A method and system for recognizing spine features based on image processing

Abstract

The present invention relates to the technical field of spinal data processing, and in particular to a spinal feature recognition method and system based on image processing. The method comprises the following steps: acquiring user optical data and spinal image data, and performing external morphological analysis based on the user optical data to obtain external morphological data; performing spinal support analysis on the external morphological data to obtain spinal support data; performing spinal bone feature extraction based on spinal image data to obtain spinal bone feature data; performing spinal multimodal feature fusion based on spinal support data and spinal bone feature data to obtain spinal feature data for spinal feature recognition auxiliary operations. The present invention is based on multimodal data fusion, combined with the geometric morphology, support force analysis and spinal image data of the spine, to perform accurate spinal assessment and feature recognition.

Description

Technical Field

[0001] The present invention relates to the technical field of spine data processing, and in particular to a spine feature recognition method and system based on image processing. Background Art

[0002] The spine is an important supporting structure of the human body, and its feature recognition is of great significance for anatomical analysis, medical research, rehabilitation treatment and health monitoring. However, the accuracy and comprehensiveness of spine feature recognition has always been one of the difficulties in the technical field. Traditional spine image processing methods mostly use grayscale feature-based segmentation algorithms or rule-based key point recognition technology. However, due to the complexity of spinal morphology (such as morphological differences between vertebrae, blurred boundaries caused by intervertebral disc degeneration, etc.), the accuracy of segmentation and key point recognition is difficult to guarantee, especially when there is degeneration or abnormality. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a spine feature recognition method and system based on image processing to solve at least one of the above technical problems.

[0004] The present application provides a method for recognizing spine features based on image processing, comprising the following steps:

[0005] Step S1: acquiring user optical data and spinal image data, and performing external morphology analysis based on the user optical data to obtain external morphology data;

[0006] Step S2: Performing spinal support analysis on the external morphological data to obtain spinal support data;

[0007] Step S3: Extracting vertebral bone features according to the vertebral image data to obtain vertebral bone feature data;

[0008] Step S4: Performing spinal multimodal feature fusion according to the spinal support data and the spinal bone feature data to obtain spinal feature data for spinal feature recognition auxiliary work.

[0009] In the present invention, the user's external morphological data (such as posture and body shape) and spinal imaging data (such as X-rays and CT images) are deeply integrated. By combining optical data and imaging data, it is possible to comprehensively identify information on the geometric morphology, mechanical characteristics, and stress conditions of the spine, so that the identification of spinal features is not only limited to the geometric structure, but also takes into account the stress and specific conditions of the spine. By analyzing the user's external morphology, personalized spinal support analysis can be performed. Since the spinal support force analysis results of each user are customized based on their specific body shape instead of using a fixed template, the adaptability of the method between different individuals is improved, avoiding the problem of "one size fits all". Support analysis can comprehensively analyze the stress conditions of the spine based on external morphological data, generate spinal support data, provide accurate data based on biomechanics, and reveal the mechanical load of the spine in daily activities, thereby achieving accurate identification of the spine. The fusion of multimodal features of the spine enables the geometric features, mechanical features, and support force data of the spine to be processed under a unified framework, which can achieve cross-modal feature information complementarity and ensure more accurate feature recognition of the spine, especially when dealing with complex situations (such as spinal injury, deformation, etc.), and can effectively extract key features.

[0010] Preferably, step S1 specifically includes:

[0011] Acquire user optical data and spinal imaging data;

[0012] Performing body shape analysis based on user optical data to obtain body shape feature data;

[0013] Generate a standard spine template according to the body shape feature data and a preset spine template library to obtain spine template data;

[0014] Perform spine template fitting according to the body shape feature data and the spine template data to obtain fitted spine template data;

[0015] Extract joint parameters from the fitted spine template data to obtain joint parameter data;

[0016] Perform posture adaptation according to the user's optical data and the spine template data to obtain posture adaptation data;

[0017] The user's optical data, joint parameter data and posture adaptation data are integrated to obtain external morphological data.

[0018] In the present invention, body shape analysis is performed according to the user's optical data, and the user's body shape characteristics (such as weight distribution, posture, spinal curve, etc.) can be identified in detail. Compared with the traditional "one-size-fits-all" modeling method, the spinal model has personalized characteristics and is more suitable for each user's unique body shape. The user's body shape feature data and the spinal template library are used to generate a standard template to provide a reference framework for spinal modeling. The template library can generate a standard spinal template that meets individual differences by considering the changes in different body shapes and postures, avoiding the limitations of the traditional template method. According to the body shape feature data and the spinal template data are fitted, it can be further accurately adjusted according to the user's specific body shape on the basis of the template. Through fitting, the accuracy and personalization of the spinal model are significantly improved, making the analysis and evaluation more accurate. According to the user's optical data and spinal template data, posture adaptation can be performed, and precise adjustments can be made according to the user's actual posture, so that the spinal model can more accurately reflect the user's actual posture state.

[0019] Preferably, step S2 specifically includes:

[0020] The spinal mechanical model is constructed according to the external morphological data to obtain the spinal mechanical model;

[0021] Perform load analysis based on the spinal mechanical model to obtain spinal load data;

[0022] The support force area is divided according to the spinal load data and the spinal mechanical model to obtain the support force area data;

[0023] Generate a force distribution diagram according to the support force area data to obtain the spine force distribution diagram data;

[0024] The force of the fitted spine template data is optimized according to the spine force distribution diagram data to obtain the spine support data.

[0025] In the present invention, by constructing a spinal mechanical model based on external morphological data, the spinal structure of each user can be personalized and modeled. Unlike the traditional standard model, it can accurately consider the body shape, posture and movement mode of each user, making the mechanical model more individual, thereby providing more accurate basic data for spinal status analysis. Load analysis based on the spinal mechanical model can reveal the behavioral characteristics of the spine under different loads, including the load bearing conditions of various parts of the spine. Generating spinal force distribution map data can clearly show the distribution of support forces in different areas.

[0026] Preferably, the construction of the spinal mechanical model is specifically as follows:

[0027] Reconstructing the spine geometry model according to the external morphological data to obtain the spine geometry model;

[0028] Assigning mechanical properties of the spine to the spine geometric model to obtain a spine physical property model;

[0029] A virtual load environment is established according to the physical property model of the spine to obtain a spinal mechanical model.

[0030] In the present invention, a personalized spine geometry model is constructed through external morphological data (such as body shape characteristics, posture, etc.). The geometric model not only contains the basic morphology and structural characteristics of the spine, but also can be tailored to the individual differences of the user (such as skeleton differences, posture characteristics, etc.). Unlike the traditional general model, the modeling can more accurately simulate the spine structure of each user, thereby providing a more accurate basis for mechanical analysis. On the basis of the spine geometry model, the mechanical properties of the spine (such as bone density, toughness, elastic modulus, stiffness, etc.) are assigned according to biomechanical knowledge and the specific data of each user (such as bone density, muscle distribution, etc.), ensuring that the mechanical model can more accurately reflect the physical properties of each person's spine. For example, for users with different bone qualities, the stiffness and compressive resistance of the spine will be different, and after assignment, the stress conditions can be simulated more realistically. On the basis of the spine geometry model and the physical property model, a virtual load environment is established to simulate the stress behavior of the spine under different load conditions, including static loads (such as standing, sitting, etc.) and dynamic loads (such as walking, running, etc.).

[0031] Preferably, the load analysis is specifically as follows:

[0032] According to the spinal mechanical model, the three-dimensional mechanical model is converted to obtain a preliminary mechanical model;

[0033] Perform dynamic simulation of spine load on the preliminary mechanical model to obtain dynamic load data;

[0034] Finite element meshing is performed based on the dynamic load data to obtain the spinal load data.

[0035] In the present invention, according to the geometric model of the spine and its physical properties, the mechanical model of the spine is converted from a two-dimensional or simple geometric representation to a complex three-dimensional mechanical model through conversion technology, ensuring that the force condition of the spine in the actual environment can be reflected more realistically and in detail. The three-dimensional model takes into account the interaction and mechanical behavior of various parts of the spine (such as vertebral bodies, intervertebral discs, ligaments, etc.), and can analyze the health status of the spine in multiple dimensions. The dynamic simulation of the spinal load is carried out through the three-dimensional mechanical model to simulate the dynamic force condition of the spine in the actual environment. Dynamic simulation not only considers static loads (such as body weight, posture, etc.), but also includes the dynamic loads borne by the spine in daily activities (such as walking, bending, jumping, etc.). The changes in the force on the spine can be tracked in real time, providing extremely detailed dynamic data for understanding the condition analysis of the spine in different movements and activity states. After obtaining preliminary dynamic load data, the spine is meshed by finite element analysis (FEA), and the three-dimensional structure of the spine is converted into a discrete model composed of a large number of small units. Each grid unit calculates the corresponding mechanical parameters such as stress and strain according to the load condition, thereby accurately simulating the performance of the spine under load. Finite element analysis can simulate the stress conditions of the spine in different parts and under different activity states, and optimize the support area according to load changes.

[0036] Preferably, the supporting force area is divided into:

[0037] Perform geometric area division according to the spinal load data and the spinal mechanical model to obtain first support force area data;

[0038] Divide the mechanical area according to the spinal load data and the spinal mechanical model to obtain the second support force area data;

[0039] Perform load distribution balancing according to the first supporting force area data and the second supporting force area data to obtain supporting force area data;

[0040] The specific geometric area division is as follows:

[0041] Surface segmentation is performed based on the spinal load data and the spinal mechanical model to obtain preliminary support force area data;

[0042] Perform load analysis and support force distribution calculation based on preliminary support force area data to obtain support force distribution data;

[0043] Extract local features of the supporting force according to the supporting force distribution data to obtain local supporting force feature data;

[0044] Performing clustering calculation according to the local support force characteristic data to obtain local support force characteristic clustering data;

[0045] A graph is constructed according to the local support force feature clustering data to obtain local support force feature clustering graph data;

[0046] Perform local optimization on the preliminary support force area data according to the local support force feature clustering graph data to obtain local optimization data of the support force area;

[0047] Perform global optimization based on the local optimization data of the support force area to obtain the global optimization data of the support force area;

[0048] Perform regional smoothing and transition optimization on the global optimization data of the supporting force region to obtain the first supporting force region data.

[0049] In the present invention, by combining the spinal load data with the mechanical model, the surface segmentation technology is used to geometrically divide the spine, which can finely identify different areas of the spine. Combined with the mechanical model, the force conditions of each area can accurately reflect the changes of the spine under load. Based on the preliminary support force area data, through load analysis and support force distribution calculation, the support force distribution of each area of ​​the spine can be accurately obtained, and the load influence of the spine in different activities, postures and environments will be considered to ensure that the support force calculation of each area is more real and comprehensive. Local support force feature extraction and clustering calculation can refine the definition of the spinal support force area, so that the support force area can better reflect the actual situation, avoiding the single global calculation in the traditional support force analysis, and ensuring the accuracy and scientificity of each support force area through the clustering of local features. The local support force feature data is constructed into a graph, and the relationship between the support force areas can be effectively captured through the graph theory method, especially the load transition between different areas. The graph-based analysis method provides a new perspective to optimize the support force area, making the transition between areas more natural and reasonable. The graph construction and optimization process can make the division of the support force area more in line with the actual load requirements. The combination of local optimization and global optimization ensures the rationality of each support force area, which not only optimizes the regional morphology, but also reflects the balance of support force in the entire spine. Transition optimization ensures the continuity between support force areas, making the overall mechanical model of the spine more natural, more accurately reflecting the actual spinal condition, and reducing the problem of data distortion caused by different optimizations.

[0050] Preferably, the mechanical region division is specifically as follows:

[0051] Perform load transfer path analysis based on spinal load data and spinal mechanical model to obtain load transfer path diagram data;

[0052] Performing a vertebral material property analysis based on the load transfer path diagram data to obtain vertebral material property data;

[0053] Perform multi-load mode analysis based on the spine material property data to obtain multi-load mode response data;

[0054] The support force density is calculated according to the multi-load mode response data to obtain the support force density data;

[0055] According to the load transfer path diagram data, spinal material property data and support force density data, the spinal mechanical model is divided into mechanical regions to obtain preliminary mechanical region data;

[0056] The boundary fitting and regional reconstruction are performed based on the preliminary mechanical region data to obtain the second support force region data.

[0057] In the present invention, by combining and analyzing the spinal load data and the mechanical model, the mechanical transfer path of the spine under the load can be accurately obtained, which lays the foundation for the division of the support force area, can reveal the stress conditions of each area of ​​the spine under different loads, and reveal the relationship between different parts in the spinal structure. Through the refined assignment of the spinal material properties, the differences of various parts of the spine can be truly reflected in the model. The analysis of the material properties makes the support force area division not only rely on the geometric form, but also takes into account the physical properties of the spine, thereby enhancing the reliability and practical application value of the simulation results. Based on the spinal material property data, through multi-load mode analysis, the response of the spine under various load conditions can be evaluated. It can simulate the stress state of the spine under different activities (such as standing, sitting, walking, etc.) and different external load conditions, and provide data support for the support force distribution under different load modes. The calculation of the support force density can quantify the load conditions of each area of ​​the spine, and provides a clear standard for optimizing the division of the support force area. Through this data, the mechanical model of the spine can be optimized on the basis of ensuring uniform load distribution, reducing the generation of excessive stress and uneven stress areas, thereby improving the overall stability of the spine. Boundary fitting and regional reconstruction can further improve the accuracy and scientificity of the support area division, ensure that the mechanical area division is more detailed, and can accurately reflect the load distribution in different areas. Through this process, the mechanical model of the spine becomes more consistent with the load conditions in the real world.

[0058] Preferably, step S3 is specifically:

[0059] Perform image segmentation according to the spine image data to obtain segmentation area data;

[0060] Position the key points according to the segmented area data to obtain the key point data;

[0061] Extracting the vertebral morphological features from the vertebral image data according to the key point data to obtain the vertebral morphological feature data;

[0062] Perform three-dimensional reconstruction based on the morphological characteristic data of the spine to obtain a three-dimensional spine model;

[0063] Performing spinal curvature analysis on the three-dimensional spinal model to obtain spinal curvature data;

[0064] Evaluate the spinal posture according to the three-dimensional spinal model and spinal curvature data to obtain spinal posture data;

[0065] The vertebral morphological feature data and vertebral posture data are feature vectorized to obtain vertebral bone feature data.

[0066] The accurate image segmentation in the present invention can avoid the errors caused by noise, blur or other factors in traditional methods, so that the key areas of the spine (such as bone structure, intervertebral disc, etc.) can be accurately identified, providing reliable basic data for key point positioning and morphological feature extraction. Accurate key point positioning provides necessary information for subsequent morphological feature extraction and three-dimensional modeling, making the reconstruction process of the spine skeleton more accurate, thereby improving the quality of accurate identification of the spine. Based on the key point data, the morphological feature extraction of the spine image can capture the overall and local morphological features of the spine (such as vertebral size, morphology, curvature, etc.). Through three-dimensional reconstruction technology (such as 3D reconstruction based on deep learning or multi-view stereo vision method), the two-dimensional image data of the spine can be converted into a three-dimensional spine model, which can comprehensively and three-dimensionally display the structure of the spine. Accurate curvature analysis can identify spinal abnormalities, such as scoliosis, kyphosis, etc. Based on the three-dimensional spine model and curvature data, the spine posture assessment can calculate the posture state of the spine (such as flexion, extension, lateral flexion, etc.), and obtain multi-dimensional spine information.

[0067] Preferably, step S4 is specifically:

[0068] Perform geometrical mechanical coupling according to the spine support data and the spine bone characteristic data to obtain the spine geometrical mechanical coupling data;

[0069] Feature extraction is performed based on the spine geometry and mechanical coupling data to obtain spine feature data for auxiliary spine feature recognition operations.

[0070] In the present invention, the geometric structure of the spine (such as the morphology and structural characteristics of the spinal bones) is coupled with the mechanical characteristics (such as force, support force distribution, etc.), which can provide a more comprehensive description of the spine. Geometric mechanical coupling takes into account factors such as the deformation and mechanical stability of the spine under different loads, and can further refine the analysis of the spine. After geometric mechanical coupling, through the feature extraction process, key information of the spine can be extracted from multiple angles, such as the mechanical properties, force distribution, bone morphology, etc. of the spine. Feature extraction provides more comprehensive information for spinal feature recognition and enhances the recognition ability of the model.

[0071] Preferably, the present application also provides a vertebral feature recognition system based on image processing, which is used to execute the vertebral feature recognition method based on image processing as described above, and the vertebral feature recognition system based on image processing comprises:

[0072] The spine data acquisition and preliminary analysis module is used to obtain the user's optical data and spine image data, and perform external morphological analysis based on the user's optical data to obtain external morphological data;

[0073] A spinal support analysis module is used to perform spinal support analysis on external morphological data to obtain spinal support data;

[0074] A vertebral bone feature extraction module is used to extract vertebral bone features according to vertebral image data to obtain vertebral bone feature data;

[0075] The spine multimodal feature fusion module is used to perform spine multimodal feature fusion based on spine support data and spine bone feature data to obtain spine feature data for auxiliary spine feature recognition operations.

[0076] The beneficial effect of the present invention is that the present invention combines optical data, spinal imaging data and support analysis data to perform multimodal fusion of spinal features. Traditional spinal feature recognition methods usually rely only on single imaging data or mechanical data, while ignoring factors such as individual external morphology and mechanical load. By fusing external morphology data, spinal bone feature data and support force distribution data, spinal feature data can be comprehensively reflected from multiple dimensions. The external morphology analysis and spinal support analysis in steps S1 and S2, relying on the combination of user optical data and spinal imaging data, can accurately identify the force changes, load transfer and spinal posture deviation of the spine under different load conditions. Compared with traditional spinal imaging examination methods, the present invention not only focuses on the bone morphology on the image, but also models and analyzes the dynamic load of the spine. Traditional methods mostly focus on the analysis of bone morphology or external posture, ignoring the physical response of the spine under specific load conditions. The present invention combines the mechanical model to couple the spinal bone morphology with the actual load condition, thereby obtaining more realistic spinal data. The support analysis in step S2 is combined with the bone feature extraction in step S3 to provide a more accurate spinal mechanical analysis. The force analysis of the spine not only focuses on the magnitude of the force, but also combines the geometry of the spine, the bone characteristics, and the force transmission path. The present invention can provide the system or user with more comprehensive and accurate spine feature data through the deep integration of the geometry, mechanics, and imaging data of the spine. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting implementations made with reference to the following drawings:

[0078] Figure 1 A flowchart of a method for recognizing spine features based on image processing according to an embodiment is shown;

[0079] Figure 2 A flowchart showing a method for collecting and preliminarily analyzing spinal data according to an embodiment is shown;

[0080] Figure 3 A flowchart showing the steps of a spinal support analysis method according to an embodiment is shown;

[0081] Figure 4 A flowchart of the steps of a method for extracting spine bone features according to an embodiment is shown. DETAILED DESCRIPTION

[0082] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. 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.

[0083] 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.

[0084] 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.

[0085] See also Figures 1 to 4 The present application provides a method for recognizing spine features based on image processing, comprising the following steps:

[0086] Step S1: acquiring user optical data and spinal image data, and performing external morphology analysis based on the user optical data to obtain external morphology data;

[0087] Specifically, a high-resolution optical imaging device (such as a 3D scanner, an infrared sensor, etc.) is used to scan the user's body to obtain the user's optical data of the body. The optical data includes point cloud data or depth images, through which information such as the user's body shape, posture, back curve, etc. can be obtained. The collected optical data is denoised, smoothed and corrected using image processing technology to remove noise and errors. Then, the data is converted into a 3D model or a two-dimensional depth map to construct the user's external morphological model, i.e., the user's optical data. The geometric features of the spine are extracted by analyzing the curve and posture of the user's back. The external morphological data of the spine, such as the curvature of the spine and the inclination angle, are evaluated by detecting the curvature change of the spine area, the inclination of the shoulders and waist, etc. Medical imaging equipment such as X-rays, CT scans, or MRIs are used to obtain the image data of the user's spine. The image data can be a two-dimensional slice image or a three-dimensional reconstructed spine image.

[0088] Step S2: Performing spinal support analysis on the external morphological data to obtain spinal support data;

[0089] Specifically, the support status of the spinal area is analyzed based on the external morphological data. By measuring the force distribution in different areas of the back, the user's posture is analyzed to see if it is normal. The mechanical model is used to simulate the force conditions of the spine in different postures and calculate the force distribution of each spinal segment. According to the morphological data of the user's back, a force analysis model is constructed (based on three-dimensional reconstruction, external load simulation is performed through a preset database for finite element calculation). By analyzing the bending, compression and tension forces of the spine in different postures, the support conditions of each segment of the spine are calculated, and the analysis results are converted into spinal support data, including the support strength of different spinal parts and the degree of spinal curvature.

[0090] Step S3: Extracting vertebral bone features according to the vertebral image data to obtain vertebral bone feature data;

[0091] Specifically, the image data is processed, including image enhancement, noise removal and edge detection, to ensure the quality of the image data. The image data is converted into a form suitable for analysis, such as converting CT slice data into 3D volume data, or performing three-dimensional reconstruction on MRI data. Feature extraction is performed on the preprocessed spinal image data to identify the various bone structures of the spine, including the connection between the vertebral body, intervertebral disc and vertebrae. Morphological operations, edge detection and segmentation algorithms are used to extract the characteristic data of the spinal skeleton from the image data. The features include the shape, size, gap, curvature, etc. of the vertebrae. The extracted spinal bone feature data will include the geometric dimensions of each part of the spine, the arrangement of the vertebrae, the angle and curvature of each segment of the spine, etc. These feature data provide the basis for multimodal feature fusion.

[0092] Step S4: Performing spinal multimodal feature fusion according to the spinal support data and the spinal bone feature data to obtain spinal feature data for spinal feature recognition auxiliary work.

[0093] Specifically, the spinal support data and spinal bone feature data from step S2 and step S3 are fused. The two types of data are standardized so that their numerical ranges are consistent. The two types of data are merged into a high-dimensional feature vector using weighted average, principal component analysis or other feature fusion methods. Weighting is performed according to the health status of the spine and the importance of the features to ensure that the fused data can reflect the overall situation of the spine. On the basis of the fusion of spinal support data and spinal bone feature data, the expression ability of the features is further enhanced by a multimodal learning method. A regression model, a neural network or other machine learning model is used to train a model of multimodal feature fusion. The model can identify the specific situation of the spine and the specific morphology of different parts based on the user's spinal morphology, support situation, bone features and other information. The feature recognition of the spine is performed by analyzing the multimodal fusion data. The recognition process can include posture analysis, spinal curvature analysis, spinal detection, etc. to identify the specific situation of the spine. The recognition results can be displayed through a graphical interface to help users understand the specific condition of the spine.

[0094] Preferably, step S1 specifically includes:

[0095] Step S11: Acquire user optical data and spine image data;

[0096] Specifically, a high-precision optical scanning device (such as a 3D scanner, laser scanner, or depth camera) is used to perform three-dimensional imaging of the user's body. The device can generate point cloud data or depth images of the user's body surface, including the user's external body shape and posture information. Medical imaging technology such as X-rays, CT scans, or MRI is used to obtain image data of the user's spine. The image data is a two-dimensional slice image or three-dimensional volume data, including structures such as the bones and intervertebral discs of the spine.

[0097] Step S12: performing body shape analysis based on the user's optical data to obtain body shape feature data;

[0098] Specifically, the acquired optical data of the user is subjected to morphological analysis to extract the body shape features related to the spine. Body shape features include back curve (such as the degree of lordosis or kyphosis of the spine), shoulder width, waist height, hip width, leg length, etc. These features can reflect the body shape structure of the user and provide a basis for the generation of the spine template. The point cloud data is denoised, smoothed and corrected. Through the feature point extraction algorithm (such as edge detection based on distance transformation or surface reconstruction method), the user's back area is identified, the general outline of the spine position is extracted, and the specific value of the body shape feature is calculated. These body shape feature data will be used for the subsequent generation of the spine template.

[0099] Step S13: generating a standard spine template according to the body shape feature data and a preset spine template library to obtain spine template data;

[0100] Specifically, a spine template library containing a variety of typical body shapes is preset, and the template library contains standard spine templates of different ages, genders, body shapes and health conditions. These templates are generally obtained through medical imaging data and anatomical research. According to the user's body shape characteristics (such as spinal curvature, shoulder position, waist height, etc.), they are matched with the standard templates in the spine template library. Using a geometric registration algorithm, the user's body shape characteristics are matched with the standard spine template to generate a standard spine template suitable for the user's body shape. The template includes geometric information such as the position, gap, and curvature of each vertebra of the spine. After the standard spine template is generated, a set of data including parameters such as the position, size, angle, and intervertebral gap of each vertebra of the spine will be obtained as the spine template data.

[0101] Step S14: performing spinal template fitting according to the body shape feature data and the spinal template data to obtain fitted spinal template data;

[0102] Specifically, the user's body feature data and standard spine template data are used for template fitting. The various structures of the standard spine template are optimized and matched with the user's body feature data. The spine template is adjusted through iterative optimization algorithms (such as least squares method, gradient descent method) to make it more closely fit the user's actual body shape. By adjusting the angle, position and shape of the spine template, the fitting result can reflect the actual structure of the user's spine to the greatest extent. The template's bending angle, intervertebral space width and spinal segments are slightly adjusted to ensure that the fitted spine template can meet the individual differences of the user. After the fitting is completed, the fitted spine template data is obtained, including the specific position, angle, size, intervertebral space and other information of the spine.

[0103] Step S15: extracting joint parameters from the fitted spine template data to obtain joint parameter data;

[0104] Specifically, according to the fitted vertebral template data, the parameters of each vertebral joint (mainly the connection between the intervertebral disc and the vertebra) are extracted. These joint parameters include the angle, displacement, rotation, etc. of the joint. The relative motion of each joint is calculated using a mathematical model, such as the change in the width of the intervertebral space, the rotation angle of the spine, etc. For example, the motion of each joint is divided into rotational motion, the rotation of the vertebra relative to the adjacent vertebra; translational motion, the translation of the vertebra relative to the adjacent vertebra; the change of the intervertebral space, the change of the relative position between the vertebrae, especially the change of the intervertebral disc in different postures; a local coordinate system is established for each vertebral segment, so that the motion of each vertebra can be described. The right-handed coordinate system is used, and the coordinate system of each vertebra will coincide with the coordinate system of the adjacent vertebra. The motion (translation and rotation) of the vertebra will be described by the transformation between adjacent coordinate systems, such as the calculation of the rotation angle of the vertebral joint, the change of the intervertebral space width is described by calculating the relative position between the vertebrae, and the translation and rotation of the vertebrae in three-dimensional space are described by the transformation matrix. The range of motion of each joint is obtained by statistically analyzing the rotation angle and displacement range of each joint in different postures, and the stability of the joint is calculated by analyzing the balance of joint motion. By analyzing the relative motion between the various joints of the spine, characteristic data such as the range of motion and stability of the joints are extracted. The obtained joint parameter data includes the kinematic parameters of each spinal segment, such as joint angle, joint position change, intervertebral space size and other information.

[0105] Step S16: performing posture adaptation according to the user's optical data and the spine template data to obtain posture adaptation data;

[0106] Specifically, according to the user's optical data and spine template data, the user's spine posture is adjusted so that the template is more consistent with the user's actual posture, including the degree of spinal curvature, rotation angle, flexion and backward posture adjustment. An algorithm based on constraint optimization (such as an inverse kinematics model) is used to adjust the spine posture to ensure that the skeletal position of the spine in different postures matches the template. The algorithm dynamically adjusts the angle and displacement of the spine by solving a set of nonlinear equations to achieve the goal of matching the user's actual posture. After completing the posture adaptation, the user's spine posture data is obtained, including information such as the spinal curvature angle, rotation angle, and displacement of the spinal segments.

[0107] Step S17: Integrate the user optical data, joint parameter data and posture adaptation data to obtain external morphology data.

[0108] Specifically, the user's optical data (such as body shape feature data), joint parameter data and posture adaptation data are integrated. The integration process is to combine these data according to certain weights and rules, or generate a new data group to include them, so that the result can fully reflect the user's external form. The integrated data is standardized so that different types of data can be compared and fused in the same coordinate system. The features of each data source (body shape, joints, posture, etc.) are merged in proportion to generate a complete external form data set. The generated external form data contains the user's complete body shape, spinal joint kinematic parameters and posture adaptation information.

[0109] Preferably, step S2 specifically includes:

[0110] Step S21: constructing a spinal mechanical model according to the external morphological data to obtain a spinal mechanical model;

[0111] Specifically, when constructing the mechanical model of the spine, the spine is assumed to be composed of a series of rigid bone structures (vertebrae) and elastic connections (intervertebral discs). The force on each vertebra is represented by the stiffness matrix, while the elastic modulus is used to describe the deformation and load-bearing capacity of the intervertebral disc. The vertebral template data in the external morphological data (including the geometry, size, position, etc. of the vertebrae) and the relative position of each segment of the spine are used to construct a three-dimensional geometric model of the spine. The position and size of each vertebra and intervertebral disc are associated with the external morphological data through mathematical modeling to obtain a skeletal model of the spine. Based on the known anatomical data and the mechanical properties of different vertebrae in the literature, appropriate mechanical properties are assigned to each vertebra and intervertebral disc in the spine model. These properties include vertebral stiffness. The vertebrae are rigid, and their stiffness is determined by factors such as the material properties and geometry of the vertebrae. Intervertebral disc elasticity. The elastic modulus of the intervertebral disc represents its deformation capacity under load, which can be obtained through experimental data or literature. Mechanical model establishment. Using the above data, a mechanical model of the spine is constructed through finite element analysis (FEA) and other methods. The model can simulate the deformation of each vertebra and intervertebral disc under the action of external forces. The model connects each vertebra and intervertebral disc to form a three-dimensional mechanical system.

[0112] Step S22: performing load analysis according to the spinal mechanical model to obtain spinal load data;

[0113] Specifically, multiple load conditions are applied to the spinal mechanical model, mainly including gravity load. The user's weight is applied to the spine through the upper part of the spine (such as the neck and waist), and the gravity generated will affect the force distribution of the spine. Posture-induced loads, according to the user's posture (such as standing, sitting, bending, etc.), calculate the load on each segment of the spine. Under different postures, the force of each segment of the spine will be different. External force, if the user is doing certain sports or weight-bearing activities, the external force (such as impact force or weight during exercise) also needs to be considered. Load distribution analysis, in the spinal model, the force of each vertebra and intervertebral disc is calculated. Through mechanical formulas (such as equilibrium equations, stress-strain relationships, etc.), the force of each vertebra and intervertebral disc under different load conditions is analyzed. The calculation results will provide detailed load data for each spinal segment, including the axial force, shear force and bending moment borne by each vertebra.

[0114] Step S23: dividing the support force area according to the spinal load data and the spinal mechanical model to obtain support force area data;

[0115] Specifically, according to the spinal load data, the spinal load force characteristics are extracted to obtain the spinal load force characteristic data; according to the spinal load force characteristic data, the spinal mechanical model is divided into support force areas to obtain first support force area data; according to the spinal load data, the spinal load deformation characteristics are extracted to obtain spinal load deformation characteristic data; according to the spinal load deformation characteristic data, the spinal mechanical model is divided into support force areas to obtain second support force area data;

[0116] The specific extraction of the spinal load force characteristics is as follows:

[0117] Divide the spinal region according to the spinal load data to obtain spinal region load data; classify the load data according to the spinal region load data to obtain spinal region load classification data; calculate the regional average load on the spinal region load classification data to obtain spinal region average load data; extract the load distribution characteristics and calculate the maximum stress area on the spinal region load classification data to obtain load distribution characteristic data and load maximum stress area data; perform cluster calculation on the spinal region average load data, load distribution characteristic data and load maximum stress area data according to the spinal region load classification data to obtain spinal load force characteristic data;

[0118] The spinal region load classification data includes first spinal region load classification data, second spinal region load classification data and third spinal region load classification data. The load data classification is specifically as follows:

[0119] Directional load data is classified according to the spinal regional load data to obtain first spinal regional load classification data; regional load data is classified according to the spinal regional load data to obtain second spinal regional load classification data; load transfer path processing is performed according to the spinal regional load data to obtain load transfer path data; joint load transfer characteristics are extracted according to the load transfer path data to obtain joint load transfer characteristic data; load correlation data is classified according to the spinal regional load data and the joint load transfer characteristic data to obtain third spinal regional load classification data;

[0120] The specific extraction of spine load deformation features is as follows:

[0121] The bending moment and displacement characteristics are calculated based on the spinal load data to obtain spinal load bending moment data and spinal load displacement characteristic data; the spinal load bending moment data and spinal load displacement characteristic data are integrated to obtain spinal load deformation characteristic data.

[0122] The spine is divided into five major regions: cervical region (C1 to C7), thoracic region (T1 to T12), lumbar region (L1 to L5), sacral region (S1 to S5), and coccyx region (Coccyx). Three-dimensional morphological data of the spine is obtained through scanning and CT imaging. The spine is analyzed using a mechanical model in combination with external loads from load data (such as body weight, movement, etc.). The load data is applied to each spinal segment to obtain the load distribution in each region. The load data is classified by direction (for example, up and down, horizontal, and oblique loads). For example, the cervical region is mainly subjected to loads in the direction of up and down, while the loads in the lumbar region are mostly forward or backward bending loads. Regional load classification is to classify the loads according to different spinal regions. For example, the cervical and thoracic loads are quite different in type and size, and the lumbar region has a larger load. The data in each region is calculated and analyzed separately. The load transfer path is treated as the transfer of loads between the spine is carried out segment by segment, and the spine interacts through joints to transfer loads from one segment to another. The data of the load transfer path is used to simulate the movement and load transfer process of the spine through a mechanical model. The joint load transfer feature extraction is to perform a detailed analysis on the joint parts of the intervertebral disc and vertebrae to extract the response characteristics of the joint during the load transfer process. For example, the change of the intervertebral space, the rotation angle, etc. Using the load data, the load distribution diagram of each area of ​​the spine is calculated through space. Based on the load distribution, the stress of each area is calculated, and the finite element analysis method (FEA) is used to simulate the force of the spine under load. By calculating the stress of each grid unit, the maximum stress area under the load of the spine is obtained. Through clustering algorithms (such as K-means or hierarchical clustering), the spinal regional load data, regional average load data, load distribution feature data, and maximum stress area data are clustered and calculated to obtain the spinal load force characteristics. Based on the clustering results, the stress, strain, displacement and other characteristics of each spinal area under different load conditions are calculated. For example, the lumbar region has a large deformation under load conditions, while the cervical region is relatively stable. By analyzing the load-force characteristics of the spine and combining the spinal mechanics model, the spine is divided into support force areas segment by segment. The first support force area is based on the load-force characteristic data of the spine to divide the spine area and determine which areas have a stronger support effect on the load. For example, the lumbar area is the main part of the support force area, and the cervical area is a relatively small support area. The second support force area is based on the load deformation characteristic data of the spine (such as bending, displacement, etc.), and the spine is further divided to find the area with larger deformation as the second support force area. For example, the intervertebral disc injury area, the area with larger spinal curvature, etc.

[0123] Specifically, based on the spinal load data, the load distribution borne by different spinal segments is analyzed. Spinal segments with larger loads are located in the lower back (such as the lumbar spine) or the pressure-bearing area of ​​the spine. These areas play an important role in supporting the spine. The spine is analyzed segment by segment, and the mechanical model is used to determine which spinal segments provide the main support under different loads. The spine is divided into different support force areas according to the force and deformation of the vertebrae and the degree of curvature of the spine. These areas can be classified according to their force magnitude, degree of deformation or stability. Using the load data output by the model, the support force area of ​​each spinal segment is calculated. For example, the lumbar region bears a larger load, so its support force area is larger, while the cervical region bears a smaller load and has a smaller support force area. The support force area data is presented in the form of heat maps, three-dimensional maps, etc., indicating the support force size of each area of ​​the spine.

[0124] Step S24: generating a force distribution diagram according to the support force area data to obtain spinal force distribution diagram data;

[0125] Specifically, a spinal force distribution diagram is generated based on the support force area data and the spinal load data. The force distribution diagram visualizes the force conditions of each spinal segment and displays the force conditions of each spinal segment. The data of the support force area is mapped onto the three-dimensional model of the spine. Color coding (such as heat maps) is used to represent the force intensity of each spinal segment. Areas with greater loads are displayed in red or dark colors, and areas with less loads are displayed in green or light colors. By analyzing the generated force distribution diagram, the overall force condition of the spine is evaluated. The diagram can show whether there is uneven force on the spine, such as excessive or insufficient force on certain segments. The generated force distribution diagram data is displayed through three-dimensional visualization software to help users understand the force distribution of the spine.

[0126] Step S25: Optimize the force of the fitted spine template data according to the spine force distribution diagram data to obtain spine support data.

[0127] Specifically, the force distribution map data of the spine describes the stress and support force distribution in each area of ​​the spine. It is a distribution map calculated from the load analysis to reflect the actual force situation of the spine. The fitted spine template data contains information such as the geometric position and mechanical parameters of the vertebrae and intervertebral discs, and is a theoretical template generated by fitting. The force distribution map data of the spine is compared with the fitted spine template data to calculate the force difference of each spinal segment. The goal of force optimization is to minimize the difference between the actual force distribution and the theoretical template. Maintain the overall mechanical balance of the spine: the optimized force must meet the overall mechanical balance to ensure that the sum of the force and the sum of the moment of the spine are zero: force area boundary constraints: the optimized support force distribution cannot exceed the reasonable anatomical and geometric boundaries. For high deviation areas, the force distribution is corrected segment by segment. If the force deviation of a vertebra is too large, adjust its support force strength to make it close to the theoretical value. Smooth the pressure distribution of the intervertebral disc to ensure that its deformation and force are in line with the actual situation. Globally smooth the corrected force distribution to eliminate overly abrupt force changes. Through iterative smoothing, the force jump problem caused by local abnormal points is reduced. Adopt iterative optimization method to gradually reduce the difference between actual force and theoretical template: Initialization: Set the initial optimization value to the theoretical template value. Each iteration: Calculate the difference between the current force distribution and the template. Correct the difference according to the optimization objective function. Termination condition: When the sum of the squares of the optimized force difference is less than the set threshold, stop the iteration.

[0128] Preferably, the construction of the spinal mechanical model is specifically as follows:

[0129] Reconstructing the spine geometry model according to the external morphological data to obtain the spine geometry model;

[0130] Specifically, based on external morphological data (such as the user's 3D scan data or spine template data), the geometric model of the spine needs to be reconstructed first. This process is divided into the following steps: the 3D point cloud or segmented image data obtained from optical scanning or medical imaging needs to be preprocessed, including denoising, interpolation and point cloud reconstruction, etc., to ensure that a complete and missing 3D data set is obtained, including applying filtering algorithms to the point cloud data to remove error points and outliers. The point cloud is converted into a polygonal mesh (such as a triangular mesh), and the 3D surface of the spine is constructed using a surface reconstruction algorithm (such as Poisson reconstruction). Using the 3D reconstructed mesh data, the various components of the spine are identified, including vertebrae, intervertebral discs, and joint positions. The geometric model of each vertebra is generated by matching it with the geometric model of the spine in the standard template library. These vertebrae will be connected to other vertebrae through intervertebral discs to form a complete geometric structure of the spine. The geometric shape of each vertebra can be modeled by describing the length, width, height and other dimensions of the vertebra, which can be normalized by the size and shape characteristic parameters of the vertebra. As an elastic structure, the intervertebral disc represents its morphology through the geometry of a disc or elliptical structure. The obtained spine geometric model will include the geometric position, size, angle of the vertebrae and the geometric position of the intervertebral discs. All information is converted into a data structure in a three-dimensional spatial coordinate system for subsequent analysis.

[0131] Assigning mechanical properties of the spine to the spine geometric model to obtain a spine physical property model;

[0132] Specifically, after the geometric model of the spine is established, it is necessary to assign physical properties to the model so that its mechanical behavior under different loads can be simulated. These properties usually include the stiffness of the vertebrae, the elasticity of the intervertebral disc, the friction of each joint, etc. The vertebrae are regarded as rigid objects, and their stiffness can be defined by the elastic modulus and geometric properties of the material. The stiffness of the vertebrae is determined by its material (such as the elastic modulus of bone tissue) and its geometric shape, that is, the elastic modulus is the cross-sectional area of ​​the vertebra multiplied by the length of the vertebra.

[0133] As an elastic structure, the intervertebral disc will deform when subjected to stress. The elastic modulus of the intervertebral disc can be set based on the biomechanical model, usually measured experimentally, considering its elastic response in compression and bending. The elastic modulus of the intervertebral disc can be expressed by the following formula, that is, the elastic modulus multiplied by the strain is the stress representation.

[0134] There is friction between the joints of the spine (such as the sliding surface of the intervertebral disc). The friction can be calculated by the contact mechanics model and determined according to the physical characteristics of different joints (such as surface roughness and lubrication degree), that is, different friction coefficients are obtained according to the data and multiplied by the normal force. All physical properties (stiffness, elastic modulus, friction, etc.) are assigned to the corresponding parts of the spine geometry model. These properties will serve as input data to provide a basis for load analysis and mechanical model calculation.

[0135] A virtual load environment is established according to the physical property model of the spine to obtain a spinal mechanical model.

[0136] Specifically, after the physical property model of the spine is assigned, the next step is to establish a virtual load environment for the spinal mechanical model. This process involves applying actual load conditions to the spinal model to simulate the stress state of the spine in different scenarios. A variety of load conditions are applied according to the actual use scenario of the spine. Load conditions include static loads: such as vertical loads caused by body weight, applied to the upper end of the spine (such as the cervical or lumbar spine). Dynamic loads: For example, the instantaneous impact force borne by the spine during actions such as bending and jumping. Forces caused by posture: According to the user's posture (such as sitting, standing, etc.), the forces on the spine in these postures are simulated. Load analysis can be performed through numerical simulation. Through finite element analysis (FEA) or other numerical methods, different load conditions are input into the spinal model to simulate the stress conditions of each vertebra and intervertebral disc of the spine. The goal of load calculation is to obtain mechanical parameters such as stress, strain and displacement of the spine under these load conditions. After establishing the virtual load environment, the mechanical model is used to simulate the force distribution, deformation and relative displacement of each segment of the spine. The mechanical behavior data of the spine under a virtual load environment are obtained, including the stress distribution, deformation, contact force, etc. of each vertebra and intervertebral disc.

[0137] Preferably, the load analysis is specifically as follows:

[0138] According to the spinal mechanical model, the three-dimensional mechanical model is converted to obtain a preliminary mechanical model;

[0139] Specifically, based on the construction of the mechanical model of the spine, the first step is to convert the geometric model into a three-dimensional mechanical model that can be used for mechanical analysis. The geometric features and physical properties of the spine are numerically processed to facilitate subsequent load analysis. The geometric model of the spine is converted into a mechanical model through the preprocessing module of the finite element analysis. First, the geometric model of the spine needs to be segmented, and each vertebra and intervertebral disc of the spine is regarded as a group of finite element units. Each vertebra and intervertebral disc can be regarded as composed of small three-dimensional units (such as tetrahedral or hexahedral units). Each vertebra and intervertebral disc consists of nodes and elements in the mechanical model. Nodes represent the discretized points of an object, and elements are geometric units that connect nodes. For each vertebra and intervertebral disc, the connection relationship between its corresponding nodes and elements is defined, and each element is assigned according to physical properties (such as elastic modulus, density, etc.). When converting the three-dimensional mechanical model based on the geometric model, the physical properties of each part of the spine (such as stiffness, elastic modulus, friction, etc.) are considered, and these physical properties are assigned to each unit to obtain a preliminary mechanical model. Furthermore, the spine has been discretized into a mesh model composed of multiple finite element units, providing a basis for load analysis.

[0140] Perform dynamic simulation of spine load on the preliminary mechanical model to obtain dynamic load data;

[0141] Specifically, after the preliminary mechanical model is established, the next step is to simulate the dynamic load of the spine. The dynamic simulation of the spinal load is to simulate the response of the spine in motion, posture changes, etc. by applying a time-varying load. For example, the following steps: According to the application scenario, apply the corresponding dynamic load conditions. The dynamic load can be a gravity load, bending load, impact force, etc. caused by body weight. The application of the load needs to consider the following aspects: Body weight load: Apply a static load in the vertical direction. Posture load: Apply loads of different directions and sizes according to different postures (such as bending, standing, sitting, etc.). Dynamic impact force: The instantaneous impact force on the spine during movement (such as jumping, running, etc.). For each load, define its size, position and direction, as well as the time-varying characteristics of the load (such as the change of the impact force over time, the increase or decrease of the force, etc.). When performing dynamic simulation on the preliminary mechanical model, it is necessary to solve the motion equation of the spine under the action of the load. The dynamic equation consists of the mass, stiffness matrix and damping matrix of the particle. Under the action of the load, the spine will deform, and the process is described by the following motion equation: ,in Externally applied loads, including gravity, impact, etc. is the mass matrix, which represents the mass distribution of vertebrae and intervertebral discs, is the acceleration of each part of the spinal system, describing the dynamic motion state of the system, is the damping matrix, which represents the energy dissipation, is the velocity of each part of the spinal system, is the stiffness matrix, which represents the stiffness of each part of the spine, is a displacement vector, which represents the deformation of the spine. By solving the dynamic equations, the dynamic response of the spine at different time points is obtained, including: the displacement, velocity, and acceleration of each vertebra and intervertebral disc; the stress and strain of each vertebra and intervertebral disc; and the overall deformation of the spine after the load is applied. These data constitute the response data of the spine under dynamic load, that is, dynamic load data. By comparing the responses under different loads, the force of the spine under specific conditions can be evaluated.

[0142] Finite element meshing is performed based on the dynamic load data to obtain the spinal load data.

[0143] Specifically, after obtaining the dynamic load data, finite element meshing is performed and the load data of the spine is generated. This process is mainly used to finely simulate the behavior of the spine under complex load environments and provide detailed force analysis. Finite element meshing is to divide the three-dimensional mechanical model of the spine into small discrete units for numerical calculations. The following techniques will be used: Meshing: Using meshing algorithms (such as tetrahedral or hexahedral meshes), the spine geometry model is automatically divided into smaller units. Mesh optimization: The mesh is optimized to make it more detailed and ensure that important areas (such as intervertebral disc connections, stress-bearing areas, etc.) have higher resolution. Through meshing, a mesh model consisting of multiple elements and nodes is generated. Each element represents a small geometric unit of the spine, and the node is the intersection of the element. When performing finite element analysis, the stress, strain, and displacement data of the spine on each mesh unit will be solved based on the previously calculated dynamic load data (such as the size, direction, and time variation of the load). The basic steps of finite element solution are as follows: Establish a stiffness matrix: For each mesh unit, a stiffness matrix is ​​established based on its geometric shape and physical properties. Applying loads and boundary conditions: Apply external loads and boundary conditions to the mesh model. The boundary conditions are usually fixed points or support points of the vertebrae. Solving the linear equations: Solve the stress and strain equations to obtain the force and deformation data of the spine on each mesh unit. After finite element analysis, the output spinal load data will include: stress and strain distribution of each vertebra and intervertebral disc; dynamic response such as displacement and acceleration of each mesh unit; force intensity distribution of each area, especially the pressure-bearing area of ​​the spine, such as the lumbar spine and cervical spine. The obtained spinal load data will provide an important basis for subsequent force optimization, support force area division and other steps. The meshing in the three-dimensional mechanical model conversion is to establish a discretized model. The main goal is to mechanize the geometric model and provide an input basis for subsequent analysis. Finite element meshing is a further refinement of the previous model. The purpose is to optimize the mesh to meet the simulation requirements of complex scenarios under the background of precise force analysis.

[0144] Preferably, the supporting force area is divided into:

[0145] Perform geometric area division according to the spinal load data and the spinal mechanical model to obtain first support force area data;

[0146] Specifically, the spine is divided into geometric regions according to the geometric shape and load data of the spine, with the aim of determining the main areas where the various parts of the spine (such as vertebrae, intervertebral discs, joints, etc.) are supported. First, by analyzing the three-dimensional geometric model of the spine, the geometric features of the various components of the spine (such as vertebrae, intervertebral discs, vertebral curves, etc.) are determined, including the shape of the vertebrae, the position of the intervertebral discs, the curvature of the spine, etc. These geometric features can be obtained from the established geometric model of the spine, including: the size, shape and relative position of the vertebrae; the thickness and position of the intervertebral discs; the curvature or bending angle of the spine, etc. By analyzing the geometric model of the spine, the support area of ​​the spine can be preliminarily determined based on the shape and distribution of the intervertebral discs and vertebrae. For example, the contact surface between the intervertebral disc and the vertebrae, and between the vertebrae and the vertebrae are the main areas where the spine is subjected to force. The division is performed through the following steps: Contact surface analysis: The contact surfaces of various parts of the spine (such as the contact surface between the intervertebral disc and the vertebrae) are extracted through the geometric model as the first support force area. The size of the contact area between the intervertebral disc and the vertebrae: According to the shape of the intervertebral disc, the area of ​​contact between the intervertebral disc and the vertebrae is delineated and used as the boundary of the first support force area. Calibrate the support force strength: On each contact surface, determine the support force strength it can withstand based on the mechanical model. At this time, the first support force area data obtained includes the support area of ​​each vertebra and intervertebral disc in the spine and the preliminary distribution of their support force.

[0147] Divide the mechanical area according to the spinal load data and the spinal mechanical model to obtain the second support force area data;

[0148] Specifically, further mechanical area division is performed according to the load data and mechanical model of the spine, the purpose is to accurately define the mechanical characteristics of the support area by analyzing the force of the spine under different loads. The load data of the spine is combined with the mechanical model, dynamic load is applied, and the stress and strain distribution of the spine under load is calculated. Through finite element analysis, the stress distribution of each part of the spine under different loads is obtained, such as stress analysis, using the mechanical model of the spine to calculate the stress distribution under load conditions. The stress distribution of each vertebra and intervertebral disc under load can help determine its support function under load. Strain analysis obtains strain data of vertebrae and intervertebral discs through mechanical models. Higher strain areas mean that the force is greater, which will cause deformation or damage, so these areas are also places with stronger support. According to stress distribution and strain data, the mechanical area of ​​the spine is further divided. The support force area is delineated by the following methods, such as stress threshold division, setting a stress threshold, and delineating the support area according to the stress magnitude. Higher stress areas indicate that the load on the spine is greater, so these areas should be delineated as areas with stronger support. Or, similar to stress analysis, the support force area is divided by setting a strain threshold. Areas with greater strain mean that the spine is undergoing greater deformation and are key areas of support. By analyzing the stress and strain distribution and combining the division of mechanical areas, the second support force area data is obtained. These data include: the stress and strain distribution of each vertebra and intervertebral disc under load; the support force size and force intensity of each area. These data provide a basis for load distribution balance and optimization.

[0149] Perform load distribution balancing according to the first supporting force area data and the second supporting force area data to obtain supporting force area data;

[0150] Specifically, the first support force area data and the second support force area data are combined, and the final support force area data is obtained by load distribution balance. By combining the first and second support force areas, it is analyzed whether the loads in various areas of the spine are evenly distributed. If some areas are subjected to too much or too little force, the spine will be unbalanced, which will affect its stability and health. According to the load data, a distribution model of spinal support force is established. This model needs to comprehensively consider the force conditions of each area and optimize it. By balancing the load distribution, it is ensured that the force on the spine is uniform and avoids local overload or stress concentration. If some areas are subjected to too much force (such as the lumbar area), the optimal distribution of the load can be achieved by adjusting the posture, correcting the position of the intervertebral disc, and so on. By balancing the load distribution, the final load data of each support area is obtained. These data include: distribution of the size of the support force: assigning a load coefficient to each area to indicate the support capacity of the area; uniformity of the distribution of the support force: evaluating whether the support force distribution is uniform and whether there is overload or insufficient. Through the above steps, the spinal support force area data is obtained. This data set includes: the support force areas of each vertebra, intervertebral disc and the entire spine; the force intensity and distribution in each area; the balance and optimization effect of the support force.

[0151] The specific geometric area division is as follows:

[0152] Surface segmentation is performed based on the spinal load data and the spinal mechanical model to obtain preliminary support force area data;

[0153] Specifically, starting from the geometric model of the spine, combined with the load data, the surface of the spine is segmented to identify various possible support areas. First, the curved surface of the spine is extracted from the three-dimensional geometric model of the spine, and a mesh model representing the surface of the spine is generated by calculating the boundaries of each vertebra and intervertebral disc of the spine. For example, the vertebral surface extraction uses a three-dimensional reconstruction algorithm to obtain the outer boundary of the vertebra from the CT or MRI image to generate the curved surface of the vertebra. Intervertebral disc surface extraction, similarly, the surface of the intervertebral disc can be extracted from the image data by image segmentation technology to obtain the geometric model of the intervertebral disc. According to the load data and geometric model of the spine, the system divides the surface of the spine into different areas. These areas represent different distribution states of support force. For example, high-load areas, in areas with large loads or large pressures (such as the lumbar spine, cervical spine, etc.), high-load areas can be delineated by calculating the force conditions of the spine. Low-load areas, in areas with small loads (such as the top of the spine), low-load areas are delineated by load data. The division of these areas is based on the mechanical model of the spine and the load analysis results. By segmenting the curved surface and combining the load data, the support force area of ​​the spine is preliminarily delineated. The boundary of each support area is determined by the surface segmentation result, and the load data (such as stress or pressure) is used to calibrate these areas to obtain preliminary support force area data.

[0154] Perform load analysis and support force distribution calculation based on preliminary support force area data to obtain support force distribution data;

[0155] Specifically, through the analysis of the preliminary support force area, further load analysis is carried out to calculate the distribution of support force in each area. The system uses a mechanical model to analyze the distribution of load on the surface of the spine (especially the contact area between the vertebrae and the intervertebral disc). Through numerical calculation and simulation, the force conditions of each support area are obtained. Factors considered include gravity and external forces, body weight and external loads applied to the spine. Sports loads, such as dynamic loads generated during walking and running. By calculating the load data of each support area, the spatial distribution of support force is obtained. Finite element analysis is performed, and the finite element method is used to calculate the stress, strain, and force distribution of each area. The stress value of each area is associated with the load data to obtain the distribution of support force and the pressure distribution map. A pressure distribution map is generated on the surface of the spine. The pressure value of each area reflects the strength of the support force in that area.

[0156] Extract local features of the supporting force according to the supporting force distribution data to obtain local supporting force feature data;

[0157] Specifically, based on the support force distribution data, local features of the spinal support area are extracted for further optimization. For each support area, local support force features are extracted based on its load data (such as pressure, stress, strain, etc.). These features include the maximum support force, the maximum force value in each support area. Support force intensity distribution, the force distribution curve or map in the area. Local strain is the strain data of the local area, reflecting the deformation of the area. By analyzing the local support force characteristics, we can understand which areas are under strong force and which areas have potential uneven force or overload.

[0158] Performing clustering calculation according to the local support force characteristic data to obtain local support force characteristic clustering data;

[0159] Specifically, similar local support force features are clustered together through a clustering algorithm. Based on the local support force features (such as support force strength, stress distribution, etc.), a clustering algorithm is applied to cluster similar support areas together. Clustering methods include K-means clustering, which divides all local support force features into multiple categories according to similarity. DBSCAN clustering, a density-based clustering method, clusters adjacent support areas according to spatial density. The result of the clustering calculation is which cluster category each support area belongs to, reflecting the similarity of these areas in support force distribution.

[0160] A graph is constructed according to the local support force feature clustering data to obtain local support force feature clustering graph data;

[0161] Specifically, according to the clustering results of the local support force characteristics, a relationship graph between the support areas is constructed to further optimize the support force areas. Based on the clustering results, a graph between the support areas is constructed. Each node of the graph represents a support area, and the edges between the nodes represent the similarity or load transfer relationship between the support areas. The weight of the edge can be defined according to the force transfer efficiency between the support areas. Each node (support area) in the graph is assigned its characteristic value, such as the support force size, load type, stress strain, etc. The weight of the edge represents the force transfer relationship between adjacent areas, or the load similarity between the two.

[0162] Perform local optimization on the preliminary support force area data according to the local support force feature clustering graph data to obtain local optimization data of the support force area;

[0163] Specifically, through graph optimization technology, the local support force characteristics are optimized to ensure the reasonable distribution of support force. According to the uniformity requirement of support force distribution, the support force of each area is adjusted through the optimization algorithm to eliminate the overload area and ensure that the load of each part of the spine is more uniform. For example, adjust the high-load area and optimize the area with excessive load to reduce its force. Enhance the low-load area and optimize the area with too little load to enhance its support capacity. The following methods can be used for local optimization. Linear programming, by solving a system of linear equations, optimize the support force distribution of each area. Genetic algorithm, use genetic algorithm for global optimization to ensure the balance and optimality of support force distribution.

[0164] Perform global optimization based on the local optimization data of the support force area to obtain the global optimization data of the support force area;

[0165] Specifically, on the basis of local optimization, global optimization is performed to ensure the load balance of the entire spine. The goal of global optimization is to eliminate the uneven load throughout the body and ensure that each area of ​​the spine is within a reasonable force range. Global optimization needs to consider the mechanical balance of the entire spine to make the load between each support area more uniform. Through global optimization algorithms (such as particle swarm optimization, simulated annealing, etc.), each support area is adjusted to ensure a more uniform load distribution and avoid overload or underload in local areas.

[0166] Perform regional smoothing and transition optimization on the global optimization data of the supporting force region to obtain the first supporting force region data.

[0167] Specifically, through smoothing and transition optimization, the support force area is further refined to ensure that the support force presents a natural transition on the surface of the spine. The boundaries of the support force area are smoothed to make the transition between different areas smoother and avoid overly abrupt load differences. Including Gaussian filtering, Gaussian smoothing is performed on the support force data to make the transition between different support areas more natural. Curve fitting is performed to curve fit the boundaries of the support force area for smooth transition. The transition parts between the support force areas are optimized to ensure that the transfer of support force between different areas is smoother and to avoid mutations. The first support force area data is obtained, which includes the optimized support force area and the support force distribution of each area.

[0168] Preferably, the mechanical region division is specifically as follows:

[0169] Perform load transfer path analysis based on spinal load data and spinal mechanical model to obtain load transfer path diagram data;

[0170] Specifically, through the spinal load data, the application point of the load and its mode of action are identified. The load includes body weight, vertical load caused by upper body weight, applied to the cervical or lumbar spine. Dynamic loads, such as impact force during exercise or lateral load caused by posture changes. Using the spinal mechanical model, the path of the load from the application point to the support area is simulated. The load transfer path is mainly calculated by the following equation: ,in is the force vector on the transfer path, is the spinal stiffness matrix, representing the stiffness of the vertebrae and intervertebral discs, is the external force vector at the load application point. Through numerical simulation, the path of load transfer from the application point to different support areas is obtained. The load transfer path is visualized on the three-dimensional spine model to generate load transfer path diagram data. The path diagram includes the force magnitude, direction and transfer efficiency of each vertebra and intervertebral disc in the transfer path.

[0171] Performing a vertebral material property analysis based on the load transfer path diagram data to obtain vertebral material property data;

[0172] Specifically, material properties are assigned to each vertebra and intervertebral disc of the spine based on the spinal mechanical model and anatomical data. Material properties include the elastic modulus and density of the vertebra, which are determined by the mechanical properties of the bone tissue. The elastic modulus and damping coefficient of the intervertebral disc, the elastic properties and damping performance of the intervertebral disc affect the load transfer path. The calculation formula of material properties is established through experimental data or literature. The elastic modulus of the vertebra is obtained by dividing the historical stress based on experience by the historical strain. The damping coefficient of the intervertebral disc is obtained by dividing the historical damping force based on experience by the historical velocity. Detailed material property data are generated for each vertebra and intervertebral disc, including elastic modulus, density, damping coefficient, etc.

[0173] Perform multi-load mode analysis based on the spine material property data to obtain multi-load mode response data;

[0174] Specifically, the force response of the spine is simulated under different load modes. For example, the static load mode simulates the force response of the spine in static states such as standing and sitting. The dynamic load mode simulates the load distribution of the spine during dynamic activities such as walking and running. The non-uniform load mode simulates the impact of asymmetric loads (such as lifting heavy objects on one side) on the spine. The mechanical model is used to solve the stress and strain distribution of the spine under different load modes. For each load mode, the response data of the spine is calculated, including: the stress and strain of each vertebra and intervertebral disc; the change of the load transfer path; the overall deformation of the spine. Generate spinal response data under multiple load modes, including the stress, strain and deformation under each load mode.

[0175] The support force density is calculated according to the multi-load mode response data to obtain the support force density data;

[0176] Specifically, the support force density indicates the load borne per unit area of ​​the spine surface. The calculation formula is the support force per unit area divided by the unit area. The support force density of each support area on the spine surface is calculated through the load transfer path diagram and material property data. The support force density reflects the load-bearing capacity of each area. The support force density data of each support area is output to indicate the distribution state of the support force on the spine surface.

[0177] According to the load transfer path diagram data, spinal material property data and support force density data, the spinal mechanical model is divided into mechanical regions to obtain preliminary mechanical region data;

[0178] Specifically, the load distribution of the spine is comprehensively analyzed based on the load transfer path, material properties, and support force density data. The spine is divided into mechanical regions according to the following standards. High-density regions, areas with higher support force density, correspond to the main load-bearing parts. Low-density regions, areas with lower support force density, correspond to secondary support parts. The divided mechanical region data includes information such as material properties, support force density, and load path for each region.

[0179] The boundary fitting and regional reconstruction are performed based on the preliminary mechanical region data to obtain the second support force region data.

[0180] Specifically, the boundaries of the preliminary mechanical region data are fitted to ensure the continuity and smoothness of the region boundaries. Fitting methods include curve fitting, which uses mathematical functions (such as polynomial fitting) to smooth the boundaries. Boundary optimization, optimizes the boundaries according to the gradient of the support force density to make the boundaries more consistent with the actual load distribution. The mechanical region is reconstructed according to the fitted boundaries, and the shape and range of each region are adjusted to make it more consistent with the actual load distribution of the spine. The generated second support force region data includes the reconstructed mechanical region, the support force density distribution within the region, material properties, load transfer path and other information.

[0181] Preferably, step S3 is specifically:

[0182] Step S31: performing image segmentation according to the spine image data to obtain segmentation area data;

[0183] Specifically, the input spinal image data is preprocessed to remove noise, enhance contrast and improve resolution, including grayscale normalization to standardize the grayscale value of the image to the range of 0-255. De-noising processing uses median filtering or Gaussian filtering to smooth the image and reduce noise interference. The spinal image is segmented by image segmentation methods based on region growing, threshold segmentation or convolutional neural network (CNN) to obtain different regions of the spine. These regions include vertebral regions, intervertebral disc regions, spinal cord regions, and joint regions. According to the grayscale value of the spine in the image, a suitable threshold is selected to separate the vertebral and intervertebral disc regions. Alternatively, a trained deep learning model (such as UNet) is used to perform pixel-level classification on the spinal image to automatically segment the vertebral and intervertebral disc regions. The segmentation results will generate different region labels and region information, and each region corresponds to a set of two-dimensional or three-dimensional coordinate data as the basis for subsequent analysis.

[0184] Step S32: positioning key points according to the segmented area data to obtain key point data;

[0185] Specifically, key points are usually some important structural positions in the spine, such as the center point of the vertebra, the position of the intervertebral disc, the center line of the spinal cord, etc. The key points are extracted by the following methods, such as the center point of the vertebra, using the segmented vertebral area to find the center point of the vertebra by centroid calculation. The intervertebral disc boundary is identified by image segmentation method, and its geometric center is taken as the key point. The key points in the spinal image are located by model or manual calibration method. Deep learning methods are used, such as convolutional neural networks to identify key points and locate them. The coordinates of all key points (such as coordinates in three-dimensional space) and their related attributes (such as vertebral number, intervertebral disc number, etc.) constitute the key point data.

[0186] Step S33: extracting the vertebral morphological features from the vertebral image data according to the key point data to obtain vertebral morphological feature data;

[0187] Specifically, the morphological features of the spine are extracted by analyzing the segmented area data and key point data. The morphological features include the length, width, and height of the vertebrae. By calculating the size of each vertebra, the length, width, and height values ​​are obtained. The shape of the vertebrae is analyzed by the contour line of the vertebrae to extract the shape features of the vertebrae (such as roundness, convexity, etc.). The thickness, shape, and position of the intervertebral disc are extracted. The thickness and boundary shape of the intervertebral disc area are calculated by contour fitting. The curvature of the spine is calculated based on the key points of the spine and the position of the vertebrae. The calculation method is to establish the center line of the spine and calculate the angle between each vertebra and the center line. All morphological features (such as vertebrae size, intervertebral disc morphology, spinal curvature, etc.) ultimately constitute the spinal morphological feature data.

[0188] Step S34: Perform three-dimensional reconstruction according to the spine morphological feature data to obtain a three-dimensional spine model;

[0189] Specifically, a 3D model of the spine is constructed using 3D reconstruction technology based on the spine image data and segmentation results. The technology includes volume rendering, which uses the segmented 2D image data to generate a 3D volume model of the spine. Surface reconstruction uses the segmented surface contour to reconstruct the 3D surface of the vertebrae and intervertebral discs, such as the Marching Cubes algorithm. The reconstructed 3D model is refined to ensure the accuracy of the model. For example, operations such as model smoothing and hole repair are performed to obtain a more accurate 3D spine model. A 3D spine model is generated, which contains structural information of the spine, such as the vertebrae, intervertebral discs, and spinal cord.

[0190] Step S35: performing spinal curvature analysis on the three-dimensional spinal model to obtain spinal curvature data;

[0191] Specifically, based on the three-dimensional spine model, the curvature of the spine is calculated. First, the center line of the spine is determined, and the curvature angle of each vertebra is calculated. Then, the curvature of the spine in the entire model is calculated to obtain the overall curvature of the spine. By comparing the changes in the curvature, it is evaluated whether the spine has abnormal curvature, such as scoliosis or kyphosis. The curvature is usually expressed in angular units (such as degrees). Based on the spine model and curvature calculation, the curvature data is output, indicating the degree of curvature of the spine at various positions.

[0192] Step S36: performing spinal posture assessment according to the three-dimensional spinal model and spinal curvature data to obtain spinal posture data;

[0193] Specifically, the three-dimensional model and curvature data are combined to evaluate the posture of the spine. Posture analysis mainly evaluates whether the posture of the spine is normal by calculating the angle and displacement of each vertebra of the spine. The deviation of the spine posture is evaluated by comparing the posture difference between the standard spine model and the current model. A posture with a large deviation indicates that the spine is abnormal. Based on the results of the spine posture analysis, the posture data is output, which contains information such as the position, angle and deviation of the spine in three-dimensional space.

[0194] Step S37: perform feature vectorization on the spine morphology feature data and the spine posture data to obtain the spine bone feature data.

[0195] Specifically, the vertebral morphological feature data and vertebral posture data are feature vectorized. First, each feature (such as vertebral size, spinal curvature, posture angle, etc.) is converted into a numerical feature. Then, these features are combined into a high-dimensional feature vector. The feature data of all vertebrae are merged to obtain a skeletal feature data containing all vertebral morphological and posture features.

[0196] Preferably, step S4 is specifically:

[0197] Perform geometrical mechanical coupling according to the spine support data and the spine bone characteristic data to obtain the spine geometrical mechanical coupling data;

[0198] Specifically, the geometric data of the spine contains information such as the shape, size, and posture of the spine, while the support data of the spine describes the stress state of the spine under different external loads. In order to perform geometric mechanical coupling, the two are combined. The geometric data includes the geometric shape, size (length, width, height), curvature, and other information of each vertebra of the spine. These data are obtained through image analysis or three-dimensional reconstruction. The support data includes the stress state of the spine under different conditions, such as the size and distribution of the support force, the density of the support area, etc. The support force is closely related to the external load, the posture of the spine, and the movement state. The geometric data is combined with the support data to construct a geometric mechanical coupling model of the spine. The specific method of the coupling process is as follows: the geometric data of each vertebra is spatially docked with the corresponding area in the support data to ensure that the support force corresponding to each vertebra can act on the correct area. This is achieved by spatially aligning the vertebrae and the support force area in three-dimensional space. Through mechanical simulation, the transmission process of the support force between vertebrae is calculated. Specifically, mechanical models such as stiffness matrix and mass matrix are used to model the mechanical response of the spine. Based on the geometric and support data, the mechanical response of the spine is calculated by establishing a coupling model to obtain the force distribution of the spine under different load conditions. For example, under the action of load, the bending, deformation, and stress distribution of the vertebrae will affect the morphology of the spine. The deformation and stress distribution of the spine are calculated by numerical methods (such as finite element analysis). After coupling the support data with the geometric data of the spine, the comprehensive mechanical characteristics of the spine can be obtained, such as the load distribution, deformation degree, and stress distribution between the vertebrae. Through the above steps, the geometric shape, posture, and support force of the spine are combined to obtain the geometric mechanical coupling data of the spine. These data provide a complete description of the spinal load and geometric state for subsequent feature extraction.

[0199] Feature extraction is performed based on the spine geometry and mechanical coupling data to obtain spine feature data for auxiliary spine feature recognition operations.

[0200] Specifically, by analyzing the geometric mechanical coupling data of the spine, some representative features are extracted to describe the health status and morphological characteristics of the spine. The specific extraction process is as follows: The load distribution characteristics of the vertebrae calculate the force strength and distribution of each vertebra under the load. By analyzing the force magnitude, force area and force uniformity of each vertebra, it is evaluated whether the spine has uneven load. The deformation characteristics of the intervertebral disc calculate the compression and bending of the intervertebral disc under the load. The deformation of the intervertebral disc directly affects the overall mechanical properties of the spine, so it is necessary to extract the deformation characteristics of the intervertebral disc and analyze its relationship with the health of the spine. The stability characteristics of the spine calculate the stability of the spine under various load conditions. By analyzing the deformation and bending of the spine under external loads, its stability is evaluated. For example, if the bending angle of the spine is too large, the stability will be reduced and needs to be corrected. The extracted and dimensionally reduced spine feature data is used as the model input for subsequent spine feature recognition. The feature data includes the geometry, load response, stability analysis and other contents of the spine, which can provide support for further evaluation.

[0201] Preferably, the present application also provides a vertebral feature recognition system based on image processing, which is used to execute the vertebral feature recognition method based on image processing as described above, and the vertebral feature recognition system based on image processing comprises:

[0202] The spine data acquisition and preliminary analysis module is used to obtain the user's optical data and spine image data, and perform external morphological analysis based on the user's optical data to obtain external morphological data;

[0203] A spinal support analysis module is used to perform spinal support analysis on external morphological data to obtain spinal support data;

[0204] A vertebral bone feature extraction module is used to extract vertebral bone features according to vertebral image data to obtain vertebral bone feature data;

[0205] The spine multimodal feature fusion module is used to perform spine multimodal feature fusion based on spine support data and spine bone feature data to obtain spine feature data for auxiliary spine feature recognition operations.

[0206] Therefore, from any point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the attached application documents rather than the above description, and it is intended that all changes falling within the meaning and scope of equivalent elements of the application documents are included in the present invention.

[0207] 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 method for recognizing spine features based on image processing, characterized in that: The following steps are involved: Step S1: acquiring user optical data and spinal image data, and performing external morphology analysis based on the user optical data to obtain external morphology data; Step S2: Performing spinal support analysis on the external morphological data to obtain spinal support data; Step S3: Extracting vertebral bone features according to the vertebral image data to obtain vertebral bone feature data; Step S4: performing spinal multimodal feature fusion according to the spinal support data and the spinal bone feature data to obtain spinal feature data for spinal feature recognition auxiliary work; Step S2 is specifically as follows: The spinal mechanical model is constructed according to the external morphological data to obtain the spinal mechanical model; Perform load analysis based on the spinal mechanical model to obtain spinal load data; The support force area is divided according to the spinal load data and the spinal mechanical model to obtain the support force area data; Generate a force distribution diagram according to the support force area data to obtain the spine force distribution diagram data; The force of the fitted spine template data is optimized according to the spine force distribution diagram data to obtain the spine support data.

2. The method according to claim 1, characterized in that Step S1 is specifically as follows: Acquire user optical data and spinal imaging data; Performing body shape analysis based on user optical data to obtain body shape feature data; Generate a standard spine template according to the body shape feature data and a preset spine template library to obtain spine template data; Iteratively optimize the spine template data according to the body shape feature data to obtain the fitted spine template data; Extract joint parameters from the fitted spine template data to obtain joint parameter data; Perform posture adaptation according to the user's optical data and the spine template data to obtain posture adaptation data; The user's optical data, joint parameter data and posture adaptation data are integrated to obtain external morphological data.

3. The method according to claim 1, characterized in that The construction of the spine mechanical model is as follows: Reconstructing the spine geometry model according to the external morphological data to obtain the spine geometry model; Assigning mechanical properties of the spine to the spine geometric model to obtain a spine physical property model; A virtual load environment is established according to the physical property model of the spine to obtain a spinal mechanical model.

4. The method according to claim 1, characterized in that: The load analysis is as follows: According to the spinal mechanical model, the three-dimensional mechanical model is converted to obtain a preliminary mechanical model; Perform dynamic simulation of spine load on the preliminary mechanical model to obtain dynamic load data; Finite element meshing is performed based on the dynamic load data to obtain the spinal load data.

5. The method according to claim 1, characterized in that The support area is divided into: Perform geometric area division according to the spinal load data and the spinal mechanical model to obtain first support force area data; Divide the mechanical area according to the spinal load data and the spinal mechanical model to obtain the second support force area data; Perform load distribution balancing according to the first supporting force area data and the second supporting force area data to obtain supporting force area data; The specific geometric area division is as follows: Surface segmentation is performed based on the spinal load data and the spinal mechanical model to obtain preliminary support force area data; Perform load analysis and support force distribution calculation based on preliminary support force area data to obtain support force distribution data; Extract local features of the supporting force according to the supporting force distribution data to obtain local supporting force feature data; Performing clustering calculation according to the local support force characteristic data to obtain local support force characteristic clustering data; A graph is constructed according to the local support force feature clustering data to obtain local support force feature clustering graph data; Perform local optimization on the preliminary support force area data according to the local support force feature clustering graph data to obtain local optimization data of the support force area; Perform global optimization based on the local optimization data of the support force area to obtain the global optimization data of the support force area; Perform regional smoothing and transition optimization on the global optimization data of the supporting force region to obtain the first supporting force region data.

6. The method according to claim 5, characterized in that The specific division of mechanical areas is as follows: Perform load transfer path analysis based on spinal load data and spinal mechanical model to obtain load transfer path diagram data; Performing a vertebral material property analysis based on the load transfer path diagram data to obtain vertebral material property data; Perform multi-load mode analysis based on the spine material property data to obtain multi-load mode response data; The support force density is calculated according to the multi-load mode response data to obtain the support force density data; According to the load transfer path diagram data, spinal material property data and support force density data, the spinal mechanical model is divided into mechanical regions to obtain preliminary mechanical region data; The boundary fitting and regional reconstruction are performed based on the preliminary mechanical region data to obtain the second support force region data.

7. The method according to claim 1, characterized in that Step S3 is specifically as follows: Perform image segmentation according to the spine image data to obtain segmentation area data; Position the key points according to the segmented area data to obtain the key point data; Extracting the vertebral morphological features from the vertebral image data according to the key point data to obtain the vertebral morphological feature data; Perform three-dimensional reconstruction based on the morphological characteristic data of the spine to obtain a three-dimensional spine model; Performing spinal curvature analysis on the three-dimensional spinal model to obtain spinal curvature data; Evaluate the spinal posture according to the three-dimensional spinal model and spinal curvature data to obtain spinal posture data; The vertebral morphological feature data and vertebral posture data are feature vectorized to obtain vertebral bone feature data.

8. The method according to claim 1, characterized in that Step S4 is specifically as follows: Perform geometrical mechanical coupling according to the spine support data and the spine bone characteristic data to obtain the spine geometrical mechanical coupling data; Feature extraction is performed based on the spine geometry and mechanical coupling data to obtain spine feature data for auxiliary spine feature recognition operations.

9. A spine feature recognition system based on image processing, characterized in that: For executing the spine feature recognition method based on image processing as claimed in claim 1, the spine feature recognition system based on image processing comprises: The spine data acquisition and preliminary analysis module is used to obtain the user's optical data and spine image data, and perform external morphological analysis based on the user's optical data to obtain external morphological data; The spinal support analysis module is used to construct a spinal mechanical model according to external morphological data to obtain a spinal mechanical model; perform load analysis according to the spinal mechanical model to obtain spinal load data; perform support force area division according to the spinal load data and the spinal mechanical model to obtain support force area data; generate a force distribution diagram according to the support force area data to obtain spinal force distribution diagram data; perform force optimization on the fitted spinal template data according to the spinal force distribution diagram data to obtain spinal support data; A vertebral bone feature extraction module is used to extract vertebral bone features according to vertebral image data to obtain vertebral bone feature data; The spine multimodal feature fusion module is used to perform spine multimodal feature fusion based on spine support data and spine bone feature data to obtain spine feature data for auxiliary spine feature recognition operations.

Images