Full-computer intelligent high-similarity prosthesis production method and system

Abstract

The invention discloses a full-computer intelligent high-similarity prosthesis production method and system, and the method comprises the steps: a main computer obtains a 3D model containing the segmentation information of all tissues in a missing part of a patient through a pre-constructed artificial intelligence model by collecting the information of the patient, and obtains a 2D unfolded image for the surface layer of the 3D model based on the information of the patient and the 3D model; controlling 3D printing equipment to print the 3D model to obtain each layered part of the prosthesis main body; controlling film manufacturing equipment to manufacture a surface layer film of the missing part; controlling 2D printing equipment to print a 2D unfolded image on the surface layer film of the missing part to form a skin; assembling equipment is controlled to combine all the parts of the prosthesis, the skin is attached to the surface of the prosthesis body, surface treatment is conducted, and a finished prosthesis product is obtained. According to the invention, the prosthesis production process is computerized and intelligent, the high-precision and high-similarity prosthesis can be provided in a shorter time, and the manufacturing cost is reduced; the artificial intelligence technology is used for prosthesis design and modeling, and the quality and similarity of the prosthesis can be improved.

Description

Technical Field

[0001] The present invention belongs to the field of prosthesis production and manufacturing, and in particular relates to a fully computer-intelligent high-similarity prosthesis production method and system. Background Art

[0002] For disabled patients, good prostheses can not only help users carry out daily activities independently and comfortably, but also help users build confidence and reintegrate into society, which is conducive to the stable development of society.

[0003] Traditional prosthesis production methods generally include the following separate processes: (1) 3D measurement sampling of the human body shape; (2) 3D printing of the negative mold; (3) making the positive mold using human body simulation plastic (such as silicone rubber); (4) manual surface spraying; and (5) surface protection treatment.

[0004] Although the sampling step in the current prosthesis production method can be completed with the help of digital instruments and equipment, the production process still relies extensively on manual operation. Therefore, the whole process is not only time-consuming, but also requires each operator to have rich personal experience and skills. Especially in surface painting, superb skills and exquisite artistic talents are required, similar to the treatment of body art. In addition, in the case of large-scale loss or loss of organs on both sides, it is necessary to rely on the experience mastered by craftsmen in long-term practice to conceive the shape of the patient's missing part. Since it is difficult to imagine the specific shape of the missing part of the body, the reconstruction of asymmetric parts such as lips and noses is difficult to achieve if it relies entirely on manual labor. Summary of the invention

[0005] In order to solve the above problems existing in the prior art, the present invention provides a fully computer-intelligent high-similarity prosthesis production method and system. The technical problem to be solved by the present invention is achieved through the following technical solutions:

[0006] In a first aspect, an embodiment of the present invention provides a fully computer-intelligent method for producing a highly similar pseudobody, the method comprising:

[0007] The main computer collects patient information and uses a pre-built artificial intelligence model to obtain a 3D model containing segmentation information of each tissue in the patient's missing part, and obtains a 2D unfolded image of the surface of the 3D model based on the collected patient information and the 3D model; wherein each tissue includes bones, muscles, tendons, fat layers, and skin layers;

[0008] The host computer controls the 3D printing device to print the 3D model to obtain the prosthesis body;

[0009] The main computer controls the film-making equipment to make a surface film at the missing part;

[0010] The main computer controls the 2D printing device to print the 2D expanded image onto the surface film of the missing part to form a skin;

[0011] The main computer controls the assembly equipment to attach the skin to the surface of the prosthesis body, and obtains the prepared prosthesis after surface treatment.

[0012] In one embodiment of the present invention, in the case where the missing part has no symmetrical healthy reference side, the host computer acquires patient information and uses a pre-built artificial intelligence model to obtain a 3D model containing segmentation information of each tissue in the missing part of the patient, and acquires a 2D unfolded image of the surface layer of the 3D model based on the acquired patient information and the 3D model, including:

[0013] The host computer uses the pre-constructed first model to reconstruct the neural network, collects information about the remaining parts of the patient, performs three-dimensional reconstruction of the missing part, obtains a reconstructed three-dimensional structural model, and obtains a reconstructed surface image of the missing part; wherein the first model reconstruction neural network is implemented using generative AI technology;

[0014] The three-dimensional structure model is subjected to structural segmentation of internal tissues using a pre-constructed second structural hierarchical neural network, and the model is modified using model attribute related information in the collected patient information and a pre-constructed third model modification neural network, to obtain a 3D model after processing;

[0015] The surface image reconstructed from the missing part is expanded in two dimensions according to the 3D model to obtain a 2D expanded image.

[0016] In one embodiment of the present invention, in the case where the missing part has a symmetrical healthy reference side, the host computer acquires patient information and uses a pre-built artificial intelligence model to obtain a 3D model containing segmentation information of each tissue in the missing part of the patient, and acquires a 2D unfolded image of the surface layer of the 3D model based on the acquired patient information and the 3D model, including:

[0017] The host computer collects depth information of the healthy reference side and the remaining part of the missing part by means of a depth information collection device, and collects a plurality of surface images of the healthy reference side at different angles by means of an image collection device;

[0018] The acquired depth information is used to perform three-dimensional reconstruction of the missing part to obtain a reconstructed three-dimensional structure model;

[0019] The three-dimensional structure model is subjected to structural segmentation of internal tissues using a pre-constructed second structural hierarchical neural network, and the model is modified using model attribute related information in the collected patient information and a pre-constructed third model modification neural network, to obtain a 3D model after processing;

[0020] The collected multiple surface images are expanded in two dimensions according to the 3D model to obtain a 2D expanded image.

[0021] In one embodiment of the present invention, the model attribute related information includes:

[0022] The patient's age, gender, height, weight and skin color; the model modification includes modifying the prosthesis size, shape, surface color and surface texture.

[0023] In one embodiment of the present invention, the process of obtaining a 2D unfolded image is to use a preset reverse filtering algorithm to compare the digital image of the surface of the 3D model with the printed image obtained by using a 2D printing device, compensate for the real visual difference between the digital image and its printed image, and obtain the 2D unfolded image to be printed; wherein the digital image includes: when the missing part has no symmetrical healthy reference side, the image of the surface image reconstructed from the missing part is unfolded in two dimensions according to the 3D model, or when the missing part has a symmetrical healthy reference side, the image of multiple collected surface images is unfolded in two dimensions according to the 3D model.

[0024] In one embodiment of the present invention, the process of the preset reverse filtering algorithm includes:

[0025] Step c1, reading initial parameters; wherein the initial parameters include image matrix size, similarity threshold and maximum number of iterations;

[0026] Step c2, acquiring the digital image;

[0027] Step c3, printing the digital image using a 2D printing device to obtain an original printed image;

[0028] Step c4, using the same image acquisition device used in the process of obtaining the digital image, performing image acquisition on the digital image and the original printed image to obtain corresponding acquired images respectively;

[0029] Step c5, calculating the reverse filtering function using the collected images corresponding to the current digital image and the original printed image respectively;

[0030] Step c6, applying the reverse filtering function to the current digital image to obtain a reverse filtered digital image;

[0031] Step c7, determining whether the current number of iterations is less than the maximum number of iterations; if so, executing step c8, if not, executing step c14;

[0032] Step c8, printing the currently obtained reverse filtered digital image using a 2D printing device to obtain a first printed image;

[0033] Step c9, using the same image acquisition device, performing image acquisition on the currently obtained reverse filtered digital image and the first printed image to obtain corresponding acquired images respectively;

[0034] Step c10, calculating a similarity index for the acquired images corresponding to the currently obtained reverse filtered digital image and the first printed image respectively;

[0035] Step c11, determining whether the similarity index is greater than the similarity threshold; if so, executing step c12, if not, executing step c13;

[0036] Step c12, determining the currently obtained reverse filtered digital image as the 2D expanded image to be printed;

[0037] Step c13, storing the currently obtained reverse filtered digital image and the similarity index, and replacing the current digital image with the currently obtained reverse filtered digital image, replacing the captured image corresponding to the current digital image with the captured image corresponding to the currently obtained reverse filtered digital image, and replacing the captured image corresponding to the current original printed image with the captured image corresponding to the current first printed image;

[0038] Step c14, searching for the reverse filtered digital image with the largest similarity index in all iterations, and using it as the 2D expanded image to be printed.

[0039] In one embodiment of the present invention, the host computer controls a 3D printing device to print the 3D model to obtain a prosthesis body, including:

[0040] The main computer controls the 3D printing device to print with materials matching each tissue in the 3D model to obtain a prosthesis body; wherein the materials matching each tissue include a variety of optional materials.

[0041] In one embodiment of the present invention, for each tissue, the matching material determines a suitable one from a plurality of optional materials for the tissue according to the collected patient information, so that the material has the hardness and viscoelasticity most similar to the patient's real human body.

[0042] In one embodiment of the present invention, the main computer controls the film-making equipment to make the surface film of the missing part, including:

[0043] The main computer controls the film-making device to obtain a sol-state silicone rubber solution, and places it on a printing paper with a preset thickness and a preset surface roughness to form a silicone rubber solution layer with a preset thickness;

[0044] The main computer controls the film-making equipment to cover the upper surface of the silicone rubber solution layer with a substrate sheet, and after the silicone rubber solution layer forms a gel, peels off the silicone rubber solution layer to obtain a surface film of the missing part, and the original contact surface between the surface film of the missing part and the printing paper serves as a printing surface; wherein the bonding strength of the surface of the substrate sheet to the silicone rubber is higher than the bonding strength of the surface of the printing paper to the silicone rubber, and the printing paper is suitable for the 2D printing equipment.

[0045] In one embodiment of the present invention, the preset surface roughness is determined according to the roughness of the patient's skin surface in the collected patient information; and the preset thickness is determined according to the thickness of the patient's skin in the collected patient information.

[0046] In one embodiment of the present invention, before the host computer controls the 2D printing device to print the 2D expanded image onto the surface film of the missing part to form a skin, the method further includes:

[0047] The main computer controls the printing system to treat the surface of the surface film of the missing part with a surface treatment liquid, and evenly prints the pigment ink after the treatment.

[0048] In one embodiment of the present invention, the main computer controls the assembly equipment to attach the skin to the surface of the prosthesis body, including:

[0049] The main computer controls the printing system to apply glue and lubricant to the surface of the prosthesis body, covers the skin on the coated surface of the prosthesis body, and controls the uniform pressurizing instrument to perform pressurization so that the skin and the prosthesis body are tightly fitted to form an integral structure.

[0050] In one embodiment of the present invention, the density and adhesion of the glue match the collected patient information; and the lubricant is a lubricant with biocompatibility.

[0051] In one embodiment of the present invention, the surface treatment comprises:

[0052] The main computer controls the inkjet system to coat the surface of the prosthesis body after the skin is attached with a protective layer similar to the surface of human skin.

[0053] In a second aspect, an embodiment of the present invention provides a fully computer-intelligent high-similarity pseudobody production system, the system comprising:

[0054] A main computer and other components; the main computer controls the other components to complete the method steps described in the first aspect; the other components include an image acquisition and storage system, a film making device, an assembly device, a printing system, a uniform pressurizing instrument, and a spray painting system; the image acquisition and storage system includes a depth information acquisition device and an image acquisition device; the printing system includes at least a 3D printing device and a 2D printing device.

[0055] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method provided in the first aspect of the present invention are implemented.

[0056] Beneficial effects of the present invention:

[0057] The fully computer-intelligent high-similarity prosthesis production method and system provided in the embodiments of the present invention controls the remaining components to complete the prosthesis design and production by a main computer, and all the above processes are computerized and intelligent as a whole, so that high-precision, high-similarity prostheses can be provided in a shorter time. The production process is digitized through computer operation, and there is no need to rely on manual experience or skills, which can greatly reduce working time and costs.

[0058] Moreover, the embodiment of the present invention introduces artificial intelligence technology and combines patient information for the design and modeling of the prosthesis. The use of artificial intelligence technology can not only obtain a high-precision and high-similarity 3D model containing the segmentation information of each tissue in the missing part of the patient for 3D printing output, but also realize the segmentation of the internal structure of the limb (such as bones, muscles, tendons, fat layers, skin layers, etc.) and the reconstruction of the missing part structure. It can also obtain a highly realistic 2D unfolded image of the surface layer of the 3D model for 2D printing output.

[0059] Through artificial intelligence algorithms, not only can the design and production of prostheses be completed using the healthy reference side as a data reference when there is a symmetrical healthy reference side for the missing part, but also, even when two symmetrical body parts (hands, arms, legs, feet, ears, etc.) are missing, the restoration of the limbs can be achieved, and a prosthesis model can be generated when a single part (nose, chin, etc.) is missing. Compared with traditional methods, prostheses can be tailored for individual patients more effectively, so that the prosthesis has a high degree of similarity with the patient's body, which can improve the quality of prosthetic products. The production of prostheses for missing parts is no longer restricted, and the needs of more patients can be met, which improves practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A schematic diagram of a process flow of a fully computer-intelligent high-similarity pseudobody production method provided by an embodiment of the present invention;

[0061] Figure 2a Schematic diagram of the training principle of the deep learning neural network model in an embodiment of the present invention;

[0062] Figure 2b A schematic diagram of the use principle of the deep learning neural network model in an embodiment of the present invention;

[0063] Figure 3 Schematic diagram of the reverse filtering algorithm in an embodiment of the present invention;

[0064] Figure 4 Schematic diagram of the process of manufacturing the surface film of the missing part in the embodiment of the present invention;

[0065] Figure 5 A schematic diagram of the working principle of a uniform pressurizing apparatus according to an embodiment of the present invention;

[0066] Figure 6a It is a schematic diagram of the working process of the fully computer-intelligent high-similarity prosthesis production system when the missing part has a symmetrical healthy reference side according to an embodiment of the present invention;

[0067] Figure 6b This is a schematic diagram of the working process of a fully computer-intelligent high-similarity prosthesis production system when there is no symmetrical healthy reference side for the missing part according to an embodiment of the present invention. DETAILED DESCRIPTION

[0068] The present invention is further described in detail below with reference to specific embodiments, but the embodiments of the present invention are not limited thereto.

[0069] In order to computerize and intelligentize the design and production process of the prosthesis in an overall and consistent manner, and to more effectively tailor high-similarity prostheses for individual patients compared to traditional methods, the embodiments of the present invention provide a fully computer-intelligent high-similarity prosthesis production method, system and computer-readable storage medium through the interaction between a main computer and other components and the control of the main computer over the other components.

[0070] The prosthesis in the embodiment of the present invention may include various artificial limbs and prostheses used by the human body, and the parts involved in the prosthesis may be hands, arms, legs, breasts, ears and the like.

[0071] In the first aspect, the embodiments of the present invention provide a fully computer-intelligent method for producing a highly similar pseudobody, such as Figure 1 As shown, the method may include the following steps:

[0072] S1, the main computer collects patient information and uses a pre-built artificial intelligence model to obtain a 3D model containing segmentation information of each tissue in the patient's missing part, and obtains a 2D unfolded image of the surface layer of the 3D model based on the collected patient information and the 3D model; wherein each tissue includes bones, muscles, tendons, fat layers, and skin layers;

[0073] The host computer in the embodiment of the present invention can be implemented by any existing computer, which has functions such as communication, control and data processing.

[0074] In an embodiment of the present invention, the main computer can control some acquisition devices to collect patient information, such as collecting relevant information of the patient's body three-dimensional structure and relevant information of the skin surface, etc., and use a pre-constructed (the pre-constructed referred to in the present invention refers to pre-established and trained) artificial intelligence model to obtain a 3D model containing segmentation information of various tissues in the patient's missing part. This 3D model can characterize the external structure and internal 3D structure data of the patient's missing part while segmenting and distinguishing various internal tissues, such as bones, muscles, tendons, fat layers, skin layers, etc., so as to facilitate the subsequent corresponding printing structure of these different internal tissues, realize the tissue differentiation inside the prosthesis body, so that the printed prosthesis body is closer to the patient's actual internal structure.

[0075] At the same time, based on the collected patient information and the 3D model, a 2D unfolded image of the surface of the 3D model can be obtained. The 2D unfolded image simulates the real skin appearance of the surface of the patient's 3D model area, so as to facilitate the subsequent printing of the 2D unfolded image and obtain a skin pattern similar to the patient's skin surface.

[0076] Specifically, many parts of the human body are symmetrically distributed, such as arms, legs, ears, etc., but there are also some parts that do not have a symmetrical distribution on the other side, such as the mouth and nose, or both symmetrical sides are missing. Therefore, the embodiment of the present invention mainly considers two situations, namely, the situation where the missing part has no symmetrical healthy reference side and the situation where the missing part has a symmetrical healthy reference side. The specific process of implementing S1 for these two situations is described below.

[0077] (1) In the case where the missing part has no symmetrical healthy reference side, the host computer collects patient information and uses a pre-built artificial intelligence model to obtain a 3D model containing segmentation information of each tissue in the missing part of the patient, and obtains a 2D unfolded image of the surface of the 3D model based on the collected patient information and the 3D model, including:

[0078] Step a1, the host computer reconstructs the neural network using the pre-constructed first model, performs three-dimensional reconstruction of the missing part by collecting information of the remaining parts of the patient, obtains a reconstructed three-dimensional structural model, and obtains a reconstructed surface image of the missing part;

[0079] In the case where the missing part has no symmetrical healthy reference side, for example, there is a symmetrical part but both sides are missing or there is no symmetrical part itself, it is impossible to directly use the data of the healthy reference side as a reconstruction reference for the missing side. Therefore, the embodiment of the present invention pre-constructs a first model reconstruction neural network for identifying and restoring the missing body shape. The first model reconstruction neural network collects three-dimensional information of the patient's remaining parts except the missing part, performs three-dimensional reconstruction of the missing part, and obtains a reconstructed three-dimensional structural model. At the same time, step a1 collects surface information of the patient's remaining parts except the missing part, and obtains a reconstructed surface image of the missing part for the reconstructed three-dimensional structural model, that is, an image of the three-dimensional structural surface of the missing part. For example, if the missing part is a forearm, the reconstructed surface image of the missing part is a surface image corresponding to one circumference of the forearm.

[0080] Wherein, the first model reconstruction neural network is implemented using generative AI (Artificial Intelligence) technology;

[0081] Specifically, the embodiment of the present invention establishes a human anatomy database (or called a human structure database), which contains human external morphological data related to age, gender, height, weight, skin color, etc., as well as three-dimensional structural information of the human body, specifically including three-dimensional information such as external structure and internal structure information, such as the distribution and position of bones, muscles, tendons, fat layers, and skin layers, as well as the viscoelasticity of various parts. By building a deep learning neural network model, and using the information A and information B obtained in large quantities from the database to match and train the built deep learning neural network model, the trained deep learning neural network model can output corresponding information B based on the input information A for specific purposes. For the training principle and use principle of the deep learning neural network model, please refer to Figure 2a and Figure 2b A brief description of .

[0082] Based on the concept of implementing the deep learning neural network model based on the human anatomy database, the embodiments of the present invention can implement deep learning neural network models for multiple purposes. It should be noted that the training methods of these deep learning neural network models are implemented based on the training methods of existing neural network models. The focus of the embodiments of the present invention is that for different purposes, that is, different input information A and output information B, corresponding training sets are constructed based on the corresponding data in the human anatomy database (the training samples in the training set correspond to information A as known samples, which carry label data corresponding to information B), so as to realize the training of the deep learning neural network model for this purpose. After the training is completed, the process of inputting information A and outputting information B can be realized for this purpose.

[0083] For example, the first model reconstructs the neural network, which is a deep learning neural network model. When the first model is used to reconstruct the neural network, the input information A of the first model reconstructs the neural network includes three-dimensional information of the patient's remaining parts except the missing parts, and the three-dimensional information can be collected using a depth information collection device.

[0084] Among them, the depth information acquisition equipment can be an X-ray machine, CT (Computed Tomography) equipment, MRI (Magnetic Resonance Imaging) equipment, ultrasonic echo imaging equipment, etc. These medical imaging devices can collect depth information of the human body through scanning and other methods. From these depth information, the three-dimensional shape of the human body and the structure, composition and corresponding position information of each internal part can be obtained.

[0085] In order to generate the missing part of the asymmetric healthy reference side, the embodiment of the present invention divides the entire body shape into different parts, such as ears, eyes, nose, chin, fingers, palms, forearms, upper arms, breasts, buttocks, upper limbs, lower limbs, soles of feet, etc. At this time, information B is one of the parts, and information A can be the integral of all the remaining parts. Therefore, the shape and three-dimensional information of the missing body part can be obtained by inputting the three-dimensional information of the remaining body parts into the trained deep learning neural network model (i.e., the first model reconstruction neural network), thereby obtaining a three-dimensional structural model (including external shape and internal structure) of the reconstructed missing part. For example, if a patient's left arm is missing, three-dimensional information of the patient's remaining parts can be collected and input into the trained first model reconstruction neural network. For the input information A, the first model reconstruction neural network will output a reconstructed three-dimensional structural model for the left arm as the corresponding information B. This reconstructed three-dimensional structural model represents the three-dimensional structure of the left arm. Since the first model reconstruction neural network is trained using a large number of patients' training samples and label data, it establishes a feature mapping relationship between input and output, which can enable the output reconstructed three-dimensional structural model to simulate the actual structure of the missing part of the patient to which the input information A belongs to the maximum extent, and achieve a good match with the remaining healthy parts of the patient, thereby improving the authenticity and structural similarity of the reconstruction.

[0086] The first model reconstruction neural network can adopt existing CNN, PointNet and other networks. It can be understood that for the three-dimensional reconstruction of the missing part, when the first model reconstruction neural network is trained, the training samples are derived from the three-dimensional information of the remaining healthy parts of a patient in the human anatomy database. These three-dimensional information can be preprocessed to a certain extent, such as dimensional unification, filtering, calculation of integrals, etc., to obtain training samples. The label data corresponding to the training samples is the three-dimensional information of the missing part, that is, the three-dimensional structure model. It can be understood that the construction of training sample-label data can be derived from the situation where the patient has a real missing part. For example, if the patient does have a missing left hand, the training sample is the three-dimensional information of the remaining parts except the left hand; the label data is the acquired three-dimensional information of the left hand, that is, the three-dimensional structure The model can be the three-dimensional information collected and retained before the patient's left hand is lost, or the three-dimensional information obtained by measuring the patient's left hand prosthesis, or the three-dimensional information of the patient's left hand obtained by other means; the construction of the training sample-label data can also be derived from the data of a healthy human body. For example, a large amount of three-dimensional information of various parts of a healthy human body is stored in the human anatomy database, and one of the parts can be regarded as a "missing part", such as the left ear. In this way, the training sample is obtained from the three-dimensional information of the remaining parts other than the left ear, and the three-dimensional information of the left ear is used as the label data. In this way, a large number of training sample-label data combinations for different missing parts can be obtained. Therefore, at least through the above-mentioned real missing and artificial missing methods, a number of training sample-label data combinations for different missing parts can be constructed, thereby forming a training set for training the first model to reconstruct the neural network. The embodiment of the present invention can combine the human anatomy data of a healthy human body and a human body with limb defects to construct the human anatomy database, enrich the types of data samples, and improve the robustness of the model.

[0087] When training the first model to reconstruct the neural network through the above training set, a certain loss function can be set, the network output result is compared with the label data of the training sample, and the network parameters are adjusted by back propagation and other methods, so that the loss function continues to converge after multiple iterations of training, that is, the network output result continues to approach the label data, and finally the first model reconstructed neural network can be obtained after training. The specific training process can be understood in combination with the conventional neural network training process.

[0088] Among them, the deep learning neural network model of the embodiment of the present invention, such as the pre-built first model reconstruction neural network, can be built in the main computer, and can also be built in other devices, such as being set in a processor, and the main computer calls the deep learning neural network model through a program. This is all reasonable and is not limited here.

[0089] As described above, in step a1, the host computer uses the pre-constructed first model to reconstruct the neural network, and collects information of the remaining parts of the patient to perform three-dimensional reconstruction of the missing part, so as to obtain a reconstructed three-dimensional structural model. Similarly, the pre-constructed first model can also be used to reconstruct the neural network to obtain a reconstructed surface image of the missing part.

[0090] To obtain a reconstructed surface image of the missing part, it is necessary to reconstruct the neural network using the pre-constructed first model to obtain a reconstructed three-dimensional structural model. At this time, it is also necessary to use the collected information on the rest of the patient's parts. The collected information on the rest of the patient's parts here refers to the collection of surface images of the rest of the patient's parts, that is, body surface information. This collection process can be carried out simultaneously with the collection before reconstructing the three-dimensional structural model, or it can be carried out separately.

[0091] Among them, collecting surface images of other parts of the patient can be achieved by using an image acquisition device, and the image acquisition device can be a camera (such as a digital camera), a video camera, a 3D scanner, etc., which is not limited here.

[0092] In order to improve the accuracy of the model output and obtain high-quality reconstructed surface images with high similarity to the patient's body surface, the surface images of the rest of the patient's body can be collected from multiple angles to enrich the data volume. The surface images of the rest of the patient's body also include images of the cross-sectional area.

[0093] The collected surface images of the remaining parts of the patient can be preprocessed to a certain extent. Similar to the three-dimensional reconstruction of the missing parts, the preprocessing here can also be image dimension unification, filtering, calculation of the integral of the surface images of all parts, etc. After the preprocessing, in the case of obtaining the reconstructed surface image of the missing part, the first model reconstructs part of the input information A corresponding to the neural network. At this time, the input information A should also have the three-dimensional information of the reconstruction of the missing part. After inputting the complete input information A into the first model reconstruction neural network, the reconstructed surface image of the missing part can be obtained, that is, the corresponding output information B is obtained.

[0094] That is to say, in an embodiment of the present invention, the neural network can be reconstructed using the pre-constructed first model, and the three-dimensional reconstruction of the missing part can be performed by collecting three-dimensional information of the remaining parts of the patient to obtain a reconstructed three-dimensional structural model. Then, based on the reconstructed three-dimensional structural model and the collected surface images of the remaining parts of the patient, the neural network can be reconstructed again using the first model to obtain a reconstructed surface image of the missing part.

[0095] In the above process, the first model used twice to reconstruct the neural network can have the same network structure, but with different network parameters, which are determined and saved through their respective training processes. When using the first model to reconstruct the neural network, a set of adapted network parameters can be called according to different input information A, and used for model recognition to output corresponding information B, thereby achieving the corresponding purpose. The above process can be implemented on the basis of existing AI technology, and the process of calling the adapted network parameters can be implemented using a computer program, which will not be elaborated in detail here.

[0096] In order to obtain the surface image reconstructed from the missing part, when the first model reconstructs the neural network, the training samples are derived from the surface images of the remaining healthy parts of a patient in the human anatomy database, and also include the three-dimensional information of the reconstruction of the missing part of the patient. Of course, the training samples can be pre-processed. The label data corresponding to the training sample is the surface image reconstructed from the missing part of the patient. Similar to the three-dimensional reconstruction of the missing part, it can be understood that the construction of the training sample-label data can be derived from the data of the patient with a real missing part, or from the data of a healthy human body. At least through the real missing and artificial missing methods, a data combination of several training samples-label data for different missing parts in this situation is constructed, thereby forming a training set for training the first model reconstruction neural network at this time. The specific training process can be understood in combination with the reconstruction of the missing part and the conventional neural network training process, and will not be described in detail here.

[0097] In a simplified way, the surface image reconstructed by the missing part can be obtained by utilizing the information of the rest of the patient's body, which may be only the surface image near the cross-sectional position of the missing part. It is also reasonable to obtain the surface image reconstructed by the missing part by extending the image.

[0098] An example of using the first model reconstruction neural network is given above, that is, the three-dimensional reconstruction of the missing part and the reconstruction of the surface image of the missing part are carried out successively. In an optional implementation, these two links can also be completed in one step. That is, the first model reconstruction neural network training is to use the three-dimensional information of the healthy parts of the patient in the human anatomy database, the surface images of the other healthy parts of the patient to construct samples, and the three-dimensional information of the missing parts and the surface images are used to construct label data for model training, so that after the training is completed, the three-dimensional information of the remaining parts of a patient except the missing parts (obtained by using a depth information acquisition device) and the surface images of the remaining parts of the patient collected are input, and the first model reconstruction neural network will output a three-dimensional structural model of the reconstruction of the missing part, which carries the surface image of the reconstruction of the missing part. The embodiment of the present invention can use any of the above methods to achieve the three-dimensional reconstruction of the missing part and the reconstruction of the surface image of the missing part, which is not limited here. However, it should be emphasized that, under different circumstances and with different combinations of input information A and output information B, the training process using a neural network is similar to the existing neural network training process, which is the embodiment of the functional principle of the neural network. The embodiment of the present invention focuses on constructing a large number of training samples and label data from the human anatomy database according to the purpose represented by the input information A and the output information B to complete the model training process, but this process can also refer to the existing neural network training principles and be understood in combination with the scenario purpose of the embodiment of the present invention, and will not be explained in more detail here.

[0099] Step a2, using a pre-constructed second structural hierarchical neural network to perform structural segmentation of internal tissues on the three-dimensional structural model, and using the model attribute related information in the collected patient information and a pre-constructed third model modification neural network to perform model modification, and obtain a 3D model after processing;

[0100] In the prior art, on the one hand, artificially manufactured prosthesis bodies are often limited to making the appearance of the prosthesis close to the real body when it is still. Therefore, the prepared prostheses are all drawn on the surface without internal structure, usually in a seamless structure without distinction of internal tissues, which makes the authenticity of the prosthesis body poor. After the prosthesis is installed on the missing part of the patient, the section of the missing part and the prosthesis are only installed in contact with the section, and the various tissues in the section, such as bones, cannot be continued in the prosthesis, making the authenticity, flexibility and usability of the prosthesis poor. On the other hand, in order to facilitate industrialization, the existing prosthesis body design and manufacturing often use the same prosthesis body for all patients, and configure a limited number of prosthesis bodies as sales products at most, so that patients can choose the one that is closer to them. If customization is to be achieved, it needs to be done manually by highly professional personnel, which cannot meet the current needs.

[0101] In response to the above-mentioned problem of the first aspect, the embodiment of the present invention performs structural segmentation of the internal tissues of the three-dimensional structural model reconstructed from the missing part, so that the positions of the various tissues in the three-dimensional structural model, such as bones, muscles, tendons, fat layers, skin layers, etc., can be distinguished. The purpose is to subsequently locate the positions of different internal tissues and select materials for corresponding printing structures, so that the internal tissues of the prosthesis main body can be distinguished. This not only makes the printed prosthesis main body closer to the patient's actual internal structure, but also can achieve precise alignment and matching with the tissue position corresponding to the cross-section of the missing part, so that the prosthesis can be close to the real human body during movement.

[0102] Specifically, the embodiment of the present invention pre-constructs a second structural hierarchical neural network for structural segmentation of internal tissues. This second structural hierarchical neural network is similar to the previous first model reconstruction neural network, and both are trained using data from the human anatomy database. The selected training samples contain a three-dimensional structural model (i.e., three-dimensional information) of the reconstructed missing part, in which the external morphology and internal structure can be seen, but the bones, muscles and other tissues cannot be distinguished. The selected training samples also contain at least three-dimensional information of the cross-sectional position of the missing part, or further three-dimensional information of the part connected to the missing part. For example, if the patient's left side is missing below the elbow joint, the training sample contains a three-dimensional structural model of the reconstructed missing part below the left elbow joint, and also contains three-dimensional information of the left elbow joint section and even the upper arm part above the section. The label data corresponding to the training sample is the segmentation information of each tissue in the training sample, that is, the three-dimensional information after the internal tissue structure is segmented for the three-dimensional structure model reconstructed for the missing part. This part of information represents the three-dimensional shape of the internal structure. Taking the above example as an example, the three-dimensional information of each internal tissue is obtained after the three-dimensional structure model of the missing part below the left elbow joint is manually marked or segmented and marked by other methods. The three-dimensional information of each segmented tissue is obtained by referring to the section or even further referring to the three-dimensional information of the remaining part of the upper arm above the elbow joint connected to the section. In this way, the shape and position of each tissue segmented in the three-dimensional structure model of the missing part below the left elbow joint matches the shape and position of the corresponding tissue at the section, and the shape and position of the corresponding tissue in the remaining part of the upper arm connected to the section. For example, for bones, the segmented three-dimensional information includes three-dimensional information such as the thickness, shape, and direction of the bones. Any bone and the corresponding bone in the section, and the corresponding bone in the upper arm above the section are aligned, so that the direction, shape, and size of the same bone from top to bottom conform to human anatomy, and no abnormal mutations and dislocations will occur.

[0103] Similarly, in the process of training the second structure hierarchical neural network, the construction of training sample-label data can be derived from data of patients with real missing parts, or from data of healthy human bodies. At least through real missing parts and artificial missing parts, a data combination of several training sample-label data for different missing parts in this situation is constructed, thereby forming a corresponding training set for training the second structure hierarchical neural network at this time. Specifically, it can be understood in combination with the relevant description of the first model to reconstruct the neural network and the principles and training process of the existing neural network, which will not be explained in detail here.

[0104] It should be noted that the second structure hierarchical neural network of the embodiment of the present invention can be applicable to the case where the missing part has no symmetrical healthy reference side, such as nose missing, or both arms missing, etc., and can also be applicable to the case where the missing part has symmetrical healthy reference side, such as one arm is healthy and the other is missing, etc. In these two cases, the training mechanism of the second structure hierarchical neural network is the same, the model can be completely the same, or the model parameters can be optimized according to different situations, which is reasonable.

[0105] In view of the above-mentioned second problem, the shape of the prosthesis should be adjusted according to the specific situation of the patient. In order to accurately restore individual differences, the embodiment of the present invention once again uses an artificial intelligence algorithm to modify the model using the model attribute related information in the collected patient information and a pre-built third model modification neural network. In other words, after the embodiment of the present invention performs two aspects of internal tissue structure segmentation and model modification on the three-dimensional structure model reconstructed from the missing part, a 3D model is finally obtained.

[0106] Similarly, the third model modified neural network of the embodiment of the present invention can be applicable to the case where the missing part has no symmetrical healthy reference side and the case where the missing part has a symmetrical healthy reference side.

[0107] The embodiment of the present invention does not limit the order of the two aspects of processing, i.e., ① the three-dimensional structural model reconstructed from the missing part can be firstly segmented by using the pre-constructed second structural hierarchical neural network for the internal tissue structure, and then the model attribute related information in the collected patient information and the pre-constructed third model modification neural network can be used for model modification, and finally a 3D model can be obtained after processing; or, ② the three-dimensional structural model reconstructed from the missing part can be firstly modified by using the model attribute related information in the collected patient information and the pre-constructed third model modification neural network, and then the structure segmentation of the internal tissue can be performed by using the pre-constructed second structural hierarchical neural network, and finally a 3D model can be obtained after processing; or, ③ the two aspects of processing can be performed in parallel, and the models processed from the two aspects can be fused and processed, and finally a 3D model can be obtained. These three different participation times of the third model modification neural network are all reasonable. The embodiment of the present invention can adopt any one of the methods, or select the best one from the three methods.

[0108] The model attribute related information includes:

[0109] The patient's age, gender, height, weight and skin color; the model modification includes modifying the prosthesis size, shape, surface color and surface texture. The purpose of model modification is to make the 3D model more consistent with the patient's personal situation.

[0110] Specifically, when the third model is used to modify the neural network, the input information A contains the three-dimensional structure model to be processed and the model attribute related information, and the output information B is the modification factors of the prosthesis size, shape, surface color and surface texture, such as the modification factors of bone thickness, muscle volume, fat layer thickness, skin color and texture, etc. The three-dimensional structure model to be processed is further modified by the obtained modification factors, and the above modification purpose can be achieved. The embodiment of the present invention can use a large amount of data to predefine the modification standards, such as the modification factor standards of various aspects of the human body under different ages, different skin colors and other attributes. The modification factors are used to modify the external and internal structures. For example, the model attribute related information contains that the patient is a 70-year-old Asian female, etc. The modification factors can be described by numbers or words, such as bone height ML_height_of_bone=0.82, bone width ML_width_of_bone=0.95, skin color Mcolor_RGB=0.9,1.2,1.1, etc. Then, the modification factors can be used in the prosthesis manufacturing process, such as 3D printing and inkjet coloring.

[0111] In the process of using the third model to modify the neural network processing, the three-dimensional structure model to be processed is different according to the different time when the third model modifies the neural network. Correspondingly, the training samples and label data of the training process are also different.

[0112] For example, with respect to the participation timing indicated by ① above, the three-dimensional structural model to be processed by the third model modification neural network is a three-dimensional structural model obtained after the three-dimensional structural model reconstructed from the missing parts is segmented by the internal tissue structure using the pre-built second structural hierarchical neural network; the third model modification neural network outputs a modification factor, and a 3D model can be obtained by adjusting the three-dimensional structural model to be processed using the modification factor.

[0113] In the case of ① mentioned above, the sample data of the training process includes a large number of three-dimensional structural models reconstructed from missing parts obtained based on the human anatomy database, which are obtained after internal tissue structure segmentation. The sample data also includes information related to model attributes in the corresponding patient information; the label data includes the corresponding modification factors; the construction of the training sample-label data can be derived from data of patients with real missing parts, or from data of healthy human bodies, at least through real and artificial missing methods, a data combination of several training sample-label data for different missing parts in this situation is constructed, thereby forming a training set for training the third model modified neural network at this time. The network selection of the third model modified neural network is not limited, and can be specifically combined with the relevant description of the first model reconstruction neural network and other networks as well as the principles and training process of existing neural networks, which will not be explained in detail here.

[0114] For the participation opportunity indicated by ② above, the three-dimensional structural model to be processed by the third model modification neural network is the three-dimensional structural model reconstructed from the missing part; the third model modification neural network outputs the modification factor, and after adjusting the three-dimensional structural model to be processed by using the modification factor, the three-dimensional structural model after model modification is obtained, and the structure of the internal organization is segmented by using the pre-constructed second structural hierarchical neural network to obtain the 3D model. In this case, the sample data and label data of the training process can be understood in accordance with the model usage process, and can be combined with the above ① situation and the relevant description of the first model reconstruction neural network and other networks, the principles of existing neural networks and the understanding of the training process, which will not be described in detail here.

[0115] For the participation opportunity indicated by ③ above, the three-dimensional structural model to be processed by the third model modification neural network is the three-dimensional structural model reconstructed from the missing part; the third model modification neural network outputs the modification factor, and after adjusting the three-dimensional structural model to be processed using the modification factor, the three-dimensional structural model after model modification is obtained; the three-dimensional structural model after model modification will be combined with the three-dimensional structural model reconstructed from the missing part using the pre-built second structural hierarchical neural network to perform structural segmentation of the internal tissue to obtain the three-dimensional model, and the 3D model will be obtained through model fusion and other processing methods. The specific method of the fusion processing of the two models can be implemented based on the existing technology. In this case, the sample data and label data of the training process can be understood in accordance with the model usage process, and can be combined with the above ① situation and the relevant description of the first model reconstruction neural network and other networks, the principles of existing neural networks and the understanding of the training process, which will not be described in detail here.

[0116] Step a3: Perform two-dimensional image expansion on the surface image reconstructed from the missing part according to the 3D model to obtain a 2D expanded image.

[0117] As mentioned above, the surface image reconstructed from the missing part is the surface image of the three-dimensional structure of the missing part. Therefore, it represents a three-dimensional surface image. The existing relevant technologies, such as Adobe Illustrator or 3D printer software and other common software, can be used to directly unfold this three-dimensional surface image in two dimensions, similar to the way of unfolding a scroll, to obtain a two-dimensional image, and after certain processing, a 2D unfolded image is obtained. The 2D unfolded image is an image used for printing by a 2D printing device.

[0118] Furthermore, in order to output a more realistic 2D expanded image for printing, an inverse filtering algorithm is designed in the embodiment of the present invention for the case where the missing part has no symmetrical healthy reference side and the case where the missing part has a symmetrical healthy reference side. For the sake of clarity of the architecture, this part will be explained in the following text.

[0119] (2) In the case where the missing part has a symmetrical healthy reference side, the host computer acquires patient information and uses a pre-built artificial intelligence model to obtain a 3D model containing segmentation information of each tissue in the missing part of the patient, and acquires a 2D unfolded image of the surface of the 3D model based on the acquired patient information and the 3D model, including:

[0120] Step b1, the host computer collects depth information of the healthy reference side and the remaining part of the missing part by means of a depth information collection device, and collects multiple surface images of the healthy reference side at different angles by means of an image collection device;

[0121] Specifically, the host computer controls the depth information acquisition device to acquire the depth information of the patient's healthy reference side and the remaining part of the missing part, that is, three-dimensional information. The depth information acquisition device is as described above.

[0122] Since the symmetrical healthy reference side is used to reconstruct the three-dimensional structure of the missing side in this case, it is necessary to collect the depth information of the healthy reference side as a reference. Similar to the case where there is no symmetrical healthy reference side for the missing part, it is also necessary to collect at least the three-dimensional information of the cross-sectional position of the missing part, or further the three-dimensional information (i.e., depth information) of the part connected to the missing part, so that the boundary area structure between the remaining and missing parts of the patient's body can be obtained for designing the wear part of the prosthesis body.

[0123] In step b1, the main computer controls the image acquisition device to acquire multiple surface images of the healthy reference side at different angles. The image acquisition device is as described above.

[0124] Step b2, using the collected depth information to perform three-dimensional reconstruction of the missing part to obtain a reconstructed three-dimensional structure model;

[0125] For example, if a patient's left arm is missing below the elbow joint, while the right arm is intact, in this case, the right arm is used to reconstruct a three-dimensional structural model of the missing part of the left arm below the elbow joint.

[0126] Then the complete three-dimensional information of the right arm can be collected, because this part of information reflects the complete three-dimensional appearance structure of the right arm. At least this part of the three-dimensional information can be used to reconstruct the left arm in three dimensions, and a reconstructed three-dimensional structural model of the left arm can be obtained.

[0127] In the above three-dimensional reconstruction process, since there is a symmetrical reference side, we can use the existing algorithms and models based on the principle of symmetry and in combination with human anatomy to construct a three-dimensional structural model of the reconstructed left arm. This reconstructed three-dimensional structural model is symmetrical with the right side in appearance and structure.

[0128] Step b3, using a pre-constructed second structural hierarchical neural network to perform structural segmentation of internal tissues on the three-dimensional structural model, and using the model attribute related information in the collected patient information and a pre-constructed third model modification neural network to perform model modification, and obtain a 3D model after processing;

[0129] In step b3, the three-dimensional structural model reconstructed on the missing side and the three-dimensional information of at least the cross section of the missing part are further processed to obtain the input information A of the second structural hierarchical neural network at that location, and the output information B is the three-dimensional structural model after the internal tissue structure of the input three-dimensional structural model is segmented.

[0130] The processing process of step b3 is similar to that of step a2. For the specific process, please refer to step a2 for understanding.

[0131] It should be noted that for the case where the missing part has no symmetrical healthy reference side and for the case where the missing part has a symmetrical healthy reference side, the second structure hierarchical neural network can have the same network structure but different network parameters, and the network parameters are determined and saved through the training process of the two cases respectively. For these two different cases, when using the second structure hierarchical neural network, the network parameters of the corresponding case can be called to output the corresponding information B for the input information A of the case. The process of calling the adaptation network parameters can be implemented using a computer program.

[0132] Similar to the second structure hierarchical neural network, the third model modified neural network can also be the same network structure but with different network parameters for the case where the missing part has no symmetric healthy reference side and the case where the missing part has a symmetric healthy reference side, and can be called accordingly when used.

[0133] Step b4, performing two-dimensional image expansion on the multiple surface images collected according to the 3D model to obtain a 2D expanded image.

[0134] In the case where the missing part has a symmetrical healthy reference side, the multiple surface images collected can be used according to the 3D model, combined with the principle of symmetry, to obtain a surface image of the three-dimensional structure of the missing part using existing methods, and then the same two-dimensional image expansion method as step a3 is used to directly expand the three-dimensional surface image in two dimensions to obtain a two-dimensional image, and after certain processing, a 2D expansion image is obtained. The 2D expansion image is an image for printing with a 2D printing device. The 2D printing device can be a 2D printer, etc.

[0135] One of the features of the present application is that it provides an automatic surface printing technology. In this technology, the surface image of the missing part will be printed onto the manufactured surface membrane of the missing part using a 2D printing device to form a skin, and the surface membrane of the missing part is a simulated skin membrane. Due to the differences in color filters between the camera and other devices for acquiring the surface image and the 2D printing device, it is inevitable that the printed color, texture and other aspects will be slightly different from the real one. In order to output a more realistic 2D digital image for printing, the embodiment of the present invention jointly designs a reverse filtering algorithm for the case where the missing part has no symmetrical healthy reference side and for the case where the missing part has a symmetrical healthy reference side. The reverse filtering algorithm is applied to the digital image before printing to solve the color difference problem between the printed image and the real object. The main implementation principle of the algorithm is briefly described below.

[0136] 1. Use a digital camera to capture an image of the test sample and obtain a test image, which can be expressed as:

[0137] I test =i test *f camera ;

[0138] The test sample here generally refers to a physical object, which may be a human body. It is understandable that the test image is a digital image.

[0139] In the embodiment of the present invention, i(x, y) is used to represent the visual image, I(x, y) is used to represent the digital image, (x, y) refers to the coordinates of the pixel; f(x, y) is used to represent the filter function; in the above formula, I test represents the test image; i test Represents the visual image of the test sample; f camera represents the filter function of the digital camera;

[0140] 2. Test image I test Print with a 2D printing device and get the corresponding printed image, which can be expressed as:

[0141] i test,print =I test *f printer =i test *f camera *f printer ;

[0142] Among them, i test,print Represents the test image I test Printed image obtained by printing with 2D printing equipment; f printer Represents the filter function of the 2D printing device.

[0143] 3. Use the same digital camera to take pictures of the test image and its printed image at the same time, and you can get the following relationship:

[0144] I test =i test *f camera ;

[0145] I test,print =i test,print *f camera =i test *f printer *f camera 2

[0146] Among them, I test,print It is a digital image obtained by photographing the printed image of the test image with a digital camera;

[0147] 4. Using the digital image obtained by taking the test image and its printed image with a digital camera, calculate the correction coefficient, expressed as:

[0148]

[0149] in, represents a two-dimensional deconvolution operation, and * represents a two-dimensional convolution operation.

[0150] It can be understood that the test image and its printed image are both digital images obtained after being photographed with a digital camera. The correction coefficient is calculated by the ratio of the two. α(x, y) represents the correction coefficient corresponding to the pixel (x, y).

[0151] 5. Use the same digital camera to capture the target human body part image, expressed as:

[0152] I body =i body *f camera ;

[0153] Among them, I body Represents the target human body part image, which is also a digital image; i body Indicates the corresponding visual image.

[0154] 6. Use the correction coefficient to transform the target human body part image I body After correction, the 2D printing device is used to print and obtain the corresponding printed image, which is expressed as:

[0155] i print,out =I body *α(x,y)*f printer ;

[0156] Among them, i print,out For the target human body part image, the final printed image is i print,out The compensation by the correction factor is equal to i body .

[0157] It can be seen that in the process of digital image acquisition and printing, the deviation between the printout and the real original image is inevitable, but the embodiment of the present invention measures the correction coefficient in the digital image. Finally, the correction coefficient in the computer is multiplied before printing, and the printout result becomes a real image with the difference pre-compensated.

[0158] Based on the above main implementation principles, the reverse filtering algorithm is specifically described.

[0159] According to the embodiment of the present invention, the process of obtaining the 2D unfolded image is to use a preset reverse filtering algorithm to compare the digital image of the surface layer of the 3D model with the printed image obtained by using a 2D printing device, and compensate for the difference between the digital image and the printed image in real vision, so as to obtain the 2D unfolded image to be printed;

[0160] The digital image includes: when the missing part has no symmetrical healthy reference side, the image is obtained by two-dimensionally expanding the surface image reconstructed from the missing part according to the 3D model; or, when the missing part has a symmetrical healthy reference side, the image is obtained by two-dimensionally expanding multiple collected surface images according to the 3D model.

[0161] For details, see Figure 3 In the embodiment of the present invention, the process of the preset reverse filtering algorithm includes:

[0162] Step c1, reading initial parameters;

[0163] Wherein, the initial parameters include image matrix size, similarity threshold and maximum number of iterations;

[0164] The image matrix size refers to the image matrix size of digital images and printed images, denoted by IM; the similarity threshold is denoted by ST; the number of iterations is denoted by a positive integer i, i=1 for the first iteration; the maximum number of iterations is denoted by IL, IL≥2.

[0165] Step c2, acquiring the digital image;

[0166] As mentioned above, in the case where the missing part has no symmetrical healthy reference side, the digital image is an image obtained by performing a two-dimensional image expansion on the surface image reconstructed from the missing part according to the 3D model;

[0167] In the case where the missing part has a symmetrical healthy reference side, the digital image is an image obtained by performing two-dimensional image expansion on the multiple surface images collected according to the 3D model.

[0168] For ease of understanding, Figure 3 In the figure, the digital image is represented by (aa1).

[0169] Step c3, printing the digital image using a 2D printing device to obtain an original printed image;

[0170] For ease of understanding, Figure 3 In FIG. 1 , the original printed image is represented by (AA1).

[0171] Step c4, using the same image acquisition device used in the process of obtaining the digital image, performing image acquisition on the digital image and the original printed image to obtain corresponding acquired images respectively;

[0172] The image acquisition device can be a digital camera, and the front and back images are the same.

[0173] For ease of understanding, Figure 3In the method, the digital image (aa1) and the original printed image (AA1) are captured to obtain corresponding captured images (aa2) and (b), respectively.

[0174] Step c5, calculating the reverse filtering function using the collected images corresponding to the current digital image and the original printed image respectively;

[0175] That is, the reverse filtering function is calculated using (aa2) and (b), which can be expressed as:

[0176]

[0177] in, Represents a two-dimensional deconvolution operation, Figure 3 In the relevant description of the part, * represents a two-dimensional convolution operation, and (c) represents a reverse filtering function.

[0178] Combined with the introduction of the main implementation principle of the reverse filtering algorithm in the previous text, the role of the reverse filtering function calculated in step c5 is equivalent to the correction coefficient in the previous text.

[0179] Step c6, applying the reverse filtering function to the current digital image to obtain a reverse filtered digital image;

[0180] Specifically, the reverse filtering function (c) is applied to the current digital image (aa1) to obtain the reverse filtered digital image (d1), which is represented by:

[0181] (aa1)*(c)=(d1);

[0182] Step c7, determining whether the current number of iterations is less than the maximum number of iterations;

[0183] Specifically, determine whether the current number of iterations i is less than the maximum number of iterations IL, if so, execute step c8, if not, execute step c14;

[0184] Step c8, printing the currently obtained reverse filtered digital image using a 2D printing device to obtain a first printed image;

[0185] See also Figure 3 , the currently obtained reverse filtered digital image (d1) is printed using a 2D printing device to obtain a first printed image, which is represented by (D1).

[0186] Step c9, using the same image acquisition device, performing image acquisition on the currently obtained reverse filtered digital image and the first printed image to obtain corresponding acquired images respectively;

[0187] Figure 3In the embodiment, the currently obtained reverse filtered digital image and the first printed image are subjected to image acquisition to obtain corresponding acquired images (aa3) and (d2), respectively.

[0188] Step c10, calculating a similarity index for the acquired images corresponding to the currently obtained reverse filtered digital image and the first printed image respectively;

[0189] Specifically, the similarity index SI is calculated for (aa3) and (d2). The similarity index SI can be implemented by any existing method of calculating image similarity, such as PSNR (Peak Signal-to-Noise Ratio), SSIM (structural similarity), MSE (Mean Squared Error), cosine similarity, histogram, mutual information, hash similarity, etc., without limitation. For example, the similarity index SI can be a two-dimensional correlation coefficient.

[0190] Step c11, determining whether the similarity index is greater than the similarity threshold;

[0191] Specifically, determine whether the similarity index SI is greater than the similarity threshold ST; if so, execute step c12; if not, execute step c13;

[0192] Step c12, determining the currently obtained reverse filtered digital image as the 2D expanded image to be printed;

[0193] Specifically, determining the currently obtained reverse filtered digital image (d1) as the 2D expanded image to be printed means that the current first printed image (D1) obtained by printing meets the requirements, and (D1) can be used as the final output.

[0194] Step c13, storing the currently obtained reverse filtered digital image and the similarity index, and replacing the current digital image with the currently obtained reverse filtered digital image, replacing the captured image corresponding to the current digital image with the captured image corresponding to the currently obtained reverse filtered digital image, and replacing the captured image corresponding to the current original printed image with the captured image corresponding to the current first printed image, and then returning to step c5 for a new iteration;

[0195] Specifically, store the current reverse filtered digital image (d1), similarity index SI and current iteration number i; then make the following changes: (aa1) = (d1); (aa2) = (aa3); (b) = (d2); i = i+1; then return to step c5 for a new iteration.

[0196] Step c14, searching for the reverse filtered digital image with the largest similarity index in all iterations, and using it as the 2D expanded image to be printed.

[0197] After step c12 and step c14, the 2D expanded image to be printed can be printed using a 2D printing device. It can be understood that the obtained printed image is the corresponding (D1).

[0198] The reverse filtering algorithm designed in the embodiment of the present invention can eliminate the color difference between the printed image and the real object, making the printed image more authentic.

[0199] S2, the host computer controls the 3D printing device to print the 3D model to obtain the prosthesis body;

[0200] The 3D printing device may be a 3D printer or the like.

[0201] As mentioned above, in the 3D model, the 3D model contains the three-dimensional information of the missing part, and the internal tissues such as bones, muscles, tendons, fat layers, skin layers, etc. have been segmented. Therefore, the three-dimensional information of each tissue such as shape, structure, size, and position is known, and 3D printing can be achieved using 3D printing equipment; wherein each part can be printed separately, and the main computer controls the assembly equipment to assemble the prosthesis body, that is, the prosthesis body contains the printing results of each layered part of the missing part.

[0202] In an optional implementation manner, S2 may include:

[0203] The main computer controls the 3D printing device to print with materials matching each tissue in the 3D model to obtain a prosthesis body; wherein the materials matching each tissue include a variety of optional materials.

[0204] Considering the use of the prosthesis body, the materials of each tissue should be carefully selected. The embodiment of the present invention can refer to the general (mechanical) parameter values ​​of the human body provided in the manual and literature in advance, such as the physical parameter values ​​of a large number of human tissues (hardness and viscoelasticity, etc.), and pre-set a variety of optional materials for each tissue. These materials must also be able to be reused for a long time and have biocompatibility. When printing a 3D model, for each tissue, select a material from the multiple optional materials for each tissue to print, for example, you can randomly select one, and when all tissues are printed, you will get a printed prosthesis body.

[0205] Furthermore, the material of each tissue can be selected in a targeted manner in combination with the patient's personal information. Specifically, for each tissue, the matching material determines a suitable one from a variety of optional materials for the tissue based on the collected patient information, so that it has the hardness and viscoelasticity that are most similar to the current patient's real human body. In this case, the physical parameter values ​​of the patient's tissues such as hardness and viscoelasticity can be collected in any step before printing the 3D model. Then, for each tissue, a suitable material can be selected from a variety of optional materials for each tissue using the collected patient's physical parameter values ​​to print. For example, bones, muscles, fat layers and other parts can be printed with plastic materials, but different tissues can select different plastic materials according to the patient's physical parameter values. Moreover, as subcutaneous parts, the size of bones, muscles, tendons, and fat layers can be adjusted according to the thickness of the skin.

[0206] S3, the main computer controls the film-making equipment to make a surface film at the missing part;

[0207] The surface membrane of the missing part is a simulated skin membrane, that is, a membrane that simulates real human skin. The mechanical properties of the membrane (such as hardness and viscoelasticity) must match those of the remaining part of the patient's limb and be biocompatible. Silicone rubber is one of the candidate materials.

[0208] In an optional implementation, see Figure 4 As shown in the schematic diagram of the process of manufacturing the surface film of the missing part, S3 can include the following steps:

[0209] S31, the main computer controls the film-making device to obtain a sol-state silicone rubber solution, and places it on a printing paper with a preset thickness and a preset surface roughness to form a silicone rubber solution layer with a preset thickness;

[0210] The film-making device can be implemented based on existing process equipment, and has a platform for placing printing paper. The printing paper has a preset thickness and a preset surface roughness. The preset thickness can be a thickness value determined based on the thickness of a large number of human skins, and the preset surface roughness can also be a roughness value determined based on the roughness of a large number of human skins, so as to adapt to most human bodies.

[0211] In an optional implementation, in order to customize the prosthesis for the patient who currently needs it, the preset surface roughness is determined according to the roughness of the patient's skin surface in the collected patient information, and the preset thickness is determined according to the thickness of the patient's skin in the collected patient information. In this case, the roughness of the patient's skin surface and the thickness of the skin need to be collected before S3 as the preset surface roughness and preset thickness, respectively.

[0212] After the film-making equipment is controlled to obtain the sol-state silicone rubber solution, it is placed on the above-mentioned printing paper to cover the surface of the printing paper with a uniform thickness to form a silicone rubber solution layer of a set thickness. Figure 4 In the lower part of the leftmost image, the burrs on the printed paper indicate the rough surface, and the blue part is the formed silicone rubber solution layer.

[0213] S32, the main computer controls the film-making equipment to cover the upper surface of the silicone rubber solution layer with a substrate sheet, and after the silicone rubber solution layer forms a gel, peels off the silicone rubber solution layer to obtain a surface film of the missing part, and the original contact surface between the surface film of the missing part and the printing paper serves as the printing surface; wherein, the bonding strength of the surface of the substrate sheet to the silicone rubber is higher than the bonding strength of the surface of the printing paper to the silicone rubber, and the printing paper is suitable for the 2D printing equipment.

[0214] Figure 4 The bottom paper pad is the substrate sheet. The surface of the substrate sheet has a higher bonding strength to the silicone rubber than the surface of the printing paper, which facilitates subsequent peeling and removal of the printing paper.

[0215] After the upper surface of the silicone rubber solution layer is covered with a substrate sheet, a Figure 4 The structure of the middle part. After the silicone rubber solution layer becomes gel, the silicone rubber solution layer is peeled off from the surface of the printing paper and turned over to obtain Figure 4 The structure of the right part is to obtain the surface film of the missing part, and the original contact surface between the surface film of the missing part and the printing paper is used as the printing surface. Through the above treatment, a silicone rubber film with the same surface roughness as the printing paper can be produced. Subsequently, the surface film of the missing part can be patterned and colored using ordinary 2D printing equipment to form a skin.

[0216] S4, the host computer controls the 2D printing device to print the 2D expanded image onto the surface film of the missing part to form a skin;

[0217] In an optional implementation manner, before executing S4, the method of the present invention further includes:

[0218] The main computer controls the printing system to treat the surface of the surface film of the missing part with a surface treatment liquid, and evenly prints the pigment ink after the treatment.

[0219] The printing system may include a 2D printing device and a 3D printing device, and may also include some other devices. It can be implemented based on a combination of some existing devices and can achieve the functions described in this application. No detailed description is given here.

[0220] Treating the surface of the surface film at the missing part with a surface treatment liquid is to physically / chemically treat the surface of the surface film at the missing part to increase the affinity with the ink. The surface treatment liquid can be implemented by using some existing liquids for surface treatment, which is not limited here.

[0221] After the treatment, the printing system will evenly print the pigment ink on the surface of the missing part surface film after the surface treatment liquid is treated, so that the 2D expanded image can be printed on the missing part surface film with the pigment ink in the subsequent S4 to form a skin with a pattern.

[0222] When S4 is specifically executed, the 2D printing device may be an ordinary inkjet printer. It can be seen that the above process in the embodiment of the present invention makes it possible to use an ordinary inkjet printer to print on the silicone film, which greatly reduces the production cost.

[0223] Furthermore, due to the low ink permeability of silicone rubber, pigment ink is more suitable for general printing than dye ink. After S4 printing, a layer of protective ink needs to be applied to the printed surface while ensuring that it has the same texture as real human skin. This part is the final coloring process, the purpose of which is to achieve similarity with the real body.

[0224] Silicone rubber is widely used as a prosthetic material due to its physical, chemical and biological inertness, and its stability in daily life environments. However, due to its inertness, it is not easy for ordinary inkjet printers or laser printers to print directly on the surface of silicone rubber. The embodiment of the present invention proposes a simple and practical technology to achieve high-quality silicone rubber film printing. By treating the surface of the silicone rubber film with a surface treatment liquid and evenly applying pigment ink (primer coating), a 2D unfolded image of the missing part of the patient's body can then be printed on it, which can simplify the process and reduce process costs.

[0225] S5, the main computer controls the assembly equipment to attach the skin to the surface of the prosthesis body, and performs surface treatment to obtain a completed prosthesis.

[0226] As mentioned above, the various parts of the prosthesis body have been assembled. At this time, the skin is directly attached to the surface of the prosthesis body to obtain a prepared prosthesis, that is, a finished prosthesis.

[0227] In an optional embodiment, the main computer controls the assembly equipment to attach the skin to the surface of the prosthesis body, which may include:

[0228] The main computer controls the printing system to apply glue and lubricant to the surface of the prosthesis body, covers the skin on the coated surface of the prosthesis body, and controls the uniform pressurizing instrument to perform pressurization so that the skin and the prosthesis body are tightly fitted to form an integral structure.

[0229] In the embodiment of the present invention, the skin is bonded to the prosthesis body by using an assembly device, and the assembly device may include a robot arm and the like.

[0230] In order to simulate the bonding state of real human skin, the density and adhesion of the glue match the collected patient information; specifically, the glue has the same viscoelasticity as the surface membrane of the missing part and the patient's body. The glue can be resin glue, and the density and adhesion of the resin glue can be pre-adjusted to match the patient's body data.

[0231] The lubricant is a biocompatible lubricant, such as inert oil, to achieve the natural lubrication effect of the human body.

[0232] In order to achieve a highly restored human skin surface effect, the present invention embodiment designs a uniform pressurizing instrument for applying uniform pressure on each part of the prosthesis body wrapped with the skin, so as to ensure that the skin and the prosthesis body fit firmly and avoid internal gaps and surface wrinkles. Figure 5 understand.

[0233] Figure 5 In the specification, Air flow refers to air flow; Air-tight container refers to airtight container; Flexible membrane refers to flexible membrane, i.e., the skin in the embodiment of the present invention; Air-through plate refers to air-through plate; Air / Water compressor refers to air / water compressor;

[0234] The working process of the uniform pressurization instrument may include:

[0235] 1. Place the solid model on a breathable plate and prepare a flexible film above the model. They are placed in a container with an air flow (or water flow) inlet and outlet. Apply a layer of sticky glue on the surface of the solid model or under the flexible film.

[0236] 2. By blowing air or flowing water at high pressure from the inlet, the internal air pressure is changed, and the flexible film is pressed against the surface of the solid model.

[0237] 3. Since in this structure, pressure is uniformly applied to the film, the film can be evenly and tightly adhered to the surface of the solid model.

[0238] Among them, the solid model refers to the prosthesis body, and the flexible membrane refers to the skin.

[0239] After the skin is attached to the surface of the prosthesis body, surface treatment is required to obtain the completed prosthesis.

[0240] Wherein, the surface treatment herein includes:

[0241] The main computer controls the inkjet system to coat the surface of the prosthesis body after the skin is attached with a protective layer similar to the surface of human skin.

[0242] The inkjet printing system can be implemented based on existing equipment or a combination thereof, as long as it can achieve the above functions, and there is no limitation here.

[0243] The protective layer similar to the human skin surface can be predetermined based on relevant information about the patient's skin surface. After the protective layer is applied, the outermost layer of the prosthesis has the same texture, diffuse reflection, and water / oil affinity as the real patient's skin, as well as sufficient durability for daily use. The material of the protective layer is not limited, and can be silicone-based resin, etc.

[0244] It can be understood that the embodiment of the present invention controls the operation of the remaining components by the main computer, performs data flow monitoring and control, and realizes the automation of the following process: construction of a 3D model containing the segmentation information of each tissue in the patient's missing part → 2D unfolded image of the surface of the 3D model → printing the 3D model to obtain the prosthesis body → manufacturing the surface membrane of the missing part → printing the 2D unfolded image onto the surface membrane of the missing part to form a skin → attaching the skin to the surface of the prosthesis body, and obtaining the prepared prosthesis after surface treatment.

[0245] In this process, the robotic arm controlled by the main computer can be used to move and manipulate the necessary equipment, tools and various parts of the prosthesis. The remaining components controlled by the main computer are not limited to the robotic arm, as well as the printing system including 3D printing equipment, 2D printing equipment, film making equipment, assembly equipment, depth information acquisition equipment, image acquisition equipment, uniform pressurization equipment, inkjet printing system, etc. The main computer can also control the data flow transmission of RGBD images, three-dimensional internal structure point data sets, data required for artificial intelligence networks (such as training data, etc.), data required for three-dimensional printing (OBJ, STL and other engineering files), robot arm control assembly and surface treatment, etc.

[0246] The fully computer-intelligent high-similarity prosthesis production method provided in the embodiment of the present invention computerizes and intelligentizes all the above-mentioned processes as a whole, can provide high-precision, high-similarity prostheses in a shorter time, and realizes digitization of the production process through computer operation, no longer needs to rely on manual experience or skills, and can greatly reduce working time and costs.

[0247] Moreover, the embodiment of the present invention introduces artificial intelligence technology and combines patient information for the design and modeling of the prosthesis. The use of artificial intelligence technology can not only obtain a high-precision and high-similarity 3D model containing the segmentation information of each tissue in the missing part of the patient for 3D printing output, but also realize the segmentation of the internal structure of the limb (such as bones, muscles, tendons, fat layers, skin layers, etc.) and the reconstruction of the missing part structure. It can also obtain a highly realistic 2D unfolded image of the surface layer of the 3D model for 2D printing output.

[0248] Through artificial intelligence algorithms, not only can the design and production of prostheses be completed using the healthy reference side as a data reference when there is a symmetrical healthy reference side for the missing part, but also, even when two symmetrical body parts (hands, arms, legs, feet, ears, etc.) are missing, limb restoration can be achieved, and a prosthesis model can be generated when a single part (nose, chin, etc.) is missing. Compared with traditional methods, prostheses can be tailored for individual patients more effectively, so that the prosthesis has a high degree of similarity with the patient's body, which can improve the quality of prosthetic products. The production of prostheses for missing parts is no longer limited, and the needs of more patients can be met, which improves the practicality of this method.

[0249] In a second aspect, corresponding to the above method embodiment, the embodiment of the present invention further provides a fully computer-intelligent high-similarity pseudobody production system, comprising:

[0250] A main computer and other components; the main computer controls the other components to complete the steps of the fully computer-intelligent high-similarity pseudobody production method described in the first aspect; the other components include an image acquisition and storage system, a film-making device, an assembly device, a printing system, a uniform pressurizing instrument, and a spray painting system; the image acquisition and storage system includes a depth information acquisition device and an image acquisition device; the printing system includes at least a 3D printing device and a 2D printing device.

[0251] In combination with the foregoing, it can be understood that the embodiments of the present invention can complete the production of prostheses for the missing part in the case where the missing part has a symmetrical healthy reference side and in the case where the missing part has no symmetrical healthy reference side. In order to more intuitively understand the working process of the fully computer intelligent high-similarity prosthesis production system in these two cases, the following is combined with the content of the first aspect of the foregoing to Figure 6a and Figure 6b As an example, a brief description is given. Figure 6a and Figure 6b The entire system for the digital production of prostheses (hands, arms, legs, breasts, ears, etc.) is shown. Figure 6a For cases where there is a symmetrical healthy reference side for the missing part, that is, when one side of the symmetrical body part is preserved; Figure 6bThis is the case when the missing part has no symmetrical healthy reference side, that is, the corresponding situation when both sides are missing or have no symmetrical parts.

[0252] against Figure 6a and Figure 6b , the computer legend above represents the main computer, and in the workflow below, each link involves some other components, which are not shown for simplicity. During the working process of the whole system, the main computer controls the other components and performs the deployment control of the corresponding data flow.

[0253] against Figure 6a and Figure 6b , the parts after digital 3D modeling are the same. The following is a brief description of each link.

[0254] Digital 3D modeling: On the one hand, it outputs a 3D model containing the segmentation information of each tissue in the missing part of the patient, and provides it to the 3D printing device to realize the subsequent 3D skeleton structure printing; on the other hand, it outputs a 2D unfolded image of the surface of the 3D model and provides it to the 2D printing device to realize the subsequent 2D surface printing. For details of this part, please refer to the relevant content in the previous S1.

[0255] Among them, for Figure 6a , digital 3D modeling is completed using depth images and surface images. The depth image here refers to the depth information collected from the patient's healthy reference side and the remaining part of the missing part; the surface image refers to the collection of multiple surface images of the healthy reference side at different angles; then the collected depth information is used to perform three-dimensional reconstruction of the missing part to obtain a reconstructed three-dimensional structure model;

[0256] against Figure 6b , digital 3D modeling is completed using generative AI, where generative AI mainly refers to using the pre-built first model to reconstruct the neural network, collect information about the rest of the patient's body parts to perform three-dimensional reconstruction of the missing part, obtain a reconstructed three-dimensional structural model, and obtain a reconstructed surface image of the missing part;

[0257] In the traditional prosthesis production process, the designer will refer to the remaining body parts to make symmetrical parts, such as arms, legs, hands, feet, breasts and ears. However, a completely accurate symmetrical appearance does not look natural. Therefore, the designer will make a slightly asymmetrical modification to the appearance to make it more natural. This modification requires the craftsman's superb skills and long-term experience. In the embodiment of the present invention, the deep learning principle can be used to make such modifications. In the data management of the digital system, the present invention uses a large human anatomy database to train a neural network system. The network can output a natural-looking prosthesis based on the asymmetry of the body parts. In addition, the network of the present invention has a significant advantage. Using the principle of generative artificial intelligence, the network can generate missing body parts without referring to the residual parts. Therefore, prostheses without symmetrical structures, such as noses, chins, etc., can be generated. In addition, prostheses that are missing on both sides, such as losing legs, can also be generated.

[0258] Then, for both cases, the internal tissue structure segmentation of the three-dimensional model of the reconstructed missing part is performed using the pre-constructed second structure hierarchical neural network, and the model is modified using the model attribute related information in the collected patient information and the pre-constructed third model modification neural network to obtain a 3D model after processing; and, the corresponding surface image of the missing part in each case is expanded into a two-dimensional image to obtain a 2D expanded image.

[0259] For specific details, please refer to the previous description and no detailed explanation will be given here.

[0260] 3D skeleton structure printing: Please refer to the previous section S2 for this part, in which the main computer controls the 3D printing device to print the 3D model to obtain the prosthesis body.

[0261] 2D surface printing: For this part, please refer to the above S3 and S4. First, the main computer controls the film-making device to produce a surface film of the missing part; then the main computer controls the 2D printing device to print the 2D unfolded image onto the surface film of the missing part to form a skin;

[0262] Overall structural assembly and surface treatment: For this part, please refer to the previous section S5. The main computer controls the assembly equipment to attach the skin to the surface of the prosthesis body, and performs surface treatment to obtain the completed prosthesis.

[0263] For the specific processing process of each module of the system, please refer to the relevant content of the first aspect and will not be repeated here.

[0264] The fully computer-intelligent high-similarity prosthesis production system provided by the embodiment of the present invention computerizes and intelligentizes the prosthesis production process as a whole, can provide high-precision and high-similarity prostheses in a shorter time, and realizes the digitization of the production process through computer operation, no longer needs to rely on manual experience or skills, and can greatly reduce working time and cost. Moreover, the system introduces artificial intelligence technology and combines patient information for the design and modeling of the prosthesis, which can not only complete the prosthesis design and production by using the healthy reference side as a data reference when the missing part has a symmetrical healthy reference side, but also, even when two symmetrical body parts (hands, arms, legs, feet, ears, etc.) are missing, the restoration of the limbs can be achieved, and a prosthesis model can be generated when a single part (nose, chin, etc.) is missing. Compared with traditional methods, it can more effectively tailor prostheses for individual patients, so that the prosthesis has a high degree of similarity with the patient's body, and the quality of the prosthesis product can be improved. The production of prostheses for missing parts can be no longer limited, and more patient needs can be met, which improves practicality.

[0265] In the third aspect, corresponding to the fully computer-intelligent high-similarity pseudobody production method provided in the first aspect, an embodiment of the present invention further provides a computer-readable storage medium, in which a computer program is stored, and when the computer program is executed by a processor, the steps of any fully computer-intelligent high-similarity pseudobody production method provided in the first aspect of the embodiment of the present invention are implemented.

[0266] For specific details, please refer to the fully computer-intelligent high-similarity pseudobody production method provided in the first aspect, which will not be elaborated here.

[0267] It should be noted that, in the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0268] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification.

[0269] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system and storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0270] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A fully computer-intelligent method for producing a highly similar pseudobody, characterized in that: include: The main computer collects patient information and uses a pre-built artificial intelligence model to obtain a 3D model containing segmentation information of each tissue in the patient's missing part, and obtains a 2D unfolded image of the surface of the 3D model based on the collected patient information and the 3D model; wherein each tissue includes bones, muscles, tendons, fat layers, and skin layers; The host computer controls the 3D printing device to print the 3D model to obtain the prosthesis body; The main computer controls the film-making equipment to make a surface film at the missing part; The main computer controls the 2D printing device to print the 2D expanded image onto the surface film of the missing part to form a skin; The main computer controls the assembly equipment to attach the skin to the surface of the prosthesis body, and obtains the prepared prosthesis after surface treatment.

2. The fully computer-intelligent high-similarity pseudobody production method according to claim 1, characterized in that: In case that the missing part has no symmetrical healthy reference side, the host computer acquires patient information and uses a pre-built artificial intelligence model to obtain a 3D model containing segmentation information of each tissue in the missing part of the patient, and acquires a 2D unfolded image of the surface layer of the 3D model based on the acquired patient information and the 3D model, including: The host computer uses the pre-constructed first model to reconstruct the neural network, collects information about the remaining parts of the patient, performs three-dimensional reconstruction of the missing part, obtains a reconstructed three-dimensional structural model, and obtains a reconstructed surface image of the missing part; wherein the first model reconstruction neural network is implemented using generative AI technology; The three-dimensional structure model is subjected to structural segmentation of internal tissues using a pre-constructed second structural hierarchical neural network, and the model is modified using model attribute related information in the collected patient information and a pre-constructed third model modification neural network, to obtain a 3D model after processing; The surface image reconstructed from the missing part is expanded in two dimensions according to the 3D model to obtain a 2D expanded image.

3. The fully computer-intelligent high-similarity pseudobody production method according to claim 1, characterized in that: In the case where the missing part has a symmetrical healthy reference side, the host computer acquires patient information and uses a pre-built artificial intelligence model to obtain a 3D model containing segmentation information of each tissue in the missing part of the patient, and acquires a 2D unfolded image of the surface of the 3D model based on the acquired patient information and the 3D model, including: The host computer collects depth information of the healthy reference side and the remaining part of the missing part by means of a depth information collection device, and collects a plurality of surface images of the healthy reference side at different angles by means of an image collection device; The acquired depth information is used to perform three-dimensional reconstruction of the missing part to obtain a reconstructed three-dimensional structure model; The three-dimensional structure model is subjected to structural segmentation of internal tissues using a pre-constructed second structural hierarchical neural network, and the model is modified using model attribute related information in the collected patient information and a pre-constructed third model modification neural network, to obtain a 3D model after processing; The collected multiple surface images are expanded in two dimensions according to the 3D model to obtain a 2D expanded image.

4. The fully computer-intelligent high-similarity pseudobody production method according to claim 2 or 3, characterized in that: The model attribute related information includes: The patient's age, gender, height, weight and skin color; the model modification includes modifying the prosthesis size, shape, surface color and surface texture.

5. The fully computer-intelligent high-similarity pseudobody production method according to claim 2 or 3, characterized in that: The process of obtaining the 2D unfolded image is to use a preset reverse filtering algorithm to compare the digital image of the surface of the 3D model with the printed image obtained by using a 2D printing device, compensate for the real visual difference between the digital image and its printed image, and obtain the 2D unfolded image to be printed; wherein the digital image includes: when the missing part has no symmetrical healthy reference side, the image of the surface image reconstructed from the missing part is unfolded in two dimensions according to the 3D model, or when the missing part has a symmetrical healthy reference side, the image of multiple collected surface images is unfolded in two dimensions according to the 3D model.

6. The fully computer-intelligent high-similarity pseudobody production method according to claim 5, characterized in that: The process of the preset reverse filtering algorithm includes: Step c1, reading initial parameters; wherein the initial parameters include image matrix size, similarity threshold and maximum number of iterations; Step c2, acquiring the digital image; Step c3, printing the digital image using a 2D printing device to obtain an original printed image; Step c4, using the same image acquisition device used in the process of obtaining the digital image, performing image acquisition on the digital image and the original printed image to obtain corresponding acquired images respectively; Step c5, calculating the reverse filtering function using the collected images corresponding to the current digital image and the original printed image respectively; Step c6, applying the reverse filtering function to the current digital image to obtain a reverse filtered digital image; Step c7, determining whether the current number of iterations is less than the maximum number of iterations; if so, executing step c8, if not, executing step c14; Step c8, printing the currently obtained reverse filtered digital image using a 2D printing device to obtain a first printed image; Step c9, using the same image acquisition device, performing image acquisition on the currently obtained reverse filtered digital image and the first printed image to obtain corresponding acquired images respectively; Step c10, calculating a similarity index for the acquired images corresponding to the currently obtained reverse filtered digital image and the first printed image respectively; Step c11, determining whether the similarity index is greater than the similarity threshold; if so, executing step c12, if not, executing step c13; Step c12, determining the currently obtained reverse filtered digital image as the 2D expanded image to be printed; Step c13, storing the currently obtained reverse filtered digital image and the similarity index, and replacing the current digital image with the currently obtained reverse filtered digital image, replacing the captured image corresponding to the current digital image with the captured image corresponding to the currently obtained reverse filtered digital image, and replacing the captured image corresponding to the current original printed image with the captured image corresponding to the current first printed image; Step c14, searching for the reverse filtered digital image with the largest similarity index in all iterations, and using it as the 2D expanded image to be printed.

7. The fully computer-intelligent high-similarity pseudobody production method according to claim 1, characterized in that: The host computer controls the 3D printing device to print the 3D model to obtain the prosthesis body, including: The main computer controls the 3D printing device to print with materials matching each tissue in the 3D model to obtain a prosthesis body; wherein the materials matching each tissue include a variety of optional materials.

8. The fully computer-intelligent high-similarity pseudobody production method according to claim 7, characterized in that: For each tissue, the matching material determines a suitable one from a variety of optional materials for the tissue based on the collected patient information, so that it has the hardness and viscoelasticity most similar to the patient's real human body.

9. The fully computer-intelligent high-similarity pseudobody production method according to claim 1, characterized in that: The main computer controls the film-making equipment to make a surface film of the missing part, including: The main computer controls the film-making device to obtain a sol-state silicone rubber solution, and places it on a printing paper with a preset thickness and a preset surface roughness to form a silicone rubber solution layer with a preset thickness; The main computer controls the film-making equipment to cover the upper surface of the silicone rubber solution layer with a substrate sheet, and after the silicone rubber solution layer forms a gel, peels off the silicone rubber solution layer to obtain a surface film of the missing part, and the original contact surface between the surface film of the missing part and the printing paper serves as the printing surface; wherein the bonding strength of the surface of the substrate sheet to the silicone rubber is higher than the bonding strength of the surface of the printing paper to the silicone rubber, and the printing paper is suitable for the 2D printing equipment.

10. The fully computer-intelligent high-similarity pseudobody production method according to claim 9, characterized in that: The preset surface roughness is determined according to the roughness of the patient's skin surface in the collected patient information; the preset thickness is determined according to the thickness of the patient's skin in the collected patient information.

11. The fully computer-intelligent high-similarity pseudobody production method according to claim 1, characterized in that: Before the host computer controls the 2D printing device to print the 2D expanded image onto the surface film of the missing part to form a skin, the method further includes: The main computer controls the printing system to treat the surface of the surface film of the missing part with a surface treatment liquid, and evenly prints the pigment ink after the treatment.

12. The fully computer-intelligent high-similarity pseudobody production method according to claim 1, characterized in that: The main computer controls the assembly equipment to attach the skin to the surface of the prosthesis body, including: The main computer controls the printing system to apply glue and lubricant to the surface of the prosthesis body, covers the skin on the coated surface of the prosthesis body, and controls the uniform pressurizing instrument to perform pressurization so that the skin and the prosthesis body are tightly fitted to form an integral structure.

13. The fully computer-intelligent high-similarity pseudobody production method according to claim 12, characterized in that: The density and adhesion of the glue match the collected patient information; the lubricant is a lubricant with biocompatibility.

14. The fully computer-intelligent high-similarity pseudobody production method according to claim 1, characterized in that: The surface treatment comprises: The main computer controls the inkjet system to coat the surface of the prosthesis body after the skin is attached with a protective layer similar to the surface of human skin.

15. A fully computerized intelligent high-similarity pseudobody production system, characterized in that: include: Main computer and other components; The main computer controls the remaining components to complete the method steps described in any one of claims 1-14; the remaining components include an image acquisition and storage system, a film making device, an assembly device, a printing system, a uniform pressurizing instrument, and a spray painting system; the image acquisition and storage system includes a depth information acquisition device and an image acquisition device; the printing system includes at least a 3D printing device and a 2D printing device.

16. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method steps described in any one of claims 1 to 14 are implemented.

Images