Multi-sign human body model spine positioning method combined with deep learning

Abstract

The invention belongs to the technical field of human body model construction, and particularly relates to a multi-sign human body model spine positioning method combined with deep learning, which comprises the following steps: firstly, based on preset human body characteristic parameters, collecting human body spine image data according to a preset standard posture, and carrying out medical image processing on the collected human body spine image data to generate a data set; performing data preprocessing on the data set, then constructing a prediction model based on a full-connection neural network model, training the prediction model according to the preprocessed data set, reserving model parameters with optimal performance as model parameters of the prediction model, and generating a final prediction model; and finally, deploying the final prediction model, generating a spine positioning parameter result by the deployed prediction model according to the input human body sign parameters, and guiding the development of the multi-sign human body model according to the spine positioning parameter result. According to the invention, the problem of low accuracy of human body model spine positioning in the existing grid transformation technology can be solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of human body model construction, and in particular relates to a multi-sign human body model spine positioning method combined with deep learning. Background Art

[0002] As the global vehicle safety assessment system undergoes a strategic transformation from physical testing to a fusion of virtual and real-world testing, digital simulation technology based on the finite element model of the human body has become a core support method. Using mesh transformation technology to geometrically scale the baseline human body model to people of different body types is a key technical approach for conducting injury mechanism research on diverse populations. As the core hub of mechanical conduction in the human body, the spatial positioning accuracy of the spine directly affects the accuracy of injury prediction for multiple areas, including the head, neck, chest, and abdomen. However, existing mesh transformation methods face significant technical bottlenecks in spinal positioning, which urgently need to be overcome through technological innovation.

[0003] Existing mesh transformation technologies mostly match the surfaces of a reference model and a target model, with the internal mesh calculated. This results in only accurate surface geometry, with no direct data supporting internal structures such as bone size and position. The spine, in particular, has a more complex positional relationship than the skeletal structures of the upper and lower limbs. Its position and angle also affect injuries to multiple areas of the body, including the head, neck, chest, and abdomen. Therefore, solving its positioning problem is crucial. Summary of the Invention

[0004] The technical problem solved by the present invention is to provide a multi-sign human body model spine positioning method combined with deep learning to solve the problem of low accuracy in human body model spine positioning in existing grid transformation technology.

[0005] The basic solution provided by the present invention is a multi-sign human body model spine positioning method combined with deep learning, including:

[0006] S1: Based on preset human feature parameters, human spine image data is collected in a preset standard posture, and medical image processing is performed on the collected human spine image data to generate a data set, and data preprocessing is performed on the data set to obtain a preprocessed data set;

[0007] S2: Build a prediction model based on a fully connected neural network model, train the prediction model based on the preprocessed data set, retain the model parameters with the best performance as the model parameters of the prediction model, and generate the final prediction model;

[0008] S3: The final prediction model is deployed. The deployed prediction model generates spinal positioning parameter results based on the input human sign parameters, and guides the development of a multi-sign human body model based on the spinal positioning parameter results.

[0009] Furthermore, the S1 includes:

[0010] S1-1: Preset sample information to obtain body shape parameters representing different physical features of the human body;

[0011] S1-2: Define the occupant posture based on different seat inclination angles and collect human spine images of occupant posture under different body parameters;

[0012] S1-3: According to the preset spine positioning method, the collected human spine images are aligned, segmented, and measured to obtain positioning parameters associated with the body shape parameters and generate a data set;

[0013] S1-4: Preprocess the dataset and divide the preprocessed dataset into a training set and a validation set according to a preset ratio.

[0014] Furthermore, the occupant posture is defined based on different seat inclination angles in S1-2 as follows:

[0015] Based on the backrest angle between the seat back and the vertical direction, the passenger posture at different backrest angles and the passenger posture in the standing state are preset.

[0016] Furthermore, the spine positioning method preset in S1-3 is specifically:

[0017] The spine is divided into cervical, thoracic, lumbar and sacrum, where the cervical spine is divided into C1-C7, the thoracic spine is divided into T1-T12, and the lumbar spine is divided into L1-L5;

[0018] According to the definition of the spine and based on the sagittal view of the human body, the positioning parameters of the spine position are generated.

[0019] Furthermore, the positioning parameters of the spine position include:

[0020] P C , describes the starting point of the spine, which is the midpoint of the contact plane between the occipital condyle and the cervical vertebra C1;

[0021] θ C , cervical lordosis angle, which describes the angle between the cervical C1 plane and the cervical C7 plane;

[0022] θ T , thoracic kyphosis angle, which describes the angle between the thoracic T1 plane and the thoracic T12 plane;

[0023] θ L , lumbar lordosis angle, which describes the angle between the lumbar L1 plane and the lumbar L5 plane;

[0024] θ S, sacral kyphosis angle, which describes the angle between the top plane of the sacrum and the horizontal plane;

[0025] P S , describes the sacral end point, that is, the endmost point of the sacrum.

[0026] Further, the S2 includes:

[0027] S2-1: Learning the relationship between body shape parameters and spinal positioning parameters based on a fully connected neural network model and generating a prediction model;

[0028] S2-2: Input the training set into the prediction model and adjust the model parameters according to the output of the prediction model;

[0029] S2-3: Based on the trained prediction model, use the validation set to evaluate the performance of the prediction model and adjust the model parameters according to the evaluation results until the performance requirements are met.

[0030] Furthermore, in S2-3, based on the trained prediction model, the performance of the prediction model is evaluated using the validation set as follows:

[0031] The performance of the prediction model is evaluated using the mean square error index, root mean square error index, and determination coefficient index. The mean square error index expression is:

[0032]

[0033] The root mean square error indicator expression is:

[0034]

[0035] Among them, n represents the number of samples, y i represents the i-th actual value, represents the i-th predicted value;

[0036] The expression of the coefficient of determination index is:

[0037]

[0038] Among them, SS res Represents the sum of squared errors between the actual value and the predicted value, SS tot Represents the sum of squared errors between the actual value and its mean.

[0039] Furthermore, the S3 includes:

[0040] S3-1: Deploy the final prediction model to the multi-sign human model development system;

[0041] S3-2: Determine the posture and body shape parameters that need to be adjusted, and obtain the target human finite element model before calibration through pre-simulation;

[0042] S3-3: Input the posture and body shape parameters into the final prediction model, fix the P of the target human finite element model C , according to the prediction model obtained P S Adjust the sacral position and convert the θ generated by the prediction model C ,θ T ,θ L ,θ S The parameters are input into the target human finite element model;

[0043] S3-4: Performing a mesh quality check and a penetration check on the target human body finite element model obtained in S3-3 to obtain the final adjusted target human body finite element model.

[0044] An electronic device includes a processor and a memory, wherein the memory stores programs or instructions, and the processor executes any of the above-mentioned multi-sign human body model spine positioning methods combined with deep learning by calling the programs or instructions stored in the memory.

[0045] A computer-readable storage medium stores a program or instruction, wherein the program or instruction enables a computer to execute a multi-sign human body model spine positioning method combined with deep learning as described in any one of the above items.

[0046] The principle and advantage of the present invention are as follows: In the technical solution of the present application, firstly, sample information is preset to represent the body shape data of different characteristics of the human body, and human spine medical imaging data is collected according to the body shape data, and then the positioning information of the spine is defined, including P C ,θ C ,θ T ,θ L ,θ S 、P S The positioning parameters are then associated with the body shape parameters and input into the constructed prediction model for training and performance verification. This allows the prediction model to predict the positioning parameters of the human spine after inputting the body shape parameters, providing great help for the subsequent construction of the human spine model.

[0047] Therefore, the advantages of this application are:

[0048] 1. The multi-sign human body model spine positioning method in this application can realize the positioning of the human spine under different body shapes, so that the generated human spine finite element model not only has accurate surface geometric position but also more accurate spine position information, thereby improving the biofidelity of the model;

[0049] 2. Compared with the existing technology of searching for volunteers with target body shapes and re-collecting spinal angles, the technical solution of this application has higher development efficiency and reduces repetitive costs. After the final prediction model is deployed to the development system, the sacrum position and various spinal angle parameters of the target human finite element model can be accurately adjusted according to the input posture and body shape parameters. The accuracy and reliability of the model are ensured through mesh quality inspection and penetration inspection, which effectively improves the development quality and efficiency of multi-sign human body models, provides model support that is more in line with actual human body characteristics for related fields such as ergonomics and medical research, and promotes the refined development of related applications based on human body models in design and analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A flowchart of an embodiment of the present invention;

[0051] Figure 2 A schematic diagram of spine positioning according to an embodiment of the present invention;

[0052] Figure 3 A schematic diagram of seat angle parameters according to an embodiment of the present invention;

[0053] Figure 4 Schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The following is further described in detail through specific implementation methods:

[0055] The symbols in the drawings of the specification include: electronic device 400 , processor 401 , memory 402 , input device 403 , and output device 404 .

[0056] The embodiment is basically as shown in the attached Figure 1 Shown: A multi-sign human body model spine positioning method combined with deep learning, including:

[0057] S1: Based on preset human characteristic parameters, human spine image data is collected in a preset standard posture, and the collected human spine image data is processed by medical imaging to generate a data set, and the data set is preprocessed to obtain a preprocessed data set; wherein S1 includes:

[0058] S1-1: Preset sample information to obtain body shape parameters representing different physical features of the human body;

[0059] S1-2: Define the occupant posture based on different seat inclination angles and collect human spine images of occupant posture under different body parameters;

[0060] S1-3: According to the preset spine positioning method, the collected human spine images are aligned, segmented, and measured to obtain positioning parameters associated with the body shape parameters and generate a data set;

[0061] S1-4: Preprocess the dataset and divide the preprocessed dataset into a training set and a validation set according to a preset ratio.

[0062] In this embodiment, the preset sample information includes height, weight, gender and other information. By fusing multi-dimensional sample information, different body parameters of the human body can be obtained. In other embodiments of this embodiment, the sample information may also include body fat percentage, spinal curvature, etc.

[0063] Regarding the definition of passenger posture, in the field of automobile safety, the angle of the spine is mainly related to the passenger's seat angle, so different passenger postures are defined by different seat inclinations; to avoid the influence of different shapes, the posture collection will use a simplified seat, such as Figure 3 As shown, the inclination is described by two angles α and β, which are specifically defined as:

[0064] α: backrest angle, the angle between the seat back and the vertical direction;

[0065] β: Cushion angle, the angle between the seat cushion and the horizontal direction;

[0066] In real-world scenarios, the main factor affecting the spine is the backrest angle α. Therefore, we focused on this angle and set α = 25°, 45°, and 65° to represent different occupant postures. We also defined a standing level to represent pedestrian posture. This ultimately determined the sample information required for data collection and the input parameters for subsequent model training.

[0067] Then, based on the simulated human body model application scenario, such as a collision scenario, and according to the role of the pedestrian and the role of the occupant, human spine images of the occupant under different body parameters, such as CT data or X-ray scan data, are collected.

[0068] Then, a spine positioning method is preset, and the collected human spine images are aligned, segmented, and measured. In this application, the preset spine positioning method is specifically as follows: the spine is divided into cervical vertebrae, thoracic vertebrae, lumbar vertebrae, and sacrum, wherein the cervical vertebrae are defined as C1-C7 segments, the thoracic vertebrae are divided into T1-T12 segments, and the lumbar vertebrae are divided into L1-L5 segments; this is because the cervical vertebrae are located at the top of the spine and are composed of 7 vertebrae, arranged from C1 to C7 from top to bottom; the thoracic vertebrae are located below the cervical vertebrae, in the chest area, and are composed of 12 vertebrae, numbered T1-T12; the lumbar vertebrae are located below the thoracic vertebrae and are composed of 5 vertebrae, numbered K1-L5.

[0069] According to the definition of the spine, the positioning parameters of the spine position are generated based on the sagittal view of the human body. Specifically, Figure 2 As shown:

[0070] Use P C Describe the starting point of the spine, which is the midpoint of the contact plane between the occipital condyle and the cervical vertebra C1;

[0071] θ C , cervical lordosis angle, which describes the angle between the cervical C1 plane and the cervical C7 plane;

[0072] θ T , thoracic kyphosis angle, which describes the angle between the thoracic T1 plane and the thoracic T12 plane;

[0073] θ L , lumbar lordosis angle, which describes the angle between the lumbar L1 plane and the lumbar L5 plane;

[0074] θ S , sacral kyphosis angle, which describes the angle between the top plane of the sacrum and the horizontal plane;

[0075] P S , describes the sacral end point, that is, the endmost point of the sacrum.

[0076] The parameters of this application define the position and relative angle of the spine in the human body, and leave a certain degree of freedom to adapt to the changes in intervertebral disc size caused by differences in physical signs. According to the above content, after the body shape parameters are generated, the spinal positioning parameters corresponding to the body shape parameters are also obtained. The spinal images of volunteers with different physical signs are obtained according to the spinal positioning parameters. For various sitting angles, corresponding fixed-angle chairs are used to facilitate unified acquisition. After the data is acquired, the spinal images are processed using existing medical image processing software. Specifically, the image is first preprocessed and the image noise is removed using a filter; the image is uniformly aligned. Although medical images usually have a relatively unified coordinate system and size, this step is still necessary due to the large number of data sources. Feature points in the image are identified through edge detection and key point detection. Based on the proportional benchmark of the image itself, the image is used. The image is then segmented using a global or local threshold method to separate the image into foreground and background, or a deep learning method such as a U-Net or DeepLab model is used to obtain the effective area of the spine. The activation function commonly used is the ReLU function. Finally, the image is measured using the parameters defined by the positioning parameters of the spine. The basic logic is to capture the corresponding key points based on the segmented image according to the parameter definition, record the coordinates and measure the angles. The key point capture can be achieved through a key point detection algorithm. Finally, the spine positioning parameters associated with the posture and physical sign parameters are obtained to form a data set.

[0077] The data set is preprocessed based on the obtained data set. In this application, the preprocessing is standardization. For example, for sample data that is different from the standard posture, the positioning parameters are standardized so that the positioning coordinate system has the same origin, thereby converting it into sample data under the standard posture. After the preprocessing is completed, the data set is divided into 80% training set and 20% validation set.

[0078] S2: Build a prediction model based on a fully connected neural network model, train the prediction model based on the preprocessed data set, retain the model parameters with the best performance as the model parameters of the prediction model, and generate the final prediction model; S2 includes:

[0079] S2-1: Learning the relationship between body shape parameters and spinal positioning parameters based on a fully connected neural network model and generating a prediction model;

[0080] S2-2: Input the training set into the prediction model and adjust the model parameters according to the output of the prediction model;

[0081] S2-3: Based on the trained prediction model, the performance of the prediction model is evaluated using the validation set, and the model parameters are adjusted according to the evaluation results until the performance requirements are met. Furthermore, in S2-2, the training set is input into the prediction model, and the model parameters are adjusted according to the output of the prediction model as follows:

[0082] Applying a convolutional neural network to the training set, extracting local features through convolutional layers and pooling layers, wherein the activation function of the convolutional layer adopts the ReLU function;

[0083] The extracted local features are input into the fully connected layer for integration and prediction, and the model parameters of the convolutional neural network are adjusted according to the prediction results.

[0084] In this embodiment, based on the body shape parameters and spinal positioning parameters in the preprocessed data set, a fully connected neural network model that excels at data processing in deep learning technology is used to learn the relationship between the body shape parameters and the spinal positioning parameters to obtain a prediction model. The training set is then input into the prediction model, and the model parameters, such as the learning rate and the number of hidden layers, are adjusted according to the output results of the prediction model until the model shows good performance. After the prediction model training is completed, its performance is evaluated using a validation set. The evaluation indicators in this application include the mean square error indicator, the root mean square error indicator, and the determination coefficient indicator. The mean square error indicator expression is:

[0085]

[0086] The root mean square error indicator expression is:

[0087]

[0088] Among them, n represents the number of samples, y i represents the i-th actual value, represents the i-th predicted value;

[0089] The expression of the coefficient of determination index is:

[0090]

[0091] Among them, SS res Represents the sum of squared errors between the actual value and the predicted value, SS tot Represents the sum of squared errors between the actual value and its mean.

[0092] Finally, based on the evaluation results, the model parameters of the prediction model are adjusted again until the accuracy requirements are met.

[0093] S3: Deploy the final prediction model. The deployed prediction model generates spinal positioning parameter results based on the input human sign parameters, and guides the development of a multi-sign human body model based on the spinal positioning parameter results. S3 includes:

[0094] S3-1: Deploy the final prediction model to the multi-sign human model development system;

[0095] S3-2: Determine the posture and body shape parameters that need to be adjusted, and obtain the target human finite element model before calibration through pre-simulation;

[0096] S3-3: Input the posture and body shape parameters into the final prediction model, fix the P of the target human finite element model C , according to the prediction model obtained P S Adjust the sacral position and convert the θ generated by the prediction model C ,θ T ,θ L ,θ S The parameters are input into the target human finite element model;

[0097] S3-4: Performing a mesh quality check and a penetration check on the target human body finite element model obtained in S3-3 to obtain the final adjusted target human body finite element model.

[0098] In this embodiment, according to the training optimization of the model in step S2, the final prediction model obtained by deployment can realize the input of physical sign parameters to obtain a corresponding set of spine positioning parameters, thereby guiding the development of a multi-sign human finite element model. In actual application, the posture and body shape that need to be adjusted are first determined, and the target human finite element model before verification is obtained through pre-simulation, grid transformation and other methods. Then, the posture and body shape parameters are input into the final prediction model deployed. According to the obtained spine positioning parameters, the P in the human body model is first reduced. C Fixed, as the global positioning reference of the spine, according to the obtained P S Adjust the sacrum position; then adjust the C ,θ T ,θ L ,θ S The parameters are input into the target human finite element model. The relative inclination angles between the vertebrae can be manually adjusted through the finite element pre-processing software to simulate the individualized physiological curvature of the spine. The automatic optimization platform can also be used to record command files to achieve the adjustment of corresponding planes (cervical C1 plane and cervical C7 plane, thoracic T1 plane and thoracic T12 plane, thoracic L1 plane and lumbar L5 plane, and the angle between the top plane of the sacrum and the horizontal plane). The relative angles between the vertebrae in each plane can also be adjusted through interpolation (such as spline function).

[0099] Finally, after the spine is calibrated, it may cause distortion of the surrounding mesh and contact penetration. Through mesh quality and penetration checks, such problems are found, the fill mesh is regenerated, and the contact surface is fine-tuned to complete the final adjustment of the human body model.

[0100] Therefore, the advantages of this application are:

[0101] 1. Realize the positioning of the human spine in different postures and body shapes, so that the human finite element model obtained after the secondary transformation not only has accurate surface geometric position, but also has more precise spinal position information, thereby improving the bio-fidelity of the model.

[0102] 2. Combined with deep learning models, the processing performance of large sample data will be significantly better than the traditional prediction model establishment method.

[0103] 3. Promoting the application and development of human body model tools can provide more accurate data support for virtual evaluation, reduce dependence on physical experiments, and reduce experimental costs and time.

[0104] like Figure 4 As shown, in another embodiment of this embodiment, an electronic device is also included, and the electronic device 400 includes one or more processors 401 and a memory 402.

[0105] The processor 401 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 400 to perform desired functions.

[0106] The memory 402 may include one or more computer program products, and the computer program product may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 401 may run the program instructions to implement a multi-sign human body model spine positioning method combined with deep learning and / or other desired functions of any embodiment of the present invention described above. Various contents such as initial external parameters, thresholds, etc. may also be stored in the computer-readable storage medium.

[0107] In one example, electronic device 400 may further include an input device 403 and an output device 404, which are interconnected via a bus system and / or other connection mechanisms (not shown). Input device 403 may include, for example, a keyboard, a mouse, etc. Output device 404 may output various information to the outside, including warning information, braking force, etc. Output device 404 may include, for example, a display, a speaker, a printer, a communication network, and remote output devices connected thereto.

[0108] Of course, to simplify, Figure 4 Only some of the components related to the present invention in the electronic device 400 are shown, and components such as buses, input / output interfaces, etc. are omitted. In addition, the electronic device 400 may further include any other appropriate components according to specific application scenarios.

[0109] In addition to the above-mentioned methods and devices, an embodiment of the present invention may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of a multi-sign human body model spine positioning method combined with deep learning provided by any embodiment of the present invention.

[0110] The computer program product may be written in any combination of one or more programming languages to implement the operations of embodiments of the present invention, including object-oriented programming languages such as Java, C++, and conventional procedural programming languages such as C or similar programming languages. The program code may be executed entirely on the user's computing device, partially on the user's computing device, as a stand-alone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0111] In addition, an embodiment of the present invention may also be a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, enables the processor to execute the steps of a multi-sign human body model spine positioning method combined with deep learning provided in any embodiment of the present invention.

[0112] The computer-readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can, for example, include but is not limited to a system, device or component of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0113] The above are only embodiments of the present invention. Common knowledge such as the known specific structures and characteristics in the scheme are not described in detail here. Ordinary technicians in the field are aware of all common technical knowledge in the technical field of the invention before the application date or priority date, can obtain all existing technologies in the field, and have the ability to apply conventional experimental means before that date. Ordinary technicians in the field can improve and implement this scheme in combination with their own abilities under the inspiration given by this application. Some typical known structures or known methods should not become obstacles for ordinary technicians in the field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several variations and improvements can be made, which should also be regarded as the scope of protection of the present invention. These will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

Claims

1. A multi-sign human body model spine positioning method combined with deep learning, characterized by: include: S1: Based on preset human feature parameters, human spine image data is collected in a preset standard posture, and medical image processing is performed on the collected human spine image data to generate a data set, and data preprocessing is performed on the data set to obtain a preprocessed data set; S2: Build a prediction model based on a fully connected neural network model, train the prediction model based on the preprocessed data set, retain the model parameters with the best performance as the model parameters of the prediction model, and generate the final prediction model; S3: The final prediction model is deployed. The deployed prediction model generates spinal positioning parameter results based on the input human sign parameters, and guides the development of a multi-sign human body model based on the spinal positioning parameter results.

2. The multi-sign human body model spine positioning method combined with deep learning according to claim 1, characterized in that: Said S1 comprises: S1-1: Preset sample information to obtain body shape parameters representing different physical features of the human body; S1-2: Define the occupant posture based on different seat inclination angles and collect human spine images of occupant posture under different body parameters; S1-3: According to the preset spine positioning method, the collected human spine images are aligned, segmented, and measured to obtain positioning parameters associated with the body shape parameters and generate a data set; S1-4: Preprocess the dataset and divide the preprocessed dataset into a training set and a validation set according to a preset ratio.

3. The multi-sign human body model spine positioning method combined with deep learning according to claim 2, characterized in that: The definition of the occupant posture based on different seat inclination angles in S1-2 is specifically as follows: Based on the backrest angle between the seat back and the vertical direction, the passenger posture at different backrest angles and the passenger posture in the standing state are preset.

4. The multi-sign human body model spine positioning method combined with deep learning according to claim 3 is characterized by: The specific spine positioning method preset in S1-3 is: The spine is divided into cervical, thoracic, lumbar and sacrum, where the cervical spine is divided into C1-C7, the thoracic spine is divided into T1-T12, and the lumbar spine is divided into L1-L5; According to the definition of the spine and based on the sagittal view of the human body, the positioning parameters of the spine position are generated.

5. The multi-sign human body model spine positioning method combined with deep learning according to claim 4 is characterized by: The positioning parameters of the spine position include: P C , describes the starting point of the spine, which is the midpoint of the contact plane between the occipital condyle and the cervical vertebra C1; θ C , cervical lordosis angle, which describes the angle between the cervical C1 plane and the cervical C7 plane; θ T , thoracic kyphosis angle, which describes the angle between the thoracic T1 plane and the thoracic T12 plane; θ L , lumbar lordosis angle, which describes the angle between the lumbar L1 plane and the lumbar L5 plane; θ S , sacral kyphosis angle, which describes the angle between the top plane of the sacrum and the horizontal plane; P S , describes the sacral end point, that is, the endmost point of the sacrum.

6. The multi-sign human body model spine positioning method combined with deep learning according to claim 5, characterized in that: The S2 includes: S2-1: Learning the relationship between body shape parameters and spinal positioning parameters based on a fully connected neural network model and generating a prediction model; S2-2: Input the training set into the prediction model and adjust the model parameters according to the output of the prediction model; S2-3: Based on the trained prediction model, use the validation set to evaluate the performance of the prediction model and adjust the model parameters according to the evaluation results until the performance requirements are met.

7. The multi-sign human body model spine positioning method combined with deep learning according to claim 6, characterized in that: In S2-3, based on the trained prediction model, the performance of the prediction model is evaluated using the validation set as follows: The performance of the prediction model is evaluated using the mean square error index, root mean square error index, and determination coefficient index. The mean square error index expression is: The root mean square error indicator expression is: Among them, n represents the number of samples, y i represents the i-th actual value, represents the i-th predicted value; The expression of the coefficient of determination index is: Among them, SS res Represents the sum of squared errors between the actual value and the predicted value, SS tot Represents the sum of squared errors between the actual value and its mean.

8. The multi-sign human body model spine positioning method combined with deep learning according to claim 7, characterized in that: The S3 includes: S3-1: Deploy the final prediction model to the multi-sign human model development system; S3-2: Determine the posture and body shape parameters that need to be adjusted, and obtain the target human finite element model before calibration through pre-simulation; S3-3: Input the posture and body shape parameters into the final prediction model, fix the P of the target human finite element model C , according to the prediction model obtained P S Adjust the sacral position and convert the θ generated by the prediction model C ,θ T ,θ L ,θ S The parameters are input into the target human finite element model; S3-4: Performing a mesh quality check and a penetration check on the target human body finite element model obtained in S3-3 to obtain the final adjusted target human body finite element model.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores programs or instructions, and the processor executes the multi-sign human body model spine positioning method combined with deep learning as described in any one of claims 1 to 8 by calling the programs or instructions stored in the memory.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program or instruction, which enables a computer to execute a multi-sign human body model spine positioning method combined with deep learning as described in any one of claims 1 to 8.

Images