An automatic general fracture simulation data generation method based on geometric parameter calculation

Abstract

The application relates to an automatic general fracture simulation data generation method based on geometric parameter calculation, which comprises the following steps: obtaining a three-dimensional CT of a healthy target bone and a bone tissue binary label thereof, and constructing a feature distance graph representing bone structure and bone density; adopting a three-dimensional watershed algorithm to convert the feature distance graph into a plurality of independently connected bone fracture and fragmentation regions, wherein the three-dimensional watershed algorithm controls the change of the fracture degree by adjusting a water level super parameter; for each bone fracture and fragmentation region, a random rigid transformation is performed to generate a three-dimensional CT simulating a fracture. Compared with the prior art, the application has the advantages of parameterizable control and higher morphological fidelity of the generated simulation data.

Description

Technical Field

[0001] This invention relates to the field of data processing, and in particular to an automated general fracture simulation data generation method based on geometric parameter calculation. Background Technology

[0002] Fractures are one of the most common types of trauma in clinical practice, including fractures of long bones in the limbs as well as fractures in complex locations such as the spine, pelvis, and periarticular areas. Especially for comminuted fractures, intra-articular fractures, and complex fractures caused by high-energy injuries, which are often accompanied by damage to surrounding soft tissues, nerves, blood vessels, and even adjacent organs, treatment is high-risk, difficult, and prone to long-term functional impairment. Therefore, it has long been a key challenge in clinical diagnosis, treatment, and rehabilitation management.

[0003] In traditional fracture reduction and internal fixation surgery, surgeons typically need to expose the fracture site through an incision, manually reduce the fracture fragments to the correct anatomical position, and perform internal fixation to restore limb or joint function. Due to significant differences in the geometry of different bone types and locations, and the complex shapes and irregular fracture surfaces of the fragments, it is crucial to avoid damaging important nerves, blood vessels, and soft tissue structures during reduction. The surgical procedure is highly dependent on the surgeon's experience and skill. In recent years, with the development of minimally invasive concepts and advancements in instruments and imaging technology, small-incision, closed, or percutaneous surgery guided by surgical navigation systems has gained increasing attention and application in the treatment of fractures in multiple locations. Compared to traditional open surgery, it offers potential advantages such as less trauma, lower risk of infection, and faster postoperative recovery; however, due to the smaller incision and the inability to see intraoperative tissues or limited field of vision, it places higher demands on the surgeon's spatial understanding and precise operation.

[0004] To improve surgical success rates and consistency, surgeons typically require extensive and systematic training. Currently, training for fracture diagnosis and surgical treatment often relies on simulated fracture data and in vitro models: doctors or engineers first create simulation data of different types of fractures based on healthy bone CT scans, then combine 3D reconstruction and 3D printing to form in vitro models for fracture mechanism research, surgical approach design, and training in reduction and internal fixation techniques. The creation of existing simulated fracture data often still depends on manual interaction. For example, patent application CN111261295A discloses a method and device for creating a simulation model of a burst fracture of the thoracolumbar spine. This method involves manually dividing the fracture fragments and meshing them on a healthy bone model, then combining this with finite element software (such as Abaqus) to achieve fracture simulation. Other existing work utilizes interactive modules in software such as Mimics to manually divide the fracture fragments and drag them manually to simulate fracture cases. However, the aforementioned methods generally suffer from problems such as heavy reliance on manual intervention, high time and labor costs, low generation efficiency, and poor repeatability. Furthermore, manual cutting or delineation often fails to accurately reflect the rough and irregular morphology of the fracture surface, the true geometric characteristics of the fragments, and their spatial relationship with surrounding bone / soft tissue, thus affecting the simulation effectiveness of surgical training and algorithm validation. Therefore, there is an urgent need for a general fracture simulation data generation method that is parameterizable, automatically generated, and has higher morphological realism, to provide more effective simulation models and training data support for fracture mechanism research, surgical procedures, and navigation system validation. Summary of the Invention

[0005] The purpose of this invention is to provide an automated, general fracture simulation data generation method based on geometric parameter calculation that can automatically and quickly generate simulation data for multiple types of fractures.

[0006] The objective of this invention can be achieved through the following technical solutions: An automated, general-purpose fracture simulation data generation method based on geometric parameter calculation includes the following steps: Obtain 3D CT scans of healthy target bones and their binary labels, and construct a feature distance map representing bone structure and bone density. The feature distance map is converted into multiple independent and connected fracture fragment regions using a three-dimensional watershed algorithm. The three-dimensional watershed algorithm controls the change in fracture severity by adjusting the water level hyperparameter. For each fracture fragment region, a random rigidity transformation is performed to generate a three-dimensional CT scan simulating the fracture.

[0007] Furthermore, the steps for constructing the feature distance map include: The CT values ​​of the three-dimensional CT scans of the healthy target bone are normalized to the [0,1] interval to represent the relative density characteristics of the bone tissue region, wherein the relative density characteristics are expressed as: , In the formula, The normalized CT value, i.e., the relative density feature. Voxel CT values, and These represent the minimum and maximum CT values ​​in the original 3D CT image of the healthy target bone, respectively; Perform a Euclidean distance transformation on the binary labels of the bone tissue, calculate the shortest distance from each voxel within the bone tissue region to the nearest boundary point, and obtain a structural distance map, represented as follows: , In the formula, This is a structural distance map. For the three-dimensional coordinates of the voxel, Indicates the coordinate dimension. This is the bone tissue area. For the set of boundary points; Using the relative density feature as a weight, and combining it with the structural distance map, a feature distance map representing bone structure and bone density is constructed.

[0008] Furthermore, random parameters are introduced during the construction of the feature distance map. For the weights Perturbation scaling is applied to obtain linearly scaled weights. Then, using the linearly scaled weights Multiplying the feature distance map by the structural distance map yields the feature distance map, which is represented as: , In the formula, This is the feature distance map.

[0009] Furthermore, the step of obtaining the fracture fragment region includes: Negating the feature distance map yields the negative feature distance map. and from Extract all local minimum points from the set of local minimum points. ,in For the first The nth local minimum point, as the nth Seed points; Let the voxel domain of the feature distance map be... For the voxel domain Any voxel point in Define the label of its growth region. satisfy: , In the formula, For connection With seed point The set of all paths For path Any voxel point on the surface; Based on the growth region label During the regional growth process, when two growth regions are respectively generated by seed points With seed point Expand and at voxel points When they meet, determine if the following conditions are met: , If so, then at the voxel point Establish a watershed boundary; otherwise, convert the voxel points. The seed point is assigned to the growth region where its path's maximum eigenvalue is minimized. After the region growth process is complete, a set of fracture fragment regions consisting of multiple independent and connected fracture fragment regions is obtained. ,in For the first A region of broken bone. This refers to the watershed water level threshold parameter.

[0010] Furthermore, the set of fracture fragment regions satisfies the complete partitioning and disjointness of the voxel domain: , In the formula, This represents the empty set.

[0011] Furthermore, the step of performing the random rigid transformation includes: For each fracture fragment region, the fracture fragment region with the largest volume is used as the reference block; The reference block is regarded as a stable reference. Each of the remaining fracture fragment regions is rotated and translated to obtain the homogeneous rigid transformation results of each of the remaining fracture fragment regions, forming a three-dimensional CT of the simulated fracture.

[0012] Furthermore, the rotation process is achieved by constructing a rotation matrix, wherein the steps for constructing the rotation matrix include: Set the independent random sampling parameters as follows , The threshold is a positive number; For each of the remaining fracture fragment regions, the independent random sampling parameters Define the rotation vector as the rotation angle. , is represented as: , In the formula, It is the unit axis vector; Based on the rotation vector The rotation matrix for each remaining fracture fragment region is determined using the Rodriguez formula, and is expressed as: , In the formula, For the first Rotation matrix of fracture fragment regions From the unit axis vector Constructed diagonal matrix It is an identity matrix.

[0013] Furthermore, the translation process is achieved by constructing a homogeneous translation matrix, wherein the steps for constructing the homogeneous translation matrix include: Set translation vector Set the translation vector The range of values ​​for each component is [-c, c]. The threshold is a positive number; For each remaining fracture fragment region that has completed rotation, according to the translation vector Construct a homogeneous translation matrix, represented as: , In the formula, For the first Homogeneous translation matrix for each fracture fragment region for The three components represent the first, second, third, fourth, fifth, sixth, seventh, eighth, and eleventh components, respectively. The area of ​​fractured bone fragments is in axis, shaft and Translational displacement in the axial direction.

[0014] Furthermore, the step of generating a three-dimensional CT scan simulating a fracture includes: The first Rotation matrix of fracture fragmentation areas By extending the matrix, we obtain the extended rotation matrix. , is represented as: , In the formula, It is a zero vector; For the extended rotation matrix and homogeneous translation matrix Perform matrix multiplication to obtain the first... The homogeneous rigid body transformation matrix of each fracture fragment region is expressed as: , In the formula, For the first Homogeneous rigid body transformation matrix for each fracture fragmentation region; The first Homogeneous rigid body transformation matrix for each fracture fragment region With the Multiplying all coordinates of each fracture fragment region yields a set of fracture fragment regions after random rigid transformation, forming a 3D CT scan simulating the fracture, represented as: ,in To simulate a fracture using 3D CT, For the first A fracture fragmentation region after random rigidity transformation.

[0015] Furthermore, the process of generating a three-dimensional CT scan simulating a fracture also includes: After the random rigid transformation is completed, the Marching Cube algorithm is used to perform three-dimensional reconstruction of the category labels of each fracture fragment region.

[0016] Compared with the prior art, the present invention has the following beneficial effects: (1) By introducing a three-dimensional watershed algorithm, the present invention controls the change of fracture degree through the water level hyperparameter in the algorithm. Compared with the traditional manual simulation method, the present invention can automatically and quickly generate multi-type fault simulation data.

[0017] (2) This invention directly processes three-dimensional voxel data through a three-dimensional watershed algorithm, which can completely preserve the connectivity and spatial position of the fragments. Therefore, it can output a set of independent connected fracture fragment regions with clear three-dimensional topology. By accurately reflecting the real spatial distribution of the fragments, it can provide a reliable basis for clinical assessment of the severity of fractures (such as the number, volume, and degree of displacement of fragments).

[0018] (3) In the process of region growth, this invention establishes a watershed boundary based on the expansion of seed points in two regions and the conditions satisfied when they meet. The boundary is generated at the point of equilibrium between the two regions, which corresponds exactly to the true edge of different fracture degrees. Furthermore, the voxel-level judgment further improves the segmentation accuracy. In addition, this invention establishes the boundary through growth encounter, which can determine local growth in real time, realizing the transformation of the global optimization problem into a local competition process, taking into account efficiency, accuracy and flexibility.

[0019] (4) The fracture location, number of fragments and morphological characteristics generated by the present invention are similar to those of real cases. In particular, the fracture surface can present an irregular serrated rough boundary that is closer to the real fracture than a simple regular planar cut.

[0020] (2) This invention is applicable to the generation of simulated fracture data for various bone types, and can cover areas such as the iliac bone, sacrum, and pubic ramus. It can provide more realistic training resources for clinical training, approach and operation research of fracture surgery. At the same time, this invention has good transfer and expansion capabilities, which is conducive to the construction of multi-site fracture simulation datasets.

[0021] (5) Compared with the traditional manual simulation method, the method of the present invention takes about 20 seconds to simulate each fracture data, which is about 20 times more efficient. It can significantly reduce the cost of simulation data construction and facilitate the batch generation of training data, validation data and in vitro model resources. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 The following are the binary labels of bone tissue and the corresponding target bone CT images in this invention, wherein (1) is the binary label of bone tissue of the pelvis, (2) is the CT image of the pelvis, (3) is the binary label of bone tissue of the femur, and (4) is the CT image of the femur. Figure 3 The following is a structural distance diagram of the pelvis as an example in this invention, wherein (1) is a structural distance diagram of anatomical substructure A in the pelvis, and (2) is a structural distance diagram of anatomical substructure B in the pelvis; Figure 4 The following are relative density feature diagrams of the pelvis as an example in this invention, wherein (1) is the relative density feature diagram of anatomical substructure A in the pelvis, and (2) is the relative density feature diagram of anatomical substructure B in the pelvis. Figure 5 The results of the fracture fragment region division in this invention, taking the pelvis as an example, are as follows: Region (1) and Region (4) are both first sacral fracture fragment regions, Region (2) and Region (5) are both second sacral fracture fragment regions, Region (3) and Region (6) are both third sacral fracture fragment regions, and Region (7) is the fourth sacral fracture fragment region. Figure 6 The above is a comparison chart of healthy pelvic data and pelvic fracture data of the present invention, wherein (1) is healthy pelvic data and (2) is generated pelvic fracture data. Figure 7 The above is a comparison chart of healthy femoral data and generated femoral fracture data of the present invention, wherein (1) is healthy femoral data and (2) is generated femoral fracture data. Detailed Implementation

[0023] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0024] This embodiment provides an automated, general fracture simulation data generation method based on geometric parameter calculation. This method is based on 3D CT scans of healthy target bone and its binary bone tissue labels (optionally: binary labels of anatomical substructures, such as different bone segments / cortical and cancellous bone, etc.) (see...). Figure 2 Based on user-defined fracture parameters, it automatically generates a 3D surface model (such as STL) of the target bone fracture fragments and the corresponding 3D fracture simulation CT data / labels. Figure 1 As shown, the overall process includes the following three stages: S1, calculation of bone mineral density and structural feature map.

[0025] Fracture occurrence in different bone structures is directly related to local geometric thickness and structural scale characteristics. To quantify this characteristic, an Euclidean distance transformation was performed on the binary labels of the target bone, and the shortest distance from each voxel within the bone tissue region to the nearest boundary point was calculated, resulting in a structural distance map (see...). Figure 3 A smaller distance value indicates a weaker bone structure and a higher potential risk of fracture; a larger distance value indicates a thicker structure and a lower potential risk of fracture. This process can be formally expressed as: , in, This is a structural distance map. and For the three-dimensional coordinates of the voxel, Indicates the coordinate dimension. Indicates the bone tissue area. This represents its boundary set. The calculated structure distance map... Characterizes the thickness of the bone structure.

[0026] Furthermore, studies have shown that bone mineral density (BMD) is inversely proportional to fracture risk; specifically, the lower the BMD, the higher the risk of fracture. For BMD characterization, since CT grayscale (HU) is correlated with the linear attenuation coefficient of tissue and significantly positively correlated with BMD, the method in this embodiment normalizes the CT values ​​in the pelvic structural region to the [0,1] interval to represent relative density characteristics (see...). Figure 4 ): , in The normalized CT value. Voxel CT values, and These represent the minimum and maximum CT values ​​in the original image, respectively.

[0027] In addition, to further simulate the randomness of fracture occurrence and enhance the diversity of generated data, a random parameter is introduced. Density weight Perform perturbation scaling. Specifically, linear scaling (Ensure that the weight is greater than 0, and) (This indicates that density weighting is not performed). The density factor is then multiplied by the structural distance map to obtain a unified feature map. Finally, this step outputs a feature distance map representing bone structure and density: , in, A feature distance map that integrates bone structure and density information.

[0028] S2. Fracture region segmentation based on feature distance map.

[0029] In this step, the three-dimensional watershed algorithm is used to map the feature map. The result is converted into a segmentation of fracture fragment regions (also known as fracture debris regions or fragments). First, the feature distance map is negatively evaluated, so that high-risk (weak / low-density) regions correspond to minimum grayscale values, becoming the starting points for region growth. Then, simulating a gradual rise in "water level," adjacent minimum value basins are aggregated. When the gradient difference between adjacent regions exceeds a threshold, a segmentation boundary is established at their intersection, thus obtaining topologically clear, independently connected fracture fragment regions. Specifically, the segmentation steps for this fracture fragment region are as follows: Let the negative feature distance map be... Its local minimum point set is .

[0030] Let the voxel domain (the set of all voxel points) of the 3D feature distance map be... Any voxel point is denoted as For any voxel point Define the label of its growth region. satisfy: , in Indicates connection With seed point The set of all paths For path Any voxel point on the path. The above definition reflects the "minimize the highest water level on the path" principle of the watershed: voxel 𝑚 is assigned to the region corresponding to the seed point that minimizes the maximum "height" on the path.

[0031] Combined with the growth area label During the region growth process, when two regions are respectively generated by seed points and Expand and in voxels When they meet, if the following conditions are met: , in If the watershed water level threshold parameter is used, then in voxels... Establish a watershed boundary at that location; otherwise, The region containing the seed point with the smallest maximum feature value in the path is assigned.

[0032] The algorithm ultimately outputs a set of fracture fragment regions. This satisfies the complete partitioning and disjointness of the voxel space: , In the formula, This represents the empty set.

[0033] Finally, the results of the regional division of the fracture fragments were obtained (see...). Figure 5 By adjusting the water level hyperparameters of the watershed algorithm. This can control changes in the severity of fractures. Water level parameters. The value of is an integer greater than 0. The smaller the value, the more bone fragments are generated. A higher value results in smaller fragments. Generally recommended... The value is 3.

[0034] S3. Rigid body transformation and model reconstruction of the bone fragment region.

[0035] After dividing the fracture fragment region, random rigid body transformations are applied to each connected block (fracture fragment region) to realistically simulate fracture displacement. First, a reference block is determined according to its volume: the largest fragment is regarded as a stable reference and kept stationary, while the remaining fragments are rotated and translated respectively.

[0036] Rotation uses rotation vector It means that among them The rotation angle is... For a small positive threshold, The vector is the unit axis; the corresponding rotation matrix is ​​given by the Rodriguez formula: , in For the first Rotation matrix of fracture fragment regions It is the identity matrix. From vector Construct a diagonal matrix.

[0037] Translation by vector Description: Each component takes values ​​[-c, c], where c is a small positive threshold. Its homogeneous translation matrix is ​​written as: , in, For the first Homogeneous translation matrix for each fracture fragment region for The three components represent the first, second, third, fourth, fifth, and sixth components, respectively. The area of ​​fractured bone fragments is in axis, shaft and Translational displacement in the axial direction.

[0038] Therefore, the first The homogeneous rigid body transformation of the fragments is as follows: , in, For the first The homogeneous rigid body transformation matrix of the fragments Indicates the first Rotation matrix of fracture fragmentation areas Expand to accommodate 4x4 Perform multiplication.

[0039] Except for the reference block, each fragment is sampled independently and randomly. ∈[-15°, 15°], the translation range is ∈[-5mm,5mm].

[0040] Finally, the first Homogeneous rigid body transformation matrix for each fracture fragment region With the Multiplying all coordinates of each fracture fragment region yields a set of fracture fragment regions after random rigid transformation, forming a three-dimensional CT image simulating a fracture: , For the first A fracture fragmentation region after random rigidity transformation.

[0041] After the rigid transformation is completed, the Marching Cube algorithm is used to perform three-dimensional reconstruction of the category labels of each fracture fragment region, generating a three-dimensional fracture model in STL format.

[0042] Preferably, in this embodiment, a corresponding rigid body transformation is applied based on the location of each fragment type in the CT image to generate a CT image simulating a fracture.

[0043] like Figure 6 As shown, this embodiment uses the above method to calculate and generate data for a bilateral iliac wing-pubis fracture from healthy pelvic data. The data generation takes 36 seconds, and the generated result is highly similar to the real case.

[0044] like Figure 7As shown in the figure, this embodiment uses the above method to calculate and generate a femoral fracture data from healthy femoral data. The data generation takes 32 seconds, and the generated result is highly similar to the real case.

[0045] Experimental verification shows that, compared with traditional manual simulation methods, the embodiments of this invention can automatically and quickly generate simulation data for multiple types of fractures by setting different fracture hyperparameters, and the generation process is parameterized and reproducible. Taking pelvic fracture as a representative case, when the water level parameter... When the parameter is 3, double-body fractures account for 26.1%, triple-body fractures for 22.7%, and multi-body fractures for 28.8% of the generated data. Furthermore, fractures can occur in different anatomical regions: iliac fractures account for 89.5%, acetabular fractures for 90.7%, and pubic fractures for 86.4%. These results demonstrate that the embodiments of the present invention can effectively cover fracture types and distributions through parameter adjustment, making them suitable for large-scale simulation data generation. Moreover, the generation efficiency of this embodiment is significantly improved, facilitating large-scale dataset construction. Taking pelvic fractures as a representative case, the method in this embodiment takes approximately 20 seconds per fracture data simulation, representing an efficiency improvement of approximately 20 times. This significantly reduces the cost of simulation data construction and facilitates the batch generation of training data, validation data, and in vitro model resources.

[0046] Furthermore, the simulated fracture data generated in this embodiment exhibits high realism and is suitable for clinical training and algorithm validation. The generated fracture location, number of fragments, and morphological characteristics closely resemble the distribution of real cases. In particular, the fracture surface presents an irregular, jagged, rough boundary that more closely resembles a real fracture, rather than a simple, regular planar cut. Taking pelvic fractures as a representative case, the generated fragments are widely distributed, covering areas such as the ilium, sacrum, and pubic ramus, thus providing more realistic training resources for clinical training, approach, and operational research in fracture surgery. Simultaneously, this embodiment of the invention is based on bone tissue labels and CT features, and its methodological framework has the ability to be transferred and extended to other bone types and anatomical locations, which is beneficial for constructing multi-site fracture simulation datasets.

[0047] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0048] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0049] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0050] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0051] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0052] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0053] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An automated, general-purpose fracture simulation data generation method based on geometric parameter calculation, characterized in that, Includes the following steps: Obtain 3D CT scans of healthy target bones and their binary labels, and construct a feature distance map representing bone structure and bone density. The feature distance map is converted into multiple independent and connected fracture fragment regions using a three-dimensional watershed algorithm. The three-dimensional watershed algorithm controls the change in fracture severity by adjusting the water level hyperparameter. For each fracture fragment region, a random rigidity transformation is performed to generate a three-dimensional CT scan simulating the fracture.

2. The automated general fracture simulation data generation method based on geometric parameter calculation according to claim 1, characterized in that, The steps for constructing the feature distance map include: The CT values ​​of the three-dimensional CT scans of the healthy target bone are normalized to the [0,1] interval to represent the relative density characteristics of the bone tissue region, wherein the relative density characteristics are expressed as: ， In the formula, The normalized CT value, i.e., the relative density feature. Voxel CT values, and These represent the minimum and maximum CT values ​​in the original 3D CT image of the healthy target bone, respectively; Perform a Euclidean distance transformation on the binary labels of the bone tissue, calculate the shortest distance from each voxel within the bone tissue region to the nearest boundary point, and obtain a structural distance map, represented as follows: ， In the formula, This is a structural distance map. For the three-dimensional coordinates of the voxel, Indicates the coordinate dimension. This is the bone tissue area. For the set of boundary points; Using the relative density feature as a weight, and combining it with the structural distance map, a feature distance map representing bone structure and bone density is constructed.

3. The automated general fracture simulation data generation method based on geometric parameter calculation according to claim 2, characterized in that, Random parameters were also introduced during the construction of the feature distance map. For the weights Perturbation scaling is applied to obtain linearly scaled weights. Then, using the linearly scaled weights Multiplying the feature distance map by the structural distance map yields the feature distance map, which is represented as: ， In the formula, This is a feature distance map.

4. The automated general fracture simulation data generation method based on geometric parameter calculation according to claim 1, characterized in that, The steps for obtaining the fracture fragment region include: Negating the feature distance map yields the negative feature distance map. and from Extract all local minimum points from the set of local minimum points. ,in For the first The nth local minimum point, as the nth Seed points; Let the voxel domain of the feature distance map be... For the voxel domain Any voxel point in Define the label of its growth region. satisfy: ， In the formula, For connection With seed point The set of all paths For path Any voxel point on the surface; Based on the growth region label During the regional growth process, when two growth regions are respectively generated by seed points With seed point Expand and at voxel points When they meet, determine if the following conditions are met: ， If so, then at the voxel point Establish a watershed boundary; otherwise, convert the voxel points. The seed point is assigned to the growth region where its path's maximum eigenvalue is minimized. After the region growth process is complete, a set of fracture fragment regions consisting of multiple independent and connected fracture fragment regions is obtained. ,in For the first A region of broken bone. This refers to the watershed water level threshold parameter.

5. The automated general fracture simulation data generation method based on geometric parameter calculation according to claim 4, characterized in that, The set of fracture fragment regions satisfies the complete partitioning and disjointness of the voxel domain: ， In the formula, This represents the empty set.

6. The automated general fracture simulation data generation method based on geometric parameter calculation according to claim 1, characterized in that, The steps for performing the random rigid transformation include: For each fracture fragment region, the fracture fragment region with the largest volume is used as the reference block; The reference block is regarded as a stable reference. Each of the remaining fracture fragment regions is rotated and translated to obtain the homogeneous rigid transformation results of each of the remaining fracture fragment regions, forming a three-dimensional CT of the simulated fracture.

7. The automated general fracture simulation data generation method based on geometric parameter calculation according to claim 6, characterized in that, The rotation process is achieved by constructing a rotation matrix, wherein the steps for constructing the rotation matrix include: Set the independent random sampling parameters as follows , The threshold is a positive number; For each of the remaining fracture fragment regions, the independent random sampling parameters Define the rotation vector as the rotation angle. , is represented as: ， In the formula, It is the unit axis vector; Based on the rotation vector The rotation matrix for each remaining fracture fragment region is determined using the Rodriguez formula, and is expressed as: ， In the formula, For the first Rotation matrix of fracture fragment regions From the unit axis vector Constructed diagonal matrix It is an identity matrix.

8. The automated general fracture simulation data generation method based on geometric parameter calculation according to claim 6, characterized in that, The translation process is achieved by constructing a homogeneous translation matrix, wherein the steps for constructing the homogeneous translation matrix include: Set translation vector And set the translation vector The range of values ​​for each component is [-c, c]. The threshold is a positive number; For each remaining fracture fragment region that has completed rotation, according to the translation vector Construct a homogeneous translation matrix, represented as: ， In the formula, For the first Homogeneous translation matrix for each fracture fragment region for The three components represent the first, second, third, fourth, fifth, sixth, seventh, eighth, and eleventh components, respectively. The area of ​​fractured bone fragments is in axis, shaft and Translational displacement in the axial direction.

9. The automated general fracture simulation data generation method based on geometric parameter calculation according to claim 6, 7, or 8, characterized in that, The steps for generating a three-dimensional CT scan simulating a fracture include: The first Rotation matrix of fracture fragmentation areas By extending the matrix, we obtain the extended rotation matrix. , is represented as: ， In the formula, It is a zero vector; For the extended rotation matrix Homogeneous translation matrix Perform matrix multiplication to obtain the first... The homogeneous rigid body transformation matrix of each fracture fragment region is expressed as: ， In the formula, For the first Homogeneous rigid body transformation matrix for each fracture fragmentation region; The first Homogeneous rigid body transformation matrix for each fracture fragment region With the Multiplying all coordinates of each fracture fragment region yields a set of fracture fragment regions after random rigid transformation, forming a 3D CT scan simulating the fracture, represented as: ,in To simulate a fracture using 3D CT, For the first A fracture fragmentation region after random rigidity transformation.

10. The automated general fracture simulation data generation method based on geometric parameter calculation according to claim 1, characterized in that, The process of generating a 3D CT scan simulating a fracture also includes: After the random rigid transformation is completed, the Marching Cube algorithm is used to perform three-dimensional reconstruction of the category labels of each fracture fragment region.

Images