High-fidelity human body model construction method and system

Through transfer learning and bioelectric impedance technology combined with MIMICS, ABAQUS and Python, the side effects of CT technology and the problem of low fineness of mannequin models are solved, and the construction of high-fidelity mannequin models is realized, and the accuracy and safety of finite element analysis are improved.

CN118919029APending Publication Date: 2024-11-08BEIJING INST OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410938782.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-15
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The existing CT technology has side effects in medical imaging and the human body model is low in fineness, making it difficult to meet the construction needs of high-fidelity human body model.

Method used

Transfer learning technology is used to combine MIMICS, ABAQUS and Python, and the signature data is measured based on bioelectrical impedance technology, and the high-fidelity human body model is reconstructed to reduce dependence on X-rays, and the fine construction of three-dimensional models is realized.

Benefits of technology

The calculation accuracy of the finite element model in the human body is improved, the risk of patients being exposed to X-rays is reduced, and the problem of small traditional training sets is avoided. It provides a high-fidelity human body model, suitable for the research of impact response and damage.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118919029A_ABST
    Figure CN118919029A_ABST
Patent Text Reader

Abstract

The invention discloses a high-fidelity human body model construction method and system. Comprising the steps that a human body model is read, a txt text is output through parallel operation, a ResNet network model is input, and feature extraction is carried out; measuring physical sign data of the to-be-measured person based on a bioelectrical impedance analysis technology; outputting to a ResNet network model for fuzzy and comparison with the features; accessing the trained transfer learning network model to MIMICS, and performing large-scale data conversion processing through a Python script; the csv file is converted into a binary STL file, the binary STL file is subjected to primary processing and then is output in an STP format for preprocessing, and a partial three-dimensional reconstruction model of the person to be tested is obtained; and outputting the processed STP file as an inp file, and importing the inp file into the ABAQUS for subsequent finite element modeling and calculation. The method can well solve the problems of side effects of the CT technology and low fineness of the personnel model, and is easy to implement, high in operation speed and relatively high in accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the fields of medical image computing technology and personnel impact response and injury technology, and in particular to a high-fidelity human body model construction method and system based on transfer learning and MIMICS / ABAQUS / Python data joint pre-processing. Background Art

[0002] With the development of CAE technology, model research has gradually become an important development direction of biomechanics in the future. High-fidelity finite element models of the human body are an important method for researchers to study impact response and injury. Among them, CT (computed tomography) technology provides support for three-dimensional reconstruction. In the diagnosis of major diseases, the application of imaging technology has improved the diagnosis rate and treatment efficiency of the disease. However, the dose of X-rays used will have an impact on patients. The higher the irradiation dose, the better the image quality, and the more obvious the side effects on patients; while reducing the irradiation dose will affect the quality of the film and may cause misdiagnosis and missed diagnosis, and the hazards of contrast agents have always existed. Therefore, there is an urgent need for a high-fidelity human model construction method that does not rely entirely on CT technology.

[0003] MIMICS software can quickly import CT, MRI and other tomographic sequence image data, perform tissue segmentation based on threshold and morphological algorithms, three-dimensional reconstruction of surface drawing and large-scale data conversion processing on the images, and output STL files for import into ABAQUS for pre-processing and then submit for calculation. Both software have complete Python interfaces, which provide the possibility for secondary development of joint pre-processing of the three data.

[0004] Transfer learning technology is introduced to transfer the trained model parameters to the new model, so as to accelerate the model training. Model parameters trained by other related tasks can achieve good performance even with a small data set. Based on feature transfer learning, the existing conventional high-fidelity 3D human body model is used for feature extraction, and the data features are mapped to the target domain. By extracting certain individual vital signs data, a mapping relationship is established with similar features in the source domain, and then all the individual feature data is reconstructed. The data is then imported into MIMICS and ABAQUS software for pre-processing operations such as solid model construction and meshing. Summary of the invention

[0005] The present invention aims at the defects of the side effects of the prior art CT technology and the low precision of the human model, and provides a method and system for constructing a high-fidelity human body model.

[0006] In order to achieve the above invention object, the technical solution adopted by the present invention is as follows:

[0007] A method for constructing a high-fidelity human body model comprises the following steps:

[0008] The first step is to extract the features of the complete human body reference model, use the 3D shape descriptor to establish the mapping relationship between the corresponding points on the surface, use a vector to describe the shape features of the specified points and areas on the surface of the human body model, and match the value of the vector to describe the correspondence between different surface points; use Python script to read the human body model and output txt text in parallel, input the neural network model, and perform feature extraction;

[0009] The second step is to measure the physical data of the person to be tested. Based on the bioelectrical impedance technology, the person's height, weight, body fat, bone weight, water content and protein content are measured, and the input neural network model is compared with the features extracted in the first step for learning and training to obtain the chest and lung modeling parameters of the person to be tested; including: lung length and width data, volume parameters, diameter parameters, and surface area parameters;

[0010] In the third step, the model trained in the second step is connected to MIMICS through Python for data conversion, and the key node data of the target part is reconstructed. Large-scale data conversion processing is performed through Python scripts, and the csv file is converted into a binary STL file. After preliminary processing, the STP format is output and transferred to the meshing software for denoising, filtering, smoothing and polishing. The partial three-dimensional reconstructed model of the person to be tested is obtained, and the binary STL data is output;

[0011] In the fourth step, Python and ABAQUS are used to perform joint rapid modeling based on the modeling parameters. The STP data obtained in step 3 after preliminary processing is output as an inp file and imported into ABAQUS for subsequent finite element modeling and calculation, and finally a high-fidelity three-dimensional model of the human body that can be used for finite element analysis and calculation is obtained.

[0012] Furthermore, in the first step, Python script is used to extract the features of the complete human reference model, as follows:

[0013] 11. First, introduce the required libraries in Python, including matplotlib.pyplot library, mpl_toolkits.mpplot3d library and numpy library;

[0014] 12. Read the txt data, convert it into a two-dimensional array, set the number of nodes to 200, and return a tuple;

[0015] 13. Define the information entropy function and initialize the special two-dimensional array;

[0016] 14. Set the image size, (x, y, z) = (0, 1, 2); set the coordinate axis invisible and the background network invisible;

[0017] 15. Output feature image.

[0018] Furthermore, the neural network model is a ResNet network model, which is divided into four layers. The first layer is the input feature attributes, including: ① chest and lung impedance, ② chest circumference, ③ height, ④ weight, ⑤ age, and ⑥ gender of the person to be tested; the second layer is the input quantity belonging to the cloud distribution, which classifies the first layer of feature quantities and calculates the activation degree of the belonging cloud distribution; the third layer is several layers of convolutional neural network layers, which perform rule reasoning and logic establishment; the fourth layer is the output prediction layer, which weights the various output values ​​of the neural network through the previous levels of logic outputs to obtain the main parameters of chest and lung modeling of the person to be tested.

[0019] Furthermore, the third step is specifically as follows: in the extracted area, according to a given threshold, an isosurface expressing the reconstructed tissue or organ is constructed, and based on a voxel-level reconstruction method, the measured discrete data is fitted with the original human body model data to reconstruct the three-dimensional digital morphology of the target part.

[0020] Furthermore, the reconstruction steps in the third step specifically include: ① point (error) processing; ② data interpolation; ③ data smoothing; ④ surface direct fitting; ⑥ point data gridding.

[0021] Furthermore, in the third step, the Python access MIMICS interface algorithm is as follows:

[0022] 21. Import the corresponding library in Python;

[0023] 22. Introduced MIMICS VBA interface;

[0024] 23. Determine whether the interfaces at both ends are available and visible;

[0025] 24. Escape different variables in VBA and open the secondary development source file;

[0026] 25. Grant MIMICS the highest access rights.

[0027] Furthermore, in the fourth step, the programming of inp files is realized through Python's text operation ability. The process includes: ① using ABAQUS / CAE to generate basic inp files; ② Python reads and rewrites inp files; ③ writing scripts (recording macros) to achieve rapid modeling.

[0028] The present invention also discloses a high-fidelity human body model construction system, which can be used to implement the above-mentioned high-fidelity human body model construction method, specifically, including: a human body model feature extraction module, a vital sign data measurement module, an ABAQUS module, a MIMICS module and a ResNet module;

[0029] A human body model feature extraction module, used for extracting features of a complete human body reference model;

[0030] A vital sign data measurement module is used to measure the vital sign data of the person to be tested;

[0031] The ResNet network model is used to train according to the characteristics and physical sign data of the human reference model to obtain the main parameters of the chest and lung modeling of the person to be tested.

[0032] The MIMICS module is used to perform data conversion with the ResNet module, reconstruct the key node data of the target part, and output binary STL data;

[0033] ABAQUS module: Combined with the Python module for rapid modeling, ultimately obtaining a high-fidelity three-dimensional model of the human body that can be used for finite element analysis calculations.

[0034] Compared with the prior art, the advantages of the present invention are:

[0035] Based on the original complete human body 3D model and the idea of ​​transfer learning, it is easy to implement, has fast computing speed and high accuracy. It improves the calculation accuracy of the finite element model of human body parts, no longer relies on CT technology, reduces the risk of multiple exposure to X-rays, and also avoids the problem of small training sets in traditional machine learning. A high-fidelity 3D model of the local human body to be tested can be reconstructed from the original overall 3D model of the human body and the local vital sign data of the person to be tested, which will bring great convenience to engineering and technical personnel who need to study the impact response and damage of the researchers. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a flow chart of a method for constructing a high-fidelity human body model according to an embodiment of the present invention;

[0037] Figure 2 It is an existing complete three-dimensional model of a human body;

[0038] Figure 3 This is a flowchart of an embodiment of the present invention using Python to extract features and realize visualization

[0039] Figure 4 is a schematic diagram of a typical transfer learning model network used in an embodiment of the present invention;

[0040] Figure 5 This is a flowchart of building a network model according to an embodiment of the present invention;

[0041] Figure 6 It is a flowchart of the Python and MIMICS interface algorithm of an embodiment of the present invention;

[0042] Figure 7 This is a schematic diagram of the combined action of three softwares in an embodiment of the present invention;

[0043] Figure 8This is a schematic diagram of a three-dimensional reconstructed model of the chest and lungs of a person under test obtained by using this method. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and examples.

[0045] like Figure 1 and 7 As shown, in this embodiment, in order to obtain a chest and lung model of a person to be tested, the reference material is as follows Figure 2 shown.

[0046] A method for constructing a high-fidelity human body model comprises the following steps:

[0047] Step 1: Feature extraction of the complete human body reference model.

[0048] Specifically, it is divided into the description of local features and the description of global features. The 3D shape descriptor is used to establish the mapping relationship between corresponding points between surfaces. A vector is used to describe the shape characteristics of specified points and areas on the surface of the human body model, and the value of the matching vector is used to describe the correspondence between different surface points.

[0049] like Figure 3 As shown, Python is used for feature extraction and visualization, specifically:

[0050] 1. First, introduce the required libraries in Python, including matplotlib.pyplot library, mpl_toolkits.mpplot3d library and numpy library;

[0051] 2. Read the txt data, convert it into a two-dimensional array, set the number of nodes to 200, and return a tuple;

[0052] 3. Define the information entropy function and initialize the special two-dimensional array;

[0053] 4. Set the image size, (x, y, z) = (0, 1, 2); set the coordinate axis invisible and the background network invisible;

[0054] 5. Output feature image.

[0055] Step 2: Measure the physical data of the person to be tested and input it into the neural network model for feature comparison learning and training.

[0056] Based on bioelectrical impedance technology, people's height, weight, body fat, bone weight, water content and protein content are measured and input into the ResNet network model.

[0057] The network model is divided into four layers. The first layer is the input feature attributes, including: ① chest and lung impedance, ② chest circumference, ③ height, ④ weight, ⑤ age, and ⑥ gender of the person to be tested; the second layer is the input quantity affiliation cloud distribution, which classifies the first layer feature quantities and calculates the activation degree of the affiliation cloud distribution; the third layer is several layers of convolutional neural network layers, which perform rule reasoning and logic establishment; the fourth layer is the output prediction layer, which weights the various output values ​​of the neural network through the previous levels of logic outputs to obtain the main parameters of the chest and lung modeling of the person to be tested.

[0058] The model specifications and parameters used in this method are shown in the attached Figure 4 The network model building process is as follows Figure 5 As shown, the details are as follows:

[0059] Step 3: Connect the model trained in the second step to MIMICS for data conversion, reconstruct the key node data of the target part, and output binary STL data;

[0060] Specifically, in the extracted area, according to the given threshold, an isosurface expressing the reconstructed tissue or organ is constructed, and based on the voxel-level reconstruction method, the measured discrete data is fitted with the original human body model data to reconstruct its three-dimensional digital form. The reconstruction steps are as follows: ① point (error) processing; ② data interpolation; ③ data smoothing; ④ direct surface fitting; ⑥ point data gridding. Python is used to assist in large-scale data editing and conversion. The algorithm flow of Python and MIMICS interface is as follows Figure 6 As shown, the specific process is as follows:

[0061] 1. Import the corresponding library in Python;

[0062] 2. Introduce MIMICS VBA interface;

[0063] 3. Determine whether the interfaces at both ends are available and visualized;

[0064] 4. Escape different variables in VBA and open the secondary development source file;

[0065] 5. Grant MIMICS the highest access rights.

[0066] Step 4: Use Python and ABAQUS for joint rapid modeling, and finally obtain a high-fidelity 3D model of the human body that can be used for finite element analysis and calculation ( Figure 8 ).

[0067] Specifically, the programming of inp files is achieved through Python's text operation capabilities. The process includes: ① using ABAQUS / CAE to generate basic inp files; ② Python reads and rewrites inp files; ③ writing scripts (recording macros) to achieve rapid modeling.

[0068] Through the above four steps, the reconstruction of the three-dimensional model of some parts of a general person based on the original three-dimensional model data of the human body is completed, eliminating the CT image processing and complex human body modeling process for three-dimensional reconstruction of conventional human body parts. The final fine three-dimensional model of the chest and lungs of the person to be tested has been verified by subsequent experiments and has good usability, and can be used for the study of personnel impact and injury mechanisms.

[0069] In another embodiment of the present invention, a high-fidelity human body model construction system is provided, which can be used to implement the above-mentioned high-fidelity human body model construction method, specifically, including: a human body model feature extraction module, a vital sign data measurement module, an ABAQUS module, a MIMICS module and a ResNet module;

[0070] A human body model feature extraction module, used for extracting features of a complete human body reference model;

[0071] A vital sign data measurement module is used to measure the vital sign data of the person to be tested;

[0072] The ResNet network model is used to train according to the characteristics and physical sign data of the human reference model to obtain the main parameters of the chest and lung modeling of the person to be tested.

[0073] The MIMICS module is used to perform data conversion with the ResNet module, reconstruct the key node data of the target part, and output binary STL data;

[0074] ABAQUS module: Combined with Python module for rapid modeling, finally obtaining a high-fidelity three-dimensional model of the human body that can be used for finite element analysis calculations. It should be understood by those skilled in the art that the embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0075] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0076] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0077] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0078] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the implementation methods of the present invention, and should be understood that the protection scope of the present invention is not limited to such special statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A method for constructing a high-fidelity human body model. It is characterized in that The following steps are involved: The first step is to extract the features of the complete human body reference model, use the 3D shape descriptor to establish the mapping relationship between the corresponding points on the surface, use a vector to describe the shape features of the specified points and areas on the surface of the human body model, and match the value of the vector to describe the correspondence between different surface points; use Python script to read the human body model and output txt text in parallel, input the neural network model, and perform feature extraction; The second step is to measure the physical data of the person to be tested. Based on the bioelectrical impedance technology, the person's height, weight, body fat, bone weight, water content and protein content are measured, and the input neural network model is compared with the features extracted in the first step for learning and training to obtain the chest and lung modeling parameters of the person to be tested; including: lung length and width data, volume parameters, diameter parameters, and surface area parameters; In the third step, the model trained in the second step is connected to MIMICS through Python for data conversion, and the key node data of the target part is reconstructed. Large-scale data conversion processing is performed through Python scripts, and the csv file is converted into a binary STL file. After preliminary processing, the STP format is output and transferred to the meshing software for denoising, filtering, smoothing and polishing. The partial three-dimensional reconstructed model of the person to be tested is obtained, and the binary STL data is output; In the fourth step, Python and ABAQUS are used to perform joint rapid modeling based on the modeling parameters. The STP data obtained in step 3 after preliminary processing is output as an inp file and imported into ABAQUS for subsequent finite element modeling and calculation, and finally a high-fidelity three-dimensional model of the human body that can be used for finite element analysis and calculation is obtained.

2. A method for constructing a high-fidelity human body model according to claim 1, Features: In the first step, Python script is used to extract the features of the complete human reference model, as follows: 11) First, import the required libraries in Python, including matplotlib.pyplot library, mpl_toolkits.mpplot3d library and numpy library; 12) Read the txt data, convert it into a two-dimensional array, set the number of nodes to 200, and return a tuple; 13) Define the information entropy function and initialize the special two-dimensional array; 14) Set the image size, (x, y, z) = (0, 1, 2); set the coordinate axis invisible and the background network invisible; 15) Output feature image.

3. A method for constructing a high-fidelity human body model according to claim 1, Features: The neural network model is a ResNet network model, which is divided into four layers. The first layer is the input feature attributes, including: ① chest and lung impedance, ② chest circumference, ③ height, ④ weight, ⑤ age, and ⑥ gender of the person to be tested; the second layer is the input quantity belonging to the cloud distribution, which classifies the first layer of feature quantities and calculates the activation degree of the belonging cloud distribution; the third layer is several layers of convolutional neural network layers, which perform rule reasoning and logic establishment; the fourth layer is the output prediction layer, which weights the various output values ​​of the neural network through the previous levels of logic outputs to obtain the main parameters of the chest and lung modeling of the person to be tested.

4. A method for constructing a high-fidelity human body model according to claim 1, Features: The third step is as follows: in the extracted area, according to the given threshold, an isosurface expressing the reconstructed tissue or organ is constructed, and based on the voxel-level reconstruction method, the measured discrete data is fitted with the original human body model data to reconstruct the three-dimensional digital morphology of the target part.

5. A method for constructing a high-fidelity human body model according to claim 1, Features: The reconstruction steps in the third step specifically include: ① point (error) processing; ② data interpolation; ③ data smoothing; ④ direct surface fitting; ⑥ point data gridding.

6. A method for constructing a high-fidelity human body model according to claim 1, Features: In the third step, the algorithm for Python accessing the MIMICS interface is as follows: 21) Import the corresponding library in Python; 22) Introduce MIMICS VBA interface; 23) Determine whether the interfaces at both ends are available and visualized; 24) Escape different variables in VBA and open the secondary development source file; 25) Grant MIMICS the highest access rights.

7. A method for constructing a high-fidelity human body model according to claim 1, Features: In the fourth step, the programming of inp files is realized through Python's text operation capabilities. The process includes: ① using ABAQUS / CAE to generate basic inp files; ② Python reads and rewrites inp files; ③ writing scripts to achieve rapid modeling.

8. A high-fidelity human body model construction system, Features: The system can be used to implement a high-fidelity human body model construction method as described in any one of claims 1 to 7, specifically comprising: a human body model feature extraction module, a vital sign data measurement module, an ABAQUS module, a MIMICS module and a ResNet module; A human body model feature extraction module, used for extracting features of a complete human body reference model; A vital sign data measurement module is used to measure the vital sign data of the person to be tested; The ResNet network model is used to train according to the characteristics and physical sign data of the human reference model to obtain the main parameters of the chest and lung modeling of the person to be tested; The MIMICS module is used to perform data conversion with the ResNet module, reconstruct the key node data of the target part, and output binary STL data; ABAQUS module: Combined with the Python module for rapid modeling, ultimately obtaining a high-fidelity three-dimensional model of the human body that can be used for finite element analysis calculations.