Orthopedic 3D Printing Model Construction Method and Device Based on Intelligent AI

Abstract

The present invention discloses an orthopedic 3D printing model construction method and device based on intelligent AI, which relates to the field of image technology. The method includes: collecting a user database, configuring an imaging scheme and performing multi-view imaging to establish a DICOM format image dataset; placing the image dataset in a three-dimensional coordinate system and establishing a registration mapping; after preprocessing the image dataset, performing bone segmentation and marking the segmentation confidence; performing image fusion and reconstruction according to the registration mapping and the segmentation confidence to generate a three-dimensional model marking the position complexity; performing printing placement fitting based on the three-dimensional model, determining the gravity direction, and analyzing the geometric shape to determine a preliminary support area; optimizing the support points according to the position complexity and the preliminary support area to establish a selection optimization result; and optimizing the three-dimensional model according to the selection optimization result to generate an orthopedic 3D printing model for the user. Furthermore, the technical effects of improving the modeling efficiency and accuracy and enhancing the printing stability are achieved.

Description

Technical Field

[0001] The present invention relates to the field of image technology, and particularly to a method and device for constructing an orthopedic 3D printing model based on intelligent AI. Background Art

[0002] Applying 3D printing technology in the medical field, especially in orthopedic surgeries, can help doctors more intuitively observe the patient's bone structure, perform preoperative planning and simulation, and improve the accuracy and success rate of surgeries. Existing methods for constructing orthopedic 3D printing models usually rely on fixed parameter models supplemented by manual operations, suffering from technical problems such as long modeling time, low accuracy, and poor printing stability. Summary of the Invention

[0003] The present invention provides a method and device for constructing an orthopedic 3D printing model based on intelligent AI to solve the technical problems of long modeling time, low accuracy, and poor printing stability in the prior art, and achieve the technical effects of improving modeling efficiency and accuracy and enhancing printing stability.

[0004] In a first aspect, the present invention provides a method for constructing an orthopedic 3D printing model based on intelligent AI, wherein the method includes:

[0005] Collect a user database of the user, configure an imaging scheme based on the user database, perform multi-view imaging of the user based on the imaging scheme, establish an image data set, and save the image data set in DICOM format; establish a three-dimensional coordinate system, place the image data set into the three-dimensional coordinate system, perform image registration with the image coordinates in the three-dimensional coordinate system, and establish a registration mapping; after preprocessing the image data set, perform bone segmentation on the preprocessed image data set to establish a segmentation result, wherein the segmentation result is marked with a segmentation confidence level; perform image fusion and reconstruction according to the registration mapping and the segmentation confidence level to generate a three-dimensional model, and mark the position complexity of the three-dimensional model; perform printing placement fitting based on the three-dimensional model, determine the gravity direction, perform geometric shape analysis on the three-dimensional model in the gravity direction to determine a preliminary support area; perform selection and optimization of support points based on the position complexity and the preliminary support area to establish a selection and optimization result; optimize the three-dimensional model according to the selection and optimization result to establish an orthopedic 3D printing model of the user.

[0006] In a second aspect, the present invention further provides a device for constructing an orthopedic 3D printing model based on intelligent AI, wherein the device includes:

[0007] An image data acquisition module, which is used to acquire the user's user database, configure an imaging scheme based on the user database, perform multi-view imaging of the user based on the imaging scheme, establish an image dataset, and save the image dataset in DICOM format.

[0008] A three-dimensional registration module, which is used to establish a three-dimensional coordinate system, place the image dataset into the three-dimensional coordinate system, perform image registration with the image coordinates in the three-dimensional coordinate system, and establish a registration mapping.

[0009] A bone segmentation module, which is used to perform bone segmentation on the preprocessed image dataset after preprocessing the image dataset, and establish a segmentation result, where the segmentation result is marked with a segmentation confidence level.

[0010] A fusion and reconstruction marking module, which is used to perform image fusion and reconstruction according to the registration mapping and the segmentation confidence level, generate a three-dimensional model, and mark the position complexity of the three-dimensional model.

[0011] A placement and support module, which is used to perform printing placement fitting based on the three-dimensional model, determine the gravity direction, perform geometric shape analysis of the three-dimensional model in the gravity direction, and determine a preliminary support area.

[0012] A support optimization module, which is used to select and optimize support points based on the position complexity and the preliminary support area, and establish a selection and optimization result.

[0013] An entity execution module, which is used to optimize the three-dimensional model according to the selection and optimization result, and establish an orthopedic 3D printing model of the user.

[0014] The present invention discloses a method and device for constructing an orthopedic 3D printing model based on intelligent AI, including: collecting and obtaining the user database information of the user, formulating an imaging scheme based on this database, and performing multi-view imaging of the user to generate an image data set, and saving the image data set as a DICOM format file; establishing a three-dimensional coordinate system, placing the image data set into this coordinate system, and performing image registration with the image coordinates in the three-dimensional coordinate system to form a registration mapping relationship; preprocessing the image data set, and performing bone segmentation operations on the preprocessed image to generate a segmentation result, and this segmentation result is marked with a segmentation confidence level; fusing and reconstructing the image according to the registration mapping and the segmentation confidence level mark to generate a three-dimensional model, and marking the spatial position complexity of this three-dimensional model; performing printing placement fitting on the three-dimensional model, determining the gravity direction, performing geometric shape analysis based on the gravity direction to determine a preliminary support area; based on the position complexity and the preliminary support area, optimizing the selection of support points and generating a selection optimization result; optimizing the three-dimensional model according to the selection optimization result to finally generate the orthopedic 3D printing model of the user. The method and device for constructing an orthopedic 3D printing model based on intelligent AI disclosed by the present invention solve the technical problems of long modeling time, low accuracy, and poor printing stability, and achieve the technical effects of improving modeling efficiency and accuracy and enhancing printing stability. Description of the Drawings

[0015] Figure 1 It is a schematic flowchart of the method for constructing an orthopedic 3D printing model based on intelligent AI of the present invention.

[0016] Figure 2 It is a schematic structural diagram of the device for constructing an orthopedic 3D printing model based on intelligent AI of the present invention.

[0017] Description of the reference numerals: Image data acquisition module 11, three-dimensional registration module 12, bone segmentation module 13, fusion reconstruction identification module 14, placement support module 15, support optimization module 16, entity execution module 17. Detailed Embodiments

[0018] In the embodiments of the present invention, the overall idea adopted for the technical solution provided to solve the technical problems of long modeling time, low accuracy, and poor printing stability existing in the prior art is as follows:

[0019] First, collect and obtain the user database of the user, and this database contains the basic information and medical imaging data of the user. Based on this user database, configure an imaging scheme, and this scheme may include imaging acquisition methods at different angles, such as CT, MRI, etc.

[0020] Then, based on the configured imaging scheme, multi-view imaging is performed on the user, and the collected image dataset will be saved in DICOM format for subsequent processing and analysis. Next, a three-dimensional coordinate system is established, and the collected image dataset is placed into this three-dimensional coordinate system. Image registration is performed using the image coordinates to ensure that the images from different perspectives can be correctly corresponding and mapped to form a registration mapping relationship. After the preprocessing of the image dataset is completed, a bone segmentation operation is executed to extract the structural information of the bones and label the confidence of each segmentation result, that is, the accuracy degree of each segmented region. According to the registration mapping and segmentation confidence, the images from different perspectives are fused and reconstructed to generate the user's three-dimensional bone model. At the same time, based on the segmentation results and image complexity, the position complexity of the three-dimensional model is identified, which is crucial for the selection of support points and 3D printing. After that, based on the generated three-dimensional model, a fitting for printing placement is performed to determine the gravity direction during printing, and geometric shape analysis is carried out to identify the regions that may require support to determine the preliminary support regions. Next, combining the position complexity of the three-dimensional model and the preliminary support regions, an optimal selection of support points is carried out to find the most suitable support point layout and generate the optimization result of support point selection. Finally, based on the optimization result of the support points, the three-dimensional model is optimized and ultimately the user's orthopedic 3D printing model is generated. This model will be used for surgical planning, prosthesis design or other orthopedic-related applications.

[0021] The above technical solutions will be described in detail below in conjunction with the accompanying drawings of the specification and specific embodiments to better understand the above technical solutions. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments that are only used to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present invention. In addition, it should be noted that for the sake of description, only the parts related to the present invention are shown in the drawings rather than all.

[0022] Embodiment 1

[0023] Figure 1 It is a flowchart of a method for constructing an orthopedic 3D printing model based on intelligent AI according to the present invention. Among them, the method includes:

[0024] A user database of the user is collected, an imaging scheme is configured based on the user database, multi-view imaging of the user is performed based on the imaging scheme, an image dataset is established, and the image dataset is saved in DICOM format.

[0025] Specifically, first interact with the user database to obtain the user data of the target user, which reflects the user characteristic information of the target user and is used as a reference basis for image imaging. Exemplarily, the user database includes data such as the user's body shape, body posture, position to be modeled, and modeling accuracy requirements.

[0026] Specifically, configure the image imaging scheme based on the user database. First, extract relevant information from the user database, such as the user's bone position, body shape, body posture, gender, age, etc., as the basic data for formulating the image imaging scheme. Then, determine the specific requirements for imaging according to the user's health condition or bone needs. For example, whether certain parts need to be focused on, whether there are specific imaging angles or resolution requirements, etc. Next, adjust the parameters of the imaging device according to the user's body structure data (such as bone position, body shape) to ensure that the concerned parts to be collected are covered and the acquisition clarity that meets the resolution requirements is provided. Among them, the parameters of the imaging device include the device position (the height and distance of the imaging device), imaging angle, radiation dose, etc. Through the above steps, it is ensured that the image imaging scheme is adapted to the personalized needs and physical characteristics of the user, so as to obtain clear and accurate images.

[0027] Furthermore, perform multi-view imaging on the user according to the image imaging scheme. Capture the user's image data from different angles and directions to ensure comprehensive and accurate image information is obtained. Exemplarily, take pictures from multiple angles such as front, back, left, right, and diagonal, and classify and organize all the captured image data according to the imaging order or view order to establish a complete image data set. The image data set contains all the view images, ensuring that the target area can be completely reconstructed and observed through multiple views in subsequent analysis.

[0028] Specifically, convert and save the processed image data set into the DICOM (Digital Imaging and Communications in Medicine) format. The DICOM file contains necessary user information (such as name, ID), imaging device parameters, imaging time, and other relevant metadata for subsequent access, invocation, and analysis.

[0029] Establish a three-dimensional coordinate system, place the image data set into the three-dimensional coordinate system, perform image registration based on the image coordinates in the three-dimensional coordinate system, and establish a registration mapping.

[0030] Specifically, first, establish a three-dimensional coordinate system as the reference system for the image data. This coordinate system is constructed based on the physical position of the device, the center point of the imaging part, or other spaces with fixed reference points, and each axis of the three-dimensional coordinate system represents the three directions of X, Y, and Z respectively, so as to accurately locate the position of each image in the three-dimensional space.

[0031] Specifically, an image data set obtained from multi-view imaging is read. According to the specific view angle, position information, and imaging time of the imaging device corresponding to each image data, its initial position and orientation in the three-dimensional coordinate system are determined, and each image is placed in the three-dimensional coordinate system according to its position information.

[0032] Specifically, the multi-view image data in the above three-dimensional coordinate system is the initial image data restored according to the acquisition direction. In actual acquisition, due to changes in the user's posture or movement, there are imperfect fitting situations among the multi-angle image data, and it is difficult to form accurate bone image data. Therefore, further image registration is required to align the images from different views and establish a consistent three-dimensional model. The purpose of registration is to correct the differences in position, rotation, scaling, etc. between multi-view images through a transformation matrix. Exemplarily, image registration in the three-dimensional coordinate system is performed based on registration algorithms such as RANSAC, 4PCE, and IPC.

[0033] Exemplarily, first, key feature points common to adjacent images are extracted from each image. The feature points can be determined based on physical structures, edges, textures, or other significant features in the image. Then, a set of four points is extracted from the images from different views using the 4PCS algorithm, and the geometric relationship of these four points in the images remains consistent. Furthermore, the spatial positions of the point sets in adjacent images are compared to obtain the similarity transformation between the images. Finally, these key point sets are matched to solve for the optimal rigid body transformation matrix (including rotation, translation, and scaling). This matrix will be used to transform one of the images so that it aligns with the images from other views.

[0034] Further, after obtaining the optimal transformation matrix, the matrix is applied to transform the images, and all the image data sets are registered so that they are aligned to the same reference frame in the three-dimensional coordinate system, and a registration mapping relationship for each image is established. The registration mapping relationship records the original position of each image, the transformation matrix after registration, and their coordinate positions in the three-dimensional space.

[0035] After preprocessing the image data set, bone segmentation of the preprocessed image data set is performed to establish a segmentation result, where the segmentation result is marked with a segmentation confidence level.

[0036] Specifically, Google Segmentation is used to segment the bone regions in the preprocessed image data set from other tissues (such as muscles, fats, etc.). The segmentation methods include threshold segmentation, region growing, level set methods, deep learning methods, etc.

[0037] Exemplarily, convolutional neural networks (CNNs) are also widely used in image segmentation tasks. First, a set of labeled training data is prepared, such as a set of CT images, and corresponding bone segmentation labels. Among them, the labeled labels are boolean values, used to indicate whether each pixel belongs to the bone. Then, the training data is used to train the CNN. This involves forward propagation, backpropagation, and parameter updates (such as stochastic gradient descent to update network parameters). After the network training is completed, it can be used to perform bone segmentation on new CT images and output the segmentation result with a segmentation confidence identifier. In other words, the preprocessed image is input into the network, and the network will output a segmented image (segmentation result), where the value of each pixel represents the probability that the pixel belongs to the bone (segmentation confidence).

[0038] Through the above method steps, performing bone segmentation on the image dataset helps to accurately identify the pixels belonging to the bone in the image data, thereby improving the construction accuracy of the subsequent 3D model.

[0039] Perform image fusion reconstruction according to the registration mapping and the segmentation confidence identifier, generate a 3D model, and identify the position complexity of the 3D model.

[0040] Specifically, based on the established registration mapping, the multi-view image data is aligned to a unified 3D coordinate system. Through these registration mappings, the bone regions of each view image can be accurately corresponded to the same 3D spatial position; then, the segmentation results of each view are weighted using the segmentation confidence identifier, where the parts with higher segmentation confidence will occupy a greater weight in the 3D model, thereby improving the reliability and accuracy of the model. Next, the segmented bone images of each view are fused.

[0041] Optionally, a voxel-level fusion method is adopted to combine the segmentation results of each view, eliminate the imperfect fitting problem caused by changes in the user's posture or actions, and convert the fused image data into 3D voxel data.

[0042] Optionally, based on the voxel data, a Marching Cubes or other surface reconstruction algorithm is used to generate the mesh structure of the 3D model, and the bone structure is presented in the form of a 3D geometric model. Preferably, in the generated 3D model, smoothing processing, denoising processing, or hole filling is adaptively performed to improve the accuracy and quality of the model.

[0043] Specifically, the complexity of different parts of the model is quantified by methods such as curvature analysis, shape complexity analysis, or density analysis to obtain the location complexity at multiple positions. Among them, parts with higher complexity mean that these areas are more likely to have errors during actual imaging or reconstruction, and at the same time, they are more difficult to print during the subsequent 3D printing process, requiring refined support settings and printing slices.

[0044] Through the above steps, a three-dimensional bone model with high precision and clear location complexity identification can be generated, providing a reliable model basis for subsequent printing analysis and planning.

[0045] In some embodiments, when performing image fusion reconstruction according to the registration mapping and the segmentation confidence identification, it further includes:

[0046] Configure weights according to the registration mapping, and the formula is as follows:

[0047] ;

[0048] Wherein, represents the weight of the th perspective image after registration mapping, represents the image quality score of the th perspective image, represents the segmentation confidence identification of the th perspective image, represents the number of perspectives corresponding to the registration mapping, is the perspective index.

[0049] Perform pixel fusion according to the configured weights, as follows:

[0050] ;

[0051] represents the fused pixel value, is the pixel value at the corresponding position of the th perspective image.

[0052] Complete image fusion reconstruction according to the fused pixel values.

[0053] Optionally, before performing image fusion reconstruction, weights are assigned to the images of each perspective, and the weights depend on the image quality and the confidence level of segmentation. Specifically, first, based on the above weight configuration formula, the weights for fusing images from multiple perspectives are calculated. Among them, the higher the image quality and the larger the segmentation confidence level indicator, the higher the corresponding weight. Then, pixel fusion is performed according to the configured weights. Specifically, the fusion value at each pixel position is obtained by the weighted average of all pixel values at the corresponding position and their weights, ensuring that the pixels of images with high quality and good segmentation confidence are given priority. Finally, the fused multi-perspective images are stacked to generate a three-dimensional image or pixel fusion is directly performed in three-dimensional space to complete the image fusion reconstruction. Through the above steps, it can be ensured that the image fusion process takes into account the quality and confidence level of each perspective image, thereby generating a more accurate fusion result.

[0054] In some implementation manners, according to the registration mapping and the segmentation confidence level indicator, image fusion reconstruction is performed to generate a three-dimensional model, and the position complexity of the three-dimensional model is identified. It further includes:

[0055] Extract the amount of fusion data of the image fusion reconstruction, perform an adaptation evaluation based on the amount of fusion data and the position complexity, and establish an adaptation anomaly indicator; generate an additional acquisition instruction based on the adaptation anomaly indicator; control the imaging device to perform additional data acquisition through the additional acquisition instruction, and perform image fusion reconstruction compensation with the additional data acquisition result.

[0056] Specifically, based on a three-dimensional sliding window, slide to extract the amount of fusion data and the corresponding position complexity at multiple positions of the three-dimensional model. The three-dimensional sliding window is a three-dimensional space with a preset size, which can be a finite space determined by the sizes in the three coordinate directions or an interval space in any two coordinates or any single coordinate direction, and is determined based on the shape characteristics of the target bone. Then, calculate the position complexity corresponding to the obtained amount of fusion data, including the sum of the complexities of multiple unit parts of the three-dimensional model between the three-dimensional sliding windows. Furthermore, based on the obtained amount of fusion data and the corresponding position complexity, an adaptation evaluation is performed. If the ratio of the amount of fusion data to the position complexity does not meet the preset adaptation constraint, an adaptation anomaly indicator is generated.

[0057] Exemplarily, if the ratio of the amount of fusion data to the position complexity is less than the lower limit of the preset adaptation constraint, the amount of fusion data at this position is too small relative to its complexity, and the modeling quality of the region is insufficient, and more data is needed to provide a more accurate reconstruction.

[0058] Furthermore, an additional acquisition instruction is generated according to the adaptation anomaly indicator to guide the imaging device to acquire more data in the corresponding region of the adaptation anomaly indicator, and image fusion and reconstruction are performed again according to the obtained additional data acquisition result to improve the image quality at this position.

[0059] Through the above method steps, the complexity of the position is combined to effectively ensure the quality balance of the 3D model after image fusion and reconstruction at different positions, avoid reconstruction errors caused by insufficient data at complex positions, and ensure that there is sufficient data at all positions for accurate image fusion and reconstruction, thereby improving the final image quality.

[0060] Based on the 3D model, perform printing placement fitting to determine the gravity direction, and perform geometric shape analysis of the 3D model in the gravity direction to determine the preliminary support area.

[0061] Specifically, first, place the generated 3D model into a virtual printing environment, simulate the printing placement position and angle of the model, and perform fitting analysis to determine the optimal placement angle and direction of the model on the printing platform to ensure the stability of the model during printing and minimize the support requirements.

[0062] Specifically, after completion of the fitting, determine the gravity direction of the model, that is, the acting direction of gravity during the actual printing process, which helps to determine the force condition and potential deformation area of the model during printing.

[0063] Specifically, based on the geometric shape of the 3D model, perform analysis in combination with the gravity direction. Focus on identifying areas in the model that may be deformed or require additional support due to the action of gravity. Exemplarily, it includes protrusions, suspended parts, surfaces with large inclination angles, and complex structures of the model.

[0064] Among them, according to the geometric shape analysis result, determine the preliminary support area of the model. The preliminary support area is used to provide additional support during printing to prevent the model from collapsing or deforming due to gravity or structural vulnerability. In addition, the selection of the support area also considers how to reduce the difficulty of removing the support after printing, thereby optimizing the printing efficiency.

[0065] Based on the position complexity and the preliminary support area, perform optimization of the selection of support points and establish the optimization result.

[0066] In some embodiments, the optimization of the selection of support points based on the position complexity and the preliminary support area to establish the optimization result further includes:

[0067] Regarding each preliminary support area as an independent area, establish an area objective function based on the area information of the preliminary support area. The evaluation features of the area objective function include the number of support points, material, stability, and post-processing difficulty. Conduct area evaluation on the independent area, establish area associations of the independent area, where the area associations include area collaborative associations and area competitive associations, and the area evaluation includes spatial proximity analysis, mechanical coupling analysis, and material sharing analysis. Establish limit constraints for the support points, configure the solution space with the limit constraints, use the area objective function as the evaluation function, perform selection optimization of the support points within the solution space, and perform iterative compensation for the selection optimization through area associations to establish the selection optimization result.

[0068] Specifically, the rating index dimensions for the preliminary support area include the number of support points, material, stability, and post-processing difficulty. Among them, the fewer the number of support points, the smaller the corresponding support contact area, and the smaller the impact on the surface quality of the model. Correspondingly, the less the material consumption, the less the waste generated from removing the support. Stability is another important factor for evaluating the support structure. The support structure should be able to maintain the stability of the model throughout the printing process, preventing it from moving or collapsing. Exemplarily, the stability is determined based on the rigidity of the support structure and the connection firmness between the support structure and the model body. Post-processing includes steps such as removing the support structure and cleaning and repairing the model surface. If the support structure is designed too complexly or is difficult to remove, then the difficulty of post-processing will increase, which will increase the manufacturing time and cost.

[0069] Specifically, multiple independent areas may have collaborative associations (i.e., they work together in a certain way) or competitive associations (i.e., they have conflicts in resources or space). By conducting spatial proximity analysis, mechanical coupling analysis, and material sharing analysis on multiple independent areas, the area associations of the independent areas can be obtained.

[0070] Among them, spatial proximity analysis is used to evaluate the spatial relationships between various areas, such as the distance and relative position between independent areas, which helps to determine which areas may need to share the support structure or which areas' printing may affect each other. Mechanical coupling analysis is used to evaluate the possible mechanical interactions during the printing process for various areas, such as stress and deformation, which helps to determine the support structure that can withstand these forces and predict possible printing problems. Material sharing analysis is used to evaluate the material requirements of various areas during the printing process and possible material sharing strategies. Exemplarily, if two adjacent independent areas have the same or similar orientations, a strategy of sharing the support structure can be designed to improve the stability of the support, reduce material usage, and printing time. Through the above evaluations, the influence relationships between different independent areas are established, which helps to optimize the 3D printing process, improve efficiency, reduce costs, and improve printing quality at the same time.

[0071] Optionally, the limit constraints are the restrictive conditions imposed on the variables in the optimization problem, which define the boundaries of the solution space. Exemplarily, the limit constraints include the printing platform size, the support form (such as tree-like, geometric curve, hybrid, etc.), the support spacing, the support generation threshold, etc. The solution space is the set of all possible solutions. After setting the limit constraints, the solution space represents the region restricted by these constraints. Preferably, the characteristics of the solution space are described by combining graphical representation and mathematical models. By setting the limit constraints and configuring the solution space, the scope of the optimization problem can be effectively narrowed, and the solution efficiency can be improved.

[0072] In some implementation manners, the iterative compensation for selective optimization by region association further includes:

[0073] Establishing cooperative groups and competitive groups based on the region association; constructing a joint objective function through the region objective function corresponding to the cooperative groups, and performing joint optimization of the cooperative groups with the joint objective function; sorting the priority order of the competitive groups to establish a sorting result; establishing sequential optimization based on the sorting result, and completing the iterative compensation according to the joint optimization and sequential optimization.

[0074] Specifically, first, according to the relationship between regions, multiple independent spaces are divided into cooperative groups and competitive groups. Among them, multiple cooperative groups include multiple independent spaces with cooperative relationships, such as multiple model branches hanging on the same side; the competitive groups include multiple independent spaces with competitive relationships, such as multiple model branches hanging in opposite directions or model parts with conflicts in space or resources, or their printing order may affect the printing efficiency and quality.

[0075] Specifically, a joint objective function is constructed based on the objective function of the cooperative groups to obtain a printing strategy that maximizes these cooperative effects. Then, the regions of the competitive groups are sorted by priority, and then optimization is performed based on this sorting result. Exemplarily, it includes determining which regions should be printed first and how to arrange the printing order to minimize conflicts and resource waste.

[0076] Further, according to the results of the above joint optimization and sequential optimization, iterative compensation is performed by adjusting the printing parameters or modifying the support design to further improve the printing efficiency and quality. The above method steps understand the complexity of the model by considering the relationships and interactions between regions and their impacts on the overall objective to obtain the optimal printing strategy.

[0077] Optimize the three-dimensional model according to the selective optimization result to establish the user's orthopedic 3D printing model.

[0078] In some embodiments, the method further includes:

[0079] Perform fixed-point position recognition on the 3D model, conduct scale division based on the fixed-point position recognition results, and establish M scale division results; perform spatial scale filtering from coarse scale to fine scale based on the M scale division results; after all scale division results are filtered, perform filtering fusion of each scale division result, and update the 3D model according to the filtering fusion result.

[0080] Specifically, first, identify key points or feature points of the 3D model as fixed-point positions for scale division to obtain scale division results, which include a series of sub-models or regions, and each sub-model or region has its own scale or resolution. For example, some parts of the model may be divided into coarse scales (low resolution), while other parts may be divided into fine scales (high resolution). Exemplarily, fixed-point positions include curvature discontinuity points, curvature mutation points, complexity mutation points, etc.

[0081] Specifically, filter the results of each scale division to eliminate noise, smooth the data, optimize the geometry and structure of each scale division, and improve the printing efficiency and quality. Preferably, perform filtering through adaptive parameter adjustment to process different parts of the model at different scales, thereby adapting to the complexity and diversity of the model.

[0082] Exemplarily, adjust the filtering parameters according to the curvature or complexity of different parts of the model to optimize the printing effect of each part. For the coarse-scale (smaller curvature or lower complexity) parts of the model, use larger filtering parameters to smooth large geometries or structures; while for the fine-scale (larger curvature or higher complexity) parts of the model, use smaller filtering parameters to retain details, improving both the printing quality and efficiency of the model.

[0083] Furthermore, fuse the filtering results of all scale divisions together, and then update the 3D model according to the fusion result to achieve the best printing effect. Through the above method steps, combining feature point analysis and filtering methods, multi-scale analysis and optimization of the 3D model are carried out, improving the printing efficiency and quality.

[0084] In some implementation manners, performing spatial scale filtering from coarse scale to fine scale based on the M scale division results further includes:

[0085] Perform spatial scale filtering through a formula as follows:

[0086] ;

[0087] where, represents the new position vector of vertex after filtering processing, is the vertex The original position vector, is the smoothing coefficient, characterizing the set of adjacent vertices of vertex ; characterizing any adjacent vertex of vertex ; is the weight between vertex and vertex , characterizing the influence degree of vertex on the position update of vertex ; is the position vector of vertex ; is the normal deviation control factor, and is the normal vector of vertex

[0088] Specifically, the smoothing coefficient determines the degree of position smoothing. A larger will make the vertex position closer to the average position of adjacent vertices, while a smaller retains more information about the original position. The normal deviation control factor is used to adjust the movement of the vertex in the normal direction for correcting the surface geometry.

[0089] In summary, the method for constructing an orthopedic 3D printing model based on intelligent AI provided by the present invention has the following technical effects:

[0090] By collecting and obtaining the user database information of the user, formulating an imaging scheme based on the database, and performing multi-view imaging of the user to generate an image data set, and saving the image data set as a DICOM format file; establishing a three-dimensional coordinate system, placing the image data set into the coordinate system, and performing image registration with the image coordinates in the three-dimensional coordinate system to form a registration mapping relationship; preprocessing the image data set, and performing bone segmentation operation on the preprocessed image to generate a segmentation result, which carries a segmentation confidence label; fusing and reconstructing the image according to the registration mapping and the segmentation confidence label to generate a three-dimensional model, and labeling the spatial position complexity of the three-dimensional model; performing printing placement fitting on the three-dimensional model, determining the gravity direction, performing geometric shape analysis based on the gravity direction to determine a preliminary support area; optimizing and selecting support points based on the position complexity and the preliminary support area, and generating a selection optimization result; optimizing the three-dimensional model according to the selection optimization result to finally generate the orthopedic 3D printing model of the user. Thereby achieving the technical effects of improving the modeling efficiency and accuracy and enhancing the printing stability.

[0091] Embodiment 2

[0092] Figure 2It is a schematic structural diagram of an orthopedic 3D printing model construction device based on intelligent AI of the present invention. For example, Figure 1 In the present invention, the schematic flowchart of the orthopedic 3D printing model construction method based on intelligent AI can be implemented through a structure as shown in Figure 2 shown.

[0093] Based on the same concept as the orthopedic 3D printing model construction method based on intelligent AI in the above embodiment, the orthopedic 3D printing model construction device provided by the present invention further includes:

[0094] An image data acquisition module 11, configured to acquire a user database of a user, configure an imaging scheme based on the user database, perform multi-view imaging of the user based on the imaging scheme, establish an image data set, and save the image data set in DICOM format.

[0095] A three-dimensional registration module 12, configured to establish a three-dimensional coordinate system, place the image data set into the three-dimensional coordinate system, perform image registration with the image coordinates in the three-dimensional coordinate system, and establish a registration mapping.

[0096] A bone segmentation module 13, configured to perform bone segmentation on the preprocessed image data set after preprocessing the image data set, establish a segmentation result, where the segmentation result is marked with a segmentation confidence level identifier.

[0097] A fusion reconstruction identification module 14, configured to perform image fusion reconstruction according to the registration mapping and the segmentation confidence level identifier, generate a three-dimensional model, and identify the position complexity of the three-dimensional model.

[0098] A placement support module 15, configured to perform printing placement fitting based on the three-dimensional model, determine the gravity direction, perform geometric shape analysis on the three-dimensional model in the gravity direction, and determine a preliminary support area.

[0099] A support optimization module 16, configured to perform selection and optimization of support points based on the position complexity and the preliminary support area, and establish a selection and optimization result.

[0100] An entity execution module 17, configured to optimize the three-dimensional model according to the selection and optimization result, and establish an orthopedic 3D printing model of the user.

[0101] Among them, the fusion reconstruction identification module 14 includes:

[0102] A registration mapping weight configuration unit, configured to configure weights according to the registration mapping, and the formula is as follows:

[0103] .

[0104] Among them, Characterize the weight of the th perspective image after registration mapping, Characterize the th perspective image's image quality score, Characterize the th perspective image's segmentation confidence flag, Characterize the number of perspectives corresponding to the registration mapping, is the perspective index.

[0105] The pixel fusion unit is used to perform pixel fusion according to the configured weights as follows:

[0106] .

[0107] Characterize the fused pixel value, is the pixel value at the corresponding position of the th perspective image.

[0108] The image fusion and reconstruction unit is used to complete image fusion and reconstruction according to the fused pixel values.

[0109] In some implementation manners, the fusion reconstruction identification module 14 further includes:

[0110] The fusion data volume extraction and adaptation evaluation unit is used to extract the fusion data volume of the image fusion and reconstruction, perform an adaptation evaluation based on the fusion data volume and the position complexity, and establish an adaptation anomaly flag.

[0111] The additional acquisition instruction generation unit is used to generate an additional acquisition instruction based on the adaptation anomaly flag.

[0112] The additional data acquisition and fusion reconstruction compensation unit is used to control the imaging device to perform additional data acquisition through the additional acquisition instruction, and perform image fusion and reconstruction compensation with the additional data acquisition result.

[0113] In some embodiments, the support optimization module 16 includes:

[0114] The support area independence processing unit is used to regard each preliminary support area as an independent area, and establish a regional objective function based on the regional information of the preliminary support area. The evaluation features of the regional objective function include the number of support points, material, stability, and post-processing difficulty.

[0115] The regional evaluation and association establishment unit is used to perform a regional evaluation on the independent area, establish a regional association of the independent area. The regional association includes regional cooperation association and regional competition association. The regional evaluation includes spatial proximity analysis, mechanical coupling analysis, and material sharing analysis.

[0116] The support point limit constraint and selection optimization unit is used to establish the limit constraint of the support point, configure the solution space with the limit constraint, use the regional objective function as the evaluation function, perform the selection optimization of the support point in the solution space, and perform iterative compensation for the selection optimization through regional association to establish the selection optimization result.

[0117] In some implementation manners, the support point limit constraint and selection optimization unit in the support optimization module 16 includes:

[0118] The collaborative grouping and competitive grouping construction unit is used to establish collaborative grouping and competitive grouping with the regional association.

[0119] The joint objective function construction and collaborative grouping optimization unit is used to construct a joint objective function through the regional objective function corresponding to the collaborative grouping, and perform joint optimization of the collaborative grouping with the joint objective function.

[0120] The competitive grouping priority sorting unit is used to sort the priority order of the competitive grouping to establish a sorting result.

[0121] The sequential optimization and iterative compensation unit is used to establish sequential optimization based on the sorting result, and complete iterative compensation according to the joint optimization and sequential optimization.

[0122] In some embodiments, the system further includes:

[0123] The three-dimensional model fixed-point position recognition unit is used to perform model fixed-point position recognition on the three-dimensional model, perform scale division based on the fixed-point position recognition result, and establish M scale division results.

[0124] The spatial scale filtering unit is used to perform spatial scale filtering from the coarse scale to the fine scale based on the M scale division results.

[0125] The filtering fusion and three-dimensional model update unit is used to perform filtering fusion of each scale division result and update the three-dimensional model according to the filtering fusion result after all the scale division results are filtered.

[0126] In some implementation manners, the spatial scale filtering unit in the system is used for:

[0127] Perform spatial scale filtering through the formula as follows:

[0128] .

[0129] Wherein, represents the new position vector of the vertex after the filtering process, is the original position vector of the vertex , is the smoothing coefficient. Characterize the vertex The set of adjacent vertices of Characterize the vertex Any adjacent vertex of Is the vertex And the vertex The weight between them, characterizing the vertex The influence degree of the position update of the vertex Is the vertex The position vector of Is the normal deviation control factor Is the vertex The normal vector of

[0130] It should be understood that the embodiments mentioned in this specification focus on their differences from other embodiments. The specific embodiments in the first-mentioned embodiment one are equally applicable to the orthopedic 3D printing model construction device based on intelligent AI described in embodiment two. For the sake of simplicity of the specification, no further elaboration will be made here.

[0131] It should be understood that the disclosed embodiments of the present invention and the above descriptions can enable those skilled in the art to implement the present invention using the present invention. At the same time, the present invention is not limited to the part of the embodiments mentioned above. It should be understood that those of ordinary skill in the art can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.​

Claims

1. A method for constructing an orthopedic 3D printing model based on intelligent AI, characterized in that: The method comprises: Acquire a user database of the user, configure an imaging scheme based on the user database, perform multi-view imaging of the user based on the imaging scheme, establish an image data set, and save the image data set in a DICOM format; Establishing a three-dimensional coordinate system, placing the image data set into the three-dimensional coordinate system, performing image registration with the image coordinates in the three-dimensional coordinate system, and establishing a registration mapping; After preprocessing the image data set, performing bone segmentation of the preprocessed image data set to establish a segmentation result, wherein the segmentation result carries a segmentation confidence mark; Performing image fusion reconstruction according to the registration mapping and the segmentation confidence identifier to generate a three-dimensional model, and identifying the position complexity of the three-dimensional model; Performing printing placement fitting based on the three-dimensional model to determine the direction of gravity, performing geometric shape analysis of the three-dimensional model based on the direction of gravity to determine a preliminary support area; Based on the position complexity and the preliminary support area, the support points are selected and optimized, and the selection and optimization results are established; Optimize the three-dimensional model according to the selected optimization result and establish the user's orthopedic 3D printing model; The image fusion reconstruction according to the registration mapping and the segmentation confidence mark also includes: The weights are configured according to the registration map, as follows: ； in, Characterize the registration mapping The weight of the image from each perspective, Characterization The image quality score of the image from each perspective is Characterization The segmentation confidence level of the image from each viewpoint is Characterizes the number of viewpoints corresponding to the registration map, is the perspective index, Characterization The image quality score of the image from each perspective is Characterization Segmentation confidence level identification of images from different viewpoints; Pixel fusion is performed according to the configured weights as follows: ； Represents the pixel value after fusion, For the The pixel value of the corresponding position of the perspective image; Image fusion reconstruction is completed based on the fused pixel values.

2. The method for constructing an orthopedic 3D printing model based on intelligent AI according to claim 1, characterized in that: The step of selecting and optimizing the support points based on the position complexity and the preliminary support area and establishing the selection and optimization result further includes: Taking each preliminary support area as an independent area, establishing a regional objective function based on the regional information of the preliminary support area, wherein the evaluation characteristics of the regional objective function include the number of support points, material, stability, and post-processing difficulty; Performing regional evaluation on the independent regions, establishing regional associations of the independent regions, wherein the regional associations include regional collaborative associations and regional competitive associations, and the regional evaluation includes spatial proximity analysis, mechanical coupling analysis, and material sharing analysis; Establish limit constraints on support points, configure solution space with the limit constraints, use the regional objective function as an evaluation function, perform selection optimization of support points in the solution space, perform iterative compensation of selection optimization through regional association, and establish selection optimization results.

3. The method for constructing an orthopedic 3D printing model based on intelligent AI as claimed in claim 2, characterized in that: The iterative compensation for selecting and optimizing by regional association also includes: Establishing cooperative groups and competitive groups based on the regional associations; Constructing a joint objective function through the regional objective functions corresponding to the collaborative groups, and performing joint optimization of the collaborative groups with the joint objective function; Sorting the competing groups in order of priority and establishing a sorting result; A sequential optimization is established based on the sorting result, and iterative compensation is completed according to the joint optimization and the sequential optimization.

4. The method for constructing an orthopedic 3D printing model based on intelligent AI according to claim 1, characterized in that: The image fusion reconstruction is performed according to the registration mapping and the segmentation confidence mark to generate a three-dimensional model and mark the position complexity of the three-dimensional model, and further includes: Extracting the amount of fused data of image fusion reconstruction, performing adaptation evaluation based on the amount of fused data and the position complexity, and establishing an adaptation abnormality mark; generating an additional collection instruction based on the adaptation exception identifier; The additional acquisition instruction is used to control the imaging device to perform additional data acquisition, and the additional data acquisition result is used to perform image fusion reconstruction compensation.

5. The method for constructing an orthopedic 3D printing model based on intelligent AI as claimed in claim 4, characterized in that: The method further comprises: Performing fixed-point position recognition on the three-dimensional model, performing scale division based on the fixed-point position recognition result, and establishing M scale division results; Based on the M scale division results, spatial scale filtering from coarse scale to fine scale is performed; After all scale division results are filtered, filtering fusion of each scale division result is performed, and the three-dimensional model is updated according to the filtering fusion results.

6. The method for constructing an orthopedic 3D printing model based on intelligent AI as claimed in claim 5, characterized in that: The spatial scale filtering from a coarse scale to a fine scale based on the M scale division results also includes: The spatial scale filtering is performed through the formula as follows: ； in, Represents a vertex The new position vector after filtering is Vertex The original position vector, is the smoothing coefficient, Characterizing Vertices The set of adjacent vertices of Characterizing Vertices Any adjacent vertex of Vertex With Vertex The weight between them represents the vertex Opposite Point The impact of location updates, Vertex The position vector of is the normal deviation control factor, Vertex The normal vector of .

7. An orthopedic 3D printing model construction device based on intelligent AI, characterized in that: The device is used to execute the method for constructing an orthopedic 3D printing model based on intelligent AI according to any one of claims 1 to 6, and the device comprises: An image data acquisition module, the image data acquisition module is used to acquire a user database of a user, configure an image imaging scheme based on the user database, perform multi-view imaging of the user based on the image imaging scheme, establish an image data set, and save the image data set in DICOM format; A three-dimensional registration module, the three-dimensional registration module is used to establish a three-dimensional coordinate system, place the image data set into the three-dimensional coordinate system, perform image registration with the image coordinates in the three-dimensional coordinate system, and establish a registration mapping; A skeleton segmentation module, wherein the skeleton segmentation module is used to perform skeleton segmentation of the preprocessed image data set after preprocessing the image data set, and establish a segmentation result, wherein the segmentation result carries a segmentation confidence mark; A fusion reconstruction identification module, the fusion reconstruction identification module is used to perform image fusion reconstruction according to the registration map and the segmentation confidence identification, generate a three-dimensional model, and identify the position complexity of the three-dimensional model; A placement support module, the placement support module is used to perform printing placement fitting based on the three-dimensional model, determine the gravity direction, perform geometric shape analysis of the three-dimensional model based on the gravity direction, and determine a preliminary support area; A support optimization module, the support optimization module is used to select and optimize support points based on the position complexity and the preliminary support area, and establish a selection optimization result; A physical execution module, the physical execution module is used to optimize the three-dimensional model according to the selection optimization result and establish an orthopedic 3D printing model for the user; The fusion reconstruction identification module includes: The registration mapping weight configuration unit is used to configure the weight according to the registration mapping. The formula is as follows: ； in, Characterize the registration mapping The weight of the image from each perspective, Characterization The image quality score of the image from each perspective is Characterization The segmentation confidence level of the image from each viewpoint is Characterizes the number of viewpoints corresponding to the registration map, is the perspective index, Characterization The image quality score of the image from each perspective is Characterization Segmentation confidence level identification of images from different viewpoints; The pixel fusion unit is used to perform pixel fusion according to the configured weights, as follows: ； Represents the pixel value after fusion, For the The pixel value of the corresponding position of the perspective image.

Images