimization method provided by an exemplary embodiment of the present invention;

[0055] Figure 6 A flow chart of a blood vessel structure optimization process provided for an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described 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 should fall within the scope of protection of the present invention.

[0057] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0058] In order to overcome the problems existing in the existing technology, this application introduces implicit functions to model the vascular structure. By combining deep implicit neural fields with geometric gradient constraints to accurately optimize the surface of the vascular structure, the reconstruction accuracy of the vascular tree can be improved, providing doctors with more reliable preoperative planning data. Based on this, this application proposes a biological tissue model structure optimization model, method, device and medium based on deep implicit neural fields and geometric gradient constraints, aiming to provide more efficient and accurate support for clinical interventional treatment, especially in the processing of complex vascular anatomical structures and the refined reconstruction of tiny blood vessels. It has important clinical application value.

[0059] Implicit neural fields are a method for implicitly representing complex data through neural networks. Instead of explicitly storing the geometry and appearance of a scene, shape, or signal in traditional forms like meshes, point clouds, or polygons, they encode the geometry and appearance of the data into the neural network's parameters. In other words, implicit neural fields implicitly define the structure and appearance of the data through the neural network's functional mapping.

[0060] Implicit neural fields typically map input space (such as position coordinates in 3D space) to output space (such as color, density, or surface normal, etc.). For example, for a 3D reconstruction of a scene, a function f can be defined: R 3 →R 4 , where the input is the three-dimensional coordinates P = (x, y, z) and the output is the color and density (r, g, b, σ). This function is implemented by a neural network, which learns the geometry and appearance of the scene through training data.

[0061] An implicit function is a mathematical concept used to implicitly represent geometric shapes or data structures. It typically maps the input space to a scalar or vector value, with the function's zero point (or specific value) defining the geometric boundary. For example, the signed distance function (SDF) is an implicit function that maps a point in space to the distance from that point to the nearest surface, where the surface is defined by the zero point. In recent years, implicit functions have shown great potential in modeling complex geometric shapes, especially when dealing with complex geometric shapes.

[0062] Deep Implicit Neural Fields (DIFs) are an extension of Implicit Neural Fields. They combine the power of deep learning and use deep neural networks to implicitly represent complex data structures. Deep Implicit Neural Fields can be seen as a method for implementing implicit functions using deep neural networks.

[0063] A deformable network is a neural network architecture designed to handle geometric deformations. Its core idea is to enable the network to better adapt to changes in the shape and position of the target by introducing deformable sampling grids or convolution operations. For example, Deformable Convolutional Networks introduce a learnable offset in the convolution operation, allowing the convolution kernel to adaptively deform, thereby better capturing the geometric changes of the target. Spatial Transformer Networks (STNs) enhance the network's adaptability to geometric transformations by explicitly performing spatial transformations (such as translation, rotation, and scaling) on ​​the input data.

[0064] In one embodiment of the present invention, a biological tissue model structure optimization model is provided. Taking the optimization of the blood vessel model structure as an example, see Figure 1 and Figure 2 , the model is constructed in the following way:

[0065] Obtaining a learning sample set, the learning sample set including a plurality of learning samples, each learning sample including the spatial coordinates of all points extracted from a vascular model sample, the vascular model sample being a standard vascular model. Specifically, a vascular model that conforms to the vascular structure of the human body (or other animals) and is undamaged can be considered the vascular model sample;

[0066] A base model is pre-constructed, comprising a deformation network and a template network. The deformation network is configured to transform the learning sample into a preset three-dimensional space to obtain a predicted vascular model, and the template network is configured to determine the SDF value of the predicted vascular model. The preset three-dimensional space is the template space, a normalized three-dimensional space used to represent the geometric structure of common shapes. In this space, all geometric variations of shapes are unified into a common reference frame (a fixed, unified coordinate system). The core concept of the template space is to represent the signed distance function (SDF) of the shape through an implicit function (i.e., the template network), thereby achieving unified modeling of different shapes.

[0067] The learning sample set is input into the basic model, and the basic model is optimized and trained using a preset total loss function to obtain a vascular structure optimization model; the total loss function is related to the SDF loss and / or geometric gradient loss of the predicted vascular model.

[0068] The basic model is trained based on the total loss function, and the vascular structure optimization model obtained by training satisfies: taking part of the learning sample as input, the output obtained includes not only the input but also the spatial coordinates of points that are not included in the input but included in the learning sample; in addition, it can also correct and repair points on the vascular model that deviate from the standard.

[0069] In one embodiment of the present invention, during the optimization training of the base model, a latent code, or initial latent code, is pre-configured for each learning sample. This initial latent code is a high-dimensional vector, and the initial latent codes for different learning samples can be the same or different. In each subsequent round of training, the deformable network transforms the learning sample into a normalized three-dimensional space based on the latent code to obtain a predicted vascular model. The template network determines the SDF value of the predicted vascular model based on the latent code, and during each round of training, the latent code is also optimized during the optimization of network parameters. For each trained model, each learning sample has a corresponding latent code associated with its shape. This latent code not only represents its global shape characteristics but also serves as the ID of each vascular model.

[0070] Latent encoding is a representation method that compresses data into a low-dimensional space, which is usually used in generative models or autoencoders. Through latent encoding, complex input data (such as images, point clouds, etc.) can be mapped to a low-dimensional latent space and operated in this space. The deformation network in this application includes multiple layers of fully connected layers and nonlinear activation functions, whose input is the point coordinates and latent encoding in three-dimensional space, and its output is a three-dimensional displacement vector. In this application, the latent encoding is related to the shape characteristics of the vascular model. This application uses latent encoding to represent the global features of the vascular model, and uses latent encoding as the input of the deformation network, which can help the deformation network better understand the global structure of the input data and process complex geometric models more efficiently and with high quality.

[0071] In the classic DeepSDF method, shape changes are directly represented by changes in the signed distance function (SDF) value itself, that is, the signed distance function. For a signed distance function, for any point p in space, the SDF value represents the distance from that point to the nearest surface. If the point is inside the object, the value is negative; if the point is outside the object, the value is positive; if the point is exactly on the surface, the value is zero. Different from this, the inventive concept of this application is that for a class of shapes represented by SDF, their geometric changes can be reflected by the difference relative to a common template SDF.

[0072] Therefore, in this application, based on the potential coding of each learning sample, the learning sample is transformed into a preset three-dimensional space using the deformation network and the SDF value of the predicted blood vessel model is obtained using the template network. It is expressed as follows:

[0073] F(p∈R 3 ,c)=T(D(p,c))

[0074] Where F represents the signed distance function, the output of which is the SDF value, p represents the spatial coordinates of the point on the vascular model sample, c represents the potential code of the learning sample, R 3 Represents the three-dimensional space of the learning sample, the spatial variation function D: R 3 ×X→R 3 Used to map the spatial coordinates of point p to the preset three-dimensional space to obtain the corresponding three-dimensional coordinates, implicit function T: R 3 →R is an implicit function that returns the SDF value of a point on the predicted blood vessel model in a preset three-dimensional space.

[0075] In the present application, D is a conditional spatial transformation function that performs a spatial transformation on the input point according to the latent code c. Specifically, the spatial transformation function D is a nonlinear deformation field that learns a complex mapping function through a deep neural network and can capture local geometric changes. Its mathematical form is: D(p, c) = p + Δp. The spatial transformation function D maps the input point p from the original space to the implicit template space. And T is an implicit function that is used to represent a universal SDF (i.e., an implicit template) that is independent of the latent code c. Therefore, for a set of shapes, the function T() represents their universal template SDF, which is subsequently referred to as an implicit template. In a specific query process, if it is necessary to calculate the signed distance value of a point p of a specific object, the point p is first mapped to the standard position in the implicit template through the spatial transformation function D, and then the standard position is queried through the implicit template T to obtain the corresponding SDF value. In other words, the implicit template is deformed according to the latent code c to model different SDFs. The latent code is related to the shape features of the learning sample. Specifically, the latent code is a high-dimensional vector representing the shape features of the learning sample. At the beginning of training, the latent code is a randomly initialized high-dimensional vector. This latent code is optimized during each round of training, and upon completion, it accurately represents the learning sample. By introducing this latent code, the trained deformable and template networks can accurately learn the overall structural characteristics of the learning sample. This allows for more efficient, rapid, and accurate determination of the overall structural characteristics of defective 3D tissue models to be optimized in practical applications.

[0076] In one embodiment of the present invention, the blood vessel structure is optimized based on the following total loss function, which is expressed as:

[0077] L sdf =L sdf-val +L sdf-geo

[0078] Among them, L sdf Represents the total loss function, L sdf-wal represents the SDF loss of the predicted vascular model, L sdf-geo represents the geometric gradient loss of the predicted blood vessel model.

[0079] In this embodiment, the SDF loss L of the predicted blood vessel model is calculated by the following formula: sdf-wal :

[0080]

[0081] Among them, Ω i represents the entire space of the i-th predicted vascular model, S irepresents the surface of the i-th predicted blood vessel model, p represents the spatial coordinate of the point; α i represents the feature vector of the i-th vascular model sample, which is randomly initialized at the beginning of training and subsequently updated and optimized by the deformation network; β represents the feature vector of a class of vascular model samples, which is also randomly initialized at the beginning of training and subsequently updated and optimized by the template network; s' represents the true SDF value, F(p; α i ,β) represents the SDF value predicted for p; φ(F(p;α i ,β))=exp(-δ·||F(p;α i ,β)||) is a penalty for the prediction of non-vascular model surface points, which aims to make its SDF value close to zero, so that the predicted vascular model pays more attention to the shape surface. When the predicted SDF value is larger, it means that the point is farther away from the surface and the loss function is smaller. When the predicted SDF value is closer to 0, it means that the point is closer to the surface and the loss function is larger. Among them, δ>>1, ω S Indicates that L sdf-wal The weight of the predicted SDF point, ω φ Indicates that L sdf-wal The weight of the predicted normal vector of a point in .

[0082] The SDF loss of the predicted vascular model directly supervises the SDF regression results of the predicted vascular model obtained during each round of training, and achieves more accurate surface modeling by penalizing the predicted values ​​of non-surface points.

[0083] In this embodiment, the geometric gradient loss L of the predicted blood vessel model is calculated by the following formula: sdf-geo :

[0084]

[0085] Among them, Ω i represents the entire space of the i-th predicted vascular model, S i represents the surface of the i-th predicted vascular model, p represents the spatial coordinate of the point, n' represents the normal vector of the plane where the surface p of the vascular model sample is located, S cos Represents cosine similarity, which is used to measure the consistency between the predicted gradient and the true normal vector, Represents the gradient direction of the surface point p of the vascular model sample. When the gradient direction of the predicted vascular model is closer to the direction of the normal vector, the accuracy of the reconstructed predicted vascular model is higher. Indicates that the gradient stride is consistent with the modulus of the unit normal vector. In other words, the gradient of the prediction model To be consistent with the normal vector in both direction and magnitude, this constraint enables the implicit template to better maintain geometric details when reconstructing complex vascular structures; ω n The coefficient representing the cosine similarity, ωEik The coefficient representing the amplitude, the amplitude and cosine similarity are not the same unit of measurement, through ω n and ω Eik Balance the two reasonably at the same order of magnitude, specifically ω n and ω Eik Usually experience points.

[0086] The vascular optimization model proposed in this application can jointly model the fracture area in the original data using an implicit template and a spatial transformation function during the reconstruction process, and through the optimization of geometric gradient constraints, the SDF value at the fracture site can be accurately predicted and completed. Figure 3 is the original data before reconstruction, Figure 4 The reconstructed data is shown in Figure 2. As can be seen, the reconstructed vascular model surface not only restores the geometric structure of the fractured area but also maintains a smooth transition with the surrounding area, fully demonstrating the advantages of this model and method in detail restoration and geometric consistency. Furthermore, by comparing the original data with the reconstructed results, we can observe that the model improves the accuracy and integrity of the surface reconstruction while maintaining the vascular topology.

[0087] It should be noted that the method of this application reconstructs all points on the input, not just the damaged areas. Specifically, this method performs unified SDF value prediction and surface optimization for all input points through joint modeling of implicit templates and spatial transformation functions. During the reconstruction process, the model automatically learns the global geometric features of the input data, including the geometric information of the damaged areas, and completes and repairs the damaged areas through geometric gradient constraints and loss functions, as well as correcting and repairing the deformed areas.

[0088] The innovation of this application is that by introducing the joint modeling of implicit templates and spatial transformation functions, the complex geometric features of vascular structures can be captured more accurately. When dealing with complex shapes, traditional DeepSDF methods often find it difficult to accurately reflect subtle changes in shape, while this method links the geometric changes in shape with the differences in the general template SDF. The implicit template and spatial transformation function are conditionally modeled through latent coding. Latent coding is a shape feature extracted from the input data that can represent the geometric differences of different vascular instances. During the training process, the parameters of the implicit template and the spatial transformation function are jointly optimized so that the two can work together to capture the global and local geometric features of the shape. Through conditional modeling, the implicit template can adapt to a variety of shape changes, while the spatial transformation function can capture the local geometric deformation of different instances. The joint optimization ensures the consistency between the implicit template and the spatial transformation function, avoids the problem of mismatch between global and local features that may occur in traditional methods, and effectively improves the accuracy of surface reconstruction.

[0089] In addition, the introduction of geometric gradients can ensure that the geometric properties of the reconstructed surface are consistent with the real surface, and can also better maintain the smoothness and consistency of the blood vessel surface, thus having broad application prospects in fields such as medical image processing.

[0090] In practical applications, this method can be used for 3D reconstruction of vascular trees, particularly in medical image analysis, helping doctors more accurately identify and diagnose vascular diseases. By optimizing the surface reconstruction of vascular structures, this method not only improves reconstruction accuracy but also enhances the model's robustness and generalization capabilities. Furthermore, the framework of this method is highly scalable and can be applied to surface reconstruction tasks for other complex shapes, such as biological tissue models like organs and bones, or for processing biomedical images.

[0091] In summary, this method proposes an efficient surface optimization model for vascular structures by combining implicit templates with spatial transformation functions and incorporating geometric gradient constraints. This model improves surface reconstruction accuracy while maintaining good geometric smoothness and consistency, possessing important theoretical and practical applications.

[0092] In one embodiment of the present invention, a method for optimizing the structure of a biological tissue model is proposed. Taking the structure optimization of a blood vessel model as an example, see Figure 5 , the method comprises the following steps:

[0093] Determining input data, where the input data includes spatial coordinates of all vertices on the 3D blood vessel model to be optimized and a latent code of the 3D blood vessel model to be optimized, where the latent code is related to shape features of the 3D blood vessel model to be optimized;

[0094] Inputting the input data into a pre-built vascular structure optimization model to obtain an optimized 3D vascular model;

[0095] The vascular structure optimization model is constructed in the following manner:

[0096] Acquire a learning sample set, wherein the learning sample set includes a plurality of learning samples, and each learning sample includes the spatial coordinates of all points extracted from a blood vessel model sample;

[0097] Pre-constructing a basic model, the basic model comprising a deformation network and a template network, the deformation network being configured to transform the learning sample into a preset three-dimensional space to obtain a predicted vascular model, and the template network being configured to determine an SDF value of the predicted vascular model;

[0098] The learning sample set is input into the basic model, and the basic model is optimized and trained using a preset total loss function to obtain a vascular structure optimization model; the total loss function is related to the SDF loss and / or geometric gradient loss of the predicted vascular model.

[0099] See also Figure 6 In the process of optimizing the input data using the biological tissue model structure optimization model to obtain an optimized 3D biological tissue model, based on the input data, the 3D biological tissue model is transformed into a preset three-dimensional space using the trained deformable network to obtain a current 3D biological tissue optimized model, and the SDF value of the current 3D biological tissue optimized model is determined using the trained template network.

[0100] Determine whether the preset total loss function converges according to the 3D biological tissue optimization model and the SDF value; if the total loss function converges, determine that the current 3D biological tissue optimization model is the optimized 3D biological tissue model.

[0101] If the total loss function does not converge, the neural network parameters and the current potential code are optimized, and the next round of optimization of the 3D biological tissue model is performed based on the optimized neural network parameters and the current potential code until the total loss function converges.

[0102] The current latent code corresponding to the first round of optimization is obtained through random initialization. The current latent code corresponding to each subsequent round of optimization is determined based on the optimization results of the previous round. Specifically, the latent code of the 3D biological tissue model to be optimized, i.e., the model to be repaired, is obtained using the maximum a posteriori probability estimation method in statistics. This involves performing N different random SDF spatial samplings on the model to be repaired, which can be used as observations. The latent code that satisfies the minimum average error or the highest similarity across these N observations is the latent code of the repaired model.

[0103] In one embodiment of the present invention, a biological tissue model structure optimization device is provided, in which a biological tissue model structure optimization model as described in any one of the above embodiments or a combination of multiple embodiments is provided. The biological tissue model structure optimization device is configured to optimize and reconstruct the input 3D biological tissue model to be optimized to obtain a more accurate 3D biological tissue model after repairing damage and correcting deformation.

[0104] In one embodiment of the present invention, a computer-readable storage medium is provided for storing program instructions, wherein the program instructions are configured to be called by a processor to execute the steps of the method described in any of the above embodiments.

[0105] It should be noted that the above-mentioned biological tissue model structure optimization method, biological tissue model structure optimization device, computer-readable storage medium embodiment and biological tissue model structure optimization algorithm embodiment belong to the same inventive concept, and the entire content of the biological tissue model structure optimization algorithm embodiment is incorporated into the biological tissue model structure optimization method, biological tissue model structure optimization device, and computer-readable storage medium embodiment by reference.

[0106] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0107] The above is only a specific implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for constructing a biological tissue model structure optimization model, characterized in that: Constructed by: Acquire a learning sample set, wherein the learning sample set includes a plurality of learning samples, and each learning sample includes the spatial coordinates of all points extracted from a biological tissue model sample; Pre-constructing a neural network basic model, the basic model including a deformation network and a template network, the deformation network being configured to transform the learning sample into a preset three-dimensional space to obtain a predicted biological tissue model, and the template network being configured to determine an SDF value of the predicted biological tissue model; The learning sample set is input into the basic model, and the basic model is optimized and trained using a preset total loss function to obtain a biological tissue model structure optimization model; the total loss function is related to the SDF loss and / or geometric gradient loss of the predicted biological tissue model; the total loss function is expressed as: ； in, L sdf represents the total loss function, L sdf-val represents the SDF loss of the predicted biological tissue model, L sdf-geo represents the geometric gradient loss of the predicted biological tissue model, k 1 is the first coefficient, k 1 >0, k 2 is the second coefficient, k 2 >0; The SDF loss of the predicted biological tissue model is calculated by the following formula: L sdf-val : ； Among them, Ω i Indicates the i The entire space of predictive biological tissue models, S i Indicates the i A surface that predicts biological tissue models, p represents the spatial coordinates of a point, α i Indicates the i The feature vector of the biological tissue model sample, β Represents the feature vector of a class of biological tissue model samples, s' represents the true SDF value, F ( p ; α i , β ) indicates that p The predicted SDF value, is a penalty on the surface point prediction of the non-biological tissue model, aiming to make its SDF value close to zero, where , Indicates L sdf-val The weight of the SDF error value in, Indicates L sdf-val The weight of the spatial distribution of SDF in ; The geometric gradient loss of the predicted biological tissue model is calculated by the following formula L sdf-geo : ； Among them, Ω i Indicates the i The entire space of predictive biological tissue models, S i Indicates the i A surface that predicts biological tissue models, p represents the spatial coordinates of a point, n' Represents the surface of a biological tissue model sample The normal vector of the plane where the point lies, S cos Represents cosine similarity, which is used to measure the consistency between the predicted gradient and the true normal vector, Represents the surface of a biological tissue model sample The gradient direction of the point, Indicates that the gradient stride is consistent with the modulus of the unit normal vector, ω n represents the coefficient of cosine similarity, ω Eik A coefficient representing the magnitude.

2. The method for constructing a biological tissue model structure optimization model according to claim 1, characterized in that: k 1 =1， k 2 =1。 3. The method for constructing a biological tissue model structure optimization model according to claim 1, characterized in that: It also includes optimizing and training the basic model in the following ways: A latent code is configured for the learning sample, the deformation network is configured to transform the learning sample into a preset three-dimensional space based on the latent code to obtain a predicted biological tissue model, the template network is configured to determine the SDF value of the predicted biological tissue model based on the latent code, and the latent code is optimized during each round of training.

4. The method for constructing a biological tissue model structure optimization model according to claim 3, characterized in that: Based on the following formula, the learning sample is transformed into a preset three-dimensional space using the deformation network and the SDF value of the predicted biological tissue model is obtained using the template network: ； in, F Represents a signed distance function whose output is the SDF value, p represents the spatial coordinates of points on the biological tissue model sample, c represents the potential code of the learning sample, R 3 Represents the three-dimensional space of the learning sample, function D Used to point The spatial coordinates of are mapped to the preset three-dimensional space to obtain the corresponding three-dimensional coordinates, T It is an implicit function that returns the SDF value of a point in a preset three-dimensional space.

5. The method for constructing a biological tissue model structure optimization model according to claim 3, characterized in that: The deformation network includes multiple layers of fully connected layers and nonlinear activation functions, whose input is the point coordinates in three-dimensional space and the potential code, and whose output is a three-dimensional displacement vector.

6. The method for constructing a biological tissue model structure optimization model according to claim 1, characterized in that: The biological tissue model structure optimization model satisfies: taking a portion of the learning sample as input, the output obtained includes not only the input but also the spatial coordinates of points that are not included in the input but included in the learning sample; and / or, The preset three-dimensional space is a template space, which is a space representing an implicit template of a biological tissue model structure. Based on the template space, the deformation network and the template network perform unified modeling of different shapes.

7. The method for constructing a biological tissue model structure optimization model according to claim 1, characterized in that: The biological tissue model includes one or more of a blood vessel model, a bone model and an organ model.

8. A method for optimizing the structure of a biological tissue model, characterized in that: The method for constructing a biological tissue model structure optimization model according to claim 1 comprises the following steps: Determining input data, wherein the input data includes spatial coordinates of all vertices on the 3D biological tissue model to be optimized and an initial latent code; The input data is input into a pre-built biological tissue model structure optimization model to obtain an optimized 3D biological tissue model; wherein the biological tissue model structure optimization model is constructed in the following manner: Acquire a learning sample set, wherein the learning sample set includes a plurality of learning samples, and each learning sample includes the spatial coordinates of all points extracted from a biological tissue model sample; Pre-constructing a neural network basic model, the basic model including a deformation network and a template network, the deformation network being configured to transform the learning sample into a preset three-dimensional space to obtain a predicted biological tissue model, and the template network being configured to determine an SDF value of the predicted biological tissue model; The learning sample set is input into the basic model, and the basic model is optimized and trained using a preset total loss function to obtain a biological tissue model structure optimization model; the total loss function is related to the SDF loss and / or geometric gradient loss of the predicted biological tissue model.

9. The biological tissue model structure optimization method according to claim 8, characterized in that: Optimizing the input data using the biological tissue model structure optimization model to obtain an optimized 3D biological tissue model, comprising: Based on the input data, transforming the 3D biological tissue model into a preset three-dimensional space using the trained deformable network to obtain a current 3D biological tissue optimization model, and determining an SDF value of the current 3D biological tissue optimization model using the trained template network; determining whether the preset total loss function converges according to the 3D biological tissue optimization model and the SDF value, and if the total loss function converges, determining that the current 3D biological tissue optimization model is an optimized 3D biological tissue model; If the total loss function does not converge, optimizing the neural network parameters and the current potential code, and performing a next round of optimization of the 3D biological tissue model based on the optimized neural network parameters and the current potential code until the total loss function converges; The current potential code corresponding to the first round of optimization is obtained by random initialization, and the current potential code corresponding to each subsequent round of optimization is determined according to the optimization result of the previous round.

10. A biological tissue model structure optimization device, characterized in that: The biological tissue model structure optimization device is provided with a biological tissue model structure optimization model constructed based on the method for constructing a biological tissue model structure optimization model according to any one of claims 1 to 7, and the biological tissue model structure optimization device is configured to optimize and reconstruct the input 3D biological tissue model to be optimized.

11. A computer-readable storage medium for storing program instructions, characterized in that: The program instructions are configured to be called by a processor to execute the steps of the biological tissue model structure optimization method according to any one of claims 8 to 9.

Images