Medical instrument visual simulation method based on data identification

Abstract

The invention relates to the technical field of image simulation modeling, in particular to a medical instrument visual simulation method based on data recognition. The method comprises the following steps: carrying out rotation angle calibration on a medical instrument, and carrying out rotation angle X-ray projection data acquisition to obtain multi-view projection image sequence data; performing three-dimensional skeleton model optimization according to the multi-view projection image sequence data to generate an optimized skeleton scanning three-dimensional model; performing potential fracture probability processing on the multi-view projection image sequence data based on the optimized skeleton scanning three-dimensional model to generate potential fracture probability data; and carrying out fracture area marking on the optimized bone scanning three-dimensional model through the potential fracture probability data, and carrying out part stress analogue simulation to obtain a three-dimensional visual fracture report. According to the method, accurate marking of the fracture area and visual report generation are realized through reconstruction of the three-dimensional space model of the scanned image, and the accuracy and efficiency of fracture detection are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image simulation modeling, and in particular to a medical device visualization simulation method based on data recognition. Background Art

[0002] Medical imaging technology is an important part of modern medical diagnosis. Among them, X-ray imaging technology has been widely used in bone injury diagnosis due to its advantages such as speed, convenience and low cost. Through X-ray images, doctors can observe the morphology, structure and presence of fractures and other lesions of bones, and assist doctors in diagnosing fractures. X-ray images are two-dimensional projection images, which cannot fully display the three-dimensional structure of bones. The fracture line is easily blocked by other tissues, making it difficult to accurately identify fractures. Some subtle bone fractures, such as fatigue fractures, often have very small fracture lines that are easily ignored, resulting in missed fractures. On the other hand, X-ray images are easily affected by various factors during the imaging process, such as patient position, exposure parameters, scattered rays, etc., resulting in noise and artifacts in the image. The fracture line is easily covered by noise, making it difficult to accurately identify fractures and leading to misdiagnosis. However, the existing two-dimensional image visualization simulation method can only display two-dimensional X-ray images, and cannot intuitively display the spatial position and morphology of fractures. It is difficult to handle complex bone structures, especially when identifying fractures in subtle cracks and overlapping bone areas, and the accuracy is low. Summary of the invention

[0003] Based on this, the present invention provides a medical device visualization simulation method based on data recognition to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a medical device visualization simulation method based on data recognition includes the following steps:

[0005] Step S1: calibrating the rotation angle of the medical device to obtain rotation angle sequence data of the medical device; collecting rotation angle X-ray projection data based on the rotation angle sequence data of the medical device, and integrating the angle sequence to obtain multi-view projection image sequence data;

[0006] Step S2: Optimizing the three-dimensional skeleton model according to the multi-view projection image sequence data to generate an optimized three-dimensional skeleton scan model; simulating the imaging process based on the optimized three-dimensional skeleton scan model to obtain virtual multi-view projection data; performing maximum inter-class variance image segmentation on the multi-view projection image sequence data through the virtual multi-view projection data, and performing image edge enhancement processing to generate edge enhanced scan image sequence data;

[0007] Step S3: extracting bone edge features from the edge-enhanced scan image sequence data to obtain bone edge feature data; performing potential fracture probability processing based on the bone edge feature data to generate potential fracture probability data; marking the fracture area of ​​the optimized bone scan 3D model using the potential fracture probability data to obtain a fracture marked 3D model;

[0008] Step S4: Render and visualize the fracture marked three-dimensional model, and perform force simulation on the parts to generate fracture stress and strain simulation data; generate a visualization report based on the fracture stress and strain simulation data, thereby obtaining a three-dimensional visualization fracture report.

[0009] The present invention collects multi-view X-ray projection data by rotating medical equipment and integrates the angle sequence, thus overcoming the limitation that traditional two-dimensional X-ray imaging can only provide single-view information. The three-dimensional bone model reconstruction using multi-view projection data can more completely display the three-dimensional structure of the bone, avoid the problem that the fracture line in the two-dimensional projection image is blocked by other tissues, and can clearly display the spatial position and morphology of the fracture, especially for complex bone structures. The real projection image is subjected to maximum inter-class variance segmentation and edge enhancement processing through virtual projection data, which effectively reduces the influence of noise and artifacts, highlights the bone edge features, and makes the subtle fracture lines easier to identify. Especially for subtle bone fractures such as fatigue fractures, traditional two-dimensional X-ray images are often difficult to display clearly, and this method can effectively improve the visualization of subtle fracture lines and reduce the risk of missed diagnosis through image enhancement processing. Potential fracture probability processing is performed based on bone edge feature data, and the fracture area is marked on the three-dimensional bone model, which elevates the fracture diagnosis from simple morphological observation to the level of quantitative analysis. By calculating the potential fracture probability, the possibility of fracture can be more objectively evaluated, avoiding errors caused by subjective judgment. At the same time, the fracture area marking function can intuitively display the location, range and morphology of the fracture, making it convenient for doctors to quickly locate the fracture area and improve diagnostic efficiency. Therefore, the medical device visualization simulation method based on data recognition of the present invention significantly improves the accuracy of fracture diagnosis through three-dimensional bone model optimization and imaging process simulation. By using edge enhancement processing and potential fracture probability calculation, the recognition ability of subtle fracture lines is effectively improved and the risk of missed diagnosis is reduced. By marking the fracture area and generating a three-dimensional visualization report, doctors are provided with intuitive fracture information, thereby achieving efficient and accurate fracture diagnosis.

[0010] Preferably, step S1 comprises the following steps:

[0011] Step S11: calibrating the rotation angle of the medical device to obtain rotation angle sequence data of the medical device;

[0012] Step S12: controlling the cone beam CT transmitter to rotate at a preset angle step size based on the medical device rotation angle sequence data, and collecting rotation angle X-ray projection data to generate single-view projection image sequence data;

[0013] Step S13: performing single-view projection preprocessing on the single-view projection image sequence data to generate corrected projection image sequence data;

[0014] Step S14: Angle sequential integration of the correction projection image sequence data is performed through the medical device rotation angle sequence data, thereby obtaining multi-view projection image sequence data.

[0015] The present invention accurately calibrates the rotation angle of the medical device to ensure the accuracy of subsequent image reconstruction. By controlling the cone beam CT transmitter to rotate according to a preset angle step and collect X-ray projection data, rich multi-view information is obtained, overcoming the limitation of the single-view of traditional X-ray imaging, effectively avoiding the problem of fracture line being blocked, and more comprehensively showing the morphology and spatial position of the fracture, especially for complex bone structures. The collected single-view projection image sequence is preprocessed, such as correcting geometric distortion, removing scattering noise, etc., to improve the image quality. This is crucial for subsequent three-dimensional reconstruction, because high-quality projection data is the basis for accurately reconstructing a three-dimensional model. The corrected image sequence is integrated in an angle sequence to form a complete multi-view projection data set. Through multi-view imaging and image preprocessing, the problem of missed diagnosis and misdiagnosis of fractures caused by the single projection angle and limited image quality in traditional X-ray imaging is effectively solved. In particular, for subtle fracture lines, such as fatigue fractures, after image preprocessing, their display on the projection image is clearer and easier to identify, thereby reducing the risk of missed diagnosis.

[0016] Preferably, step S13 comprises the following steps:

[0017] Step S131: repairing image bad pixels on the single-view projection image sequence data to generate repaired projection image sequence data;

[0018] Step S132: performing noise level assessment according to the restored projection image sequence data, and performing noise suppression using adaptive median filtering to generate denoised projection image sequence data;

[0019] Step S133: Acquire scanning geometric parameters of the cone-beam CT system and scanning imaging detector position data;

[0020] Step S134: establishing a geometric mapping relationship between the projection image and the real object according to the cone beam CT system scanning geometric parameters and the scanning imaging detector position data, thereby obtaining projection image-object mapping relationship data;

[0021] Step S135: performing polynomial fitting processing based on the projection image-object mapping relationship data, and performing geometric distortion correction on the denoised projection image sequence data to generate corrected projection image sequence data.

[0022] The present invention repairs bad pixels in a single-view projection image sequence, effectively eliminating the influence of bad pixels that appear during image acquisition on subsequent processing, and ensuring the integrity of image data. Then, noise suppression is performed through adaptive median filtering, which effectively reduces the image noise level, improves the signal-to-noise ratio of the image, and makes the bone structure clearer, especially for subtle fracture lines, which are more obvious in the image, reducing the risk of being masked by noise. The cone-beam CT system scanning geometric parameters and detector position data are introduced to establish an accurate geometric mapping relationship between the projection image and the real object. This correction method based on the physical model can more accurately correct the geometric distortion in the image and avoid the errors caused by the empirical model. The geometric distortion correction of the denoised projection image is performed through polynomial fitting, which further improves the geometric accuracy of the image and ensures the accuracy of subsequent three-dimensional reconstruction.

[0023] Preferably, step S2 comprises the following steps:

[0024] Step S21: performing filtered back-projection reconstruction processing according to the multi-view projection image sequence data to generate an initial bone scan three-dimensional model;

[0025] Step S22: optimizing the initial three-dimensional skeleton scan model to generate an optimized three-dimensional skeleton scan model;

[0026] Step S23: performing an X-ray imaging process simulation based on the optimized bone scanning three-dimensional model, thereby obtaining virtual multi-view projection data;

[0027] Step S24: performing multi-scale wavelet transform processing on the virtual multi-view projection data to generate a multi-scale wavelet coefficient matrix; performing image local edge intensity calculation on the multi-view projection image sequence data using the multi-scale wavelet coefficient matrix to generate image local area edge intensity data;

[0028] Step S25: using the edge intensity data of the local area of ​​the image to perform maximum inter-class variance image segmentation on the multi-view projection image sequence data, and performing image edge enhancement processing to generate edge enhanced scanned image sequence data.

[0029] The present invention uses multi-view projection image sequence data for filtered back-projection reconstruction to generate an initial three-dimensional bone model, converting two-dimensional image information into three-dimensional spatial information, so that doctors can more intuitively observe the morphology and structure of the bones. Subsequently, the initial model is optimized to further improve the accuracy and integrity of the model. Multi-scale wavelet transform is performed using virtual projection data, and combined with the calculation of local edge strength of the image, more refined image segmentation and edge enhancement are achieved. Multi-scale wavelet transform can effectively capture the local detail features of the image, while the local edge strength of the image highlights the bone edge information. Based on this information, maximum inter-class variance image segmentation is performed, which can more accurately segment the bone area and perform edge enhancement processing to make the bone edge clearer, especially for subtle fracture lines, the degree of visualization is significantly improved, reducing the risk of missed diagnosis.

[0030] Preferably, step S22 includes the following steps:

[0031] Step S221: performing scanned bone type recognition on the initial bone scanned three-dimensional model to generate scanned bone type data;

[0032] Step S222: Obtaining bone CT scan sample data according to the scanned bone type data;

[0033] Step S223: performing principal component analysis on the bone CT scan sample data and extracting bone geometric feature points to obtain bone sample feature point data;

[0034] Step S224: performing non-rigid part registration according to the bone sample feature point data, and performing bone shape change pattern statistics to generate bone shape change statistical data;

[0035] Step S225: defining shape constraint parameters and ranges based on the skeletal shape change statistics, thereby establishing a skeletal shape constraint model;

[0036] Step S226: Use the bone shape constraint model to perform shape constraint iterative optimization on the initial bone scan three-dimensional model to generate an optimized bone scan three-dimensional model.

[0037] The present invention first identifies the scanned bone type and obtains corresponding CT scan sample data according to the type. This step realizes the personalization of the model optimization process, adopts different optimization strategies for different types of bones, avoids the "one size fits all" processing method, and better retains the individual differences of bones. The geometric features of the bones are extracted from the sample data, and non-rigid part registration is performed to count the change patterns of the bone shape. This provides a data basis for establishing a bone shape constraint model, so that the model can more accurately reflect the morphological change law of the real bone. Based on the statistical data of bone shape change, shape constraint parameters and ranges are defined to establish a bone shape constraint model. The model is not based on simple geometric constraints, but is obtained based on a large amount of sample data statistical analysis, so that the shape of the three-dimensional model can be more effectively constrained. The initial three-dimensional model is iteratively optimized using the bone shape constraint model to generate a final optimized model. Through shape constraints, the optimized model is closer to the morphology of the real bone, especially in some complex bone structures, and can more accurately reflect the detailed features of the bone, such as small bone cracks or subtle concave and convex on the bone surface.

[0038] Preferably, step S226 includes the following steps:

[0039] Performing vertex three-dimensional control mesh parameterization on the initial bone scan three-dimensional model to generate scan three-dimensional model parameters;

[0040] Performing viewing angle parameter projection according to the scanned 3D model parameters, and performing projection error calculation on multi-view projection image sequence data to generate projection error data;

[0041] Calculating the model deformation vector on the projection error data to generate the model deformation vector data;

[0042] Calculate the shape regularization term according to the bone shape constraint model to generate shape regularization data;

[0043] Perform iterative optimization of model parameters based on model deformation vector data and shape regularization data to generate model adjustment optimization parameters;

[0044] The skeleton model parameters of the initial skeleton scan three-dimensional model are updated by adjusting the optimization parameters of the model, and the skeleton model deformation is adjusted to obtain an optimized skeleton scan three-dimensional model.

[0045] The present invention parameterizes the initial three-dimensional model, and links the two-dimensional projection image information with the three-dimensional model through perspective parameter projection and error calculation. The projection error data reflects the difference between the three-dimensional model and the actual X-ray image, and provides directional guidance for the optimization of the model. By calculating the model deformation vector, the projection error is converted into the adjustment amount of the three-dimensional model, and the correction of the three-dimensional model based on the two-dimensional image information is realized. Based on the pre-established bone shape constraint model, the shape regularization term is calculated, and it is combined with the model deformation vector in the iterative optimization process of the model parameters. This combination makes the optimization of the model not only driven by the projection error, but also restricted by the shape constraint, thereby ensuring that the optimized model not only conforms to the X-ray image information, but also maintains a reasonable bone morphology. Through iterative optimization, the model parameters are continuously updated, and the three-dimensional bone model is also deformed and adjusted accordingly, and finally the optimized three-dimensional model is obtained. This iterative optimization method can gradually reduce the projection error and make the shape of the model more reasonable, thereby improving the accuracy and reliability of the model.

[0046] Preferably, step S24 comprises the following steps:

[0047] Step S241: performing multi-scale wavelet transform decomposition on the virtual multi-view projection data to generate a multi-scale wavelet coefficient matrix;

[0048] Step S242: performing wavelet decomposition sub-band energy calculation according to the multi-scale wavelet coefficient matrix to generate wavelet sub-band energy data;

[0049] Step S243: sorting the wavelet subband energy data in ascending order, and calculating the energy difference between adjacent wavelet subbands to generate subband energy difference data;

[0050] Step S244: screening the maximum energy difference according to the wavelet sub-band energy data, and taking it as the sub-band frequency threshold; performing sub-band classification on the wavelet sub-band energy data according to the sub-band frequency threshold, and obtaining low-frequency sub-band data and high-frequency sub-band data respectively;

[0051] Step S245: performing smoothing processing on the low-frequency sub-band data to obtain smoothed low-frequency sub-band data; performing sub-band coefficient enhancement processing on the high-frequency sub-band data to generate enhanced high-frequency sub-band data;

[0052] Step S246: extracting wavelet coefficient features according to the enhanced high-frequency sub-band data and the smoothed low-frequency sub-band data, and constructing an edge detection operator, thereby obtaining an edge detection operator;

[0053] Step S247: Use the edge detection operator to perform edge feature guided detection on the multi-view projection image sequence data, and perform local edge strength calculation to generate edge strength data of the local area of ​​the image.

[0054] The present invention performs multi-scale wavelet transform decomposition on virtual multi-view projection data and calculates the energy of each sub-band. By analyzing the difference in sub-band energy, the sub-band with the largest energy difference is selected as the frequency threshold, and the sub-band is divided into two categories: low frequency and high frequency. The low-frequency sub-band mainly contains the smoothing information of the image, while the high-frequency sub-band contains the details and edge information of the image. Smoothing the low-frequency sub-band can effectively remove the noise in the image while retaining the overall structural information of the bone. Performing coefficient enhancement processing on the high-frequency sub-band can highlight the detailed features of the bone edge and make the subtle fracture line more obvious. The constructed edge detection operator is used to perform edge feature guided detection and local edge strength calculation on the multi-view projection image sequence data, generating a clearer and more accurate edge image. This edge enhancement method based on virtual projection data and multi-scale wavelet transform effectively improves the visualization of bone edges in X-ray images.

[0055] Preferably, step S3 comprises the following steps:

[0056] Step S31: performing bone region segmentation according to edge enhanced scan image sequence data to obtain bone region image data;

[0057] Step S32: extracting bone edge features from the bone region image data to obtain bone edge feature data;

[0058] Step S33: using a preset support vector machine to calculate the potential fracture probability of the bone edge feature data to generate potential fracture probability data;

[0059] Step S34: performing fracture area mask processing according to the potential fracture probability data, and marking the fracture area of ​​the optimized bone scan three-dimensional model, thereby obtaining a fracture marked three-dimensional model.

[0060] The present invention performs bone region segmentation on the edge-enhanced scanned image sequence and extracts the bone region image data. This effectively eliminates the interference of irrelevant areas, so that the subsequent feature extraction and analysis are more focused on the bone region, and the processing efficiency and accuracy are improved. The extracted bone edge feature data is used to calculate the potential fracture probability using a pre-trained support vector machine. The support vector machine is a powerful machine learning algorithm that can effectively identify complex patterns and perform probability predictions. By analyzing the bone edge features, the support vector machine can determine whether there is a fracture in the area and give a corresponding probability value. A fracture region mask is generated based on the potential fracture probability data, and the fracture region is marked on the optimized three-dimensional bone model, which realizes intuitive visualization of the fracture site and provides clear three-dimensional spatial positioning.

[0061] Preferably, step S32 includes the following steps:

[0062] Step S321: performing edge Gaussian difference detection on the bone region image data to generate bone edge image data;

[0063] Step S322: performing edge gradient calculation on the bone edge image data to generate bone edge gradient data;

[0064] Step S323: Calculate the edge curvature according to the bone edge gradient data to obtain the bone edge curvature data;

[0065] Step S324: performing edge direction analysis according to the bone edge gradient data to obtain bone edge direction data;

[0066] Step S325: evaluating edge point continuity according to the bone edge gradient data to obtain bone edge continuity data;

[0067] Step S326: integrating the bone edge curvature data, the bone edge direction data and the bone edge continuity data to obtain the bone edge feature data.

[0068] The present invention uses an edge Gaussian difference detection method to extract bone edge image data, which can effectively suppress noise and accurately locate the edge position. On this basis, the edge gradient, curvature, direction and continuity are further calculated to achieve a multi-dimensional description of the bone edge features. The edge gradient reflects the intensity change of the edge, the curvature describes the bending degree of the edge, the direction indicates the direction of the edge, and the continuity evaluates the integrity of the edge. These features reflect the morphological characteristics of the bone edge from different angles. Integrating them can more comprehensively describe the geometric properties of the bone edge. For example, the fracture edge usually shows characteristics such as edge gradient mutation, abnormal curvature, and discontinuous direction. By comprehensively analyzing these features, the fracture area can be more accurately identified and the risk of misdiagnosis can be reduced. The evaluation of the continuity of the bone edge can effectively distinguish the natural edge of the bone from the pseudo-edge caused by imaging artifacts or noise, further improving the accuracy and reliability of feature extraction. This is particularly important for the diagnosis of subtle fractures, because the edges of subtle fractures are often discontinuous and easily masked by noise or artifacts.

[0069] Preferably, step S4 comprises the following steps:

[0070] Step S41: Rendering and visualizing the fracture mark three-dimensional model to generate a visualized fracture three-dimensional model;

[0071] Step S42: performing meshing processing on the fracture site based on the visualized three-dimensional fracture model, and performing finite element model processing to obtain a finite element simulation model;

[0072] Step S43: performing force simulation on the parts according to the finite element simulation model to generate fracture stress and strain simulation data;

[0073] Step S44: perform a fracture risk assessment on the fracture marker three-dimensional model using the fracture stress-strain simulation data, and generate a visualization report, thereby obtaining a three-dimensional visualization fracture report.

[0074] The present invention renders and visualizes the fracture mark three-dimensional model to generate a clear and intuitive three-dimensional fracture model, so that doctors can observe the morphology and spatial position of the fracture site from any angle, making up for the defect of insufficient information in traditional two-dimensional X-ray images. The fracture site is gridded and a finite element simulation model is established. This provides a basis for subsequent force simulation. Through finite element analysis, the stress and strain distribution of bones under different stress states can be simulated, so as to more accurately evaluate the stability and risk level of the fracture. Based on the finite element simulation model, the site stress simulation is performed to generate fracture stress and strain simulation data. These data can quantitatively reflect the stress conditions of the fracture site, and judge the type of fracture, such as transverse fracture, oblique fracture or spiral fracture, according to the morphological characteristics of the fracture, such as the direction and shape of the fracture line. According to the location and range of the fracture, the degree of the fracture is judged, such as mild fracture, moderate fracture or severe fracture. According to the stress and strain data, the stability of the fracture is judged, such as stable fracture or unstable fracture. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 A schematic diagram of the steps of a medical device visualization simulation method based on data recognition according to the present invention;

[0076] Figure 2 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0077] Figure 3 for Figure 1 Detailed implementation steps of step S4 in FIG.

[0078] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0079] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are 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 technicians in this field without creative work are within the scope of protection of the present invention.

[0080] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0081] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0082] To achieve this, please refer to Figures 1 to 3 The present invention provides a medical device visualization simulation method based on data recognition, comprising the following steps:

[0083] Step S1: calibrating the rotation angle of the medical device to obtain rotation angle sequence data of the medical device; collecting rotation angle X-ray projection data based on the rotation angle sequence data of the medical device, and integrating the angle sequence to obtain multi-view projection image sequence data;

[0084] Step S2: Optimizing the three-dimensional skeleton model according to the multi-view projection image sequence data to generate an optimized three-dimensional skeleton scan model; simulating the imaging process based on the optimized three-dimensional skeleton scan model to obtain virtual multi-view projection data; performing maximum inter-class variance image segmentation on the multi-view projection image sequence data through the virtual multi-view projection data, and performing image edge enhancement processing to generate edge enhanced scan image sequence data;

[0085] Step S3: extracting bone edge features from the edge-enhanced scan image sequence data to obtain bone edge feature data; performing potential fracture probability processing based on the bone edge feature data to generate potential fracture probability data; marking the fracture area of ​​the optimized bone scan 3D model using the potential fracture probability data to obtain a fracture marked 3D model;

[0086] Step S4: Render and visualize the fracture marked three-dimensional model, and perform force simulation on the parts to generate fracture stress and strain simulation data; generate a visualization report based on the fracture stress and strain simulation data, thereby obtaining a three-dimensional visualization fracture report.

[0087] In the embodiment of the present invention, reference Figure 1 FIG. 1 is a schematic diagram of a process flow of a medical device visualization simulation method based on data identification according to the present invention. In this embodiment, the medical device visualization simulation method based on data identification includes the following steps:

[0088] Step S1: calibrating the rotation angle of the medical device to obtain rotation angle sequence data of the medical device; collecting rotation angle X-ray projection data based on the rotation angle sequence data of the medical device, and integrating the angle sequence to obtain multi-view projection image sequence data;

[0089] In an embodiment of the present invention, a mechanical arm equipped with an X-ray transmitter and a receiver is used as an example to calibrate the rotation angle of a medical device. This can be achieved by installing an angle sensor on the medical device, which can accurately measure the rotation angle of the device. A series of rotation angles are preset, for example, starting from 0 degrees, rotating every 1 degree until 360 degrees. After the angle calibration is completed, the medical device is aligned with the area to be scanned. The X-ray machine is started, and the medical device is rotated in sequence according to the preset angle sequence and X-ray projection data is collected. Each time it rotates to a preset angle, the X-ray machine emits X-rays, which penetrate the area to be scanned, and are received and recorded by the detector. For example, at an angle of 0 degrees, the first frame of X-ray projection data is collected; at an angle of 1 degree, the second frame of X-ray projection data is collected, and so on. The collected X-ray projection data will be accompanied by corresponding angle information. After the data collection is completed, the angle sequence needs to be integrated. Due to time delays or data transmission problems in the collection process, the data sequence is disordered. Therefore, according to the angle information accompanying the acquisition, all X-ray projection data are reordered in order from small to large angles to generate a continuous multi-view projection image sequence data, which represents the X-ray image of the area to be scanned observed from different angles.

[0090] Step S2: Optimizing the three-dimensional skeleton model according to the multi-view projection image sequence data to generate an optimized three-dimensional skeleton scan model; simulating the imaging process based on the optimized three-dimensional skeleton scan model to obtain virtual multi-view projection data; performing maximum inter-class variance image segmentation on the multi-view projection image sequence data through the virtual multi-view projection data, and performing image edge enhancement processing to generate edge enhanced scan image sequence data;

[0091] In an embodiment of the present invention, after obtaining multi-view projection image sequence data, the three-dimensional skeleton model is first optimized. The initial three-dimensional skeleton model can come from a pre-established general skeleton model database, or a rough model obtained by other imaging technologies (such as CT). The multi-view projection image sequence data is registered with the initial skeleton model, and the shape and position of the skeleton model are adjusted by an iterative algorithm to make it as consistent as possible with the projection data. The optimization process can adopt an optimization algorithm based on image gray value differences, or an optimization algorithm based on bone anatomical constraints. For example, using the gradient descent method, the model parameters are continuously adjusted to minimize the difference between the model projection and the actual projection. The optimized skeleton model is called an optimized skeleton scanning three-dimensional model. Then, the imaging process is simulated based on the optimized skeleton scanning three-dimensional model. Computer graphics technology is used to simulate the process of an X-ray machine scanning the optimized skeleton model at different angles to generate virtual multi-view projection data. The virtual projection data is the same as the actual collected projection data in terms of imaging principle, and can be used to assist image processing. Next, the virtual multi-view projection data is used to perform maximum inter-class variance image segmentation on the multi-view projection image sequence data. The maximum inter-class variance method is a commonly used image segmentation method that can automatically select the optimal threshold to divide the image into the target area (bone) and the background area. By calculating the inter-class variance under different thresholds, the threshold with the maximum variance is selected as the segmentation threshold. After the segmentation is completed, the segmented bone image is edge enhanced. Edge enhancement can use various image filtering algorithms, such as Sobel operator, Canny operator, etc., to highlight the bone edge. The processed image sequence data is called edge enhanced scanned image sequence data.

[0092] Step S3: extracting bone edge features from the edge-enhanced scan image sequence data to obtain bone edge feature data; performing potential fracture probability processing based on the bone edge feature data to generate potential fracture probability data; marking the fracture area of ​​the optimized bone scan 3D model using the potential fracture probability data to obtain a fracture marked 3D model;

[0093] In an embodiment of the present invention, edge features may include information such as edge gradient, direction, and curvature. These features may be extracted by various image processing algorithms. For example, the image gradient is calculated using the Sobel operator to obtain edge strength and direction; the edge contour is extracted using the Canny operator, and the curvature of the edge is calculated. The extracted edge feature data is quantified and encoded to form bone edge feature data. Then, potential fracture probability processing is performed based on the bone edge feature data. A fracture probability model is established, which can map the edge feature data to the fracture probability. The fracture probability model can be obtained by training based on a machine learning algorithm, such as a support vector machine (SVM), a neural network, etc. The training model requires a large amount of bone image data, which includes images of normal bones and fractured bones, and the fracture area is marked. After the training is completed, the extracted bone edge feature data is input into the fracture probability model, the fracture probability of each bone position is calculated, and the potential fracture probability data is generated. The higher the probability value, the greater the possibility of fracture at the position. Finally, the fracture area of ​​the optimized bone scan three-dimensional model is marked by the potential fracture probability data. A fracture probability threshold is set. When the fracture probability at a certain position exceeds the threshold, it is considered that there is a fracture at that position and marked on the 3D model. For example, the fracture area can be marked in red and the fracture probability value can be displayed on the model. The marked 3D model is called a fracture marked 3D model.

[0094] Step S4: Render and visualize the fracture marked three-dimensional model, and perform force simulation on the parts to generate fracture stress and strain simulation data; generate a visualization report based on the fracture stress and strain simulation data, thereby obtaining a three-dimensional visualization fracture report.

[0095] In an embodiment of the present invention, the fracture mark three-dimensional model is rendered to make it realistic and three-dimensional. The color, lighting, material and other parameters of the model can be adjusted to make the fracture area more eye-catching. For example, the bone can be rendered translucent, the fracture area can be rendered red, and light and shadow effects can be added to make the model more three-dimensional. After the rendering is completed, the model is displayed on the screen, and the user can rotate, scale, translate and other operations on the model through interactive devices such as a mouse or a touch screen to observe the fracture from different angles. Then, the force simulation of the part is performed. The force simulation of the fracture mark three-dimensional model is performed using methods such as finite element analysis (FEA). First, a finite element model of the bone is established according to the anatomical structure and material properties of the bone. Then, virtual loads and boundary conditions are applied to the model to simulate the mechanical response of the bone under different force conditions. For example, the force conditions of the bone in activities such as walking and jumping can be simulated, and the stress and strain distribution inside the bone can be calculated. Through simulation, the influence of the fracture on the mechanical properties of the bone can be analyzed and the stability of the fracture can be evaluated. For example, stable fractures or unstable fractures. The risk assessment results can be expressed in the form of values ​​or levels. For example, fracture risk levels can be divided into low risk, medium risk, and high risk according to the type, degree, and stability of the fracture. Then, the risk assessment results, stress-strain distribution diagram, and visualized fracture 3D model are integrated into a visual report. The report can be in the form of a combination of graphics and text for easy understanding and analysis.

[0096] Preferably, step S1 comprises the following steps:

[0097] Step S11: calibrating the rotation angle of the medical device to obtain rotation angle sequence data of the medical device;

[0098] Step S12: controlling the cone beam CT transmitter to rotate at a preset angle step size based on the medical device rotation angle sequence data, and collecting rotation angle X-ray projection data to generate single-view projection image sequence data;

[0099] Step S13: performing single-view projection preprocessing on the single-view projection image sequence data to generate corrected projection image sequence data;

[0100] Step S14: Angle sequential integration of the correction projection image sequence data is performed through the medical device rotation angle sequence data, thereby obtaining multi-view projection image sequence data.

[0101] In an embodiment of the present invention, a high-precision rotary encoder is installed on the medical device, and the encoder can accurately measure the rotation angle of the medical device. A rotation angle range is set, for example, from 0 degrees to 360 degrees, and the angle step is determined, for example, it rotates once every 0.5 degrees. The medical device is started to rotate according to the set angle step within the preset angle range. During the rotation process, the rotary encoder records the rotation angle of the medical device in real time. The recorded rotation angle data is stored in chronological order to form the rotation angle sequence data of the medical device. For example, the rotation angle sequence data can be expressed as: [0.0, 0.5, 1.0, 1.5, ..., 359.5, 360.0]. These data represent the angular position of the medical device at each moment during the rotation process. The medical device is mechanically connected to the cone beam CT transmitter to ensure that the two rotate synchronously. The cone beam CT transmitter rotates according to the received angle data at a preset angle step. For example, when the received angle data is 10.0 degrees, the cone beam CT transmitter rotates to the 10.0 degree position. At each angle position, the cone beam CT transmitter emits X-rays to penetrate the area to be scanned. After passing through the area to be scanned, the X-rays are received by the detector. The detector converts the received X-ray signals into electrical signals and performs analog-to-digital conversion to generate digital X-ray projection data. Each X-ray projection data collected corresponds to a viewing angle to form a single-view projection image. For example, if the angle step is 0.5 degrees, 720 frames of single-view projection images will be collected. Due to the non-ideal characteristics of the X-ray source and detector, the collected images have artifacts and noise. Correction algorithms such as air correction and background correction can be used to remove these artifacts and noise. Secondly, image filtering is performed. In order to further reduce image noise, various image filtering algorithms such as median filtering and Gaussian filtering can be used. The selection of filtering algorithms needs to be adjusted according to the characteristics of image noise and actual needs. Due to time delays or data transmission problems in the data acquisition process, the order of image sequence data is disordered. Therefore, it is necessary to re-sort all corrected projection images in order from small to large angles according to the rotation angle information recorded during acquisition. For example, if the rotation angle sequence data shows that the angle corresponding to the 10th frame image is 5.0 degrees, then the frame image is placed at the corresponding 5.0 degree position in the sequence. After the sorting is completed, a continuous multi-view projection image sequence data is generated, which represents the X-ray image of the area to be scanned observed from different angles.

[0102] Preferably, step S13 comprises the following steps:

[0103] Step S131: repairing image bad pixels on the single-view projection image sequence data to generate repaired projection image sequence data;

[0104] Step S132: performing noise level assessment according to the restored projection image sequence data, and performing noise suppression using adaptive median filtering to generate denoised projection image sequence data;

[0105] Step S133: Acquire scanning geometric parameters of the cone-beam CT system and scanning imaging detector position data;

[0106] Step S134: establishing a geometric mapping relationship between the projection image and the real object according to the cone beam CT system scanning geometric parameters and the scanning imaging detector position data, thereby obtaining projection image-object mapping relationship data;

[0107] Step S135: performing polynomial fitting processing based on the projection image-object mapping relationship data, and performing geometric distortion correction on the denoised projection image sequence data to generate corrected projection image sequence data.

[0108] In an embodiment of the present invention, each frame of a single-view projection image is traversed to check the grayscale value of each pixel. A grayscale value threshold range is set. For example, for 16-bit image data, the grayscale value range is 0-65535, and the threshold range can be set to 100-60000. If the grayscale value of a pixel exceeds this threshold range, the pixel is considered to be a bad pixel. For the detected bad pixel, the grayscale value of the neighboring pixels is interpolated and repaired. For example, the average or median grayscale value of 8 neighboring pixels around the bad pixel can be used as the repair value of the bad pixel. A more advanced repair method can perform weighted interpolation based on the gradient information of the pixels around the bad pixel. The repaired pixel grayscale value will be replaced to the corresponding position in the original image. Traverse all pixels and all image frames to complete the bad pixel repair. The noise level of the repaired projection image sequence data is evaluated. The global noise level of the image can be calculated, for example, the mean square error or standard deviation of the image is calculated. The local noise level can also be calculated, the image is divided into multiple small blocks, and the noise level of each small block is calculated separately. The noise level assessment results are used to guide the subsequent noise suppression parameter selection. Then, adaptive median filtering is used for noise suppression. Adaptive median filtering is a nonlinear filtering method that can effectively remove impulse noise and salt and pepper noise while retaining image edge details. The window size of the adaptive median filter is adjusted according to the local noise level. For example, in areas with high noise levels, a larger filter window is used, while in areas with low noise levels, a smaller filter window is used. Each frame of the image is processed by adaptive median filtering to generate denoised projection image sequence data. The scanning geometric parameters include the distance from the X-ray source to the rotation center (isocenter), the distance from the rotation center to the detector, the pixel size of the detector, the array size of the detector, etc. These parameters can be obtained from the calibration file or system parameter settings of the cone-beam CT system. The position data of the scanning imaging detector refers to the position and posture information of the detector relative to the X-ray source and the rotation center during the scanning process. These data can be obtained through the position sensor on the detector or system records. For example, the scanning geometric parameters may include: the source to isocenter distance is 1000mm, the isocenter to detector distance is 500mm, the detector pixel size is 0.2mm×0.2mm, and the detector array size is 1000×1000. According to the cone beam CT system scanning geometric parameters and the scanning imaging detector position data obtained in step S133, a geometric mapping relationship between the projection image and the real object is established. First, a world coordinate system is established, and the rotation center is usually defined as the origin of the world coordinate system. Then, a detector coordinate system is established, and the detector center is usually defined as the origin of the detector coordinate system. Based on the scanning geometric parameters, the position of the X-ray source in the world coordinate system and the position and posture of the detector in the world coordinate system can be determined. For each pixel on the projection image, the direction of the corresponding X-ray in the world coordinate system can be calculated based on its position in the detector coordinate system.The point where this X-ray intersects with the real object is the object space position corresponding to the pixel. In this way, the mapping relationship between the projection image pixel coordinates and the object space coordinates is established to generate projection image-object mapping relationship data. Select some known feature points, for example, place some metal balls with known positions in the scanning phantom, and these metal balls will appear as circles or ellipses on the projection image. According to the positions of these feature points on the projection image and the known positions in the object space, a distortion model can be established. The distortion model usually adopts the method of polynomial fitting, such as quadratic polynomial or cubic polynomial. The polynomial coefficients are determined by fitting algorithms such as least squares method. Then, the distortion model obtained by fitting is used to perform geometric distortion correction on the denoised projection image sequence data. For each pixel on the projection image, its corrected coordinates are calculated according to its coordinates and the distortion model. According to the corrected coordinates, the corrected pixel grayscale value is interpolated from the original image. All pixels and all image frames are traversed to generate corrected projection image sequence data.

[0109] Preferably, step S2 comprises the following steps:

[0110] Step S21: performing filtered back-projection reconstruction processing according to the multi-view projection image sequence data to generate an initial bone scan three-dimensional model;

[0111] Step S22: optimizing the initial three-dimensional skeleton scan model to generate an optimized three-dimensional skeleton scan model;

[0112] Step S23: performing an X-ray imaging process simulation based on the optimized bone scanning three-dimensional model, thereby obtaining virtual multi-view projection data;

[0113] Step S24: performing multi-scale wavelet transform processing on the virtual multi-view projection data to generate a multi-scale wavelet coefficient matrix; performing image local edge intensity calculation on the multi-view projection image sequence data using the multi-scale wavelet coefficient matrix to generate image local area edge intensity data;

[0114] Step S25: using the edge intensity data of the local area of ​​the image to perform maximum inter-class variance image segmentation on the multi-view projection image sequence data, and performing image edge enhancement processing to generate edge enhanced scanned image sequence data.

[0115] In an embodiment of the present invention, each frame of projection image is filtered, and commonly used filters include Ram-Lak filter, Shepp-Logan filter, etc. The selection of the filter needs to be adjusted according to the image noise level and the reconstruction quality requirements. Then, the filtered projection image is back-projected, that is, the gray value of each pixel is back-projected into the three-dimensional space along the X-ray path. During the back-projection process, it is necessary to use the medical device rotation angle sequence data calibrated in step S11 to determine the X-ray direction corresponding to each projection image. Finally, the results of the back-projection of all viewing angles are superimposed and averaged to obtain the initial bone scanning three-dimensional model. The initial bone model is projected to each viewing angle to generate a virtual projection image. Then, the difference between the virtual projection image and the actual collected projection image is calculated. According to the difference value, the shape and position of the bone model are adjusted so that the virtual projection image is as consistent as possible with the actual projection image. Repeat the above projection and adjustment process until the model converges to the optimal solution. For example, an iterative reconstruction algorithm based on the gradient descent method can be used to minimize the mean square error between the model projection and the actual projection by continuously adjusting the model parameters. The optimized bone model is called the optimized bone scanning three-dimensional model. According to the scanning geometric parameters of the cone beam CT system (such as the parameters obtained in step S133), the position and posture of the virtual X-ray source and the virtual detector are set. Then, the virtual X-ray is emitted from the virtual X-ray source and passes through the optimized bone scanning three-dimensional model. According to the attenuation characteristics of different tissues in the bone model to X-rays, the intensity of the X-ray after passing through the bone model is calculated. The X-ray intensity is projected onto the virtual detector to generate virtual multi-view projection data. The process of generating virtual projection data is similar to the actual X-ray imaging process, but the whole process is simulated in a computer. Select a suitable wavelet basis function, such as Daubechies wavelet, Haar wavelet, etc., to perform multi-scale decomposition on the virtual projection data and generate a multi-scale wavelet coefficient matrix. Then, the multi-scale wavelet coefficient matrix is ​​used to calculate the local edge strength of each frame of the multi-view projection image sequence data. In wavelet transform, the high-frequency subband coefficient reflects the edge and detail information of the image. Maximum inter-class variance (Otsu) image segmentation is performed on the multi-view projection image sequence data. The Otsu algorithm is a commonly used image segmentation method. Its basic principle is to select an optimal threshold based on the image grayscale histogram so that the inter-class variance of the image foreground and background is maximized. First, the edge intensity data of the local area of ​​the image is used as the input of the Otsu algorithm to calculate the inter-class variance under different thresholds. Then, the threshold when the inter-class variance is maximum is selected as the segmentation threshold. According to the segmentation threshold, each frame image in the multi-view projection image sequence data is divided into a bone area and a non-bone area to generate a binary image. Then, the segmented image is edge enhanced to highlight the bone edge. Edge enhancement can be performed in a variety of ways, such as gradient operator, Laplacian operator, Gaussian filtering, etc.Select a suitable edge enhancement algorithm to process the binary image to make the bone edge clearer. The processed image sequence data is called edge enhanced scan image sequence data. For example, after Otsu segmentation, the image gradient can be calculated using the Sobel operator, and then the gradient amplitude can be superimposed on the original image to enhance the bone edge.

[0116] Preferably, step S22 includes the following steps:

[0117] Step S221: performing scanned bone type recognition on the initial bone scanned three-dimensional model to generate scanned bone type data;

[0118] Step S222: Obtaining bone CT scan sample data according to the scanned bone type data;

[0119] Step S223: performing principal component analysis on the bone CT scan sample data and extracting bone geometric feature points to obtain bone sample feature point data;

[0120] Step S224: performing non-rigid part registration according to the bone sample feature point data, and performing bone shape change pattern statistics to generate bone shape change statistical data;

[0121] Step S225: defining shape constraint parameters and ranges based on the skeletal shape change statistics, thereby establishing a skeletal shape constraint model;

[0122] Step S226: Use the bone shape constraint model to perform shape constraint iterative optimization on the initial bone scan three-dimensional model to generate an optimized bone scan three-dimensional model.

[0123] In an embodiment of the present invention, the initial skeleton model is preprocessed, for example, isolated points and noise in the model are removed, and the model surface is smoothed. Then, the geometric features of the skeleton model are extracted, for example, the overall shape of the skeleton, the relative size of each skeleton part, the curvature of the skeleton surface, etc. These features can be extracted using morphological methods, statistical methods, or machine learning methods. Next, the extracted skeleton features are compared with a pre-stored skeleton type feature library to determine the skeleton type to which the initial skeleton model belongs, and generate scanned skeleton type data. Corresponding skeleton CT scanning sample data are obtained, which can be from a public medical imaging database or from a pre-established skeleton sample library. Representative geometric feature points are selected in the skeleton sample data. The selection of feature points can be based on the anatomical structure of the skeleton or based on the results of principal component analysis. For example, positions such as the articular surface, bone protrusion, and bone corner of the skeleton can be selected as feature points. The three-dimensional coordinates of the feature points in each sample data are extracted to obtain skeleton sample feature point data. For example, principal component analysis can be performed on the femoral sample data, and then the femoral head center, femoral neck midpoint, femoral condyle and other positions are selected as feature points, and the three-dimensional coordinates of these feature points in each femoral sample are extracted. Due to the differences in bone shape between different individuals, the feature points of all sample data need to be aligned to a common coordinate system. Use the thin plate spline interpolation (TPS) algorithm or the Demons algorithm for non-rigid registration. After the registration is completed, the feature points of all sample data have a corresponding relationship in space. Then, the bone shape change pattern statistical analysis is performed on the registered feature point data. The mean and variance of the feature point position can be calculated, and the range of variation of geometric parameters such as the distance and angle between the feature points can also be calculated. Through statistical analysis, the law and pattern of the shape change of this type of bone can be understood, and the bone shape change statistics can be generated. According to the bone shape change statistics, the reasonable range of variation of each shape constraint parameter is determined. For example, according to the statistical distribution of the femoral neck length, the constraint range of the femoral neck length can be determined as the mean plus or minus two standard deviations. Then, based on the defined shape constraint parameters and ranges, a bone shape constraint model is established. The skeletal shape constraint model can take many forms, such as a parametric model, a statistical shape model (SSM), a physical model, etc. The initial skeletal model is aligned with the skeletal shape constraint model, and the shape parameters of the initial skeletal model are mapped to the shape constraint model. Then, it is calculated whether the shape parameters of the initial skeletal model meet the shape constraint conditions. If not, the shape parameters of the initial skeletal model are adjusted according to the shape constraint model so that it gradually approaches the constraint range. During the adjustment process, optimization algorithms such as the gradient descent method and the simulated annealing method can be used. After each adjustment, the shape parameters are recalculated and constraint judgment is performed until all shape parameters meet the constraint conditions or the preset number of iterations is reached.During the adjustment process, the image grayscale value difference and shape constraints can be considered at the same time, so that the optimized bone model not only conforms to the image data, but also has a reasonable bone shape. For example, an energy function can be used to measure the image grayscale value difference and shape constraints at the same time, and then the bone model can be optimized by minimizing the energy function. The final generated model is called the optimized bone scan 3D model.

[0124] Preferably, step S226 includes the following steps:

[0125] Performing vertex three-dimensional control mesh parameterization on the initial bone scan three-dimensional model to generate scan three-dimensional model parameters;

[0126] Performing viewing angle parameter projection according to the scanned 3D model parameters, and performing projection error calculation on multi-view projection image sequence data to generate projection error data;

[0127] Calculating the model deformation vector on the projection error data to generate the model deformation vector data;

[0128] Calculate the shape regularization term according to the bone shape constraint model to generate shape regularization data;

[0129] Perform iterative optimization of model parameters based on model deformation vector data and shape regularization data to generate model adjustment optimization parameters;

[0130] The skeleton model parameters of the initial skeleton scan three-dimensional model are updated by adjusting the optimization parameters of the model, and the skeleton model deformation is adjusted to obtain an optimized skeleton scan three-dimensional model.

[0131] In an embodiment of the present invention, the initial skeleton model is represented as a three-dimensional mesh model composed of vertices and faces. Then, some control vertices are selected on the surface of the model. These control vertices can be evenly distributed on the surface of the model, or can be selected according to the anatomical structure of the bone. Next, the shape change of the model is represented as a function of the position of the control vertices. For example, the model can be parameterized using a free deformation (FFD) method or a radial basis function (RBF) method. The FFD method embeds the model into a three-dimensional mesh and controls the deformation of the model by controlling the position of the mesh vertices. The RBF method uses a radial basis function to interpolate the position of the control vertices to control the deformation of the model. For each viewing angle, the projection of the three-dimensional model on the detector is calculated according to the position of the X-ray source, the position and posture of the detector, and the shape parameters of the three-dimensional model. Then, the generated virtual projection image is compared with the multi-view projection image sequence data obtained in step S1 to calculate the difference between the two. The difference can be measured using a variety of indicators, such as mean square error (MSE), peak signal-to-noise ratio (PSNR), structural similarity (SSIM), etc. The gradient of the projection error relative to the model parameters is calculated. The gradient represents the rate of change of the projection error as the model parameters change. Then, the negative of the gradient is calculated according to the skeletal shape constraint model established in step S225, and the shape regularization term is calculated. The shape regularization term is used to constrain the shape change of the model so that it conforms to the anatomical structure and biomechanical properties of the bone. The shape regularization term can be defined as a function of the distance between the model parameters and the shape constraint model. For example, the sum of the squares of the differences between the model parameters and the corresponding parameters in the shape constraint model can be used as the shape regularization term. The direction is used as the direction of the model deformation vector. The size of the deformation vector can be adjusted according to the size and step size of the gradient. The optimization goal of the model parameters is to minimize the projection error and the shape regularization term at the same time. A variety of optimization algorithms can be used, such as gradient descent method, conjugate gradient method, quasi-Newton method, etc. In each iteration, the model parameters are adjusted according to the model deformation vector and the shape regularization term. For example, the weighted sum of the model deformation vector and the shape regularization term can be used as the direction of parameter adjustment. The weight coefficient can be adjusted according to the relative importance of the projection error and the shape regularization term. The iterative process continues until the model parameters converge to an optimal solution and generate model adjustment optimization parameters. Update the bone model parameters of the initial bone scan 3D model and adjust the bone model deformation. Apply the optimized model parameters to the parametric model and update the position of the model control vertices. The update of the control vertex position will drive the deformation of the entire model, and finally obtain the optimized bone scan 3D model.

[0132] Preferably, step S24 comprises the following steps:

[0133] Step S241: performing multi-scale wavelet transform decomposition on the virtual multi-view projection data to generate a multi-scale wavelet coefficient matrix;

[0134] Step S242: performing wavelet decomposition sub-band energy calculation according to the multi-scale wavelet coefficient matrix to generate wavelet sub-band energy data;

[0135] Step S243: sorting the wavelet subband energy data in ascending order, and calculating the energy difference between adjacent wavelet subbands to generate subband energy difference data;

[0136] Step S244: screening the maximum energy difference according to the wavelet sub-band energy data, and taking it as the sub-band frequency threshold; performing sub-band classification on the wavelet sub-band energy data according to the sub-band frequency threshold, and obtaining low-frequency sub-band data and high-frequency sub-band data respectively;

[0137] Step S245: performing smoothing processing on the low-frequency sub-band data to obtain smoothed low-frequency sub-band data; performing sub-band coefficient enhancement processing on the high-frequency sub-band data to generate enhanced high-frequency sub-band data;

[0138] Step S246: extracting wavelet coefficient features according to the enhanced high-frequency sub-band data and the smoothed low-frequency sub-band data, and constructing an edge detection operator, thereby obtaining an edge detection operator;

[0139] Step S247: Use the edge detection operator to perform edge feature guided detection on the multi-view projection image sequence data, and perform local edge strength calculation to generate edge strength data of the local area of ​​the image.

[0140] In an embodiment of the present invention, each frame of projection image in the virtual multi-view projection data is subjected to wavelet transform decomposition. Multiscale decomposition refers to decomposing an image into multiple sub-bands of different resolutions, each sub-band containing image information of different frequency ranges. For example, three-layer wavelet decomposition can be performed to decompose the image into a low-frequency sub-band (LL), a horizontal high-frequency sub-band (LH), a vertical high-frequency sub-band (HL) and a diagonal high-frequency sub-band (HH). The decomposed wavelet coefficients are organized into a multi-scale wavelet coefficient matrix, in which each row corresponds to a sub-band and each column corresponds to a pixel position. Subband energy refers to the sum of squares or absolute values ​​of the wavelet coefficients in the sub-band. For each sub-band, the energy values ​​of all its coefficients are calculated. For example, for the horizontal high-frequency sub-band LH, the sum of squares of all coefficients in the LH sub-band is calculated as the energy value of the LH sub-band. Sorting can be arranged from small to large according to the energy value. Then, the energy difference of adjacent wavelet sub-bands is calculated for the sorted wavelet sub-band energy data. For example, assuming that the sorted energy data is E1, E2, E3, ..., En, then the adjacent energy differences are E2-E1, E3-E2, ..., En-En-1, and sub-band energy difference data is generated. According to the sub-band energy difference data generated in step S243, the maximum energy difference is screened. The position with the largest energy difference is found, and the corresponding energy value is used as the sub-band frequency threshold. The sub-band frequency threshold is used to distinguish between low-frequency sub-bands and high-frequency sub-bands. Sub-bands with energy values ​​less than the threshold are classified as low-frequency sub-bands, and sub-bands with energy values ​​greater than the threshold are classified as high-frequency sub-bands. For example, if the maximum energy difference occurs between E3 and E4, E3 is used as the sub-band frequency threshold. Sub-bands E1 and E2 with energy values ​​less than E3 are classified as low-frequency sub-bands, and sub-bands E4, E5, ..., En with energy values ​​greater than E3 are classified as high-frequency sub-bands. Low-frequency sub-band data and high-frequency sub-band data are obtained respectively. The low-frequency sub-band data obtained in step S244 is smoothed. Smoothing can be performed using a low-pass filter, such as a Gaussian filter, to obtain smooth low-frequency sub-band data. The high-frequency sub-band data obtained in step S244 is subjected to sub-band coefficient enhancement processing. The enhancement processing can be performed using a variety of methods, such as multiplying the high-frequency coefficient by a coefficient greater than 1, or performing a nonlinear transformation on the high-frequency coefficient to generate enhanced high-frequency sub-band data. Extract local variance, gradient and other features of the high-frequency sub-band coefficient. Then, based on the extracted wavelet coefficient features, construct an edge detection operator. The edge detection operator can be a function or a matrix used to detect edges in an image. Apply the edge detection operator to each frame of projection image in the multi-view projection image sequence data. According to the response value of the edge detection operator, it is determined whether each pixel in the image belongs to an edge. Then, the local edge strength of the detected edge is calculated. For example, the gray value gradient of the neighborhood pixels around each edge pixel can be calculated as the local edge strength of the pixel.

[0141] As an example of the present invention, refer to Figure 2As shown, Figure 1 Detailed implementation steps of step S3 in the flowchart, in this example, step S3 includes:

[0142] Step S31: performing bone region segmentation according to edge enhanced scan image sequence data to obtain bone region image data;

[0143] In an embodiment of the present invention, the grayscale histogram of the edge-enhanced image is analyzed to find the optimal threshold between the bone region and the background region. The threshold can be automatically determined by the Otsu method or other adaptive threshold methods. Then, the pixels whose grayscale values ​​in the edge-enhanced image are greater than the threshold are marked as the bone region, and the remaining pixels are marked as the background region. In addition, morphological operations, such as corrosion and expansion operations, can be combined to remove the small noise areas or the bone regions with broken connections in the segmentation results to make the bone regions more complete. The segmented bone region images are stored as bone region image data. For example, for a group of edge-enhanced scanned image sequences of femurs, the femur region can be separated from the background by threshold segmentation to obtain an image sequence containing only the femur region.

[0144] Step S32: extracting bone edge features from the bone region image data to obtain bone edge feature data;

[0145] In an embodiment of the present invention, a Canny edge detector or other edge detection algorithms are used to extract the edge contour of the bone area. Then, for each pixel on the extracted edge contour, its gradient, direction, curvature and other features are calculated. The gradient can be calculated by the Sobel operator or other gradient operators; the direction can be calculated by the angle of the gradient vector; the curvature can be calculated by the second-order derivative of the edge contour. In addition, more complex features can be extracted, such as local binary patterns (LBP), histograms of directional gradients (HOG), etc. The extracted features are stored as bone edge feature data. For example, the coordinates, gradient, direction and curvature of each edge pixel can be stored in a feature vector, and the feature vectors of all edge pixels constitute the bone edge feature data.

[0146] Step S33: using a preset support vector machine to calculate the potential fracture probability of the bone edge feature data to generate potential fracture probability data;

[0147] In an embodiment of the present invention, a preset support vector machine (SVM) is used to calculate the potential fracture probability of the bone edge feature data extracted in step S32. The preset SVM model needs to be trained in advance with a large amount of training data. The training data should include positive samples (fracture cases) and negative samples (normal cases), and extract the same bone edge features as step S32. The trained SVM model can map the bone edge features to the fracture probability. The bone edge feature data extracted in step S32 is input into the trained SVM model to obtain the fracture probability of each edge pixel or region. For example, for each edge pixel in the femur image, the SVM model can output a probability value between 0 and 1, indicating the possibility of fracture in the pixel. The fracture probabilities of all pixels or regions are stored as potential fracture probability data.

[0148] Step S34: performing fracture area mask processing according to the potential fracture probability data, and marking the fracture area of ​​the optimized bone scan three-dimensional model, thereby obtaining a fracture marked three-dimensional model.

[0149] In an embodiment of the present invention, a fracture area mask is generated based on potential fracture probability data. A probability threshold is set, such as 0.5. Pixels or areas where the potential fracture probability is greater than the threshold are marked as fracture areas, and a mask is generated in the corresponding area of ​​the three-dimensional model. Then, the fracture area mask is applied to the optimized bone scan three-dimensional model generated in step S22. For example, the three-dimensional model vertices corresponding to the pixels marked as fracture areas in the mask can be set to red, or other special marks can be assigned to finally obtain a fracture marked three-dimensional model, which clearly shows the potential fracture area. For example, if the fracture probability of the femoral neck area is greater than the threshold, the femoral neck area of ​​the three-dimensional femoral model is marked, for example, the color of the area is set to red to highlight the potential fracture area.

[0150] Preferably, step S32 includes the following steps:

[0151] Step S321: performing edge Gaussian difference detection on the bone region image data to generate bone edge image data;

[0152] Step S322: performing edge gradient calculation on the bone edge image data to generate bone edge gradient data;

[0153] Step S323: Calculate the edge curvature according to the bone edge gradient data to obtain the bone edge curvature data;

[0154] Step S324: performing edge direction analysis according to the bone edge gradient data to obtain bone edge direction data;

[0155] Step S325: evaluating edge point continuity according to the bone edge gradient data to obtain bone edge continuity data;

[0156] Step S326: integrating the bone edge curvature data, the bone edge direction data and the bone edge continuity data to obtain the bone edge feature data.

[0157] In an embodiment of the present invention, DoG is an edge detection method, which detects edges by calculating the difference of Gaussian filter images of different scales. Use two Gaussian kernels with different standard deviations to perform convolution operation on the bone region image to obtain two Gaussian filter images of different scales. Then, subtract the two Gaussian filter images to obtain a DoG image. In the DoG image, the pixel value of the edge area is larger, and the pixel value of the non-edge area is smaller. The DoG image is thresholded, and the pixels greater than the threshold are marked as edge pixels to generate bone edge image data. For example, using the Sobel operator, the gradient of the image in the x direction and the y direction are calculated respectively. Then, the amplitude and direction of the gradient are calculated. The gradient amplitude represents the intensity of the edge, and the gradient direction represents the normal direction of the edge. The calculated gradient amplitude and direction are stored as bone edge gradient data. For example, in the femoral edge image, the gradient amplitude at the femoral edge is larger, while the gradient amplitude in other areas is smaller. Calculate the derivative of the edge gradient direction relative to the edge arc length. Then, the amplitude of the derivative is used as the edge curvature. The larger the curvature value, the greater the degree of edge curvature. The calculated curvature value is stored as bone edge curvature data. For example, in the femoral edge image, the curvature of the femoral head and condyle is larger, while the curvature of the femoral shaft is smaller. The edge direction refers to the normal direction of the edge. The edge gradient direction can be used as the edge direction. For example, if the edge gradient direction is (gx,gy), the edge direction can be expressed as arctan(gy / gx). The calculated edge direction is stored as bone edge direction data. For example, the difference in gradient directions between adjacent edge pixels is calculated. The smaller the difference, the more continuous the edge. A curvature-based continuity evaluation method can also be used, for example, calculating the rate of change of the edge curvature. The smaller the rate of change, the more continuous the edge. The calculated continuity index is stored as bone edge continuity data. The values ​​of different features are concatenated into a feature vector, or the values ​​of different features are combined into a new feature. For example, the curvature, direction, and continuity index of each edge pixel can be combined into a three-dimensional feature vector, and the feature vectors of all edge pixels constitute the bone edge feature data.

[0158] As an example of the present invention, refer to Figure 3 As shown, Figure 1 Detailed implementation steps of step S4 in the embodiment are shown in the flowchart. In this embodiment, step S4 includes:

[0159] Step S41: Rendering and visualizing the fracture mark three-dimensional model to generate a visualized fracture three-dimensional model;

[0160] In an embodiment of the present invention, various three-dimensional rendering technologies are used, such as rendering based on ray tracing, rendering based on rasterization, etc. First, appropriate rendering algorithms and parameters are selected as needed, such as lighting model, material properties, viewing angle, etc. Then, the fracture mark three-dimensional model is input into the rendering engine for rendering. In order to highlight the fracture area, the fracture area can be marked with different colors or textures. For example, the fracture area can be rendered red, and the normal bone area can be rendered gray. A transparency effect can also be added to observe the structure inside the fracture area. After the rendering is completed, a visual fracture three-dimensional model is generated, which can clearly display information such as the location, shape, and range of the fracture.

[0161] Step S42: performing meshing processing on the fracture site based on the visualized three-dimensional fracture model, and performing finite element model processing to obtain a finite element simulation model;

[0162] In an embodiment of the present invention, a visualized fracture three-dimensional model is imported into a finite element analysis software. Then, the fracture site is meshed and the three-dimensional model is discretized into many small units. The density and shape of the mesh will affect the accuracy of the simulation results. For example, a denser mesh can be used near the fracture area. Next, material properties, such as Young's modulus, Poisson's ratio, etc., are defined. They can be set according to the actual material properties of the bone or data in the literature. For the fracture area, the material properties can be adjusted according to the degree of fracture, such as reducing the Young's modulus to simulate the strength reduction of the fracture. Finally, boundary conditions and loads, such as fixed constraints, force loads, etc., are defined. These conditions and loads can be set according to actual conditions or simulation requirements. For example, the load on the femur when the human body is standing can be simulated.

[0163] Step S43: performing force simulation on the parts according to the finite element simulation model to generate fracture stress and strain simulation data;

[0164] In an embodiment of the present invention, the finite element simulation model is input into the finite element analysis software for solving. During the simulation process, the software calculates the stress and strain of each unit in the model according to the defined material properties, boundary conditions and loads. For example, the dynamic load on the femur when the human body walks or jumps can be simulated, and the stress and strain distribution of the femur at different times can be calculated. After the simulation is completed, the fracture stress and strain simulation data are generated, including the stress and strain values ​​of each unit in the model.

[0165] Step S44: perform a fracture risk assessment on the fracture marker three-dimensional model using the fracture stress-strain simulation data, and generate a visualization report, thereby obtaining a three-dimensional visualization fracture report.

[0166] In an embodiment of the present invention, the type of fracture is determined according to the morphological characteristics of the fracture, such as the direction and shape of the fracture line, such as a transverse fracture, an oblique fracture or a spiral fracture. The degree of the fracture is determined according to the location and range of the fracture, such as a mild fracture, a moderate fracture or a severe fracture. The stability of the fracture is determined according to the stress-strain data, such as a stable fracture or an unstable fracture. The risk assessment result can be expressed in the form of a numerical value or a grade. For example, the fracture risk level can be divided into low risk, medium risk and high risk according to the type, degree and stability of the fracture. Then, the risk assessment results, stress-strain distribution map, visualized fracture three-dimensional model and other information are integrated into a visualization report. The report can be in the form of a combination of graphics and text for easy understanding and analysis. For example, the report can contain information such as the fracture type, degree, stability, stress-strain cloud map, fracture three-dimensional model, and finally generate a three-dimensional visualization fracture report. For example, a visualization report of a femoral neck fracture can contain information such as a transverse fracture type, a moderate fracture degree, an unstable fracture stability, and a stress-strain cloud map of the femoral neck area and a femoral three-dimensional model.

[0167] The present application is to collect multi-view X-ray projection data by rotating medical equipment and integrate the angle sequence, thereby overcoming the limitation that traditional two-dimensional X-ray imaging can only provide single-view information, thereby more completely showing the three-dimensional structure of the bone. This step not only enhances the visualization of key features such as fracture lines in the image, but also avoids the problem of fracture lines being blocked by other tissue structures, especially for the diagnosis of complex bone structures, and can clearly display the spatial position and morphology of fractures. By reconstructing and optimizing the three-dimensional bone model of multi-view projection data, the accuracy and integrity of the model are further improved. The maximum inter-class variance segmentation and edge enhancement processing of the real projection image by virtual projection data effectively reduces the influence of noise and artifacts, highlights the bone edge features, and makes subtle fracture lines easier to identify. This process is particularly important for the diagnosis of subtle bone fractures such as fatigue fractures. Traditional two-dimensional X-ray images are often difficult to display clearly, and this method can effectively improve the visualization of subtle fracture lines and reduce the risk of missed diagnosis through image enhancement processing. Potential fracture probability processing is performed based on bone edge feature data, and fracture area marking is performed on the three-dimensional bone model, which elevates fracture diagnosis from simple morphological observation to the level of quantitative analysis. By calculating the potential fracture probability, the possibility of fracture can be evaluated more objectively, avoiding errors caused by subjective judgment. At the same time, the fracture area marking function can intuitively display the location, range and morphology of the fracture, judge the stability of the fracture based on the stress-strain data, and help doctors quickly locate the fracture area and improve diagnostic efficiency.

[0168] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0169] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A medical device visualization simulation method based on data recognition, characterized in that: The following steps are involved: Step S1: calibrating the rotation angle of the medical device to obtain rotation angle sequence data of the medical device; collecting rotation angle X-ray projection data based on the rotation angle sequence data of the medical device, and integrating the angle sequence to obtain multi-view projection image sequence data; Step S2: Optimizing the three-dimensional skeleton model according to the multi-view projection image sequence data to generate an optimized three-dimensional skeleton scan model; simulating the imaging process based on the optimized three-dimensional skeleton scan model to obtain virtual multi-view projection data; performing maximum inter-class variance image segmentation on the multi-view projection image sequence data through the virtual multi-view projection data, and performing image edge enhancement processing to generate edge enhanced scan image sequence data; Step S3: extracting bone edge features from the edge-enhanced scan image sequence data to obtain bone edge feature data; performing potential fracture probability processing based on the bone edge feature data to generate potential fracture probability data; marking the fracture area of ​​the optimized bone scan 3D model using the potential fracture probability data to obtain a fracture marked 3D model; Step S4: Render and visualize the fracture marked three-dimensional model, and perform force simulation on the parts to generate fracture stress and strain simulation data; generate a visualization report based on the fracture stress and strain simulation data, thereby obtaining a three-dimensional visualization fracture report.

2. The medical device visualization simulation method based on data recognition according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: calibrating the rotation angle of the medical device to obtain rotation angle sequence data of the medical device; Step S12: controlling the cone beam CT transmitter to rotate at a preset angle step size based on the medical device rotation angle sequence data, and collecting rotation angle X-ray projection data to generate single-view projection image sequence data; Step S13: performing single-view projection preprocessing on the single-view projection image sequence data to generate corrected projection image sequence data; Step S14: Angle sequential integration of the correction projection image sequence data is performed through the medical device rotation angle sequence data, thereby obtaining multi-view projection image sequence data.

3. The medical device visualization simulation method based on data recognition according to claim 2 is characterized in that: Step S13 includes the following steps: Step S131: repairing image bad pixels on the single-view projection image sequence data to generate repaired projection image sequence data; Step S132: performing noise level assessment according to the restored projection image sequence data, and performing noise suppression using adaptive median filtering to generate denoised projection image sequence data; Step S133: Acquire scanning geometric parameters of the cone-beam CT system and scanning imaging detector position data; Step S134: establishing a geometric mapping relationship between the projection image and the real object according to the cone beam CT system scanning geometric parameters and the scanning imaging detector position data, thereby obtaining projection image-object mapping relationship data; Step S135: performing polynomial fitting processing based on the projection image-object mapping relationship data, and performing geometric distortion correction on the denoised projection image sequence data to generate corrected projection image sequence data.

4. The medical device visualization simulation method based on data recognition according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing filtered back-projection reconstruction processing according to the multi-view projection image sequence data to generate an initial bone scan three-dimensional model; Step S22: optimizing the initial three-dimensional skeleton scan model to generate an optimized three-dimensional skeleton scan model; Step S23: performing an X-ray imaging process simulation based on the optimized bone scanning three-dimensional model, thereby obtaining virtual multi-view projection data; Step S24: performing multi-scale wavelet transform processing on the virtual multi-view projection data to generate a multi-scale wavelet coefficient matrix; performing image local edge intensity calculation on the multi-view projection image sequence data using the multi-scale wavelet coefficient matrix to generate image local area edge intensity data; Step S25: using the edge intensity data of the local area of ​​the image to perform maximum inter-class variance image segmentation on the multi-view projection image sequence data, and performing image edge enhancement processing to generate edge enhanced scanned image sequence data.

5. The medical device visualization simulation method based on data recognition according to claim 4 is characterized in that: Step S22 includes the following steps: Step S221: performing scanned bone type recognition on the initial bone scanned three-dimensional model to generate scanned bone type data; Step S222: Obtaining bone CT scan sample data according to the scanned bone type data; Step S223: performing principal component analysis on the bone CT scan sample data and extracting bone geometric feature points to obtain bone sample feature point data; Step S224: performing non-rigid part registration according to the bone sample feature point data, and performing bone shape change pattern statistics to generate bone shape change statistical data; Step S225: defining shape constraint parameters and ranges based on the skeletal shape change statistics, thereby establishing a skeletal shape constraint model; Step S226: Use the bone shape constraint model to perform shape constraint iterative optimization on the initial bone scan three-dimensional model to generate an optimized bone scan three-dimensional model.

6. The medical device visualization simulation method based on data recognition according to claim 5, characterized in that: Step S226 includes the following steps: Performing vertex three-dimensional control mesh parameterization on the initial bone scan three-dimensional model to generate scan three-dimensional model parameters; Performing viewing angle parameter projection according to the scanned 3D model parameters, and performing projection error calculation on multi-view projection image sequence data to generate projection error data; Calculating the model deformation vector on the projection error data to generate the model deformation vector data; Calculate the shape regularization term according to the bone shape constraint model to generate shape regularization data; Perform iterative optimization of model parameters based on model deformation vector data and shape regularization data to generate model adjustment optimization parameters; The skeleton model parameters of the initial skeleton scan three-dimensional model are updated by adjusting the optimization parameters of the model, and the skeleton model deformation is adjusted to obtain an optimized skeleton scan three-dimensional model.

7. The medical device visualization simulation method based on data recognition according to claim 4 is characterized in that: Step S24 includes the following steps: Step S241: performing multi-scale wavelet transform decomposition on the virtual multi-view projection data to generate a multi-scale wavelet coefficient matrix; Step S242: performing wavelet decomposition sub-band energy calculation according to the multi-scale wavelet coefficient matrix to generate wavelet sub-band energy data; Step S243: sorting the wavelet subband energy data in ascending order, and calculating the energy difference between adjacent wavelet subbands to generate subband energy difference data; Step S244: screening the maximum energy difference according to the wavelet sub-band energy data, and taking it as the sub-band frequency threshold; performing sub-band classification on the wavelet sub-band energy data according to the sub-band frequency threshold, and obtaining low-frequency sub-band data and high-frequency sub-band data respectively; Step S245: performing smoothing processing on the low-frequency sub-band data to obtain smoothed low-frequency sub-band data; performing sub-band coefficient enhancement processing on the high-frequency sub-band data to generate enhanced high-frequency sub-band data; Step S246: extracting wavelet coefficient features according to the enhanced high-frequency sub-band data and the smoothed low-frequency sub-band data, and constructing an edge detection operator, thereby obtaining an edge detection operator; Step S247: Use the edge detection operator to perform edge feature guided detection on the multi-view projection image sequence data, and perform local edge strength calculation to generate edge strength data of the local area of ​​the image.

8. The medical device visualization simulation method based on data recognition according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing bone region segmentation according to edge enhanced scan image sequence data to obtain bone region image data; Step S32: extracting bone edge features from the bone region image data to obtain bone edge feature data; Step S33: using a preset support vector machine to calculate the potential fracture probability of the bone edge feature data to generate potential fracture probability data; Step S34: performing fracture area mask processing according to the potential fracture probability data, and marking the fracture area of ​​the optimized bone scan three-dimensional model, thereby obtaining a fracture marked three-dimensional model.

9. The medical device visualization simulation method based on data recognition according to claim 8, characterized in that: Step S32 includes the following steps: Step S321: performing edge Gaussian difference detection on the bone region image data to generate bone edge image data; Step S322: performing edge gradient calculation on the bone edge image data to generate bone edge gradient data; Step S323: Calculate the edge curvature according to the bone edge gradient data to obtain the bone edge curvature data; Step S324: performing edge direction analysis according to the bone edge gradient data to obtain bone edge direction data; Step S325: evaluating edge point continuity according to the bone edge gradient data to obtain bone edge continuity data; Step S326: integrating the bone edge curvature data, the bone edge direction data and the bone edge continuity data to obtain the bone edge feature data.

10. The medical device visualization simulation method based on data recognition according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: Rendering and visualizing the fracture mark three-dimensional model to generate a visualized fracture three-dimensional model; Step S42: performing meshing processing on the fracture site based on the visualized three-dimensional fracture model, and performing finite element model processing to obtain a finite element simulation model; Step S43: performing force simulation on the parts according to the finite element simulation model to generate fracture stress and strain simulation data; Step S44: perform a fracture risk assessment on the fracture marker three-dimensional model using the fracture stress-strain simulation data, and generate a visualization report, thereby obtaining a three-dimensional visualization fracture report.

Images